(Problem Books in Mathematics) Vladimir V. Tkachuk (Auth.) - A Cp-Theory Problem Book - Special Features of Function Spaces-Springer International Publishing (2014) PDF

Problem Books in Mathematics
Vladimir V.Tkachuk
A Cp-Theory
Problem
Book
Special Features of Function Spaces
Problem Books in Mathematics
Series Editors:
Peter Winkler
Department of Mathematics
Dartmouth College
Hanover, NH 03755
USA
For further volumes:

http://www.springer.com/series/714
Vladimir V. Tkachuk
A Cp-Theory Problem Book

Special Features of Function Spaces
123
Vladimir V. Tkachuk
Departamento de Matematicas
Universidad Autonoma Metropolitana-Iztapalapa
San Rafael Atlixco, Mexico City, Mexico
ISSN 0941-3502
ISBN 978-3-319-04746-1 ISBN 978-3-319-04747-8 (eBook)
DOI 10.1007/978-3-319-04747-8
Springer Cham Heidelberg New York Dordrecht London
Library of Congress Control Number: 2014933677
Mathematics Subject Classification (2010): 54C35
Springer International Publishing Switzerland 2014

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publishers location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
This is the second volume of the series of books of problems in Cp -theory

entitled A Cp -Theory Problem Book, i.e., this book is a continuation of the first
volume subtitled Topological and Function Spaces. The series was conceived as an
introduction to Cp -theory with the hope that each volume will also be used as a
reference guide for specialists.
The first volume provides a self-contained introduction to general topology and
Cp -theory and contains some highly nontrivial state-of-the-art results. For example,
Sect. 1.4 presents Shapirovskys theorem on the existence of a point-countable -
base in any compact space of countable tightness and Sect. 1.5 brings the reader
to the frontier of the modern knowledge about realcompactness in the context of
function spaces.
This present volume introduces quite a few topics from scratch but dealing
with topology and Cp -theory is already a professional endeavour. The objective
is to study the behaviour of general topological properties in function spaces and
establish the results on duality of cardinal functions and classes with respect to
the Cp -functor. The respective background includes a considerable amount of top-
notch results both in topology and set theory; the authors obsession with keeping
this work self-contained implied that an introduction to advanced set theory had to
be provided in Sect. 1.1. The methods developed in this section made it possible to
present a very difficult example of Todorcevic of a compact strong S -space.
Of course, it was impossible to omit the famous Baturovs theorem on coinci-
dence of the Lindelf number and extent in subspaces of Cp .X / for any Lindelf
-space X and the result of Christensen on -compactness of X provided that
Cp .X / is analytic. The self-containment policy of the author made it obligatory for
him to give a thorough introduction to Lindelf -spaces in Sect. 1.3 and to the
descriptive set theory in Sect. 1.4.
We use all topological methods developed in the first volume, so we refer to its
problems and solutions when necessary. Of course, the author did his best to keep
every solution as independent as possible, so a short argument could be repeated
several times in different places.
v
vi Preface
The author wants to emphasize that if a postgraduate student mastered the

material of the first volume, it will be more than sufficient to understand every
problem and solution of this book. However, for a concrete topic, much less might
be needed. Finally, let me outline some points which show the potential usefulness
of the present work:
The only background needed is some knowledge of set theory and real numbers;
any reasonable course in calculus covers everything needed to understand this
book.
The student can learn all of general topology required without recurring to any
textbook or papers; the amount of general topology is strictly minimal and is
presented in such a way that the student works with the spaces Cp .X / from the
very beginning.
What is said in the previous paragraph is true as well if a mathematician working
outside of topology (e.g., in functional analysis) wants to use results or methods
of Cp -theory; he (or she) will find them easily in a concentrated form or with full
proofs if there is such a need.
The material we present here is up to date and brings the reader to the frontier
of knowledge in a reasonable number of important areas of Cp -theory.
This book seems to be the first self-contained introduction to Cp -theory. Although
there is an excellent textbook written by Arhangelskii (1992a), it heavily depends
on the readers good knowledge of general topology.
Mexico City, Mexico Vladimir V. Tkachuk

Contents
1 Duality Theorems and Properties of Function Spaces . . . . . . . . . . . . . . . . . . . 1

1.1 Some Additional Axioms and Hereditary Properties .. . . . . . . . . . . . . . . . . 2
1.2 Monolithity, Stability and Their Generalizations ... . . . . . . . . . . . . . . . . . . . 11
1.3 Whyburn Spaces, Calibers and Lindelf -Property . . . . . . . . . . . . . . . . . 19
1.4 A Glimpse of Descriptive Set Theory .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 27
1.5 Additivity of Properties: Mappings Between Function Spaces . . . . . . . 35
1.6 Bibliographic Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 42
2 Solutions of Problems 001500 . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 45
3 Bonus Results: Some Hidden Statements . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 527
3.1 Standard Spaces .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 528
3.2 Metrizable Spaces .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 529
3.3 Compact Spaces and Their Generalizations . . . . . . .. . . . . . . . . . . . . . . . . . . . 530
3.4 Properties of Continuous Maps. . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 531
3.5 Completeness and Convergence Properties .. . . . . . .. . . . . . . . . . . . . . . . . . . . 532
3.6 Product Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 533
3.7 Cardinal Invariants and Set Theory . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 535
3.8 Raznoie (Unclassified Results) . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 536
4 Open Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 539
4.1 Analyticity and Similar Properties . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 540
4.2 Whyburn Property in Function Spaces . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 541
4.3 Uniformly Dense Subspaces . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 542
4.4 Countable Spread and Similar Properties .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . 544
4.5 Metacompactness and Its Derivatives .. . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 547
4.6 Mappings Which Involve Cp -Spaces . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 551
4.7 Additivity of Topological Properties . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 552
4.8 Raznoie (Unclassified Questions) . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 554
vii
viii Contents
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 557
List of Special Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 573
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 5 7 7
Detailed Summary of Exercises
1.1. Tampering with Additional Axioms: Some Hereditary Properties

Equivalent conditions for hl.X / . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 001.
Characterizing hereditarily/perfectly normal spaces . . . . . . . . . . Problems 002003.
Right-separated and left-separated spaces . . . . . . . . . . . . . . . . . . . Problems 004009.
Invariants s; hl; hd in finite and countable powers . . . . . . . . . . Problems 010013.
Spread of X X is an upper bound for minfhd.X /; hl.X /g . . . . . . . Problem 014.
A bound on cardinality of X using hl.X / . . . . . . . . . . . . . . . . . . . . . . . . . Problem 015.
.X X / '.Cp .X // if ' 2 fs; hl; hd g and is its dual . . . . Problems 016018.
Hedgehogs and duality theorems for s ; hl ; hd . . . . . . . . . . . . Problems 019035.
Properties s ! and hl ! can be multiplicative in Cp .X / . . . . . . . Problem 036.
Non-separable spaces of countable spread . . . . . . . . . . . . . . . . . . . . . . . . Problem 037.
The Delta-lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 038.
CH constructions for distinguishing s; hl and hd . . . . . . . . . . . . . Problems 039040.
Compact spaces and their points of ! under CH . . . . . . . . . . . . . . Problem 041.
There are P -points in !n! under CH . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 042.
Some results on Luzin spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 043046.
Martins axiom and its applications . . . . . . . . . . . . . . . . . . . . . . . . . Problems 047063.
Stationary and club subsets of !1 . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 064067.
Trees and Jensens axiom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 068071.
Souslin tree and Souslin continuum . . . . . . . . . . . . . . . . . . . . . . . . . Problems 072076.
Condensing onto hereditarily separable spaces . . . . . . . . . . . . . . . Problems 077079.
Around perfect normality in function spaces . . . . . . . . . . . . . . . . . Problems 080089.
Metrizability and perfect normality of compact spaces . . . . . . . . Problems 090096.
Preparing ground for example of a compact strong S -space . . . . . . . . .Problem 097.
Todorcevics example and its applications . . . . . . . . . . . . . . . . . . . Problems 098100.
1.2. Monolithity, Stability and Their Generalizations
Monolithity and stability in metrizable spaces . . . . . . . . . . . . . . . Problems 101106.
Monolithity and stability in general spaces . . . . . . . . . . . . . . . . . . Problems 107126.
Scattered spaces, P -spaces and -simple spaces . . . . . . . . . . . . . Problems 127130.
Versions of !-stability and to pseudocompactness . . . . . . . . . . . . . . . . . Problem 131.
ix
x Detailed Summary of Exercises
Versions of !-monolithity in Cp .X / implying that X is finite . . . . . . . Problem 132.

Scattered spaces and FrchetUrysohn property in Cp .X / . . . . Problems 133136.
On !-relativization of Cech-completeness in Cp .X / . . . . . . . . . Problems 137138.
On !-relativization of normality in Cp .X / . . . . . . . . . . . . . . . . . . Problems 139142.
Stability and monolithity in Cp .X / vs those in Cp .X; I/ . . . . . . Problems 143144.
.!/-stability in X and X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 145.
Some generic duality theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 146151.
Some applications of generic duality theorems . . . . . . . . . . . . . . . Problems 152177.
'.!/-monolithity in Cp .X / for ' 2 fi w; ; g. . . . . . . . . . . . . . Problems 178180.
Duality for relativizations of tightness and similar properties . . Problems 181184.
Relativizations of tightness and FrchetUrysohn property . . . . Problems 185186.
A dual property for being -scattered . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 187.
A property of Cp .X / for Hurewicz d.!/-stable spaces X . . . . . . . . . Problem 188.
Some monolithity properties and countable spread . . . . . . . . . . . Problems 189197.
Free sequences in Cp .X / under MAC:CH . . . . . . . . . . . . . . . . . . . . . . . Problem 198.
!-monolithity and countable spread in X under MAC:CH . . . . . . . . Problem 199.
Hereditary stability of Cp .X / can imply nw.X / D ! . . . . . . . . . . . . . . Problem 200.
1.3. Whyburn Spaces, Calibers and Lindelf -Property
Two statements about hereditarily normal compact spaces . . . . Problems 201202.
If X is compact and X 2 n is paracompact, then X is metrizable . . . . Problem 203.
Whyburn spaces and weakly Whyburn spaces . . . . . . . . . . . . . . . Problems 204213.
Hereditarily k-spaces are FrchetUrysohn . . . . . . . . . . . . . . . . . . . . . . . Problem 214.
More of Whyburn spaces and weakly Whyburn spaces . . . . . . . Problems 215220.
p-spaces vs -spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 221232.
Some characterizations of Lindelf -property by extensions . . . . . . Problem 233.
-property in Cp .X / . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problems 234239.
Compact-valued maps and Lindelf -property . . . . . . . . . . . . . Problems 240242.
Categorical properties of Lindelf -spaces . . . . . . . . . . . . . . . . . Problems 243259.
Lindelf -spaces and with small compact subsets . . . . . . . . . . Problems 260262.
Baire property together with Lindelf (-)property of Cp .X / . Problems 263265.
Stability in Lindelf -spaces and their products . . . . . . . . . . . . Problems 266268.
Baturovs theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Problem 269.
Hereditary Lindelf -property implies countable netweight . .Problems 270274.
Basic properties of calibers and precalibers . . . . . . . . . . . . . . . . . . Problems 275287.
The Souslin property and caliber !1 under additional axioms . . Problems 288289.
Duality results for small diagonals and calibers . . . . . . . . . . . . . . Problems 290293.
Compact spaces and Lindelf -spaces with small diagonal . . Problems 294300.
1.4. A Glimpse of Descriptive Set Theory
Zero-dimensional spaces and their properties . . . . . . . . . . . . . . . . Problems 301314.
A theorem on continuous selections and its applications . . . . . . Problems 315317.
Universal Borel sets of given class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 318.
Borel complexity of subsets of uncountable Polish spaces . . . . . Problems 319321.
A countable space not embeddable into any Cp .Borel set/ . . . . . . . . . . Problem 322.
Detailed Summary of Exercises xi
Absolute F - and absolute G -spaces . . . . . . . . . . . . . . . . . . . . . . . Problems 323324.

Open, closed and perfect maps on Polish spaces . . . . . . . . . . . . . Problems 325329.
Capturing all Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 330.
Categorical properties of Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . Problems 331333.
Some trivial properties of analytic spaces . . . . . . . . . . . . . . . . . . . Problems 334337.
Inverse images of analytic sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 338.
Separation theorem and Borel sets as analyticCcoanalytic . . . . Problems 339340.
Condensations of Borel sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 341.
A non-analytic subspace of the reals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Problem 342.
Some properties of K-analytic spaces . . . . . . . . . . . . . . . . . . . . . . . Problems 343346.
Axiomatic characterizations for the Cantor set, Q and RnQ . . . . Problems 347349.
Q is a universal space for countable metrizable spaces . . . . . . . . . . . . . Problem 350.
Closed embeddings of Q in non-Polish spaces . . . . . . . . . . . . . . . . . . . . Problem 351.
Closed embeddings in analytic spaces . . . . . . . . . . . . . . . . . . . . . . Problems 352353.
Pytkeevs theorem on condensations of Borel sets . . . . . . . . . . . . . . . . . Problem 354.
Condensations and continuous maps onto nice spaces . . . . . . . . Problems 355359.
-compactness of X and analyticity of Cp .X / . . . . . . . . . . . . . . . Problems 360370.
A countable space X with Cp .X / non-analytic . . . . . . . . . . . . . . . . . . . . Problem 371.
Countable spaces X with Cp .X / of any Borel complexity . . . . . . . . . . Problem 372.
Some results on absolute F -property . . . . . . . . . . . . . . . . . . . . . . . Problem 373375.
Embedding second countable spaces into Cp .compactum/ . . . . Problems 376378.
The functions of first Baire class and Rosenthal compacta . . . . .Problems 379387.
K-analytic spaces and P-directed compact covers . . . . . . . . . . . . Problems 388394.
K-analytic spaces with small compact subsets . . . . . . . . . . . . . . . . . . . . Problem 395.
P-directed sets and metrization of a compact space . . . . . . . . . . . . . . . . Problem 396.
K-analyticity in general and in spaces Cp .X / . . . . . . . . . . . . . . . Problems 397400.
1.5. Additivity of Properties: Mappings Between Function Spaces
Some properties are not finitely/countably additive. . . . . . . . . . . Problems 401407.
Additivity in squares and countable powers . . . . . . . . . . . . . . . . . . Problems 408418.
A metric which generates the uniform topology on C .X / . . . . . . . . . Problem 419.
Finite additivity in Cp .X /: reduction to dense summands . . . . . Problems 420421.
Some properties are finitely additive in Cp .X /. . . . . . . . . . . . . . . Problems 422430.
Decompositions of Cp .X / into countable unions . . . . . . . . . . . . . Problems 431432.
Closed countable additivity of hereditary properties . . . . . . . . . . . . . . . Problem 433.
Some properties are countably additive in arbitrary Cp .X /. . . . Problems 434450.
Some properties are countably additive in nice Cp .X /. . . . . . Problems 451455.
Some results on uniformly dense subspaces of Cp .X /. . . . . . . . Problems 456461.
Extent of X and tightness of compact subspaces of Cp .X /. . . . . . . . Problem 462.
Around the theorem of GerlitsPytkeev . . . . . . . . . . . . . . . . . . . . . Problems 463466.
Extending continuous maps from X to Cp .Cp .X // . . . . . . . . . . Problems 467469.
When Cp .X / maps continuously onto Cp .Y /? . . . . . . . . . . . . . . .Problems 470491.
Mapping between products of metric spaces and Cp .X / . . . . . . Problems 493499.
If Cp .X / is a retract of a G -subset of R , then X is discrete . . . . . . Problem 500.
Introduction
The term Cp -theory was invented to abbreviate the phrase The theory of function
spaces endowed with the topology of pointwise convergence. The credit for the
creation of Cp -theory must undoubtedly be given to Alexander Vladimirovich
Arhangelskii. The author is proud to say that Arhangelskii also was the person
who taught him general topology and directed his Ph.D. thesis. Arhangelskii was
the first to understand the need to unify and classify a bulk of heterogeneous results
from topological algebra, functional analysis and general topology. He was the first
to obtain crucial results that made this unification possible. He was also the first
to formulate a critical mass of open problems which showed this theorys huge
potential for development.
Later, many mathematicians worked hard to give Cp -theory the elegance and
beauty it boasts nowadays. The author hopes that the work he presents for the
readers judgement will help to attract more people to this area of mathematics.
The main text of this volume consists of 500 statements formulated as problems;
it constitutes Chap. 1. These statements provide a gradual development of many
popular topics of Cp -theory to bring the reader to the frontier of the present-day
knowledge. A complete solution is given to every problem of the main text.
The material of Chap. 1 is divided into five sections with 100 problems in each
one. The sections start with an introductory part where the definitions and concepts
to be used are given. The introductory part of any section never exceeds two pages
and covers everything that was not defined previously. Whenever possible, we try
to save the reader the effort of ploughing through various sections, chapters and
volumes, so we give the relevant definitions in the current section not caring much
about possible repetitions.
Chapter 1 ends with some bibliographical notes to give the most important
references related to its results. The selection of references is made according to
the authors preferences and by no means can be considered complete. However,
a complete list of contributors to the material of Chap. 1 can be found in our
bibliography of 300 items. It is my pleasant duty to acknowledge that I consulted
the paper of Arhangelskii (1998a) to include quite a few of its 375 references in my
bibliography.
xiii
xiv Introduction
Sometimes, as we formulate a problem, we use without reference definitions and

constructions introduced in other problems. The general rule is to try to find the
relevant definition not more than ten problems before.
The first section of Chap. 1 deals with hereditary properties in Cp .X /. To under-
stand the respective results, the reader needs a topological background including
the ability to manage additional axioms of ZFC and apply strong and difficult
methods of set theory. The pursuit of self-containment obliged the author to give
an introduction to advanced set theory. In this section the reader can find the
applications of continuum hypothesis, Martins axiom, Jensens axiom, Souslin trees
and Luzin spaces.
The non-Cp material presented in Chap. 1 also includes an introduction to
descriptive set theory and Lindelf -spaces. This helped to keep this work self-
contained when we gave the proofs of Baturovs theorem on Cp .X / for a Lindelf
-space X and Christensens theorem on -compactness of X provided that Cp .X /
is analytic. There are many topics in Chap. 1 which are developed up to the
frontier of the present-day knowledge. In particular, Sect. 1.5 includes the famous
GerlitsPytkeev theorem about coincidence of the FrchetUrysohn property and
k-property in any space Cp .X /.
The complete solutions of all problems of Chap. 1 are given in Chap. 2. Chapter 3
begins with a selection of 100 statements which were proved as auxiliary facts in
the solutions of the problems of the main text. This material is split into six sections
to classify the respective results and make them easier to find. Chapter 4 consists
of 100 open problems presented in ten sections with the same idea: to classify this
bulk of problems and make the readers work easier.
Chapter 4 also witnesses an essential difference between the organization of our
text and the book of Arhangelskii and Ponomarev (1974): we never put unsolved
problems in the main text as is done in their book. All problems formulated in
Chap. 1 are given complete solutions in Chap. 2 and the unsolved ones are presented
in Chap. 4.
There is little to explain about how to use this book as a reference guide. In this
case the methodology is not that important and the only thing the reader wants is
to find the results he (or she) needs as fast as possible. To help with this, the titles
of chapters and sections give the first approximation. To better see the material of a
chapter, one can consult the second part of the Contents section where a detailed
summary is given; it is supposed to cover all topics presented in each section.
Besides, the index can also be used to find necessary material.
To sum up the main text, I believe that the coverage of Cp -theory will be
reasonably complete and many of the topics can be used by postgraduate students
who want to specialize in Cp -theory. Formally, this book can also be used as
an introduction to general topology. However, it would be a somewhat biased
introduction, because the emphasis is always given to Cp -spaces and the topics are
only developed when they have some applications in Cp -theory.
To conclude, let me quote an old saying which states that the best way for one to
learn a theorem is to prove it oneself. This text provides a possibility to do this. If
the readers wish is to read the proofs, then they are concentrated immediately after
the main text.
Chapter 1
Duality Theorems and Properties
of Function Spaces
This chapter presents some fundamental aspects of set theory, descriptive set theory,
general topology and Cp -theory.
Section 1.1 introduces some advanced concepts of set theory. We give the state-
ments and applications of the continuum hypothesis, Martins axiom and Jensens
axiom. The next thing under the study is the behavior of spread, hereditary Lindelf
number and hereditary density in function spaces. The most important results of
this section are the duality theorems for s ; hd and hl (Problems 025030) and
Todorcevics example of a strong S -space (Problem 098).
In Sect. 1.2 we deal with monolithity, stability and their generalizations. The
principal results are presented as several generic theorems on duality between
./-monolithity and ./-stability, formulated in Problems 146151.
Section 1.3 starts with Whyburn spaces and their properties. Next, we introduce
Lindelf -spaces and their most important characterizations. The rest of the
section is devoted to calibers, precalibers and small diagonals. The most important
results include Baturovs theorem on extent in subspaces of Cp .X / for a Lindelf
-space X (Problem 269) and Gruenhages theorem on Lindelf -spaces with a
small diagonal (Problem 300).
In Sect. 1.4 we introduce the basic notions of descriptive set theory and give their
applications to Cp -theory. This section features three main results: Christensens
theorem on analyticity of Cp .X / (Problem 366), Fremlins theorem on K-analytic
spaces whose compact subspaces are metrizable (Problem 395) and Pytkeevs
theorem on condensations of Borel sets (Problem 354).
The first part of Sect. 1.5 comprises some results on decompositions of Cp .X /
into a finite or countable union of subspaces with nice properties. The second
part is devoted to the study of the existence of good mappings between Cp .X /
and Cp .Y / and the simplest implications this has for the spaces X and Y . We also
have two main results in this section: GerlitsPytkeev theorem on k-property in
Cp .X / (Problem 465) and Tkachuks theorem on discreteness of X if Cp .X / is
homeomorphic to a retract of a G -subspace of RX (Problem 500).
V.V. Tkachuk, A Cp-Theory Problem Book: Special Features of Function Spaces, 1

Problem Books in Mathematics, DOI 10.1007/978-3-319-04747-8__1,
2 1 Duality Theorems and Properties of Function Spaces
1.1 Some Additional Axioms and Hereditary Properties
A space X is left-separated (right-separated) if there exists a well-order < on

X such that the set fy 2 X W y < xg is closed (open) in X for any x 2 X .
A space X is scattered if any subspace Y X has an isolated point. Recall that
' .X / D supf'.X n / W n 2 Ng and h'.X / D supf'.Y / W Y X g for any cardinal
invariant '. All results of this book are proved assuming that ZFC axioms hold. The
abbreviation ZFC stands for ZermeloFraenkelChoice. This axiomatic system is
the most accepted one at the present moment. We wont need to have the knowledge
of what the axioms of ZFC exactly say. It is sufficient to know that all they do is
to postulate some very natural properties of sets. For the reader who wants to learn
more, the book of Kunen (1980) is an excellent introduction to this subject.
In the twentieth century topologists and set-theorists discovered that there were
some very natural problems which could not be solved using ZFC axioms only;
to fix this, quite a few additional axioms have been created. Practically all of
those axioms are proved to be consistent with ZFC which means that if ZFC has
no contradiction, then ZFC, together with the axiom in question, does not have
one. In this section we formulate the most popular additional axioms and their
applications. All results of this book are proved in ZFC if no additional assumptions
are formulated explicitly; however, we sometimes emphasize this.
The statement CH (called Continuum Hypothesis) says that the first uncountable
ordinal is equal to the continuum, i.e., !1 D c. The statement C D 2 for any
infinite cardinal is called Generalized Continuum Hypothesis (GCH).
A partial order on a set P is a relation on P with the following properties:
(PO1) p p for any p 2 P;
(PO2) p q and q r imply p r;
(PO3) p q and q p imply p D q.
The pair .P; / is called a partially ordered set. If the order is clear, we will write
P instead of .P; /. Let .P; / be a partially ordered set. The elements p; q 2 P
are called compatible if there is r 2 P such that r p and r q. If p and q are
not compatible, they are called incompatible. A set A P is an antichain if the
elements of A are pairwise incompatible. We say that .P; / has the property ccc
if any antichain of P is countable. A set D P is called dense in P if, for every
p 2 P, there is q 2 D such that q p.
A non-empty set F P is a filter if it has the following properties:
(F1) for any p; q 2 F , there is r 2 F such that r p and r q;
(F2) if p 2 F and p q, then q 2 F .
Given an infinite cardinal , we denote by MA./ the following statement: for
any ccc partial order P and any family D of dense subsets of P with jDj , there
is a filter F P such that F \ D ; for any D 2 D. Now, Martins axiom, MA,
says that MA./ holds for any infinite < c.
A subset C !1 is called club (closed and unbounded) if C is uncountable
and closed in the order topology on !1 . A set S !1 is stationary if S \ C ;
1.1 Some Additional Axioms and Hereditary Properties 3
for any club C !1 . Jensens axiom } is the statement: for each < !1 , there is a
set A such that, for any A !1 , the set f 2 !1 W A \ D A g is stationary.
The principle }C is the following statement: for each 2 !1 , there is a countable
family A exp./ such that, for any A !1 , there is a club C !1 for which
A \ 2 A and C \ 2 A for any 2 C . The sequence fA W < !1 g is called
a }C -sequence.
A space X is called zero-dimensional if X has a base consisting of clopen sets. A
point x 2 X is called a P -point if any countable intersection of neighborhoods of x
is a neighborhood of x. An uncountable dense-in-itself space X is called Luzin (also
written Lusin) if any nowhere dense subspace of X is countable. Say that X is an L-
space if hl.X / D ! < d.X /; if hd.X / D ! < l.X /, then X is called an S -space.
The axiom SA says that there are no S -spaces, i.e., that every regular hereditarily
separable space is Lindelf. Furthermore, X is a strong S -space if hd .X / D ! <
l.X /; if hl .X / D ! < d.X /, then X is called strong L-space.
A tree is a partially ordered set .T ; / such that, for every x 2 T , the set Lx D
fy 2 T W y < xg is well ordered by . We will often write T instead of .T ; /.
If T is a tree and x 2 T , then the height of x in T or ht.x; T / is the ordinal
isomorphic to Lx . For each ordinal , the -th level of T or Lev .T / is the set
fx 2 T W ht.x; T / D g. The height ht.T / of the tree T is the least such that
Lev .T / D ;. A subset T 0 T is called a subtree of T if Lx T 0 for every
x 2 T 0 . A subset C 2 T is called a chain if C is linearly ordered by , i.e.,
every two elements of C are comparable. An antichain of T is a set A T such
that x; y 2 A and x y implies x 6 y and y 6 x. For every infinite cardinal , a
-Souslin tree is a tree T such that jT j D and every chain and every antichain have
cardinality < . An !1 -Souslin tree is called Souslin tree. If is a regular cardinal,
a -tree is a tree of height with levels of cardinality < . A -Aronszajn tree is
a -tree with no chains of cardinality . An !1 -Aronszajn tree is called Aronszajn
tree.
If f W X ! Y and Z X , we denote the restriction of f to Z by f jZ or f jZ.
If we have maps f; g W X ! Y , then f g if the set fx 2 X W f .x/ g.x/g
is finite. Given functions f W X ! Y and g W X1 ! Y1 , we say that f g if
X X1 ; S Y Y1 and gjX D f . Now, ! is the set of all maps from to ! and
! <!1
D f! W < !1 g. Any !1 -sequence fs W < !1 g ! <!1 such that
s 2 ! is an injective map and s j s for all < < !1 is called Aronszajn
coding. Denote by P the set of all monotonically increasing functions from ! ! , i.e.,
P D ff 2 ! ! W f .i / < f .j / whenever i < j g. Given f; g 2 ! ! , we say that
f < g if there exists m 2 ! such that f .n/ < g.n/ for all n m. A sequence
ff W <
g ! ! is called strictly < -increasing if f < f for all < <
.
A set S ! ! is < -cofinal in ! ! if, for any f 2 ! ! , we have f < g for some
g 2 S.
001. Given an infinite cardinal prove that the following properties are equivalent
for any space X :
(i) hl.X / ;
(ii) l.X / and every U 2 .X/ is a union of -many closed subsets
of X ;
(iii) l.X / and every closed F X is a G -set in X ;
(iv) l.U / for any open U X .
In particular, a space X is hereditarily Lindelf if and only if it is Lindelf
and perfect.
002. Prove that a space X is hereditarily normal if and only if any open subspace
of X is normal.
003. Prove that if X is perfectly normal, then any Y X is also perfectly normal.
004. Let X be any space. Prove that hd.X / D supfjAj W A is a left-separated
subspace of X g. In particular, the space X is hereditarily separable if and only
if every left-separated subspace of X is countable.
005. Let X be any space. Prove that hl.X / D supfjAj W A is a right-separated
subspace of X g. In particular, the space X is hereditarily Lindelf if and only
if every right-separated subspace of X is countable.
006. Prove that a space is right-separated if and only if it is scattered.
007. Let X be a left-separated space. Prove that hl.X / s.X /. In particular, any
left-separated space of countable spread is hereditarily Lindelf.
008. Let X be a right-separated space. Prove that hd.X / s.X /. In particular, any
right-separated space of countable spread is hereditarily separable.
009. Prove that any space has a dense left-separated subspace.
010. Suppose that s.X / D !. Prove that X has a dense hereditarily Lindelf
subspace.
011. Prove that for any space X , we have hl .X / D hl.X ! /. In particular, if all
finite powers of X are hereditarily Lindelf, then X ! is hereditarily Lindelf.
012. Prove that for any space X , we have hd .X / D hd.X ! /. In particular, if all
finite powers of X are hereditarily separable, then X ! is hereditarily separable.
013. Prove that for any space X , we have s .X / D s.X ! /.
014. Suppose that s.X X / . Prove that hl.X / or hd.X / . In particular,
if s.X X / D !, then X is hereditarily separable or hereditarily Lindelf.
015. Prove that jX j 2hl.X / for any space X . In particular, any hereditarily
Lindelf space has cardinality c.
016. Prove that s.X X / s.Cp .X // s .X / for any space X .
017. Prove that hd.X X / hl.Cp .X // hd .X / for any space X .
018. Prove that hl.X X / hd.Cp .X // hl .X / for any space X .
019. For an arbitrary n 2 N, let Jn D J.n/ be the hedgehog with n spines. Prove
that s.X n / s.Cp .X; Jn // s.Cp .X / Cp .X // for any space X .
that hd.X n / hl.Cp .X; Jn // hl.Cp .X / Cp .X // for any space X .
that hl.X n / hd.Cp .X; Jn // hd.Cp .X / Cp .X // for any space X .
022. For any space X prove that s.Cp .X / Cp .X // D s .Cp .X //.
023. For any space X prove that hl.Cp .X / Cp .X // D hl .Cp .X //.
024. For any space X prove that hd.Cp .X / Cp .X // D hd .Cp .X //.
025. Prove that s .X / D s .Cp .X // for any space X .
026. Prove that hl .X / D hd .Cp .X // for any space X .
027. Prove that hd .X / D hl .Cp .X // for any space X .
028. For an infinite cardinal , suppose that s.Cp .X // D and .X / . Prove
that s .Cp .X // and hence s .Cp .X // D s.Cp .X //. In particular, if X is
a space with a G -diagonal, then s .Cp .X // D s.Cp .X // D s .X /.
029. For an infinite cardinal , suppose that hl.Cp .X // D and .X / . Prove
that hl .Cp .X // and hence hl .Cp .X // D hl.Cp .X //. In particular, if X
is a space with a G -diagonal, then hl .Cp .X // D hl.Cp .X // D hd .X /.
030. (Velichkos theorem) Prove that hd .Cp .X // D hd.Cp .X // D hl .X / for
any space X .
031. Prove that s.Cp .Cp .X /// D s .X / for any space X .
032. Prove that hd.Cp .Cp .X /// D hd .X / for any space X .
033. Prove that hl.Cp .Cp .X /// D hl .X / for any space X .
034. Prove that for a zero-dimensional space X , we have s .X / D s.Cp .X //.
035. Prove that hd .X / D hl.Cp .X // for any zero-dimensional space X .
036. Prove that under SA, the following conditions are equivalent:
(i) s.Cp .X // D !;
(ii) hl..Cp .X //! / D !;
(iii) hd..Cp .X //! / D !.
In particular, if SA holds, then hl.Cp .X // D ! implies hl..Cp .X //! / D !
and s.Cp .X // D ! implies s..Cp .X //! / D !.
037. Prove that the following statements are equivalent (remember that all spaces
are assumed to be Tychonoff):
(i) there is a space X with s.X / D ! and d.X / > !;
(ii) there is a space X with hl.X / D ! and d.X / > !;
(iii) there is a left-separated space X with s.X / D ! and jX j D !1 .
038. (-system lemma) Prove that for any regular uncountable cardinal , if U is a
family of finite sets with jUj D , then there exists a set F (called the -root
for U) and a family V U (called the -system for U) such that jVj D and
A \ B D F for any distinct A; B 2 V.
039. Prove that under CH, there exists a hereditarily Lindelf non-separable dense
subspace X of the space f0; 1g!1 . In particular, L-spaces exist under CH.
040. Prove that under CH, there exists a hereditarily separable non-Lindelf dense
subspace X of the space f0; 1g!1 . Thus, S -spaces exist under CH.
041. Prove that under CH, any sequential compact space has points of countable
character.
042. Prove that under CH, there is a P -point in !n!.
043. Let X be a Luzin space. Prove that hl.X / D ! and hd.X / !1 .

044. Prove that if a Luzin space X is separable, then all closed subsets of X are
separable.
045. Prove that no Luzin space can be condensed onto a compact space.
046. Prove that under CH, there is a Luzin second countable space as well as a
Luzin non-separable space.
047. Prove that MA(!) holds in ZFC (and hence CH implies MA), while MA(c) is
false in ZFC.
048. Prove that MA./ is equivalent to MA./ restricted to ccc partially ordered
sets of cardinality , i.e., if MA./ is true for all ccc partial orders of
cardinality , then it is true for all ccc partial orders.
049. Let .P; / be a partially ordered set. Call a subset A P centered if for any
n 2 N and any p1 ; : : : ; pn 2 A, there is r 2 P such that r pi for all i n.
Assume MAC:CH and take any ccc partially ordered set P. Prove that for
any uncountable R P, there exists an uncountable centered Q R. In
particular, all elements of Q are pairwise compatible.
050. Assume MAC:CH. Let Xt be Q a space with c.Xt / D ! for every t 2 T . Prove
that c.X / D !, where X D fXt W t 2 T g.
051. Given families
S A; B exp.!/ such that jAj ; jBj and < c, suppose
that Bn. A0 / is infinite for every B 2 B and any finite family A0 A. Prove
that MA./ implies that there exists M ! such that BnM is infinite for any
B 2 B while AnM is finite for any A 2 A. T
052. (Booth lemma) Let C exp.!/ be a family such that jCj D < c and C 0 is
infinite for every finite C 0 C. Prove that MA./ implies that there exists an
infinite L ! such that LnC is finite for any C 2 C.
053. Let A exp.!/ be an almost disjoint family ( all elements of A are infinite
while A \ B is finite whenever A and B are distinct elements of A). Suppose
that is an infinite cardinal and jAj D < c. Prove that MA./ implies that
A is not maximal.
054. Assume MAC:CH. Let X be a space such that .x; X / < c for some x 2 X .
Prove that for any countable A X with x 2 A, there exists a sequence
fan gn2! A such that an ! x.
055. Let X be a second countable space of cardinality < c. Prove that under MA,
any subset of X is a G in X .
056. Prove that MA implies 2 D c for any infinite < c.
057. Let X be a second countable dense-in-itself space. Given a cardinal such
that 0 < < c, suppose that N is a nowhere S dense subset of the space X for
any < . Prove that under MA, the set fN W < g is of first category
in X .
058. Prove that Martins axiom is equivalent to the following statement: Given a
compact space X such that c.X / D S !, for any family
of nowhere dense
subsets of X with j
j < c, we have
X .
059. Show that under MAC:CH, if s .X / D !, then hd.X ! / D hl.X ! / D !.
In particular, neither strong S -spaces nor strong L-spaces exist under
MAC:CH.
060. Prove that under MAC:CH, every compact space of countable spread is
hereditarily separable.
061. Prove that under MAC:CH, every compact space of countable spread is
perfectly normal.
062. Suppose that MAC:CH hold. Prove that every compact space X with
s.X X / ! is metrizable.
063. Prove that MAC:CH implies that there are no Luzin Tspaces.
064. Let Cn !1 be a club for any natural n. Prove that fCn W n 2 !g is a club.
065. Prove that
(i) every stationary subset of !1 is uncountable;
(ii) not all uncountable subsets of !1 are stationary;
(iii) if A !1 contains a stationary set, then A is stationary;
(iv) any intersection of a stationary subset
S of !1 with a club is stationary;
(v) if An !1 for each n 2 ! and fAn W n 2 !g is stationary, then An is
stationary for some n 2 !.
066. Let A be a stationary subset of !1 . Prove that there exists a disjoint family of
stationary subsets fA W < !1 g such that A A for each 2 !1 .
067. (Fodors Lemma, also called Pressing-Down Lemma) Let A !1 be a
stationary subset of !1 . Suppose that f W A ! !1 is a map such that f ./ <
for any 2 A. Prove that there is 0 2 !1 such that the set f 1 .0 / is
stationary (and, in particular, jf 1 .0 /j D !1 ).
068. Given anySordinal < !1 , let S D ff 2 ! W f is an injectiong. In the
set S D <!1 S , consider the following partial order: f g if and only
if f g. Observe that .S; / is a tree with all its chains countable. Prove
that there exists an !1 -sequence fs W < !1 g S such that s 2 S
and sSj s for all < < !1 . Deduce from this fact that the subtree
T D <!1 fs 2 S W s s g of the tree S is an Aronszajn tree. Hence
Aronszajn codings and trees exist in ZFC.
069. Observe that Jensens axiom implies CH; prove that it is equivalent to any of
the following statements:
(i) for every < !1 , there exists a countable family A exp./ such that
for any A !1 , the set f W A \ 2 A g is stationary;
(ii) for any < !1 , there is a set B such that for any B !1 !1 ,
the set f W B \ . / D B g is stationary;
(iii) for any < !1 , there is a function f W ! such that for any
f W !1 ! !1 , the set f W f j D f g is stationary;
(iv) for any < !1 , there is a function g W ! !1 such that for any
g W !1 ! !1 , the set f W gj D g g is stationary;
(v) there exists a set S of cardinality !1 and a set of functions fh W < !1 g
such that h W ! S for all < !1 and, for any h W !1 ! S , the set
f W hj D h g is stationary;
(vi) for any set T of cardinality !1 , there exists a family fk W < !1 g of
functions such that k W ! T for all < !1 , and, for any function
k W !1 ! T , the set f W kj D k g is stationary.
070. Prove that if Jensens axiom holds, then there is a Souslin tree.
071. Prove that MAC:CH implies there are no Souslin trees.
072. Given two topologies and on the same set X , say that is weaker than
if . If and , then is said to be strictly weaker than . If
.X; / is a linearly ordered set and Y X , then Y is the order considered
only on the points of Y . Let .L; / be any linearly ordered space.
(i) Prove that for any M L, the topology .M / on M generated by the
order M is weaker than the topology M L
of the subspace on M induced
from L.
L
(ii) Show that .M / can be strictly weaker than M even if M is a dense or
a clopen subspace of L.
(iii) Assume that the set M is order dense in L, i.e., for any a; b 2 L, if
a < b, then a p < q b for some p; q 2 M . Prove that M is dense
in L and .M / D M L
.
(iv) Prove that there exists a compact linearly ordered space .K; / such that
for some order dense N K, there exists an order isomorphism between
.N; N / and .L; /. In particular, .N / D NK , the space N is dense in
K and .L; / is order isomorphic to .N; N /.
073. Prove that under }, there exists a linearly ordered hereditarily Lindelf non-
separable compact space.
074. Prove that a linearly ordered compact L-space exists if and only if there exists
a Souslin tree.
075. Let L be a non-separable linearly ordered space such that c.L/ !. Prove
that c.L L/ > !. In particular, if X is a perfectly normal non-separable
linearly ordered compact space, then c.X / D ! but c.X X / > !.
076. Let X be a linearly ordered hereditarily Lindelf non-separable compact
space. Prove that Cp .X / is not Lindelf.
077. Suppose that s .X / D !. Prove that X condenses onto a hereditarily
separable space.
078. Suppose that Cp .X / has countable spread. Prove that it can be condensed onto
a hereditarily separable space.
079. Prove that under Jensens axiom, there is a space X of countable spread which
does not condense onto a hereditarily separable space.
080. For an arbitrary space X , assume that Y is a second countable space such that
the space Cp .X; Y / is dense in Y X . Fix any base B .Y / in the space Y ; an
open set U Cp .X; Y / is called B-standard (or standard with respect to X; Y
and B) if there exist n 2 N, points x1 ; : : : ; xn 2 X and sets O1 ; : : : ; On 2 B
such that U D ff 2 Cp .X; Y / W f .xi / 2 Oi for all i D 1; : : : ; ng. Prove that
Cp .X; Y / is perfectly normal if and only if any open subset of Cp .X; Y / is a
union of countably many B-standard open subsets of Cp .X; Y /.
081. Suppose that Cp .X / is perfectly normal. Prove that all closed subsets of X X
are separable.
082. Let X be a compact space with Cp .X / perfectly normal. Prove that X X is
083. Let X be a compact space with Cp .X / perfectly normal. Prove that under
MA+:CH, the space X is metrizable.
084. Prove that the following properties are equivalent for any space X :
(i) Cp .X / Cp .X / is perfectly normal;
(ii) .Cp .X //n is perfectly normal for any natural n 1;
(iii) .Cp .X //! is perfectly normal;
(iv) Cp .X; Y / is perfectly normal for any second countable space Y ;
(v) for every n 2 N, all closed subsets of X n are separable.
085. Prove that for any compact space X , the space Cp .X / Cp .X / is perfectly
normal if and only if .Cp .X //! is hereditarily Lindelf.
086. Prove that under SA, if Cp .X / is perfectly normal, then .Cp .X //! is
hereditarily Lindelf.
087. Let X be a space with a G -diagonal. Prove that Cp .X / is perfectly normal if
and only if .Cp .X //! is perfectly normal.
088. Prove that under MA+:CH, all closed subspaces of Cp .X / are separable if
and only if .Cp .X //! is hereditarily separable.
089. Prove that under CH, there exists a subspace X of f0; 1g!1 such that for
all n 2 N, all closed subsets of X n are separable and X is not hereditarily
separable. Therefore, Cp .X / is a perfectly normal non-Lindelf space.
090. Prove that a compact space X is metrizable if and only if X 3 is hereditarily
normal.
091. Prove that w.X / D .X / for any infinite compact space X . Deduce from this
fact that a compact space X is metrizable if and only if the diagonal of X is a
G -subspace of X X .
092. Prove that a countably compact space X is metrizable if and only if every
subspace Y X with jY j !1 is metrizable.
093. Give an example of a non-metrizable pseudocompact space P such that every
Y P with jY j !1 is metrizable.
094. Let X be a non-metrizable compact space. Prove that there exists a continuous
map of X onto a (non-metrizable compact) space of weight !1 .
095. Let X be a perfectly normal compact space. Prove that d.X / !1 .
096. Let Y be a subspace of a perfectly normal compact space X . Prove that
nw.Y / D w.Y /.
097. Prove that under Continuum Hypothesis, there exists a strictly < -increasing
!1 -sequence S D ff W < !1 g P which is < -cofinal in ! ! .
098. Assuming CH, prove that there exists a space Y with the following
properties:
(i) jY j D !1 and Y is scattered;
(ii) Y is locally compact and locally countable, i.e., every point of Y has a
countable neighborhood (and hence Y is not Lindelf);
(iii) Y k is hereditarily separable for every k 2 N.
In particular, strong S -spaces exist under CH.
099. Prove that under CH, there exists a scattered compact space X which is not
first countable (and hence not metrizable), while X ! is hereditarily separable,
and hence, .Cp .X //! is hereditarily Lindelf. This implies that under CH,
there exist strong L-spaces and strong compact S -spaces. Observe that under
MAC:CH, any compact space X is metrizable whenever .Cp .X //! is
100. Prove that under CH, there exists a non-normal X such that the space
.Cp .X //! is hereditarily Lindelf.
1.2 Monolithity, Stability and Their Generalizations 11
1.2 Monolithity, Stability and Their Generalizations
As usual, all spaces are assumed to be Tychonoff. Given a cardinal , a space X

is -monolithic if A X and jAj imply nw.A/ . A space is monolithic
if it is -monolithic for any infinite cardinal . A space X is -stable if for any
continuous onto map f W X ! Y , we have nw.Y / whenever iw.Y / . A
space is called stable if it is -stable for any infinite cardinal . A space X is said
to be strongly -monolithic if for any A X with jAj , we have w.A/ .
If a space is strongly -monolithic for any infinite cardinal , it is called strongly
monolithic. A space X is -scattered if jAj for any A X with jAj . The
space X is -simple if for any continuous onto map f W X ! Y , we have jY j
if w.Y / . The space X is simple if it is -simple for any infinite cardinal .
Let P be a topological property. A space X is called P./-monolithic if for any
set A X with jAj , the subspace A has the property P. For example, a space
X is normal.!/-monolithic if the closure of any countable subset of X is normal.
Analogously, if is a cardinal function, then a space X is called ./-monolithic
if for any A X with jAj , we have .A/ . For example, a space X is
said to be s .!/-monolithic or spread .!/-monolithic if for any countable A X ,
we have s .A/ ! (recall that this means that spread of all finite powers of A is
countable). Note that a space X is -monolithic if and only if it is nw./-monolithic
and X is strongly -monolithic if and only if it is w./-monolithic.
Let P be a topological property. A space X is called P./-stable if for any
continuous onto map f W X ! Y , the space Y has P whenever iw.Y / .
Analogously, if is a cardinal function, a space X is called ./-stable if for any
continuous onto map f W X ! Y , we have .Y / whenever iw.Y / . For
example, a space X is FrchetUrysohn.!/-stable if for any continuous onto map
f W X ! Y , the space Y is FrchetUrysohn if it can be condensed onto a second
countable space. Note that in this notation, a space is -stable if and only if it is
nw./-stable.
Now, if F is a class of continuous maps, a space X is called P./-F -stable if for
every map f W X ! Y with f 2 F , the space Y has the property P in case it can
be condensed onto a space of weight . In an analogous way, a space X is said to be
./-F -stable if for every continuous onto map f W X ! Y with f 2 F , we have
.Y / if Y can be condensed onto a space of weight . For example, a space X
is tightness ./-R-quotient-stable, if for any R-quotient map f W X ! Y , all finite
powers of Y have tightness whenever iw.Y / . In all the above definitions
of monolithity and stability, if the cardinal is omitted, then the respective property
is defined to be fulfilled for all infinite . For example, a space X is -open-stable
if for any infinite cardinal and any open continuous onto map f W X ! Y , the
space Y has character if it can be condensed onto a space of weight .
All linear spaces in this book are considered over the space R of the real numbers.
A linear space L, equipped with a topology , is called linear topological space
if the linear operations .x; y/ ! x C y and .t; x/ ! tx are continuous with
respect to . A subset A of a linear space L is called convex if x; y 2 A implies
tx C .1 t/y 2 A for any number t 2 0; 1 . The convex hull conv.A/ of a subset A

of a linear space L is the set ft1 x1 C

Ctn xn W n 2 N; x1 ; : : : ; xn 2 A; t1 ; : : : ; tn 2
0; 1 ; t1 C

C tn D 1g. A linear topological space L is called locally convex if it
has a base which consists of convex sets.
A space X is called a P -space if every G -subset of X is open in X . Given
an infinite cardinal , a -modification of a space X is the space with the same
underlying set, whose topology is generated by all G -subsets of X . A subset fx W
< g of a space X is called free sequence if for any < , we have fx W < g\
fx W g D ;. The point-finite cellularity p.X / of a space X is the supremum
of cardinalities of point-finite families of non-empty open subsets of X . If X is a
space, then a.X / D supf W A./ embeds in X g. The cardinal function a is called
the Alexandroff number. Note that a.X / makes sense only if X has a nontrivial
convergent sequence. A space X is Hurewicz if for any sequence fUn W n 2 !g
S open covers of X , we can choose, for each n 2 !, a finite Vn Un such that
of
fVn W n 2 !g is a cover of QX .
Given a product X D fXt W t 2 T g of the spaces Xt , and a point x 2 X ,
let .X; x/ D fy 2 X W jft 2 T W y.t/ x.t/gj !g. The space .X; x/ is
called the -product of spaces fXt W t 2 T g with the center x. If X is the same
product, then the set .X; x/ D fy 2 X W jft 2 T W y.t/ x.t/gj < !g is called
the -product of spaces fXt W t 2 T g with the center x. If some statement about
-products or -products is made with no center specified, then this statement holds
for an arbitrary center. The symbols .A/ and .A/ are reserved for the respective
- and -products of real lines with the center zero, i.e., .A/ D fx 2 RA W jfa 2
A W x.a/ 0gj !g and .A/ D fx 2 RA W jfa 2 A W x.a/ 0gj < !g.
A space X is a k-space if for any non-closed A X , there exists a compact
subspace K X such that A \ K is not closed in K. A subspace Y of a space X
is a retract of X if there exists a continuous map r W X ! Y (called a retraction)
such that r.y/ D y for all y 2 Y . If X is a space and Y X , then Cp .Y jX / D
Y .Cp .X // D ff 2 Cp .Y / W there exists g 2 Cp .X / such that gjY D f g
Cp .Y /.
Given a class F of mappings and a class X of spaces, we might say that a space
X is an F -preimage of a space from X . This means that there exists a surjective
map f W X ! Y such that f 2 F and Y 2 X . For example, X is a perfect
preimage of a second countable space if there is a second countable space Y and
a perfect map of X onto Y . The reader could also have noticed that we often say
phrases like Y is an open continuous image of X or open continuous onto map.
Why say so if we defined an open map as a continuous and surjective one with
some additional property? Well, the answer is that in many books, the continuity
and surjectivity are not always assumed for open, closed and even perfect maps.
Since this book is also supposed to be a reference guide for specialists (who almost
never read definitions), the author rather prefers some excess in assumptions than a
possible misunderstanding.
101. Prove that any metrizable space is strongly monolithic.

102. Prove that if X is a metrizable space with w.X / c, then X condenses onto
a second countable space.
103. Suppose that A is a proper closed subset of a metric space .X; d / and let
d.x; A/ D inffd.x; a/ W a 2 Ag for any point x 2 X . Prove that there exists a
family fUs ; as W s 2 S g (called a Dugundji system for X nA) such that
(i) Us X nA and as 2 A for any s 2 S ;
(ii) fUs W s 2 S g is an open locally finite (in X nA) cover of X nA;
(iii) d.x; as / 2d.x; A/ for any s 2 S and any x 2 Us .
104. Let A be a closed subspace of a metrizable space X . Suppose that f W A ! L
is a continuous map of A into a locally convex linear topological space L.
Prove that there exists a continuous map F W X ! L such that F jA D f and
F .X / conv.f .A//.
105. Prove that every metrizable space of uncountable weight can be mapped
continuously onto a metrizable space of weight !1 .
106. Prove that a metrizable space is !-stable if and only if it is separable.
107. Prove that any -product of spaces with a countable network is monolithic.
In particular, any -product of second countable spaces is monolithic.
108. Prove that any space X is -stable for any nw.X /. In particular, any space
with a countable network is stable.
109. Prove that any product of spaces with a countable network is stable. In partic-
ular, any product of second countable spaces is stable.
110. Prove that any -product of spaces with countable network is stable. In par-
ticular, any -product of second countable spaces is stable.
111. Prove that any -product of spaces with countable network is stable. In par-
ticular, any -product of second countable spaces is stable.
112. Prove that any -product of Lindelf P -spaces is !-stable.
113. Prove that for any cardinal , if X is -monolithic, then any subspace Y X
is -monolithic.
114. Let be an infinite cardinal.
Q Suppose that X is a -monolithic space for each
< . Prove that fX W < g is also -monolithic. In particular, any
finite product of -monolithic spaces is -monolithic.
115. Prove that if a space X is covered with a locally finite family of closed
monolithic subspaces, then X is monolithic.
116. Give an example of a space which is not !-monolithic being a union of two
monolithic subspaces.
117. Show that R!1 is not !-monolithic. Hence the product of uncountably many
monolithic spaces can fail to be !-monolithic.
118. Prove that any compact space is stable.
119. Prove that any pseudocompact space is !-stable.
120. Show that any compact !-monolithic space of countable tightness must
be FrchetUrysohn. Give an example of a monolithic compact space of
uncountable tightness.
121. Let f W X ! Y be a closed continuous onto map. Prove that if the space X is
-monolithic, then so is Y .
122. Give an example showing that a continuous image of a monolithic space is not
necessarily !-monolithic.
123. Given an infinite cardinal , prove that any continuous image of a -stable
space is -stable. InSparticular, any retract of a -stable space is -stable.
124. Suppose that X D fX W < g and X is a -stable space for any < .
Prove that X is -stable. In particular, any space which is a countable union
of stable subspaces is stable.
125. Give an example of an !-stable space X such that some closed subspace of X
is not !-stable.
126. Give an example of an !-stable space whose square is not !-stable.
127. Prove that any Lindelf P -space is !-simple.
128. Prove that for any scattered space X with l.X / , the Lindelf degree of
the -modification of X does not exceed . In particular, if X is a Lindelf (or
compact!) scattered space, then the !-modification of X is Lindelf.
129. Prove that any Lindelf scattered space (and, in particular, any compact
scattered space) is a simple space.
130. Give an example of a pseudocompact scattered non-simple space.
131. Prove that the following are equivalent for any space X :
(i) the space X is weight.!/-stable;
(ii) the space X is character.!/-stable;
(iii) the space X is FrchetUrysohn.!/-stable;
(iv) the space X is sequential.!/-stable;
(v) The space X is k-property.!/-stable.
(vi) the space X is w.!/-stable;
(vii) the space X is .!/-stable;
(viii) the space X is pseudocompact.
132. Prove that the following conditions are equivalent for any space X :
(i) for any countable A Cp .X / the space A is -compact;
(ii) for any countable A Cp .X / the space A is -countably compact;
(iii) for any countable A Cp .X / the space A is -pseudocompact;
(iv) for any countable A Cp .X / the space A is locally compact;
(v) Cp .X / is a Hurewicz space;
(vi) Cp .X / is a -locally compact space;
(vii) for any countable A Cp .X /, the space A is Hurewicz;
(viii) for any countable A Cp .X /, the space A is -locally compact;
(ix) X is finite.
133. Prove that any non-scattered countably compact space can be continuously
mapped onto I.
134. Prove that the following conditions are equivalent for any compact X :
(i) Cp .X / is a FrchetUrysohn space;
(ii) A is a FrchetUrysohn space for any countable A Cp .X /;

(iii) X is scattered.
135. Prove that for any Lindelf P -space X , the space Cp .X / must be strongly
!-monolithic and FrchetUrysohn.
136. Prove that for any Lindelf scattered space X , the space Cp .X / is a strongly
monolithic FrchetUrysohn space.
137. Let X be a Lindelf P -space. Prove that the set A is Cech-complete for any
countable A Cp .X /.
138. Let X be a pseudocompact space. Suppose that A is Cech-complete for any
countable A Cp .X /. Prove that X is finite.
139. Suppose that A is normal for any countable A Cp .X /. Prove that A is
collectionwise-normal for any countable A Cp .X /, i.e., if the space Cp .X /
is normal.!/-monolithic, then it is collectionwise-normal.!/-monolithic.
140. Prove that under MAC:CH, there exists a normal.!/-monolithic space which
is not collectionwise-normal.!/-monolithic.
141. Suppose that A is normal for any countable A Cp .X /. Prove that A is
countably paracompact for any countable A Cp .X /.
142. Suppose that A is hereditarily normal for any countable A Cp .X /. Prove
that A is perfectly normal for any countable A Cp .X /.
143. Given an infinite cardinal , prove that Cp .X; I/ is -stable if and only if
Cp .X / is -stable.
144. Given an infinite cardinal , prove that Cp .X; I/ is -monolithic if and only if
Cp .X / is -monolithic.
145. Let be any cardinal function. Prove that the Hewitt realcompactification X
of a space X is .!/-stable if and only if X is .!/-stable. In particular, X
is !-stable if and only if X is !-stable.
146. Let and be cardinal functions such that .Z/ D .Cp .Z// for any
space Z. Suppose that is hereditary, i.e., for any space Z and any subspace
Y Z, we have the inequality .Y / .Z/. Prove that for any infinite
cardinal , a space X is ./-stable if and only if Cp .X / is ./-monolithic.
147. Let and be cardinal functions such that .Z/ D .Cp .Z// for any space
Z. Suppose that the function is closed-hereditary, i.e., for any space Z
and any closed Y Z, we have .Y / .Z/. Prove that for any infinite
cardinal , a space X is ./-R-quotient-stable if and only if Cp .X / is ./-
monolithic.
148. Suppose that and are cardinal functions such that
(i) for any space Z, if a space Y is a continuous image of Z, then .Y /
.Z/;
(ii) for any space Z and any closed A Z, we have .A/ D .Cp .AjZ//.
Prove that for an arbitrary space X , the space Cp .X / is ./-stable if and
only if X is ./-monolithic.

(i) for any space Z, if a space Y is a quotient image of Z, then .Y / .Z/;
(ii) for any space Z and any closed A Z we have .A/ D .Cp .AjZ//.
Prove that for an arbitrary space X , the space Cp .X / is ./-quotient-
stable if and only if X is ./-monolithic.
(i) for any space Z, if a space Y is an R-quotient continuous image of Z,
then .Y / .Z/;
(ii) for any space Z and any closed A Z we have .A/ D .Cp .AjZ//.
Prove that for an arbitrary space X , the space Cp .X / is ./-R-quotient-
stable if and only if X is ./-monolithic.
(i) for any space Z, if Y is an open continuous image of Z, then .Y /
.Z/;
Prove that for an arbitrary space X , the space Cp .X / is ./-open-stable if
and only if X is ./-monolithic.
152. Given a space X and an infinite cardinal prove that X is -monolithic if and
only if Cp .X / is -stable. In particular, X is monolithic if and only if Cp .X /
is stable.
153. Prove that if Cp .X / is a stable space, then .Cp .X // is also a stable space for
any cardinal .
154. Suppose that X is an arbitrary space and is an infinite cardinal. Prove that
Cp .X / is -monolithic if and only if X is -stable. In particular, Cp .X / is
monolithic if and only if X is stable.
155. Prove that X is a monolithic space if and only if so is Cp Cp .X /.
156. Prove that X is a stable space if and only if so is Cp Cp .X /.
157. Prove that X is -simple if and only if Cp .X / is strongly -monolithic.
(i) Cp .X / is strongly -monolithic;
(ii) Cp .X / is w./-monolithic;
(iii) Cp .X / is ./-monolithic;
(iv) Cp .X / is ./-monolithic.
159. Given a space X and an infinite cardinal prove that X is s ./-monolithic if
and only if Cp .X / is s ./-stable.
160. Given an arbitrary space X and an infinite cardinal prove that if Cp .X / is
spread./-stable, then X is spread./-monolithic.
161. Give an example of a spread-monolithic space X such that Cp .X / is not
spread(!)-stable.
162. Given a space X and an infinite cardinal prove that X is s ./-stable if and
only if Cp .X / is s ./-monolithic.
163. Prove that if Cp .X / is spread()-monolithic, then X is spread()-stable.
164. Show that there exists a spread-stable space X such that Cp .X / is not
spread(!)-monolithic.
165. Given a space X and an infinite cardinal prove that X is hd ./-monolithic
if and only if Cp .X / is hl ./-stable.
hl./-stable, then X is hd./-monolithic.
167. Show that there exists an hd-monolithic space X such that Cp .X / is not hl.!/-
stable.
168. Given a space X and an infinite cardinal prove that X is hd ./-stable if and
only if Cp .X / is hl ./-monolithic.
hl./-monolithic, then X is hd./-stable.
170. Give an example of an hd-stable space X such that the space Cp .X / is not
hl.!/-monolithic.
171. Given a space X and an infinite cardinal prove that X is hl ./-stable if and
only if Cp .X / is hd./-monolithic.
172. Give an example of an hl-stable space X such that the space Cp .X / is not
hd.!)-monolithic.
173. Given a space X and an infinite cardinal prove that X is hl ./-monolithic
if and only if Cp .X / is hd./-stable.
174. Show that there exists an hl-monolithic space X such that Cp .X / is not
hd.!)-stable.
175. Given a space X and an infinite cardinal prove that X is p./-stable if and
only if Cp .X / is a./-monolithic.
176. Given a space X and an infinite cardinal prove that X is l ./-stable if and
only if Cp .X / is t./-monolithic.
177. Given a space X and an infinite cardinal prove that X is d./-stable if and
only if Cp .X / is iw./-monolithic.
178. Given an arbitrary space X and an infinite cardinal prove that the following
conditions are equivalent:
(i) Cp .X / is iw./-monolithic;
(ii) Cp .X / is ./-monolithic;
(iii) Cp .X / is ./-monolithic.
179. Give an example of a space X which is pseudocharacter-monolithic but not
diagonal-number-monolithic.
180. Give an example of a space X which is diagonal-number-monolithic but not
i -weight-monolithic.
181. Suppose that X is an arbitrary space and is an infinite cardinal. Prove that
X is tm ./-R-quotient-stable if and only if Cp .X / is q./-monolithic.
182. Give an example of a space X for which Cp .X / is q-monolithic and X is not
tm .!/-stable.
183. Given an infinite cardinal , prove that X is l ./-monolithic if and only if

Cp .X / is t./-quotient-stable.
184. Given an infinite cardinal , suppose that Cp .X / is l./-monolithic. Prove that
the space X is tightness ./-R-quotient-stable.
185. Given an arbitrary space X , suppose that Cp .X / is t./-monolithic for some
infinite cardinal . Prove that it is t ./-monolithic.
186. Suppose that Cp .X / is FrchetUrysohn./-monolithic for some infinite
cardinal . Prove that it is FrchetUrysohn./-monolithic.
187. Given an infinite cardinal , prove that X is -scattered if and only if Cp .X /
is w./-open-stable.
188. Let X be a d.!/-stable space such that X n is Hurewicz for all n 2 N. Prove
that for any A Cp .X / and any f 2 AnA, there is a discrete D A such
that f is the only accumulation point of D.
189. Let X be an l.!/-monolithic space of countable spread. Prove that X is
Lindelf.
190. Let X be an hl.!/-monolithic space of countable spread. Prove that X is
hereditarily Lindelf. In particular, if X is an !-monolithic space of countable
spread, then X is hereditarily Lindelf.
191. Let X be an !-monolithic space such that Cp .X / has countable spread. Prove
that hl .X / D ! and hence hd .Cp .X // D !.
192. Suppose that a space X is !-monolithic, !-stable and s.Cp .X // D !. Prove
that nw.X / D !.
193. Assume SA. Prove that for any space X , if Cp .X / is s.!/-monolithic, then
it is hl .!/-monolithic. In particular, if the space Cp .X / is hl.!/-monolithic,
then it is hl .!/-monolithic.
194. Give an example of an hl.!/-monolithic non-hl .!/-monolithic space.
195. Assume SA. Let X be a !-stable space such that s.Cp .X // D !. Prove that
X has a countable network.
196. Assume SA. Prove that if s.Cp .X // D !, then for any Y X we have
s.Cp .Y // D !.
197. Assume MAC:CH. Let X be an !-monolithic space such that the spread of
Cp .X / is countable. Prove that X has a countable network.
198. Assume MAC:CH and suppose that Cp .X / contains no uncountable free
sequences. Prove that for any Y X , the space Y is hereditarily Lindelf if
and only if it is hereditarily separable.
199. Assume MAC:CH and let X be an !-monolithic space with l .X / D !.
Suppose that Cp .X / is Lindelf and Y X has countable spread. Prove that
Y has a countable network.
200. Assuming MAC:CH prove that if Cp .X / is hereditarily stable, then X has a
countable network.
1.3 Whyburn Spaces, Calibers and Lindelf -Property 19
1.3 Whyburn Spaces, Calibers and Lindelf -Property
All spaces are assumed to be Tychonoff. Given two families A and B of subsets of
a space X , say that A is a network with respect to B if for any B 2 B and any open
U B, there is A 2 A such that B A U . A space X is a -space, if X has
a closed cover C such that all elements of C are countably compact and there exists
a -discrete family F which is a network with respect to C. A Lindelf -space is a
-space which has the Lindelf property. A space X is locally compact if every
x 2 X has a compact neighborhood.
A space X is called a p-space if there is a sequence
T fUn W n 2 !g of open (in X )
covers of X such that for every x 2 X , the set fSt.x; Un / W n 2 !g is contained in
X . A space X is disconnected if there exist U; V 2 .X / such that U \ V D ; and
U [ V D X . A space is connected if it is not disconnected. A space X is a k-space
if for any non-closed A X , there exists a compact subspace K X such that
A \ K is not closed. Recall that ext.X / D supfjDj W D is a closed discrete subspace
of X g. The cardinal ext.X / is called the extent of the space
T X . Say that a space X is
K if there exists a space Y such that X Y and X D fYn W n 2 !g where each
Yn is a -compact subset of Y . A compact space K is called a compactification of
a space X if X is homeomorphic to a dense subspace of K.
A subset F of a space X is called almost closed if F nF is a singleton. A space
X is Whyburn or a Whyburn space if for any non-closed A X and any x 2 AnA,
there exists an almost closed F A with x 2 F . The space X is weakly Whyburn
or a weakly Whyburn space if for any non-closed A X , there exists x 2 AnA
and an almost closed F A such that x 2 F . A space X is called submaximal if
it has no isolated points and every dense subspace of X is open in X . A space X is
Hurewicz if for any sequence fUn W n 2 S !g of open covers of X , we can choose, for
each n 2 !, a finite Vn Un such that fVn W n 2 !g is a cover of X . A space X
is radial if for any A X and any x 2 AnA, there exist a regular cardinal and
a transfinite sequence S D fx W < g A such that S ! x, i.e., for any open
U 3 x, there is < such that for each , we have x 2 U . The space X
is pseudoradial if A X and A A imply that there is a regular cardinal and a
transfinite sequence S D fx W < g A such that S ! x A.
Let X be a space with Y X and Z X . We say that a family F exp.X /
separates Y from Z if for any y 2 Y and any z 2 Z, there exists F 2 F such
that y 2 F and z F . S If Z D ;, then the agreement is that F separates Y from
Z if and only if Y F . Given spaces X and Y , a map p W X ! exp.Y /
is called compact-valued if p.x/ is a compact subset of Y for each x 2 X . We
will often write p W X ! Y instead of p W X ! exp.Y /. A compact-valued
map p W X ! Y is called upper semicontinuous if for any open U Y , the set
1
S .U / D fx 2 X W p.x/ U g is open in X ; say that p is an onto map, if
p
fp.x/ W x 2 X g D Y .
An uncountable regular cardinal is a caliber of a space X if for any indexed
family
T U D fU W < g .X /, there exists a set A such that jAj D and
fU W 2 Ag ;. An uncountable regular cardinal is called a precaliber of a
space X if for any indexed family U D fU W < g .X /, there exists A

suchT that jAj D and the family U 0 D fU W 2 Ag has finite intersection property,
i.e., V ; for any finite V U 0 . The diagonal D f.x; x/ W x 2 X g X 2 of a
space X is -small if for any A X 2 n with jAj D , there exists B A such that
jBj D and B \ D ;. The !1 -small diagonal is called small. A continuous onto
map f W X ! Y is called pseudo-open or hereditarily quotient if for any y 2 Y ,
we have y 2 IntY .f .U // whenever f 1 .y/ U 2 .X/.
For any uncountable cardinal , let L./ D [ fag, where a . If x 2 , let
Bx D ffxgg. If x D a, then Bx D ffag [ .nB/ W B is a countable subset of g.
The families fBx W x 2 L./g generate a topology .L.// for which Bx is a local
base at x for any x 2 L./. We call L./ the Lindelfication of the discrete space
of cardinality . A class of spaces P is invariant under some operation if the result
of this operation is in P when applied to elements of P. For example, P is invariant
under finite products if for any X1 ; : : : ; Xn 2 P, we have X1

Xn 2 P.
Analogously, P is invariant with respect to continuous images (closed subspaces) if
for any X 2 P, any continuous image (any closed subspace) of X belongs to P.
Suppose that X is a set without topology and
is an operator on exp.X / such
that ; D ;; A [ B D A [ B , A A and A D A for all A; B X . Then
there exists a unique topology on X such that A D cl .A/ for any A X . We
will say that is generated by the closure operator
. Now, if h
i is an operator on
exp.X / such that h;i D ;; hA \ Bi D hAi \ hBi, hAi A and hhAii D hAi for
all A; B X , then there exists a unique topology on X such that hAi D Int .A/
for any A X . We will say that is generated by the interior operator h
i.
Assume
S that X is a set without topology and B is a family of subsets of X such
that B D X and for any U; V 2 B, if x 2 U \ V then there exists W 2 B such
that x 2 W U \ V . Then there exists a unique topology on the set X such that
B is a base of .X; /. The topology is said to Sbe generated by the family B as a
base. If S is a family of subsets of X such that S D X , then there exists a unique
topology on X such that S is a subbase of .X; /; we will say that the topology
is generated by S as a subbase. Next, suppose that Bx exp.X / for any x 2 X
and the family fBx W x 2 X g has the following properties:
T
(LB1) Bx ; and Bx 3 x for every x 2 X ;
(LB2) if x 2 X and U; V 2 Bx , then there is W 2 Bx such that W U \ V ;
(LB3) if x 2 U 2 By , then there is V 2 Bx such that V U .
Then there exists a unique topology on X such that Bx is a local base of
.X; / at x for any x 2 X . We will call the topology generated by the families
fBx W x 2 X g as local bases. Finally assume that F is a family of maps defined on
X such that every f 2 F maps X into a topological space Yf . There exists a unique
topology on the set X such that the family ff 1 .U / W f 2 F and U 2 .Yf /g is
a subbase of .X; /; we will say that the topology is generated by the family F .
201. Prove that any continuous image of a hereditarily normal compact space is
hereditarily normal.
202. Let X be a compact space for which X 2 is hereditarily normal. Prove that X
is perfectly normal and hence first countable.
203. Given a compact space X , let D f.x; x/ W x 2 X g X 2 be the diagonal of
X . Prove that if X 2 n is paracompact, then X is metrizable.
204. Observe that any FrchetUrysohn space must be Whyburn. Prove that any
countably compact Whyburn space is FrchetUrysohn.
205. Give an example of a pseudocompact Whyburn space which fails to be
FrchetUrysohn.
206. Observe that a continuous image of a Whyburn space need not be Whyburn.
Prove that any image of a Whyburn space under a closed map is a Whyburn
space. Prove that the same is true for weakly Whyburn spaces.
207. Prove that every space with a unique non-isolated point is Whyburn. In
particular, there exist Whyburn spaces of uncountable tightness.
208. Prove that any submaximal space is Whyburn.
209. Prove that any radial space is weakly Whyburn.
210. Prove that every Whyburn k-space is FrchetUrysohn. In particular, any
sequential Whyburn space as well as any Cech-complete Whyburn space must
be FrchetUrysohn.
211. Prove that every compact weakly Whyburn space is pseudoradial but not
necessarily sequential.
212. Give an example of a Whyburn space which is not pseudoradial.
213. Prove that any scattered space is weakly Whyburn.
214. Observe that any sequential space is a k-space. Prove that any hereditarily
k-space (and hence any hereditarily sequential space) is FrchetUrysohn.
215. Prove that there exist hereditarily weakly Whyburn spaces which are not
Whyburn.
216. Prove that if X is d.!/-stable and X n is a Hurewicz space for each natural n,
then Cp .X / is a Whyburn space. In particular, if X is a -compact space, then
Cp .X / is a Whyburn space.
217. Prove that for a paracompact space X , if Cp .X / is Whyburn, then X is a
Hurewicz space. In particular, if X is metrizable and Cp .X / is Whyburn, then
X is separable.
218. Given a space X such that Cp Cp .X / is Whyburn, prove that X has to be finite
if it is either countably compact or has a countable network.
219. Prove that there exists a separable metrizable space X such that Cp .X / is not
weakly Whyburn.
220. Prove that there exists a compact weakly Whyburn space which fails to be
hereditarily weakly Whyburn.
221. Prove that any metrizable space is a p-space and a -space at the same time.
222. Prove that Cp .X / is a p-space if and only if X is countable.
223. Prove that every Lindelf p-space is a Lindelf -space. Give an example of
a p-space which is not a -space.
224. Prove that

(i) Any closed subspace of a -space is a -space. In particular, any closed
subspace of a Lindelf -space is a Lindelf -space.
(ii) Any closed subspace of a p-space is a p-space. In particular, any closed
subspace of a Lindelf p-space is a Lindelf p-space.
225. Prove that X is a Lindelf -space if and only if X has a countable network
with respect to a compact
S cover C.
226. Suppose that X D fXn W n 2 !g where Xn is countably compact and closed
in X for any n 2 !. Prove that X is a -space. As a consequence, every
-compact space is a Lindelf -space.
227. Prove that the Sorgenfrey line and the space L.!1 / are examples of Lindelf
spaces which are not Lindelf .
228. Prove that any space with a -discrete network is a -space. In particular, if
nw.X / !, then X is a Lindelf -space.
229. Let X be a metrizable space. Prove that Cp .X / is a -space if and only if X
is second countable.
230. Prove that any p-space is a k-space. Give an example of a Lindelf -space
which is not a k-space.
231. Give an example of a countable space which is not a p-space. Note that this
example shows that not every Lindelf -space is a p-space.
232. Prove that any Cech-complete space is a p-space. Give an example of a
p-space which is not Cech-complete.
(i) for an arbitrary compactification bX of the space X , there exists a
countable family of compact subspaces of bX which separates X from
bX nX ;
(ii) there exists a compactification bX of the space X and a countable family
of compact subspaces of bX which separates X from bX nX ;
(iii) there exists a compactification bX of the space X and a countable family
of Lindelf -subspaces of bX which separates X from bX nX ;
(iv) there exists a space Z such that X is a subspace of Z and there is a
countable family of compact subspaces of Z which separates X from
ZnX ;
(v) X is a Lindelf -space.
234. Let X be a space of countable tightness such that Cp .X / is a -space. Prove
that if Cp .X / is normal, then it is Lindelf.
235. Let X be a -space with a G -diagonal. Prove that X has a -discrete
network.
236. Prove that for any Lindelf -space X with .X / !, we have jX j c.
237. Prove that under CH, there exists a hereditarily separable compact space X
such that Cp .X / does not have a dense -subspace.
238. Prove that under CH, the space Cp .!/ is not a -space.
239. Prove that for any metrizable X , the space Cp .X / has a dense Lindelf
-subspace.
240. Let p W X ! Y be compact-valued upper semicontinuous onto map. Prove
that l.Y / l.X /.
that if X is compact, then so is Y .
that if X is a Lindelf -space, then so is Y .
243. Prove that
(i) any continuous image of a Lindelf -space is a Lindelf -space;
(ii) any perfect preimage of a Lindelf -space is a Lindelf -space.
244. Prove that w.X / D nw.X / D iw.X / for any Lindelf p-space X . In
particular, any Lindelf p-space with a countable network has a countable
base.
245. Prove that any perfect image and any perfect preimage of a Lindelf p-space
is a Lindelf p-space. Give an example of a closed continuous onto map f W
X ! Y such that X is a Lindelf p-space and Y is not a p-space.
246. Suppose that Cp .X / is a closed continuous image of a Lindelf p-space. Prove
that X is countable.
247. Show that an open continuous image of a p-space need not be a p-space.
Supposing that Cp .X / is an open continuous image of a p-space, prove that
X is countable (and hence Cp .X / is a p-space).
248. Prove that a space X has the Lindelf -property if and only if there exist
a second countable space M and a compact K such that X is a continuous
image of a closed subspace of K M .
(i) there exist a second countable space M and a space Y such that Y can
be mapped perfectly onto M and continuously onto X ;
(ii) there exists an upper semicontinuous compact-valued onto map ' W
M ! X for some second countable space M ;
(iii) X is a Lindelf -space.
250. Give an example of a space X which embeds into Cp .Y / for some Lindelf
p-space Y and is not embeddable into Cp .Z/ for any K -space Z.
251. Give an example of a p-space Y and a pseudo-open (and hence quotient) map
' W Y ! Cp .X / of Y onto Cp .X / for an uncountable space X .
252. Prove that X is a Lindelf p-space if and only if there is a perfect map of X
onto a second countable space.
253. Prove that X is a Lindelf -space if and only if it is a continuous image of a
Lindelf p-space.
254. Prove that the class of Lindelf -spaces is the smallest one which contains
all compact spaces, all second countable spaces, and is invariant with respect
to finite products, continuous images and closed subspaces.
255. Suppose
Q that Xn is a Lindelf p-space for each n 2 !. Prove that the product
fXn W n 2 !g is a Lindelf p-space.
Suppose that Xn is a Lindelf -space for each n 2 !. Prove that the product
256. Q
fXn W n 2 !g is a Lindelf -space.
S
257. Let X be a space such that X D fXn W n 2 !g, where each Xn is a Lindelf
-space. Prove that X is a Lindelf -space.
T Z is a space and Xn Z is Lindelf for each n 2 !. Prove
258. Suppose that
that X D fXn W n 2 !g is a Lindelf -space.
259. Let X be a Lindelf -space such that each compact subset of X is finite.
Prove that X is countable.
260. Give an example of a Lindelf p-space X such that nw.X / > ! and all
compact subsets of X are countable.
261. Prove that any K -space is Lindelf . Show that there exists a K -space
which is not Lindelf p.
262. Let X be a K -space such that all compact subsets of X are countable and
.X / !. Prove that X is countable.
263. Suppose that X is a Lindelf -space and Cp .X / has the Baire property.
Prove that X is countable. In particular, if X is a space with a countable
network and Cp .X / is Baire, then X is countable.
264. Prove that there exists an uncountable Lindelf space X for which Cp .X / has
the Baire property.
265. Suppose that Cp .X / is a Lindelf -space and has the Baire property. Prove
266. Prove that every Lindelf -space is stable, and hence, for every Lindelf
-space X , the space Cp .X / is monolithic.
267. Prove that if X is a Lindelf -space, then X is !-stable.
268. Prove that any product and any -product of Lindelf -spaces is stable.
Show that any -product of Lindelf -spaces is !-stable.
269. (Baturovs theorem). Let X be a Lindelf -space. Prove that for any set
Y Cp .X /, we have ext.Y / D l.Y /.
270. Prove that every subspace of X is a Lindelf -space if and only if X has a
countable network.
271. Prove that every subspace of X is a Lindelf p-space if and only if X is second
countable.
272. Observe first that there exist hereditarily Cech-complete non-metrizable
spaces. Therefore a hereditarily p-space need not be metrizable. Prove that
any hereditarily Cech-complete space is scattered.
273. Prove that !1 C 1 is a scattered compact space which is not hereditarily Cech-
complete.
274. Prove that every subspace of X is -compact if and only if X is countable.
275. Prove that
(i) If an uncountable regular cardinal is a caliber of a space X , then is a
precaliber of X .
(ii) if an infinite successor cardinal is a precaliber of a space X , then

c.X / < . In particular, if !1 is a precaliber of X , then c.X / D !.
276. Let be an uncountable regular cardinal. Prove that if is a precaliberS
(caliber) of Xn for every n 2 !, then is a precaliber (caliber) of fXn W
n 2 !g.
277. Let be an uncountable regular cardinal. Prove that if is a precaliber
(caliber) of X , then is a precaliber (caliber) of every continuous image of X .
278. Suppose that is an uncountable regular cardinal and Y is a dense subspace
of X . Prove that
(i) if is a caliber of Y , then it is a caliber of X ;
(ii) is a precaliber of Y if and only if it is a precaliber of X .
279. Show that an uncountable regular cardinal is a caliber of a compact space X
if and only if it is a precaliber of X .
280. Let be an uncountable regular cardinal.QProve that if is a precaliber of Xt
for every t 2 T , then is a precaliber of fXt W t 2 T g.
281. Let be an uncountable regular cardinal.
Q Prove that if is a caliber of Xt for
every t 2 T , then is a caliber of fXt W t 2 T g.
282. Prove that any product X of separable spaces satisfies the Shanin condition,
i.e., every uncountable regular cardinal is a caliber of X .
283. Prove that any uncountable regular cardinal is a precaliber of Cp .X / for any
space X .
284. Prove that there exists a space X such that !1 is a precaliber of X while the
point-finite cellularity of X is uncountable. Observe that if !1 is a caliber of
X , then p.X / D !.
285. Let X be a metrizable space. Prove that any regular uncountable cardinal is a
caliber of Cp .X /.
286. Prove that an uncountable regular cardinal is a precaliber of X if and only if
it is a caliber of X .
287. Suppose that X is a compact space of countable tightness. Prove that if !1
is a caliber of X , then X is separable. Give an example of a non-separable
compact space X such that !1 is a caliber of X .
288. Assuming MAC:CH, prove that !1 is a precaliber of any space which has
the Souslin property.
289. Assume the axiom of Jensen (}). Prove that there exists a space X with
c.X / D ! while !1 is not a precaliber of X .
290. Prove that for any uncountable regular cardinal , the diagonal of Cp .X / is
-small if and only if is a caliber of X . In particular, !1 is a caliber of X if
and only if Cp .X / has a small diagonal.
291. Prove that an uncountable regular cardinal is a caliber of X if and only if it
is a caliber of Cp .Cp .X //.
292. Suppose that an uncountable regular cardinal is a caliber of Cp .X /. Prove
that for any Y X the cardinal is a caliber of Cp .Y /.
293. Let be an uncountable regular cardinal. Prove that if is a caliber of Cp .X /,

then the diagonal of X is -small. In particular, if !1 is a caliber of Cp .X /,
then the diagonal of X is small.
294. Let be an uncountable regular cardinal. Prove that if all finite powers of X
are Lindelf and X has a -small diagonal, then is a caliber of Cp .X /. As
a consequence, if l .X / D !, then X has a -small diagonal if and only if
is a caliber of Cp .X /. In particular, if X is compact, then the diagonal of X is
small if and only if !1 is a caliber of Cp .X /.
295. Prove that any compact space of weight !1 with a small diagonal is
metrizable.
296. Let X be a compact space with a small diagonal. Prove that if X is !-
monolithic and has countable tightness, then it is metrizable.
297. Let X be a compact space with a small diagonal. Prove that if X is monolithic,
then it is metrizable.
298. Prove that under CH, any compact space with a small diagonal is metrizable.
299. Assume that 2!1 D !2 . Prove that any compact X , with !1 and !2 calibers of
Cp .X /, is metrizable.
300. Note that every Lindelf -space with a diagonal G must have a countable
network. Prove that under CH, any Lindelf -space with a small diagonal
has a countable network.
1.4 A Glimpse of Descriptive Set Theory 27
1.4 A Glimpse of Descriptive Set Theory
All spaces are assumed to be Tychonoff. A space X is called zero-dimensional if it

has a base which consists of clopen sets. The space X is strongly zero-dimensional
if any finite open cover of X has a disjoint open refinement. We say that the large
inductive dimension of X is equal to zero (denoting it by IndX D 0) if for any closed
F X and open U F , there is a clopen V X such that F V U . The
symbol P stands for the space of the irrational numbers with its topology inherited
from R. A space X is called submetrizable if it can be condensed onto a metrizable
space.
Given spaces X and Y , the expression X ' Y says that X is homeomorphic to
Y ; a map ' W X ! exp.Y / is called lower semicontinuous if for any open U Y ,
the set 'l1 .U / D fx 2 X W '.x/ \ U ;g is open in X . The map ' is called upper
semicontinuous if for any open U Y , the set 'u1 .U / D fx 2 X W '.x/ U g
is open in X . If '.x/ is a compact subset of Y for each x 2 X , the map ' is
called compact-valued. Given a set X , an arbitrary function d W X X ! R is
called a pseudometric on X if d.x; x/ D 0; d.x; y/ 0 for any x; y 2 X and
d.x; z/ d.x; y/ C d.y; z/ for any x; y; z 2 X .
Denote by A the set of numeric sequences D fi W i 2 Ng such P thatii D 0 or
i D 2 for all i 2 N. Given D fi W i 2 Ng 2 A, let x./ D 1 i D1 3
i . Let
K D fx./ W 2 Ag. The set K is called the Cantor perfect set or simply the Cantor
set. If we refer to the Cantor set as a topological space, its topology is considered
to be inherited from R. The space D is the set f0; 1g with the discrete topology.
A Tychonoff cube is the space I for some infinite cardinal . Analogously, a Cantor
cube is the space D for some infinite cardinal . Recall that a space X is said to
be pseudocompact if C.X / D C .X /, i.e., every continuous function is bounded
on X .
Given a space X , denote by 00 .X / the family of all open subsets of X . If, for
some countable ordinal , we have 0 .X /, let 0 .X / D fX nU W U 2 0 g. In
particular, 00 .X / is the family of all closed subsets of X . Now, if is a countable
ordinal and we have 0 .X / for all < , we let 0 .X / D fU X W there exists a
S
sequence fn W n 2 !g such that U D fUn W n 2 !g, where Un 2 0n .X / for
each n 2 !g. This makes it possible to define the families 0 .X / and 0 .X / for
all countable ordinals . We will also need the families 0 .X / D 0 .X / \ 0 .X /
for all countable ordinals S.
The family B.X / D f0 .X / [ 0 .X / W < !1 g is called the family of Borel
subsets of the space X . For any ordinal < !1 , the collection 0 .X / is called the
family of Borel subsets of X of additive class and 0 .X / is the family of all Borel
subsets of X of multiplicative class . In the classical terminology the elements
of the family 10 .X / are called F -sets, the elements of 10 .X / are G -sets and
the family 20 .X / consists of G -sets; the sets that belong to 20 .X / are called
F -subsets of X , the family 30 .X / consists of F -sets, the family 30 .X / is
referred to as the family of G -sets and so on.
A space is called Polish if it is second countable and completely metrizable.

A second countable space is called a Borel set if it is homeomorphic to a Borel
subspace of some Polish space. A space is analytic if it is a continuous image of P.
A metrizable space X is an absolute Borel set of additive class > 0, if X , when
embedded into an arbitrary metric space M , belongs to 0 .M /. Analogously, X is
an absolute Borel set of multiplicative class , if, for any embedding ' of X into a
metric space M , the set '.X / belongs to 0 .M /. In particular, X is an absolute G
if X is a G -subset whenever it is embedded into a metric space; X is an absolute
F if for any metric space M , any subspace of M homeomorphic to X is F -set
in M .
A space X is called a K -space if it is homeomorphic to a subspace Y of
some space Z such that Y is a countable intersection of -compact subspaces of
Z. A space is called K-analytic if it is a continuous image of a K -space. Given a
family U of topologies
S on a set X , the family sup U is the topology whose subbase
is the family U.
A sequence fUn W n 2 !g of covers of a space X is called Tcomplete if, for any
filter F on X such that F \ Un ; for all n 2 !, we have fF W F 2 F g ;.
If X is a space, let B1 .X / D ff 2 RX W fn ! f for some sequence ffn W n 2
!g Cp .X /g. We will consider the set B1 .X / with the topology inherited from
RX . A compact space K is called Rosenthal compact if K is homeomorphic to a
subspace of B1 .M / for some Polish space M .
We denote by ! <! the set of all finite sequencesS of elements of !, including the
empty one. In other words, ! 0 D f;g and ! <! D f! n W n 2 !g. If f 2 ! ! and
n 2 N, then f jn D f jf0; : : : ; n 1g; this function is sometimes identified with
the sequence .f .0/; : : : ; f .n 1//. If n D 0, then f jn D ;. Given two functions
f; g 2 ! ! , we write f g, if f .n/ g.n/ for every n 2 !.
A family B exp.X / is P-directed if B D fBf W f 2 ! ! g, and, for any
f; g 2 ! ! , we have Bf Bg whenever f g. Thus, a P-directed compact cover
of a space X is aSP-directed family B such that the elements of B are compact
subsets of X and B D X .
If f is a function, then dom.f / is its domain; given another function g, the
expression f g says that dom.f / dom.g/ and gjdom.f / D f . If we
have a set of functions ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D
fj j.dom.fi /S\ dom.fj // for any i; j 2 I , then we can define a function f with
dom.f / D i 2I dom.fi / as follows: given any x 2 dom.f /, find any i 2 I with
x 2 dom.fi / and let f .x/ D fi .x/. It is easy to check that the value of f at x does
not depend on the choice
S of i so we have consistently defined a function f which
will be denoted by ffi W i 2 I g.
Given two linearly ordered sets .A; A / and .B; B / the lexicographic order <
on the set A B is introduced as follows: given distinct points p D .a; b/ and
q D .c; d / of the set A B, we let p < q if a A c; if c A a, then q < p. Now,
if a D c and b B d , then p < q; if d B b, then q < p. It is easy to see that if
A and B are well-orders, then < is also a well-order.
301. Let X be a zero-dimensional space. Prove that any subspace of X is also zero-
dimensional.
302. Prove that any product of zero-dimensional spaces is a zero-dimensional
space.
303. Given a cardinal and an infinite space X with w.X / , prove that X is
zero-dimensional if and only if it can be embedded in D .
304. Prove that any space X is a perfect image of a zero-dimensional space Y such
that w.Y / w.X /.
305. Prove that any non-zero-dimensional space can be continuously mapped
onto I.
306. Prove that any Lindelf space is zero-dimensional if and only if it is strongly
zero-dimensional. In particular, any compact and any second countable zero-
dimensional space is strongly zero-dimensional.
307. Let X be a space with jX j < c. Prove that X is zero-dimensional. In particular,
any countable space is strongly zero-dimensional.
308. For an arbitrary space X , prove that X is strongly zero-dimensional if and only
if IndX =0. Observe that as a consequence, any strongly zero-dimensional
space is normal.
309. Prove that any strongly zero-dimensional space is zero-dimensional. Give
an example of a normal zero-dimensional space which is not strongly zero-
dimensional.
310. Prove that a closed subspace of a strongly zero-dimensional space is strongly
zero-dimensional. Give an example of a strongly zero-dimensional space X
such that some Y X is not strongly zero-dimensional.
S
311. Let X be a normal space such that X D fXn W n 2 !g, where each Xn
is strongly zero-dimensional and closed in X . Prove that X is strongly zero-
dimensional.
312. Prove that there exists aSspace X which is not zero-dimensional and we
have the equality X D fXn W n 2 !g, where each Xn is strongly zero-
dimensional and closed in X .
313. Prove that the space P of the irrationals is homeomorphic to ! ! and hence P
is zero-dimensional.
314. Let X be a paracompact space. Prove that X is strongly zero-dimensional if
and only if every open cover of X has a disjoint open refinement.
315. Let P be a strongly zero-dimensional paracompact space and suppose that M
is a completely metrizable space. Denote by CL.M / the set of all closed non-
empty subsets of M and let ' W P ! CL.M / be a lower semicontinuous
map. Prove that ' has a continuous selection, i.e., there exists a continuous
map f W P ! M such that f .x/ 2 '.x/ for any x 2 P .
316. Let M be a strongly zero-dimensional completely metrizable space. Prove
that any closed non-empty F M is a closed retract of M , i.e., there exists a
closed continuous map f W M ! F such that f .x/ D x for any x 2 F .
317. Let F be a closed subspace of P and suppose that a space X is a continuous
image of F . Prove that X is also a continuous image of P.
318. Given a second countable space X and an arbitrary ordinal < !1 , prove that
there exists a set U 2 0 .X K/ such that 0 .X / D fU y W y 2 Kg,
where U y D fx 2 X W .x; y/ 2 U g for any y 2 K. Observe that as an easy
consequence, for any second countable space X and every countable ordinal
, there exists a set V 2 0 .X K/ such that 0 .X / D fV y W y 2 Kg.
319. Prove that for any uncountable Polish space X and every countable ordinal ,
the classes 0 .X / and 0 .X / do not coincide.
320. Let X be a second countable space. Given countable ordinals and > ,
prove that 0 .X / [ 0 .X / 0 .X /. Show that if X is an uncountable
Polish space, then 0 .X / [ 0 .X / 0 .X /.
321. Suppose that X is a second S countable space. Prove that for every countable
limit ordinal , we have f0 .X / W < g 0 .X /. Show that if X is
S
uncountable and Polish, then the inclusion is strict, i.e., f0 .X / W < g
0 .X /.
322. Prove that there exists a countable space which cannot be embedded into
Cp .B/ for any Borel set B.
323. Prove that a second countable space X is an absolute F if and only if X is
-compact.
324. Prove that a second countable space X is an absolute G if and only if X is
Cech-complete.
325. Suppose that X is a Polish space and f W X ! Y is a perfect map. Prove that
Y is Polish (remember that any perfect map is onto).
326. Let X be a Polish space. Suppose that f W X ! Y is a continuous surjective
open map. Prove that there is a closed F X such that f .F / D Y and f jF
is a perfect map.
327. Prove that any open continuous image of a Polish space is a Polish space.
328. Prove that X is a Polish space if and only if it is an open continuous image
of P.
329. Prove that a second countable space is Polish if and only if it is a closed
continuous image of P. Show that a closed continuous image of P is not
necessarily first countable.
330. Prove that X is homeomorphic to a Borel subset of some Polish space if and
only if it is homeomorphic to a Borel subset of R! .
331. Let X be a Borel set. Prove that every Y 2 B.X / is also a Borel set. In
particular, any closed and any open subspace of a Borel set is a Borel set.
332. Given second countable spaces X and Y and a continuous map f W X ! Y ,
prove that for every Borel subset A of the space Y , the set f 1 .A/ is Borel
in X .
333. Prove that any countable product of Borel S sets is a Borel set. Show that for
any second countable space X , if X D fXi W i 2 !g and each Xi is a Borel
set, then X is also a Borel set.
334. Prove that every Borel set is an analytic space.
335. Prove that
(i) any closed subspace of an analytic space is an analytic space;
(ii) any open subspace of an analytic space is an analytic space;

(iii) any countable product of analytic spaces is an analytic space.
336. Assume thatT Y is a space and Xi Y is an analytic space for all i 2 !. Prove
that X D fXi W iS2 !g is also an analytic space.
337. Assume that X D fXi W i 2 !g and Xi is an analytic space for every i 2 !.
Prove that X is an analytic space.
338. Let X and Y be Polish spaces. Suppose that f W X ! Y is a continuous map.
Prove that for any analytic set B Y , the set f 1 .B/ is also analytic.
339. Let A and B be two disjoint analytic subsets of a Polish space M . Prove that
there exist Borel subsets A0 and B 0 of the space M such that A A0 ; B B 0
and A0 \ B 0 D ;.
340. Let X be a subspace of a Polish space M . Prove that X is Borel if and only if
X and M nX are analytic.
341. Prove that X is a Borel set if and only if there exists a closed subspace of P
which condenses onto X . As a consequence, if a Borel set X can be condensed
onto a second countable space Y , then Y is a also Borel set.
342. Show that there exists a subspace X R which is not analytic (and hence not
Borel).
343. Prove that
(i) any closed subset of a K-analytic space is a K-analytic space;
(ii) any countable product of K-analytic spaces is a K-analytic space.
344. Assume that Y is Ta space and Xi Y is a K-analytic space for all i 2 !.
Prove that X D fX S i W i 2 !g is also a K-analytic space.
345. Assume that X D fXi W i 2 !g and Xi is a K-analytic space for all i 2 !.
Prove that X is a K-analytic space.
346. Observe that there exist K-analytic non-analytic spaces. Show that any
analytic space is a K-analytic space. Prove that for any space X with a
countable network, X is analytic if and only if it is K-analytic.
347. Prove that a non-empty Polish space X is homeomorphic to P if and only if X
is zero-dimensional and any compact subspace of X has empty interior.
348. Prove that a metrizable compact X is homeomorphic to the Cantor set if and
only if X is zero-dimensional and has no isolated points.
349. Prove that a countable metrizable space X is homeomorphic to Q if and only
if X has no isolated points.
350. Prove that every countable metrizable space is homeomorphic to a closed
subspace of Q.
351. Let X be a second countable -compact space. Prove that X is not Polish if
and only if it contains a closed subspace homeomorphic to Q.
352. Let X be an analytic non--compact space. Prove that some closed subspace
of X is homeomorphic to P.
353. Prove that any uncountable analytic space contains a closed subspace which
is homeomorphic to the Cantor set.
354. Prove that any non--compact Borel set can be condensed onto I! as well as
onto R! .
355. Give an example of a non--compact subspace of the real line R which cannot
be condensed onto a compact space.
356. Prove that Q cannot be condensed onto a compact space.
357. Prove that for any metrizable compact X , the space Cp .X / condenses onto a
compact space.
358. Prove that a Polish space X is dense-in-itself if and only if P condenses
onto X .
359. Prove that for any metrizable compact X , the space P condenses onto Cp .X /.
360. Prove that Cp .X / is analytic if and only if R! maps continuously onto Cp .X /.
Observe that not every analytic space is a continuous image of R! .
361. Suppose that X is an infinite space such that Cp .X / is analytic. Prove that
Cp .X / can be continuously mapped onto R! . Deduce from this fact that if X
and Y are infinite spaces such that Cp .X / and Cp .Y / are analytic, then each
of the spaces Cp .X / and Cp .Y / maps continuously onto the other one.
362. Prove that for any second countable -compact space X , the space Cp .X /
is K .
363. Let X be a space with a countable network. Prove that X is analytic if and
only if every second countable continuous image of X is analytic.
364. Let X be a space with a countable network. Prove that X is -compact if and
only if every second countable continuous image of X is -compact.
365. Prove that a second countable space M is Polish if and only if there exists a
map f W P ! exp.M / with the following properties:
(a) f .s/ is compact for any s 2 P;
(b) for any s; t 2 P, if s t, then f .s/ f .t/;
(c) for any compact K M , there is s 2 P such that K f .s/.
366. Prove that if Cp .X / is analytic, then X is -compact.
367. Prove that the following are equivalent for an arbitrary second countable
space X :
(i) Cp .X / is analytic;
(ii) Cp .X / is a K -space;
(iii) X is -compact.
Observe that, as a consequence, the spaces Cp .P/ and Cp .R! / are not analytic.
368. For a second countable space X let A be a countable dense subspace of X .
Prove that the following conditions are equivalent:
(i) Cp .AjX / is a Borel set;
(ii) Cp .AjX / is analytic;
369. Given a countable space X prove that Cp .X / is analytic if and only if Cp .X; I/
is analytic.
370. Prove that a countable space X embeds into Cp .P/ if and only if Cp .X / is
analytic.
371. Take any point 2 !n! and consider the space X D ! [ fg with the
topology inherited from !. Prove that neither Cp .X / nor RX nCp .X / is
analytic. As a consequence, X cannot be embedded into Cp .P/.
372. Prove that if < !1 , then there exists a countable
S space X with a unique
non-isolated point such that Cp .X / 2 B.RX /n. < 0 .RX //, i.e., the space
Cp .X / can have an arbitrarily high Borel complexity for a countable space X
with a unique non-isolated point.
373. Prove that the following are equivalent for any metrizable space X :
(i) X is an absolute F ;
(ii) there is a completely metrizable space M such that X is an F -subset
of M ;
(iii) X has a complete sequence of -discrete closed covers.
374. Prove that Cp .X / is an absolute F for any countable metrizable X .
375. Let K be a compact space. Given a countable X Cp .K/, prove that Cp .X /
is an absolute F .
376. Prove that any second countable space embeds into Cp .K/, where K is the
Cantor set.
377. Give an example of a second countable X such that for any compact K, the
space X cannot be embedded in Cp .K/ as a closed subspace.
378. Prove that any countable second countable space embeds into Cp .K/ as a
closed subspace.
379. Given a function f W X ! R, consider the following conditions:
(i) for any open U R the set f 1 .U / is an F -set in X ;
(ii) there exists a sequence ffn W n 2 !g Cp .X / which converges to f .
Prove that (ii)H)(i) for any space X . Show that if X is second countable,
then also (i)H)(ii) and hence (i) (ii).
380. Prove that if X D R, then B1 .X / RX .
381. Prove that a compact space X is countable if and only if B1 .X / D RX .
382. Prove that under MAC:CH, there exists an uncountable X R such that
B1 .X / D RX .
383. Prove that the two arrows space is Rosenthal compact.
384. Prove that every Rosenthal compact space is FrchetUrysohn.
385. Let X be a separable compact space. Prove that X is Rosenthal compact if and
only if for any dense countable A X , the space Cp .AjX / is analytic.
386. Let X be a compact space. Assume that A and B are dense countable subsets
of X such that Cp .AjX / is analytic and Cp .BjX / is not. Prove that X contains
a subspace homeomorphic to !.
387. Suppose that X is a compact space and A is a dense countable subset of X
such that Cp .AjX / is analytic. Prove that X is Rosenthal compact or else !
embeds in X .
388. Prove that a space X is K-analytic if and only if X is an image of P under a

compact-valued upper semicontinuous map, i.e., whenS there exists a compact-
valued upper semicontinuous ' W P ! X such that f'.p/ W p 2 Pg D X .
389. Prove that an arbitrary space X is K-analytic if and only if there exists a
family K D fKf W f 2 ! ! g of compact subsets of X with the following
properties:
S
(i) the family K is a cover of X , i.e., K D X ;
(ii) if a sequence fn 2 ! ! converges to f 2 ! ! and xn 2 Kfn for all n 2 !,
then the sequence fxn g has an accumulation point which belongs to Kf .
390. Prove that an arbitrary space X is K-analytic if and only if there exists a space
Y which maps perfectly onto P and continuously onto X .
391. Prove that an arbitrary space X is K-analytic if and only if X is realcompact
and has a P-directed compact cover.
392. Suppose that X can be condensed onto a metrizable space. Prove that X is
analytic if and only if it has a P-directed compact cover.
393. Let X be a compact space. Prove that Cp .X / is K-analytic if and only if it has
a P-directed compact cover.
394. Give an example of a non-K-analytic space which has a P-directed compact
cover.
395. Assuming MAC:CH, prove that if X is a K-analytic space such that every
compact subspace of X is metrizable, then X has a countable network.
Observe that if every compact subspace of an analytic space X is countable,
then X is countable.
396. Suppose that X is a compact space and some outer base B of its diagonal
D f.x; x/ W x 2 X g in X X can be indexed as B D fOp W p 2 ! ! g in
such a way that Op Oq whenever p; q 2 ! ! and q p. Prove that X is
metrizable.
397. Suppose that Cp .X / is K-analytic and X is separable. Prove that Cp .X / is
analytic.
398. Let X be a compact space such that Cp .X / is K-analytic. Prove that X is a
FrchetUrysohn space.
(i) Cp .Cp .X // is K-analytic;
(ii) Cp .Cp .X // is analytic;
(iii) X is finite.
(i) X is hereditarily K-analytic;
(ii) X is hereditarily analytic;
(iii) X is countable.
1.5 Additivity of Properties: Mappings Between Function Spaces 35
1.5 Additivity of Properties: Mappings Between

Function Spaces
All spaces are assumed to be Tychonoff. A topological property P will be called

finitely additive if a space X has P provided that X is a finite union of its subspaces
with the property
S P. We say that P is countably additive if a space X has P given
that X D fXn W n 2 !g and Xn ` P (this is read: Xn has P) for every n 2 !.
The finite and countable additivity of a cardinal function are defined analogously,
i.e., a cardinal function ' is finitely additive if the conditions X D X1 [ : : : [ Xn
and '.Xi / for each S i n imply '.X / . We say that ' is countably additive
if the conditions X D fXn W n 2 !g and '.Xn / for every n 2 ! imply
'.X / . A cardinal function ' is called completely additive ifSfor any infinite
cardinal and any space X , we have '.X / whenever X D fX W < g
and '.X / for each < .
A property P is called finitely (countably) additive in some class C of spaces if
for any X 2 C, the space X has P whenever X is a finite (countable) union of its
subspaces with the property P. Analogously, a cardinal function ' is called finitely
additive in a class C if for any space X 2 C, the conditions X D X1 [ : : : [ Xn and
'.Xi / for each i n imply '.X / . We say that ' is countably
S additive in
a class C if for any X 2 C, we have '.X / whenever X D fXn W n 2 !g and
'.Xn / for every n 2 !. A cardinal function ' is called completely additive in
a classSC if for any infinite cardinal and any X 2 C, we have '.X / whenever
X D fX W < g and '.X / for each < .
If X is a space and Cn .X/ for each n 2 !, the sequence fCn W n 2 !g is
called pseudocomplete if for any family T fUn W n 2 !g such that U nC1 Un and
Un 2 Cn for every n 2 !, we have fUn W n 2 !g ;. A space X is called
pseudocomplete if there is a pseudocomplete sequence fBn W n 2 !g of -bases in
X . The space X is homogeneous if for any x; y 2 X , there is a homeomorphism
h W X ! X such that h.x/ D y. A subset F of a space X is called a zero-set in X
if there exists a function f 2 C.X / such that F D f 1 .0/. Say that U X is a
cozero-set in X if X nU is a zero-set in X .
Let M be an infinite maximal almost disjoint family in ! and M D ! [ M. If
x 2 !, let Bx D fxg. Given x 2 M, define Bx D ffxg [ .xnA/ W A is a finite
subset of xg (remember that for any x 2 M, we can consider x to be a point of
M or a subset of !). The families fBx W x 2 M g generate a topology M on M
as local bases and the space .M; M / is called a Mrowka space. Given a cardinal
, let A./ D [ fag, where a . If x 2 , let Bx D ffxgg. If x D a, then
Bx D ffag [ .nB/ W B is a finite subset of g. The families fBx W x 2 A./g
generate a topology .A.// for which Bx is a local base at x for any x 2 A./. The
space A./ is called Alexandroff one-point compactification of the discrete space of
cardinality . For any uncountable cardinal , let L./ D [ fag, where a . If
x 2 , let Bx D ffxgg. If x D a, then Bx D ffag [ .nB/ W B is a countable subset
of g. The families fBx W x 2 L./g generate a topology .L.// for which Bx is
a local base at x for any x 2 L./. The space L./ is called the Lindelfication of
the discrete space of cardinality .
u
Given a space X and A C.X /, denote by A the set ff 2 C.X / W there
exists a sequence ffn W n 2 !g A such that fn ! !f g. The closure operator
u
A ! A generates a topology u (called the uniform convergence topology) on the
u
space C.X /, such that A D clu .A/ for every A C.X /. The space .C.X /; u /
will be denoted Cu .X /. If O is a subset of R, then Cu .X; O/ is the set C.X; O/
with the topology inherited from Cu .X /. If the set C .X / of continuous bounded
functions on X is considered with the topology inherited from Cu .X /, it is denoted
by Cu .X /. A set A Cp .X / is strongly dense (or uniformly dense) in Cp .X / if
u
A D Cp .X /. We will identify any space X with the subspace E.X / D fex W x 2
X g Cp .Cp .X //, where ex .f / D f .x/ for any x 2 X .
As usual, the expression X ' Y says that the spaces X and Y are homeo-
morphic. A space X is metacompact if every open cover of X has a point-finite
open refinement. A space X is called Dieudonn complete if it embeds as a closed
subspace into a product of metrizable spaces. Now, X is realcompact if it embeds
as a closed subspace into a product of real lines. Given a space X , the extent of
X is defined as follows: ext.X / D supfjDj W D is a closed discrete subspace of
X g. A space X is FrchetUrysohn if for any A X and any x 2 A, there is a
sequence fan W n 2 !g A such that an ! x. A space X is sequential if for any
non-closed A X , there is a sequence .an / A which converges to some point of
X nA. A space X is a k-space if for any non-closed A X , there exists a compact
subspace K X such that A \ K is not closed. We say that X is -scattered if
jAj for any A X with jAj .
Given aSsequence
T S D fAn W n 2 !g of subsets of a space X , consider the set
lim S D f fAm W m ng W n 2 !g. The set lim S is called the limit of the
sequence S . The fact that A D lim S will also be expressed as S ! A or An ! A.
Recall that a set X is !-covered by a family U if for any finite A X , there is
U 2 U such that A US. The Gerlits property ' is formulated as follows: X has '
if for any family U D fUn W n 2 !g .X/ with Un UnC1 for each n 2 !,
which !-covers the space X , there exists a sequence fXn W n 2 !g of subsets of X
such that Xn ! X and Xn is !-covered by Un .
A map f W X ! Y is finite-to-one if the set f 1 .y/ is finite for any y 2 Y .
Let X be a space. Given an infinite cardinal , a function f W X ! R is called
strictly -continuous if for any A X with jAj , there exists g 2 C.X /
with f jA D gjA . Let tm .X / D minf W any strictly -continuous function on X is
continuousg. The cardinal tm .X / is called weak functional tightness of the space X .
A subspace Y of a space X is called -placed in X if for any x 2 X nY , there exists
a G -set H in X such that x 2 H X nY . Let q.X / D minf W X is -placed
in X g. The cardinal q.X / is called the HewittNachbin number of the space X .
401. Observe first that the space A.!1 / is the union of two discrete (and hence
metrizable) subspaces of countable i -weight. Deduce from this observation
that the first axiom of countability, metrizability, i -weight, P -property and
pseudocharacter are not finitely additive.
402. Representing L.!1 / as a union of two metrizable subspaces, observe that
sequentiality, -character, the FrchetUrysohn property, Cech-completeness
and k-property are not finitely additive.
403. Let 2 !n! and observe that the space ! [ fg is a union of two second
countable spaces while w.! [ fg/ > !. Therefore weight is not finitely
additive.
404. Give an example of a non-realcompact space which is the union of two
hereditarily realcompact subspaces.
405. Prove that if ' is a cardinal function and ' 2 fnetwork weight, spread,
Lindelf number, hereditary Lindelf number, density, hereditary density,
extent, Souslin number, point-finite cellularityg, then ' is completely additive
and hence countably additive.
406. Prove that pseudocompleteness, -weight and the Baire property are finitely
additive but not countably additive.
407. Considering any Mrowka space, prove that normality is not finitely additive.
408. Suppose that X X D Y [ Z, where Y and Z are first countable. Prove that
X is also first countable.
409. Suppose that X X D Y [Z, where Y and Z have countable pseudocharacter.
Prove that .X / !.
410. Suppose that X X D Y [Z, where Y and Z have countable tightness. Prove
that t.X / !.
411. Suppose that X X D Y [ Z, where Y and Z have countable weight. Prove
that w.X / !.
412. Suppose that X is a separable space such that X X D Y [ Z, where Y and
Z are metrizable. Prove that X is metrizable.
413. Suppose that X is a compact space such that X X D Y [ Z, where Y and
414. Give an example of a non-metrizable space X such that X X is the union of
two metrizable subspaces.
S
415. Suppose that X ! D fXn W n 2 !g. Prove that for some n 2 !, there is Y
Xn such that there exists an open continuous map of Y onto X ! , and hence,
there exists an open continuous map of Y onto X . As a consequence, if X ! is
the countable union of first countable subspaces, then X is first countable.
416. Given an arbitrary space X , suppose that X ! is the finite union of metrizable
subspaces. Prove that X is metrizable.
417. Given a countably compact space X , suppose that X ! is a countable union of
metrizable subspaces. Prove that X is metrizable.
418. Give an example of a non-metrizable space X such that X ! is a countable
union of its metrizable subspaces.
419. For an arbitrary space X and any f; g 2 C .X /, let
d.f; g/ D supfjf .x/ g.x/j W x 2 X g:
Prove that d is a complete metric on the set C .X / and the topology,

generated by d , coincides with the topology of Cu .X /.
420. Let P be an F -hereditary property, i.e., X ` P implies Y ` P whenever Y is
an F -subspace of X . Suppose that Cp .X / is a finite union of subspaces which
have the property P. Prove that Cp .X / is a finite union of dense subspaces,
each one of which has the property P.
421. Let P be a hereditary property. Suppose that Cp .X / is a finite union of
subspaces which have the property P. Prove that there is n 2 N and " > 0 such
u
that Cp .X; . "; "// D Y1 [ : : : [ Yn , where Yi ` P and Yi C.X; . "; "//
for each i 2 f1; : : : ; ng.
422. Suppose that Cp .X / is a finite union of its paracompact (not necessarily
closed) subspaces. Prove that Cp .X / is Lindelf and hence paracompact.
423. Suppose that Cp .X / D Y1 [ : : : [ Yn , where Yi is realcompact for each i n.
Prove that Cp .X / is realcompact.
424. Suppose that Cp .X / D Y1 [ : : : [ Yn , where Yi is homeomorphic to Ri for
each i n. Prove that X is discrete.
425. Suppose that Cp .X / D Y1 [ : : : [ Yn , where Yi is hereditarily realcompact
for each i n. Prove that iw.Cp .X // D .Cp .X // D ! and hence Cp .X / is
hereditarily realcompact.
426. Given an infinite cardinal , suppose that the space Cp .X / is a finite union of
its -monolithic (not necessarily closed) subspaces. Prove that Cp .X / must be
-monolithic.
427. Given an infinite cardinal suppose that the space Cp .X / is a finite union of
its spread()-monolithic (not necessarily closed) subspaces. Prove that Cp .X /
is spread()-monolithic.
428. Given an infinite cardinal suppose that the space Cp .X / is a finite union
of its hd./-monolithic (not necessarily closed) subspaces. Prove that Cp .X /
must be hd./-monolithic.
429. Given an infinite cardinal suppose that the space Cp .X / is a finite union
of its hl./-monolithic (not necessarily closed) subspaces. Prove that Cp .X /
must be hl./-monolithic.
430. Suppose that Cp .X / is a finite union of its Dieudonn complete subspaces.
Prove that Cp .X / is realcompact and hence Dieudonn Scomplete.
431. Let X be an arbitrary space. Suppose that Cp .X / D fZn W n 2 !g. Prove
that there exists a function f 2 Cp .X / and " > 0 such that for some n 2 !,
the set .Zn C f / \ C.X; . "; "// is dense in Cu .X; . "; "// and hence also in
Cp .X; . "; "//. S
432. Suppose that Cp .X / D fZn W n 2 !g and each Zn is closed in Cp .X /.
Prove that some Zn contains a homeomorphic copy of S Cp .X /.
433. Let P be a hereditary property. Suppose that Cp .X / D fZn W n 2 !g, where
each Zn is closed in Cp .X / and has P. Prove that Cp .X / also has P.
S
434. Suppose that Cp .X / D fZn W n 2 !g, where each Zn is locally compact.
Prove that X is finite. S
435. Suppose that Cp .X / D fZn W n 2 !g and each Zn is locally pseudocom-
pact. Prove that Cp .X / isS-pseudocompact.
436. Suppose that Cp .X / D fZn W n 2 !g, where each Zn is realcompact and
closed in Cp .X /. Prove that Cp .X / is realcompact.
437. Prove that any metacompact collectionwise-normal space is paracompact.
438. Prove that if Cp .X / is normal and metacompact, then it is Lindelf.
439. Prove that Cp .!/ is not metacompact.
440. Prove that Cp .L.// is not metacompact for any uncountable .
441. Prove that neither the Baire property nor pseudocompleteness is countably
additive in spaces Cp .X /.
442. Prove that -weight and -character are not countably additive in spaces
Cp .X /.
443. Suppose that the space Cp .X / is a countable union of its Cech-complete (not
necessarily closed) subspaces. Prove that X is countable and discrete (and
hence Cp .X / is Cech-complete).
444. Given an infinite cardinal suppose that Cp .X / is a union of countably many
(not necessarily closed) subspaces of character . Prove that .Cp .X //
and hence jX j .
445. Prove that weight is countably additive in spaces Cp .X /.
446. Prove that metrizability is countably additive in spaces Cp .X /.
447. Prove that tightness is countably additive in spaces Cp .X /.
448. Prove that pseudocharacter is countably additive in spaces Cp .X /.
449. Prove that i -weight and diagonal number are countably additive in spaces
Cp .X /.
450. Prove that the FrchetUrysohn property is countably additive in spaces
Cp .X /. S
451. Suppose that X is a metrizable space and Cp .X / D fYi W i 2 !g, where Yi
is hereditarily realcompact (not necessarily closed) for every i 2 !. Prove that
nw.Cp .X // D iw.Cp .X // D ! and hence Cp .X / is hereditarily
S realcompact.
452. Suppose that X is a pseudocompact space and Cp .X / D fZn W n 2 !g,
where each Zn is paracompact and closed in Cp .X /. Prove that Cp .X / is
Lindelf.
453. Give an example of a non-normal space which is a countable union of its
closed normal subspaces. S
454. Let X be a compact space. Suppose that Cp .X / D fZn W n 2 !g, where
each Zn is normal and closed in Cp .X /. Prove that Cp .X / is Lindelf.
455. Let X be a metrizable space. Suppose that Cp .X / is a countable union of its
(not necessarily closed) normal subspaces. Prove that X is second countable
and hence Cp .X / is normal.
456. Let X be an arbitrary space. Given a uniformly dense Y Cp .X /, prove that
t.X / l.Y /.
457. For an arbitrary space X and a uniformly dense Y Cp .X / prove that
nw.Y / D nw.Cp .X // and d.Y / D d.Cp .X //.
458. For an arbitrary space X and a uniformly dense Y Cp .X / prove that

hd.Y / D hd.Cp .X //; hl.Y / D hl.Cp .X // and s.Y / D s.Cp .X //.
459. Suppose that X is a space and Y Cp .X / is uniformly dense in Cp .X /.
Prove that if Y is a Lindelf -space, then Cp .X / is also Lindelf .
460. Suppose that X is a space and Y Cp .X / is uniformly dense in Cp .X /.
Prove that
(i) if Y is K-analytic, then Cp .X / is K-analytic;
(ii) if Y is analytic, then Cp .X / is analytic.
461. Given a space X and a uniformly dense subspace Y Cp .X /, prove that
t.Y / D t.Cp .X //.
462. Let X be a space with ext .X / . Prove that t.Y / for any compact
Y Cp .X /.
463. Suppose that X has the Gerlits property '. Prove that all continuous images
and all closed subspaces of X have '.
464. Prove that Cp .X / is a FrchetUrysohn space if and only if X has the Gerlits
property ' and t.Cp .X // D !.
465. (GerlitsPytkeev theorem) Prove that the following conditions must be equiv-
alent for any space X :
(ii) Cp .X / is a sequential space;
(iii) Cp .X / is a k-space.
S
466. Suppose that X is a -compact space such that Cp .X / D n2! Yn where Yn
is a k-space for every n 2 !. Prove that Cp .X / is a FrchetUrysohn space.
In particular, if X is -compact and Cp .X / is a countable union of sequential
spaces, then Cp .X / is a FrchetUrysohn space.
467. Given spaces X and Y suppose that ' W X ! Y is a continuous map and let
' .f / D f ' for any f 2 Cp .Y /; then ' W Cp .Y / ! R D ' .Cp .Y //
Cp .X /. We can define a map r' W Cp .Cp .X // ! Cp .Cp .Y // by the equality
r' ./ D .jR/ ' for any 2 Cp .Cp .X //. Prove that r' is a continuous
ring homomorphism such that r' jX D ' (here we identify X and Y with their
canonical copies in Cp .Cp .X // and Cp .Cp .Y //, respectively). Prove that '
is the unique continuous ring homomorphism with this property, i.e., if s W
Cp .Cp .X // ! Cp .Cp .Y // is a continuous ring homomorphism such that
sjX D ', then s D r' .
468. Suppose that X is an !-monolithic compact space. Prove that for every
surjective continuous map ' W X ! Y , the map r' W Cp .Cp .X // !
Cp .Cp .Y // is surjective.
469. Given spaces X and Y let ' W X ! Y be a continuous onto map. Prove that
the mapping r' W Cp .Cp .X // ! r' .Cp .Cp .X /// Cp .Cp .Y // is open if
and only if ' is R-quotient.
470. Suppose that there exists a continuous map of Cp .X / onto Cp .Y /. Prove that
nw.Y / nw.X /.
iw.Y / iw.X /.
s .Y / s .X /; hl .Y / hl .X / and hd .Y / hd .X /.
if X is -monolithic, then Y is also -monolithic.
474. Suppose that there exists a quotient map of Cp .X / onto Cp .Y /. Prove that
l .Y / l .X / and q.Y / q.X /.
475. Suppose that there exists a quotient map of Cp .X / onto Cp .Y /. Prove that if
X is l ./-monolithic, then Y is also l ./-monolithic.
476. Suppose that there exists a continuous open map of Cp .X / onto Cp .Y /. Prove
that jY j jX j.
477. Suppose that there exists a continuous open map of Cp .X / onto Cp .Y /. Prove
that if X is -scattered, then Y is also -scattered.
478. Suppose that there exists a continuous closed map of Cp .X / onto Cp .Y /.
Prove that if X is -stable, then Y is also -stable.
479. Give an example of spaces X and Y for which there is a continuous map of
Cp .X / onto Cp .Y / while jY j > jX j.
Cp .X / onto Cp .Y / while l .Y / > l .X /.
Cp .X / onto Cp .Y / while q.Y / > q.X /.
Cp .X / onto Cp .Y / while X is compact and Y is not -compact.
483. Give an example of spaces X and Y for which there is an open continuous
map of Cp .X / onto Cp .Y / while d.Y / > d.X /.
map of Cp .X / onto Cp .Y / while tm .Y / > tm .X /.
map of Cp .X / onto Cp .Y / while c.Y / > c.X / and p.Y / > p.X /.
Cp .X / onto Cp .Y / while X is discrete and Y is not discrete.
487. Suppose that there exists a perfect map of Cp .X / onto Cp .Y /. Prove that
d.X / D d.Y /.
nw.X / D nw.Y /.
jX j D jY j.
hd .X / D hd .Y /.
491. Can Cp .!n!/ be mapped continuously onto Cp .!1 /?
492. Suppose that there exists a perfect irreducible map ' W Cp .X / ! R for some
cardinal . Prove that X is discrete.
Q
493. Let M D t 2T Mt where Mt is a metrizable space for all t 2 T ; assume that
' W M ! Cp .X / is a closed continuous onto map. Prove that for Q every t 2 T ,
we can choose a closed separable Nt Mt in such a way that '. t 2T Nt / D
Cp .X /. In particular, a space Cp .X / is a closed continuous image of a product
of (completely) metrizable spaces if and only if it is a closed continuous image
of a product of separable (completely) metrizable spaces.
494. Let M be a product of completely metrizable spaces. Suppose that there exists
a continuous closed onto map ' W M ! Cp .X /. Prove that X is discrete. In
particular, X is discrete if Cp .X / is a closed continuous image of R for some
cardinal .
495. Let X be a pseudocompact space. Suppose that Cp .X / contains a dense
subspace which is a continuous image of a product of separable spaces. Prove
that X is compact and metrizable.
496. Let X be a pseudocompact space. Suppose that Cp .X / is a closed continuous
image of a product of metrizable spaces. Prove that X is countable.
497. Let X be a pseudocompact space. Suppose that Cp .X / is an open continuous
image of a product of separable metrizable spaces. Prove that X is countable.
498. Let M be a product of separable completely metrizable spaces. Assuming that
there is a finite-to-one open map ' W M ! Cp .X /, prove that X is discrete.
499. Let M be a product of separable completely metrizable spaces. Assuming that
there is a finite-to-one open map ' W Cp .X / ! M prove that X is discrete.
500. Let H be a G -subspace of R for some . Prove that if Cp .X / is
homeomorphic to a retract of H , then X is discrete. In particular, if Cp .X / is
homeomorphic to a retract of R for some cardinal , then X is discrete.
1.6 Bibliographic Notes
The main text of Chap. 1 consists of problems of the following types:

(i) textbook statements which give a gradual development of some topic;
(ii) folkloric statements that might not be published but are known by specialists;
(iii) famous theorems cited in textbooks and well-known surveys;
(iv) comparatively recent results which have practically no presence in textbooks.
We will almost never cite original papers for the results of the first three types. We
will cite them for a very small selection of results of the fourth type. This selection
is made according to the preferences of the author and does not mean that all
statements of the fourth type will be mentioned. I bring my apologies to readers who
might think that I did not select something more important than what is selected.
The point is that such a selection has to be subjective because it is impossible to
mention all contributors. As a consequence, there are many statements which are
published as results in papers, but this fact is never mentioned in this book. There
are statements of the main text which constitute published or unpublished results
1.6 Bibliographic Notes 43
of the author. However, they are treated exactly like the results of others: some are
mentioned and some arent. On the other hand, the bibliography contains (to the best
knowledge of the author) the papers and books of all contributors to the material of
this book.
Section 1.1 contains quite a few nontrivial facts from set theory. This material
is best covered in Kunens book (1980). The general topology part is a group of
results (sometimes quite difficult ones) which are present in all major surveys on
cardinal invariants. See, e.g., Arhangelskii (1978), Arhangelskii and Ponomarev
(1974) and Juhsz (1980). The Cp -theory part is concerned with the behavior
of spread, hereditary Lindelf number and hereditary density. The main duality
theorems (Problems 025030) were proved in Velichko (1981) and Zenor (1980).
The results on perfect normality of Cp .X / (Problems 081089) were published in
Tkachuk (1995). The multiplicativity of countable spread and countable hereditary
Lindelf number under SA was proved in Arhangelskii (1989b). A highly nontrivial
example of a compact strong S -space (Problem 099) was constructed in Todorcevic
(1989).
Section 1.2 basically consists of textbook results of Arhangelskii on monolithity
and stability and their development in Tkachuk (1991). The books of Arhangelskii
(1992a) and Bessaga and Pelczinski (1975) give all background material and more.
The SA results in Problems 195 and 196 were published by Arhangelskii in 1996b.
Section 1.3 contains an introduction to Lindelf -spaces. Most results on this
class are folkloric and their formulations are dispersed in many papers and books.
However, the author knows no source where they would all be presented with
complete proofs and systematized. The introductory material culminates in a famous
theorem of Baturov (Problem 269) proved in 1987. We postpone the applications
of Baturovs theorem until the next chapter. The investigations of small diagonals
were triggered by a paper of Huek (1977). One of the most important results here
is Juhsz theorem (Problem 298) published in Juhsz (1992).
Section 1.4 constitutes a brief introduction to descriptive set theory and its
applications in Cp -theory. The textbook results can be found in Kuratowski (1966).
Problem 354 is, in fact, a very deep result of Pytkeev (1976). The statement of
Problem 366 was proved by Christensen (1974). The examples from Problem 372
were constructed in Lutzer et al. (1985). The statement of Problem 395 was proved
in Fremlin (1977).
In Sect. 1.5, the results on countable additivity of properties in Cp .X / were
proved in Tkachuk (1994). The equivalencies in Problem 465 were proved in Gerlits
(1983) and Pytkeev (1992a). The result of Problem 500 was established by Tkachuk
in the paper (1994).
Chapter 2
Solutions of Problems 001500
This chapter brings the reader to the cutting edge of many areas of Cp -theory, so
the treatment of topology and Cp -theory is already professional. When you read
a solution of a problem of the main text, it has more or less the same level of
exposition as a published paper on a similar topic. Recall that we already gave
500 solutions of the problems of the main text and proved a total of more than 200
statements inside the solutions. These inside statements (called facts) are sometimes
quite difficult and are often a particular case or the whole of some famous theorems.
A reader who mastered this material is more than prepared for reading the most
advanced papers in Cp -theory.
The author hopes, however, that reading our solutions is more helpful than
ploughing through the proofs in published papers; the reason is that we are not
so constrained by the amount of the available space as a journal contributor, so we
take much more care about all details of the proof. It is also easier to work with the
references in our solutions than with those in research papers because in a paper
the author does not need to bother about whether the reference is accessible for the
reader whereas we only refer to what we have proved in this book apart from some
very simple facts of calculus and set theory.
Another important difference between this chapter and the first one is that we use
the textbook facts from general topology without giving a reference to them. This
book is self-contained so all necessary results are proved in the first chapter, but the
references to standard things have to stop sometime. This makes it difficult for a
beginner to read the second chapter results without reading the first one. However, a
reader who mastered the material of the first four chapters of Engelking book (1977)
will have no problem with this.
We also stopped giving references to some very simple facts of Cp -theory. The
reader can easily find the respective proofs using the index. Our reference omission
rule can be expressed as follows: we omit references to textbook results from
topology proved in the first volume. We omit references to some very simple and
fundamental facts of Cp -theory also proved in the first volume. We denote the first

46 2 Solutions of Problems 001500
volume by [TFS], so a reference Problem 225 of [TFS] says that we refer to

the statement of Problem 225 of A Cp -Theory Problem Book. Topological and
Function Spaces.
When we refer to a solution, please, keep in mind that the solutions of this volume
are denoted by the letter T. Therefore, T.337 means the solution of Problem 337
of this book. The reference S.118 stands for solution of Problem 118 of [TFS].
There are quite a few phrases like it is easy to see or it is an easy exercise;
the reader should trust the authors word and experience that the statements like
that are really easy to prove as soon as one has the necessary background. On the
other hand, the highest percentage of errors comes exactly from omissions of all
kinds, so my recommendation is that, even though you should trust the authors
claim that the statement is easy to prove or disprove, you shouldnt take just his
word for the truthfulness of any statement. Verify it yourself and if you find any
errors communicate them to me to correct the respective parts.
T.001. Given an infinite cardinal prove that the following properties are equivalent
for any space X :
(i) hl.X / ;
(ii) l.X / and every U 2 .X/ is a union of -many closed subsets of X ;
(iii) l.X / and every closed F X is a G -set in X ;
(iv) l.U / for any open U X .
In particular, a space X is hereditarily Lindelf if and only if it is Lindelf and
perfect.
Solution. Say that H X is an F -set if H is a union of -many closed subsets
of X .
(i)H)(ii). Assume that hl.X / and take any U 2 .X/. For any point
x 2 U there is Ux 2 .x; X/ such that U x U . Since l.U / , the open cover
fUx W x 2 U g of the space S U has a subcover of cardinalityS , i.e., there is A U
such that jAj and fUx W x 2 Ag D U . Therefore fU x W x 2 Ag D U and
hence U is an F -set.
It is a matter of passing to complements to see that (ii) (iii).
(ii)H)(iv).
S If (ii) holds, then take any U 2 .X/; then U is an F -set, i.e.,
U D F where jF j and each F 2 F is closed in X ; thus l.F / for
any F 2 F . It is an easy exercise to see that a union -many spaces of Lindelf
number has the Lindelf number so l.U / .
(iv)H)(i). Suppose that l.U / Sfor each open U X and take an arbitrary
set Y X ; given any
.Y / with
SD Y , for each U 2
choose OU 2 .X/
such that OU \ Y D U . The set O D fOU W U 2
g is open in X and hence
0
l.O/ ; since S fOU W U 2 0
g is an open cover of O, thereS 0is

such that
0
j
j and fOU W U 2
g D O. It is immediate that
D Y which proves
that l.Y / .
T.002. Prove that a space X is hereditarily normal if and only if any open subspace
of X is normal.
2 Solutions of Problems 001500 47
Solution. We must only prove that X is hereditarily normal if every open subspace
of X is normal. Assume that every W 2 .X/ is normal; take any Y X and
any disjoint sets F; G Y which are closed in Y . It is easy to see that P D
F \ G X nY (the bar denotes the closure in X ). Therefore the open set U D
X nP contains Y . The space U is normal; it is immediate that F 0 D F \ U and
G 0 D G \ U are disjoint closed subsets of U . Choose any open (in U and hence
in X ) sets O1 and O2 such that F 0 O1 ; G 0 O2 and O1 \ O2 D ;. Finally
observe that O10 D O1 \ Y and O20 D O2 \ Y are disjoint open subsets of Y such
that F O1 and G O2 , so Y is normal.
T.003. Prove that if X is perfectly normal, then any Y X is also perfectly normal.
Solution. Any open U X must be an F -set in X because X is perfect. Since
any F -subset of a normal space is normal (Fact 1 of S.289), the space U is normal.
It turns out that any open subset of X is normal. Hence any Y X is normal by
Problem 002. Since it is trivial that any subspace of X is perfect, we showed that
any Y X is perfectly normal.
T.004. Let X be any space. Prove that hd.X / D supfjAj W A is a left-separated
subspace of X g. In particular, the space X is hereditarily separable if and only if
every left-separated subspace of X is countable.
Solution. If is an ordinal and Y D fy W < g is a set indexed by the set ,
say that a well-order < on Y is induced by the indexation if y < y0 is equivalent
to < 0 for any ; 0 < . We say that a well-order < on a space Y witnesses
left-separation of Y (or left-separates Y , or Y is left-separated by <) if the set
LYx D fy 2 Y W y < xg is closed in Y for any x 2 Y .
Fact 1. Suppose that Y is a left-separated space with its left-separation witnessed
by a well-order <. Then any subspace Z Y is also left-separated by the order
restricted to Z from Y .
Proof. Denote by the restriction of the order < to Z. Then, for any x 2 Z, we
have LZx D fz 2 Z W z xg D Lx \ Z. Therefore Lx is closed in Z for any
Y Z
x 2 Z, so Z is left-separated by . Fact 1 is proved. t

u
Fact 2. Let be a regular infinite cardinal. If Y is a space which has a left-separated
subspace of cardinality , then there exists a subspace Z D fz W < g Y such
that Z is left-separated by its indexation.
Proof. Take any T Y which is left-separated by an order < and jT j D . There
0
is T 0 T such that jT 0 j D and the cardinality of the set LTx is less than for
each x 2 T 0 . Indeed, if jLTx j < for each x 2 T , we can take T 0 D T . If this is
not true, then let x0 D minfy 2 T W jLTy j D g. It is evident that T 0 D LTx0 is as
promised.
Now pick z0 2 T 0 arbitrarily; assume that < and we have constructed a set
Z D fz W < g T 0 such that
(*) < 0 implies z < z0 whenever < 0 < .
S 0
If Z is cofinal in T 0 , then T 0 D fLTx W x 2 Z g which is impossible because
0
jLTx j < for each x 2 Z and the cardinal is regular. Thus there is z 2 T 0 such
that z < z for all < . It is immediate that . / holds for all and hence
our inductive construction can be continued to provide a set Z D fz W < g T 0
with the property . / for all < . It is evident that the order < on Z is induced
by the indexation of Z. Since the order < left-separates Z by Fact 1, the space Z is
left-separated by its indexation, so Fact 2 is proved. t
u
Returning to our solution assume that hd.X / and X has a left-separated
subspace of cardinality C ; by Fact 2 there is Y D fx W < C g X which is
left-separated by its indexation. Since d.Y / , there is A Y with jAj and
Y A (the bar denotes the closure in X ). There exists an ordinal < such that
A X D fx W < g; the set X being closed in Y , we have x X , whereas
x 2 A X . This contradiction proves that the cardinality of any left-separated
Y X does not exceed hd.X /.
Now suppose that jAj for any left-separated A X . Assume that there is
Y X such that d.Y / C and pick any y0 2 Y . Suppose that < C and we
have chosen points fy W < g in the set Y so that y
fy W <
g for each
< . Since d.Y / > , the set fy W < g cannot be dense in the space Y ;
choose any y 2 Y nfy W < g. It is clear that this inductive construction gives
us a set A D fy W < C g X such that y fy W < g for each < C .
Therefore A is a left-separated (by its indexation) subspace of X with jAj > which
is a contradiction. This proves that hd.X / , so our solution is complete.
T.005. Let X be any space. Prove that hl.X / D supfjAj W A is a right-separated
subspace of X g. In particular, the space X is hereditarily Lindelf if and only if
every right-separated subspace of X is countable.
Solution. If is an ordinal and Y D fy W < g is a set indexed by the set ,
say that a well-order < on Y is induced by the indexation if y < y0 is equivalent
to < 0 for any ; 0 < . We say that a well-order < on a space Y witnesses
right-separation of Y (or right-separates Y , or Y is right-separated by <) if the set
LYx D fy 2 Y W y < xg is open in Y for any x 2 Y .
Fact 1. Suppose that Y is a right-separated space with its right-separation witnessed
by a well-order <. Then any subspace Z Y is also right-separated by the order
restricted to Z from Y .
Proof. Denote by the restriction of the order < to Z. Then, for any x 2 Z, we
have LZx D fz 2 Z W z xg D Lx \ Z. Therefore Lx is open in Z for any x 2 Z,
Y Z
so Z is right-separated by . Fact 1 is proved. t

u
Fact 2. Let be a regular infinite cardinal. If a space Y has a right-separated
subspace of cardinality , then there exists a subspace Z D fz W < g Y
such that Z is right-separated by its indexation.
Proof. Take any T Y which is right-separated by an order < and jT j D . There

0
is T 0 T such that jT 0 j D and the cardinality of the set LTx is less than for
each x 2 T 0 . Indeed, if jLTx j < for each x 2 T , we can take T 0 D T . If this is
not true, then let x0 D minfy 2 T W jLTy j D g. It is evident that T 0 D LTx0 is as
promised.
Now pick z0 2 T 0 arbitrarily; assume that < and we have constructed a set
Z D fz W < g T 0 such that
(*) < 0 implies z < z0 whenever < 0 < .
S 0
If Z is cofinal in T 0 , then T 0 D fLTx W x 2 Z g which is impossible because
0
jLTx j < for each x 2 Z and the cardinal is regular. Thus there is z 2 T 0 such
that z < z for all < . It is immediate that . / holds for all and hence
our inductive construction can be continued to provide a set Z D fz W < g T 0
with the property . / for all < . It is evident that the order < on Z is induced by
the indexation of Z. Since the order < right-separates Z by Fact 1, the space Z is
right-separated by its indexation, so Fact 2 is proved. t
u
Returning to our solution assume that hl.X / and X has a right-separated
subspace of cardinality C . By Fact 2 we can assume, without loss of generality,
that there exists a subset Y D fx W < C g X right-separated by its indexation.
If U D fx W < g, then U D fU W < C g is an open coverS of the space
Y . Since l.Y / , there is A C with jAj such that Y D fU W 2
Ag. There exists an ordinal < such that AS ; it is immediate that U
U Y nfy g for all 2 A. Therefore x fU W 2 Ag U which is a
contradiction. This contradiction proves that the cardinality of any right-separated
Y X does not exceed hl.X /.
Now suppose that jAj for any right-separated A X . Assume that there
is Y X such that l.Y / C and take any open cover U of the space Y which
has no subcover of cardinality . Pick any y0 2 Y and any U0 2 U for which
y0 2 U0 . Suppose that < C and we have chosen aSset fy W < g Y
and a family fU W < g U such that y
2 U
n fU W <
g for each
< . Since S l.Y / > , the set fU W < g cannot be a cover of Y ; choose
any y 2 Y n fU W < g and any U 2 U with y 2 U . It is clear that
this inductive construction gives us a set ASD fy W < C g X and a family
fU W < C g U such that y 2 U n fU W < g for each < C . It is
easy to see that the set fy W < g is open in A for any < C and therefore
A is a right-separated (by its indexation) subspace of X with jAj > which is a
contradiction. This proves that hl.X / , so our solution is complete.
T.006. Prove that a space is right-separated if and only if it is scattered.
Solution. If X is a right-separated space, take any well-order < on X which

witnesses this property. Given any Y X , let y 2 Y be the <-minimal point
of Y . If y 2 Y nfyg, then Y nfyg ; so we can choose a <-minimal element
x 2 Y nfyg. The set U D fz 2 X W z < xg is open in X and U \ Y D fyg which
contradicts the fact that y 2 Y nfyg. This proves that y is an isolated point of Y and
hence X is scattered.
Given a space Z we denote by IP .Z/ the set of isolated points of Z. It is clear

that the set ZnIP .Z/ is closed in Z. Now assume that X is a non-empty scattered
space and let U0 D ;. Suppose that < D jX jC and we have chosen U 2 .X/
for all < so that the following conditions are satisfied:
(1) U U0 if < 0 < ; S
(2) if
< is a limit ordinal, then U
D fU W <
g;
(3) if C 1 < , then UC1 D U [ IP .X nU /.
S
If S fU W < g D X , then our inductive construction stops. If not, we let
U D fU W < g if is a limit ordinal; if D
C 1, then let U D U
[
IP .X nU
/. It is worth mentioning that, in case of a successor , the set U is open
in X because the set X nU D .X nU
/nIP .X nU
/ is closed in X nU
and hence
in X . Observe also that X is scattered so it follows from (3) that UC1 nU ;
whenever C 1 .
It is evident that the properties (1)(3) are now satisfied for all so our
inductive construction can be continued as long as the union of chosen sets is
not equal to X . The family U D fUC1 nU W < g is disjoint for any limit
ordinal . Since U consists of non-empty
S sets, our construction cannot involve
steps because jX j < . Therefore fU W < g D X for some < .
Given any x 2 X , let .x/ D minf < W x 2 U g. It follows from (2) that
.x/ is a successor ordinal for all x 2 X . The properties (1) and (2) imply that the
family fUC1 nU W C 1 g is a disjoint cover of X ; fix a well-order < on the
set UC1 nU for each with C 1 .
We are now in position to define the required well-order on X . For any distinct
x; y 2 X , let x < y if .x/ < .y/; if .x/ > .y/, then y < x. Now, if we
have the equality .x/ D .y/ D C 1, then fx; yg AC1 nA ; we let x < y if
x < y and y < x if y < x. It is routine to verify that < is indeed a well-order
on X . Let us show that it witnesses the fact that X is right-separated.
Given any point x 2 X , there is < such that C 1 D .x/. Observe that
W D fy 2 X W y < xg D U [P where P D fy 2 UC1 nU W y < xg. Therefore
X nW D .X nU /nP is a closed set in X nU because P consists of isolated points
of X nU by (3). Any closed subset of X nU is closed in X so W is open in X . This
proves that X is right-separated.
T.007. Let X be a left-separated space. Prove that hl.X / s.X /. In particular, any
left-separated space of countable spread is hereditarily Lindelf.
Solution. The following fact is very useful in any consideration where spread is
involved.
Fact 1. Given a space Z with s.Z/ and an open cover U of the space Z,
0 0
S 0 a discrete D Z and a family U U such that jU j and Z D
there exists
D [ . U /.
Proof. Take any z0 2 Z and any U0 2 U with z0 2 U0 ; suppose that < C and we
have chosen a set fz W < g Z and a family fU W < g U such that
S
(*) z
2 U
n.fz W <
g [ . fU W <
g// for all
< .
Observe that . / implies that the set D D fz W < g is discrete because

S
U
\ D D fz
g for each
< . If Z D fz W <
g [ . fU W <
g/ Z,
then take any z 2 ZnZ and U 2 U with z 2 U . It is immediate that . / holds
for all
, so our inductive construction can go on as long as Z Z. If we
carry it out for all < C , then we obtain a discrete set D C Z of cardinality C
which contradicts s.Z/ . Thus we have Z D Z for some < C , so D D D
and U 0 D fU W < g are as promised. Fact 1 is proved. t
u
Returning to our solution let s.X / D ; fix any order on X which
witnesses that X is left-separated. Given any Z X and z 2 Z, let LZ z D
fy 2 Z W y zg. To prove that hl.X / assume the contrary; then
H D fH X W l.H / > g ;. We claim that there is Y 2 H such that l.LYx /
for any x 2 Y . Indeed, take any H 2 H and consider A D fz 2 H W l.LH z / > g.
If A D ;, then we can take Y D H . If A ;, then, for the point y D min.A/, the
set Y D LH y is as promised.
Take any open cover U of the space Y which has no subcover of cardinality .
It follows from our choice of Y S that for any x 2 Y , there is a family Ux U
such that jUx j and LYx Ux . Apply Fact 1S to find a family U 0 U and a
discrete set D Y such that jU 0 j and Y D U 0 [ D (the bar denotes the
closure in Y ). We have jDj because s.YS / s.X / . The set D cannot
be cofinal in Y ; for otherwise, the family V D fUx W x 2 Dg is a subfamily of U
of cardinality which covers Y . Thus there is z 2 Y such that D LYz . Since
the order left-separates Y (Fact
S 01 of T.004), the set Lz 0 is closed in Y so D Lz .
Y Y
As a consequence, Y nLz U and therefore Uz [ U U is a subcover of Y of

Y
cardinality which is a contradiction. This shows that H D ;, so our solution is

complete.
T.008. Let X be a right-separated space. Prove that hd.X / s.X /. In particular,
any right-separated space of countable spread is hereditarily separable.
Solution. Let s.X / D ; suppose that hd.X / > . Then there is a left-separated
Y X with jY j D C by Problem 004. Since s.Y / s.X / D and Y is left-
separated, we can apply Problem 007 to conclude that hl.Y / . The space Y is
a subspace of a right-separated space X , so Y is right-separated by Fact 1 of T.005;
since jY j D C , we have hl.Y / D C by Problem 005. This contradiction shows
that hd.X / D s.X /.
T.009. Prove that any space has a dense left-separated subspace.
Solution. Take any space X and fix a well-order < on X . For any U 2 .X / let xU
be the <-minimal point of U . We claim that Y D fxU W U 2 .X /g is a dense left-
separated subspace of X . The density of Y follows from the fact that xU 2 Y \ U
for any U 2 .X /. Now pick any y 2 Y and U 2 .X / such that y D xU .
To prove that the set Ly D fz 2 Y W z < yg is closed in Y take any x 2 Y nLy ;
then x D xW for some W 2 .X /. Given any z 2 Ly , we have z < y x, so
z cannot belong to W because x is the <-minimal point of W . This shows that W
is a neighborhood of x with W \ Ly D ; so x Ly . The point x 2 Y nLy has

been chosen arbitrarily, so Ly is closed in Y for any y 2 Y . Therefore Y is a dense
left-separated subspace of X .
T.010. Suppose that s.X / D !. Prove that X has a dense hereditarily Lindelf
subspace.
Solution. There is a dense left-separated Y X . Since s.Y / s.X / !, we can
apply Problem 007 to conclude that hl.Y / s.Y / D ! and therefore Y is a dense
hereditarily Lindelf subspace of X .
T.011. Prove that for any space X , we have hl .X / D hl.X ! /. In particular, if all
finite powers of X are hereditarily Lindelf, then X ! is hereditarily Lindelf.
Solution. It is evident that X n embeds in X ! for all n 2 N, so it suffices to show
that hl.X ! / hl .X /. For each n 2 N let n W X ! ! X n be the natural projection
defined by n .x/ D xjn for all x 2 X ! (recall that we identify any ordinal with the
set of its predecessors and, in particular, n D f0; : : : ; n 1g). Note that the family
Bn D fU0

Un1 X n W Ui 2 .X/ W for all i < ng is aSbase in X n for all
n 2 N. Let Cn D fn1 .U / W U 2 Bn g for all n 2 N; then C D fCn W n 2 Ng is a
base in X ! .
Let hl .X / D ; suppose that hl.X ! / > . Then there is a right-separated
subspace of X ! of cardinality C (see Problem 005). Apply Fact 2 of T.005 to find
a subspace Y D fy W < C g X ! which is right-separated by its indexation.
The set L D fy W < C 1g is open in Y for all < C ; since C is a base in X ! ,
there is n 2 N and U 2 Cn such that y 2 U \ Y L for each < C . The
cardinal C is regular, so there is m 2 N such that the set A D f < C W n D mg
has cardinality C ; let Y 0 D fy W 2 Ag.
Consider the set Z D m .Y 0 / D fz D m .y / W 2 Ag X m . For each 2 A
take V 2 Bm such that m1 .V / D U . Observe that z 2 V for all 2 A; besides,
if > and 2 A, then y U . Since U D m1 .V /, we have z V for each
> with 2 A. In particular, the map m W Y 0 ! Z is a bijection.
Thus we can define a well-order on Z letting z z if and only if < .
Given an arbitrary z 2 Z, observe that z D z for some 2 A and therefore
Qz D fz0 2 Z W z0 zg D fz 2 Z W 2 S A and < g. Since z V
for each < with 2 A, for the set G D fV W 2 A and < g, we
have G \ Z D fz W 2 A and < g D Qz and hence the set Qz is open in
Z for each z 2 Z, i.e., Z is a right-separated subspace of X m . Since jZj D C ,
this contradicts hl.X m / hl .X / D (see Problem 005), so we proved that
hl.X ! / D hl .X / and therefore hl.X ! / D hl .X /.
T.012. Prove that for any space X , we have hd .X / D hd.X ! /. In particular, if
all finite powers of X are hereditarily separable, then X ! is hereditarily separable.
that hd.X ! / hd .X /. For each n 2 N let n W X ! ! X n be the natural
projection defined by n .x/ D xjn for all x 2 X ! (recall that we identify any
ordinal with the set of its predecessors and, in particular, n D f0; : : : ; n 1g). Note
that the family Bn D fU0

Un1 X n W Ui 2 .X/ W for all i < ng is
1
Sin X for all n 2 N. Let C!n D fn .U / W U 2 Bn g for all n 2 N; then
n
a base
C D fCn W n 2 Ng is a base in X .
Let hd .X / D ; suppose that hd.X ! / > . Then there is a left-separated
subspace of X ! of cardinality C (see Problem 004). Apply Fact 2 of T.004 to find
a subspace Y D fy W < C g X ! which is left-separated by its indexation. The
set L D fy W < g is closed in Y for all < C ; since C is a base in X ! , there is
n 2 N and U 2 Cn such that y 2 U \Y Y nL for each < C . The cardinal
C is regular, so there is m 2 N such that the set A D f < C W n D mg has
cardinality C ; let Y 0 D fy W 2 Ag.
if > and 2 A, then y U . Since U D m1 .V /, we have z V for each
> with 2 A. In particular, the map m W Y 0 ! Z is a bijection.
Thus we can define a well-order on Z letting z z if and only if < .
Given an arbitrary z 2 Z, observe that z D z for some 2 A and therefore
Qz D fz0 2 Z W z0 zg D fz 2 Z W S 2 A and < g. Since z V for
each < with 2 A, for the set G D fV W 2 A and g, we have
G \ Z D fz W 2 A and g D Y nQz , and hence the set Y nQz is open
in Z for each z 2 Z. As a consequence, Qz is closed in Z for all z 2 Z, so Z
is a left-separated subspace of X m . Since jZj D C , this contradicts hd.X m /
hd .X / D (see Problem 004), so we proved that hd.X ! / D hd .X / and
therefore hd.X ! / D hd .X /.
T.013. Prove that for any space X , we have s .X / D s.X ! /.
that s.X ! / s .X /. For each n 2 N let n W X ! ! X n be the natural projection
defined by n .x/ D xjn for all x 2 X ! (recall that we identify any ordinal with the
set of its predecessors and, in particular, n D f0; : : : ; n 1g). Note that the family
Bn D fU0

Un1 X n W Ui 2 .X/ W for all i < ng is aSbase in X n for all
n 2 N. Let Cn D fn1 .U / W U 2 Bn g for all n 2 N; then C D fCn W n 2 Ng is a
base in X ! .
Let s .X / D ; suppose that s.X ! / > . Then there is a discrete subspace
Y D fy W < C g X ! . Since C is a base in X ! , there is n 2 N and U 2 Cn
such that U \ Y D fy g for each < C . The cardinal C is regular, so there
is m 2 N such that the set A D f < C W n D mg has cardinality C ; let
Y 0 D fy W 2 Ag.
if and 2 A, then y U . Since U D m1 .V /, we have z V for each
with 2 A. In particular, the map m W Y 0 ! Z is a bijection. Besides,
z V for each implies that V \ Z D fz g for each 2 A and hence Z is
a discrete subspace of X m . Since jZj D C , this contradicts s.X m / s .X / D ,
so we proved that s.X ! / D s .X / and therefore s.X ! / D s .X /.
T.014. Suppose that s.X X / . Prove that hl.X / or hd.X / .

In particular, if s.X X / D !, then X is hereditarily separable or hereditarily
Lindelf.
Solution. Assume that hd.X / > and hl.X / > . It follows from hd.X / >
that there exists a left-separated subspace of X of cardinality C (see Problem 004).
Apply Fact 2 of T.004 to find a subspace Y D fy W < C g X which is
left-separated by its indexation. Analogously, hl.X / > implies that there exists a
subspace Z D fz W < C g X which is right-separated by its indexation (see
Fact 2 of T.005).
Let D D f.y ; z / W < g X X ; for any < C the set Y D fy W < g
is closed in Y , so there exists U 2 .y ; X / such that U \ Y D ;. Analogously,
the set fz W < C 1g is open in Z so the set Z D fz W < g is closed in Z
and therefore there exists V 2 .z ; X / such that V \ Z D ;. As a consequence,
W D U V 2 ..y ; z /; X X / and W \D D f.y ; z /g for each < C . This
shows that D is a discrete subspace of X X ; since jDj D C , this is a contradiction
with s.X X / . Thus it is impossible that hd.X / > and hl.X / > , so we
have hd.X / or hl.X / .
T.015. Prove that jX j 2hl.X / for any space X . In particular, any hereditarily
Lindelf space has cardinality c.
Solution. Given any set A, denote by P .A; / the family of all subsets of A of
cardinality . Then P .P .A; /; / is the collection of all families of cardinality
of subsets of A of cardinality each. If Z is an arbitrary space and we
have familiesSV .Z/S and D exp.Z/, say that the pair .V; D/ is incomplete if
O.V; D/ D . V/ [ . fD W D 2 Dg/ Z.
Fact 1. For any space Z, we have jZj 2 .Z/s.Z/
.
Proof. Let T .Z/
s.Z/ D . For any z 2 Z fix a family Vz .z; Z/ such that
jVz j and Vz D fzg. Take any z0 2 Z and let A0 D fz0 g. Suppose that < C
and we have constructed a family fA W < g of subsets of Z with the following
properties:
(1) jA j 2 for each < ;
(2) A A
whenever <
S< ; S
(3) for any
< , if A0
D fA W <
g and C
D fVz W z 2 A0
g, then for
any families D 2 P .P .A0
; /; / and V 2 P .C
; /, such that the pair .V; D/
is incomplete, we have A
nO.V; D/ ;.
S S
To construct A , let A0 D fA W < g and C D fVz W z 2 A0 g. Observe
that we have jP .P .A0 ; /; /j 2 and jC j 2 because jP .B; /j D 2 for any
set B of cardinality 2 . For any family V 2 P .C ; / and any D 2 P .P .A0 ; /; /
such that the pair .V; D/ is incomplete, choose a point p.V; D/ 2 ZnO.V; D/.
Now, let A D A0 [ fp.V; D/ W D 2 P .P .A0 ; /; /; V 2 P .C ; / and the pair
.V; D/ is incompleteg. It is immediate that the properties (1)(3) hold for the family
fA W g, so we can continue our inductive construction to obtain a family
A D fA W < C g with the properties (1)(3).
S
Assume first that A D A Z and fix any t 2 ZnA.SFor any U 2 Vt , and
any z 2 AnU , choose Vz 2 Vz with t Vz ; the family H D fVz W z 2 AnU g is an
open cover of AnU , so we can apply Fact 1 of T.007 to find a family HU H and
a set DU AnUS such that jHU j , the set DU isS discrete (and hence jDU j
as well) and . HU / [ D U S AnU . Since A D fAnU W U 2 Vt g, we have
A O.V;
S D/ where V D fHU W U 2 Vt g and D D fDU W U 2 Vt g. Let
W D fVz W z 2 Ag; since jVj ; jDj and V W, there exists < such
D P .P .A0 ; /; / and V 2 P .C ; /. Observe that D U ZnU for each U 2 Vt
S
and therefore t D U forS all U 2 Vt . In St
S addition, HU for each U 2 Vt
which implies that t V. Thus t . V/ [ . fD U W U 2 Vt g/ D O.V; D/
which shows that the pair .V; D/ is incomplete and therefore p D p.V; D/ 2 A .
However, A A O.V; D/ so p 2 ZnO.V; D/ ZnA; this contradiction
shows that A D Z and hence jZj C
2 D 2 D 2 .X /s.X /, so Fact 1 is proved.
t
u
Finally, observe that s.X / hl.X / and .X / hl.X /, so we can apply
Fact 1 to conclude that jX j 2s.X / .X / 2hl.X /hl.X / D 2hl.X / , so our solution
is complete.
T.016. Prove that s.X X / s.Cp .X // s .X / for any space X .
Solution. Given points x1 ; : : : ; xn 2 X and O1 ; : : : ; On 2 .R/, recall that the set
x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng is called a
standard open subset of Cp .X /. Standard open sets x1 ; : : : ; xn I O1 ; : : : ; On where
n 2 N; x1 ; : : : ; xn 2 X and O1 ; : : : ; On are rational intervals form a base in Cp .X /
(see Problem 056 of [TFS]).
Let s.Cp .X // D ; we will prove first that s.X / . Assuming the contrary
we can find a discrete subspace D X with jDj D C . For each d 2 D choose
Ud 2 .d; X/ such that U \ D D fd g and a function fd 2 C.X; 0; 1 / such that
fd .d / D 1 and fd j.X nUd / 0. Let Od D ff 2 Cp .X / W f .d / > 0g; it is
clear that Od is an open subset of Cp .X / and fd 2 Od for each d 2 D. The set
F D ffd W d 2 Dg is discrete; to see this, take any a 2 Dnfd g. Then d Ua
and hence fa .d / D 0 which shows that fa Od . This implies Od \ F D ffd g for
each d 2 D and therefore the correspondence d ! fd is a bijection between D
and F . As a consequence, F is a discrete subspace of Cp .X / of cardinality C ; this
contradiction with s.Cp .X // shows that s.X / .
Now assume that s.X X / > and fix a discrete subspace E X X with
jEj D C . The set D f.x; x/ W x 2 X g is a subspace of X X homeomorphic to
X so s./ . Thus jE \ j and hence jEnj D C , i.e., we have a discrete
subspace D D En .X X /n of cardinality C . For each d D .x; y/ 2 D
there exist Ud 2 .x; X/ and Vd 2 .y; X/ such that .Ud Vd / \ D D fd g. Since
x y, we lose no generality assuming that Ud \ Vd D ; for each d 2 D. Take
any functions gd ; hd 2 C.X; 0; 1 / such that gd .x/ D hd .y/ D 1; gd j.X nUd /
0; hd j.X nVd / 0 and let fd D gd hd . Then Od D ff 2 Cp .X / W f .x/ > 0
and f .y/ < 0g is an open subset of Cp .X / and fd 2 Od for all d 2 D.
We claim that the set F D ffd W d 2 Dg is discrete; to see this, assume that d D
.x; y/ 2 D and a D .s; t/ 2 Dnfd g. Since d Ua Va , we have x Ua or y Va .
In the first case we have fa .x/ 0 and in the second one we obtain fa .y/ 0;
thus, in both cases fa Od and therefore Od \ D D fd g and, in particular, the
correspondence d ! fd is a bijection between D and F . As a consequence, F is a
discrete subspace of Cp .X / of cardinality C ; this contradiction with s.Cp .X //
shows that s.X X / and therefore s.X X / s.Cp .X //.
To prove that s.Cp .X // s .X / assume that s .X / < s.Cp .X // and
fix a discrete subspace F Cp .X / with jF j D C . For each f 2 F , there
exists open standard set Of 2 .f; Cp .X // such that Of \ F D ff g. We have
f f f f f f
Of D x1 ; : : : ; xnf I O1 ; : : : ; Onf where x1 ; : : : ; xnf are distinct points of X
f f
and O1 : : : ; Onf are rational intervals. Since there are only countably many of all
possible n-tuples of rational intervals for any n 2 N, there exist n 2 N, a set G F
and rational intervals O1 ; : : : ; On such that jGj D C ; nf D n; Oi D Oi ; i D
f
f f
1; : : : ; n for each f 2 G. Let xf D .x1 ; : : : ; xn / 2 X n for all f 2 G. For the set
1 1
O.f / D f .O1 /

f .On / we have f 2 O.f / for all f 2 G. We claim
that the set D D fxf W f 2 Gg X n is discrete.
g
To show this, pick any f 2 G; if g 2 Gnff g, then f Og so f .xi / Oi
g 1
for some i n. This implies xi f .Oi / and therefore xg O.f /. An evident
consequence is that O.f / \ D D fxf g. Consequently, the subspace D is discrete
and the correspondence f ! xf is a bijection between G and D. This gives us
a discrete D X n of cardinality C which contradicts s.X n / s .X / D .
We established that s.Cp .X // D s .X /, so our solution is complete.
T.017. Prove that hd.X X / hl.Cp .X // hd .X / for any space X .
Solution. Given points x1 ; : : : ; xn 2 X and sets O1 ; : : : ; On 2 .R/, the set
x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng
is called a standard open subset of Cp .X /. Standard open sets x1 ; : : : ; xn I O1 ; : : : ;

On where n 2 N; x1 ; : : : ; xn 2 X and O1 ; : : : ; On are rational intervals form a
base in Cp .X / (see Problem 056 of [TFS]).
Let hl.Cp .X // D ; we will prove first that hd.X / . Assuming the contrary
we can find a left-separated subspace D X with jDj D C (see Problem 004).
Denote by < an order that left-separates D; for any point d 2 D, consider the set
Dd D fz 2 D W z < d g. For each d 2 D the set Dd is closed in D and d Dd ,
so we can choose a set Ud 2 .d; X/ with Ud \ Dd D ; and a function fd 2
C.X; 0; 1 / such that fd .d / D 1 and fd j.X nUd / 0. Let Od D ff 2 Cp .X / W
f .d / > 0g; it is clear that Od is an open subset of Cp .X / and fd 2 Od for each
d 2 D. Take any a 2 Dnfd g; if a < d , then a Ud so fd .a/ D 0 fa .a/ D 1.
Therefore fd fa . If d < a, then d Ua , and hence fa .d / D 0 fd .d / D 1
which shows again that fa fd . Consequently, the correspondence d ! fd is a
bijection, so we can define a well-order on the set F D ffd W d 2 Dg by letting
fd fa if and only if d < a.
We claim that the well-order right-separates F ; to see this, take any f 2 F .

To prove that the set Ff D fg 2 F W g f g is open in F , take any g f . There
is a 2 D such that g D fa ; given any f 0 2 F nFf , we have f 0 D fd for some
d 2 D. Since f 0 Ff , we have a < d . It follows from a Ud that we have
fd .a/ D 0 whence f 0 D fd Oa . The point f 0 2 F nFf was taken arbitrarily, so
we established that Oa \ F Ff . It is evident that g 2 Oa , so every g D fa 2 Ff
has a neighborhood W .g/ D Oa such that W .g/ \ F Ff . Thus Ff is open in
F for each f 2 F and therefore F is a right-separated subspace of Cp .X /. Since
jF j D jDj D C , this is a contradiction with hl.Cp .X // (see Problem 005)
which shows that hd.X / .
Now assume that hd.X X / > and fix a left-separated subspace E X X
with jEj D C . The set D f.x; x/ W x 2 X g is a subspace of X X
homeomorphic to X so hd./ . Thus jE \ j and hence jEnj D C ,
i.e., we have a left-separated subspace D D En .X X /n of cardinality C .
Denote by < an order that left-separates D and let Dz D fd 2 D W d < zg for any
z 2 D. For each point d D .x; y/ 2 D there exist Ud 2 .x; X/ and Vd 2 .y; X/
such that .Ud Vd / \ Dd D ;. Since x y, we lose no generality assuming that
Ud \ Vd D ; for every point d 2 D. Take any functions gd ; hd 2 C.X; 0; 1 / such
that gd .x/ D hd .y/ D 1; gd j.X nUd / 0; hd j.X nVd / 0 and let fd D gd hd .
Then Od D ff 2 Cp .X / W f .x/ > 0 and f .y/ < 0g is an open subset of Cp .X /
and fd 2 Od for all d 2 D.
If a; d 2 D and a d , we can assume, without loss of generality that a < d . If
a D .x 0 ; y 0 / and d D .x; y/, then a Ud Vd which implies x 0 Ud or y 0 Vd .
In the first case we have fd .x 0 / 0 while fa .x 0 / D 1 > 0; in the second case
we obtain fd .y 0 / 0 while fa .y 0 / D 1 < 0. Therefore in both cases we have
fa fd which proves that the correspondence d ! fd is a bijection. This makes
it possible to define a well-order on the set F D ffd W d 2 Dg by letting fa fd
if and only if a < d .
We prove next that the well-order right-separates F . Take any f 2 F ; to
prove that the set Ff D fg 2 F W g f g is open in F , take any g f . There
is a D .x; y/ 2 D such that g D fa ; given any f 0 2 F nFf , we have f 0 D fd
for some d D .x 0 ; y 0 / 2 D with a < d . It follows from a Ud Vd that x Ud
or y Vd ; in the first case we have fd .x/ 0 and in the second one we obtain
fd .y/ 0 so in both cases f 0 D fd Oa .
The point f 0 2 F nFf was taken arbitrarily, so we established that Oa \ F Ff
for any g D fa 2 Ff . Thus any g D fa 2 Ff has a neighborhood W .g/ D Oa
such that W .g/ \ F Ff ; this proves that Ff is open in F for each f 2 F
and therefore F is a right-separated subspace of Cp .X /. Since jF j D jDj D C ,
this is a contradiction with hl.Cp .X // (see Problem 005) which shows that
hd.X X / D hl.Cp .X //.
To prove that hl.Cp .X // hd .X / assume that hd .X / < hl.Cp .X //
and fix a right-separated subspace F Cp .X / with jF j D C . Let < be the well-
order that right-separates F . For each g 2 F , the set Fg D ff 2 F W g < f g is
closed in F . This is evident if Fg D ;; if not, then for the function g 0 D min Fg we
have Fg D F nff 2 F W f < g 0 g, so Fg is closed in F being the complement of
the set ff 2 F W f < g 0 g which is open in F because F is right-separated.
Therefore, for each f 2 F , there exists a standard open set Of 2 .f; Cp .X //

f f f f
such that Of \ Ff D ;. We have Of D x1 ; : : : ; xnf I O1 ; : : : ; Onf where
f f f f
x1 ; : : : ; xnf are distinct points of X and O1 ; : : : ; Onf are rational intervals. Since
there are only countably many of all possible n-tuples of rational intervals for any
n 2 N and C is a regular uncountable cardinal, there exist n 2 N, a set G F
and rational intervals O1 ; : : : ; On such that jGj D C ; nf D n; Oi D Oi ; i D
f
f f
1; : : : ; n for each f 2 G. Let xf D .x1 ; : : : ; xn / 2 X n for all f 2 G. For the set
1 1
O.f / D f .O1 /

f .On /, we have xf 2 O.f / for all f 2 G. We claim
that the set D D fxf W f 2 Gg X n is left-separated.
To show this, pick any f; g 2 G such that g < f . Since f 2 Fg , we have
f Og , so f .xi / Oi for some i n. This implies xi f 1 .Oi / and therefore
g g
xg O.f /. Thus we have proved that

(*) for any f; g 2 G such that g < f , we have xf 2 O.f / and xg O.f /.
An evident consequence is that xf xg for any distinct f; g 2 G, so the
correspondence f ! xf is a bijection between G and D. This makes it possible
to well-order the set D by letting xf xg if and only if f < g. Fix x 2 D;
we must prove that the set Dx D fy 2 D W y xg is closed in D. Take any
y 2 Dx ; there are f; g 2 G such that x D xf and y D xg . Since y x, we have
g < f and hence xg O.f / by . /. This shows that every x D xf 2 D has a
neighborhood Wx D O.f / such that Wx \ Dx D ;; an easy consequence is that Dx
is closed in D for any x 2 D, so D is left-separated by the well-order . However,
jDj D jGj D C and G X n which contradicts hd.X n / hd .X / . Thus
hl.Cp .X // D hd .X /, so our solution is complete.
T.018. Prove that hl.X X / hd.Cp .X // hl .X / for any space X .
Solution. Given points x1 ; : : : ; xn 2 X and sets O1 ; : : : ; On 2 .R/, the set
x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng
is called a standard open subset of Cp .X /. Standard open sets x1 ; : : : ; xn I O1 ; : : : ;

On where n 2 N; x1 ; : : : ; xn 2 X and O1 ; : : : ; On are rational intervals form a
base in Cp .X / (see Problem 056 of [TFS]).
Let hd.Cp .X // D ; we will prove first that hl.X / . Assuming the contrary
we can find a right-separated subspace D X with jDj D C (see Problem 005).
Denote by < an order that right-separates D; for any d 2 D, let Dd D fz 2 D W
d < zg. For each d 2 D the set Dd is closed in D. This is evident if Dd D ;; if not,
then d 0 D min Dd is well-defined and the set Dd D Dnfz 2 D W z < d 0 g is closed
being the complement of the set fz 2 D W z < d 0 g which is open in D because D is
right-separated.
Since d Dd we can choose a set Ud 2 .d; X/ with Ud \ Dd D ; and a
function fd 2 C.X; 0; 1 / such that fd .d / D 1 and fd j.X nUd / 0. Consider the
set Od D ff 2 Cp .X / W f .d / > 0g; it is clear that Od is an open subset of Cp .X /
and fd 2 Od for each d 2 D. Take any distinct points a; d 2 D; without loss of
generality we can assume that a < d . Then d Ua so fa .d / D 0 fd .d / D 1.

This proves that fd fa and therefore the correspondence d ! fd is a bijection.
Thus we can define a well-order on the set F D ffd W d 2 Dg by letting fd fa
if and only if d < a. Given any f 2 F , take any g f . There are a; d 2 D such
that g D fa and f D fd . Since g f , we have a < d . It follows from d Ua that
we have fa .d / D 0 whence g D fa Od . This proves that g Od for any g f
and therefore Od \ Ff D ; where Ff D fg 2 F W g f g. As a consequence, each
f D fd 2 D has an open neighborhood O.f / D Od such that O.f / \ Ff D ;;
this implies that Ff is closed in F for any f 2 F and hence F is a left-separated
subspace of Cp .X / with the well-order witnessing this. Since jF j D jDj D C ,
this is a contradiction with hd.Cp .X // (see Problem 004) which shows that
hl.X / .
Now assume that hl.X X / > and fix a right-separated subspace E X X
with jEj D C . The set D f.x; x/ W x 2 X g is a subspace of X X
homeomorphic to X so hl./ . Thus jE \ j and hence jEnj D C ,
i.e., we have a right-separated subspace D D En .X X /n of cardinality
C . Denote by < an order that right-separates D and let Dd D fz 2 D W d < zg
for any d 2 D. Since every Dd is closed in D, for each d D .x; y/ 2 D there
exist Ud 2 .x; X/ and Vd 2 .y; X/ such that .Ud Vd / \ Dd D ;. Since
x y, we lose no generality assuming that Ud \ Vd D ; for each d 2 D. Take any
functions gd ; hd 2 C.X; 0; 1 / such that gd .x/ D hd .y/ D 1; gd j.X nUd /
0; hd j.X nVd / 0 and let fd D gd hd . Then Od D ff 2 Cp .X / W f .x/ > 0
and f .y/ < 0g is an open subset of Cp .X / and fd 2 Od for all d 2 D.
If a; d 2 D and a d , we can assume without loss of generality that d < a.
If a D .x 0 ; y 0 / and d D .x; y/, then a Ud Vd which implies x 0 Ud or
y 0 Vd . In the first case we have fd .x 0 / 0 while fa .x 0 / D 1 > 0; in the second
case we obtain fd .y 0 / 0 while fa .y 0 / D 1 < 0. Since in both cases we have
fa fd , the correspondence d ! fd is a bijection. This makes it possible to
define a well-order on the set F D ffd W d 2 Dg by letting fd fa if and only
if d < a.
We prove next that the well-order left-separates F . Take any f 2 F ; to prove
that the set Ff D fg 2 F W g f g is closed in F , take any g f . There is a point
a D .x 0 ; y 0 / 2 D such that g D fa ; we have f D fd for some d D .x; y/ 2 D
with a < d . It follows from d Ua Va that x Ua or y Va ; in the first
case we have fa .x/ 0 and in the second one we obtain fa .y/ 0, so in both
cases g D fa Od . The point f 2 F was taken arbitrarily, so we established
that Od \ Ff D ; for any f D fd 2 F . This proves that Ff is closed in F
for each f 2 F and therefore F is a left-separated subspace of Cp .X /. Since
jF j D jDj D C , this is a contradiction with hd.Cp .X // (see Problem 004)
which shows that hl.X X / D hd.Cp .X //.
To prove that hd.Cp .X // hl .X / assume that hl .X / < hd.Cp .X // and
fix a left-separated subspace F Cp .X / with jF j D C . Let < be a well-order that
right-separates F . For each f 2 F , the set Ff D fg 2 F W g < f g is closed in F .
Therefore, for each f 2 F , there exists a standard open set Of 2 .f; Cp .X // such
f f f f f f
that Of \ Ff D ;. We have Of D x1 ; : : : ; xnf I O1 ; : : : ; Onf where x1 ; : : : ; xnf
f f
are distinct points of X and O1 ; : : : ; Onf are rational intervals. Since there are only
countably many of all possible n-tuples of rational intervals for any n 2 N and C is
a regular uncountable cardinal, there exist n 2 N, a set G F and rational intervals
O1 ; : : : ; On such that
jGj D C ; nf D n; Oi D Oi ; i D 1; : : : ; n
f
f f
for each f 2 G. Let xf D .x1 ; : : : ; xn / 2 X n for all f 2 G. For the set O.f / D
f 1 .O1 /

f 1 .On /, we have xf 2 O.f / for all f 2 G. We claim that the
set D D fxf W f 2 Gg X n is right-separated.
To show this, pick any f; g 2 G such that g < f . Since g 2 Ff , we have
g Of so g.xi / Oi for some i n. This implies xi g 1 .Oi / and therefore
f f
xf O.g/. Thus we have proved that

(*) for any f; g 2 G such that g < f , we have xf 2 O.f / and xf O.g/.
An evident consequence is that xf xg for any distinct f; g 2 G, so the
correspondence f ! xf is a bijection between G and D. This makes it possible to
well-order the set D by letting xf xg if and only if f < g. Fix x 2 D; we must
prove that the set Dx D fy 2 D W y xg is open in D. Take any y 2 Dx and any
x 0 2 DnDx ; there are f; g 2 G such that x 0 D xf and y D xg . Since y x 0 , we
have g < f and hence x 0 D xf O.g/ by . /. This shows that every y D xg 2 Dx
has a neighborhood Wy D O.g/ such that Wy \ D Dx ; an easy consequence is
that Dx is open in D for any x 2 D so D is right-separated by the well-order .
However, jDj D jGj D C and G X n which contradicts hl.X n / hl .X / .
Thus hd.Cp .X // D hl .X /, so our solution is complete.
T.019. For an arbitrary n 2 N, let Jn D J.n/ be the hedgehog with n spines. Prove
that s.X n / s.Cp .X; Jn // s.Cp .X / Cp .X // for any space X .
Solution. For an arbitrary n 2 N, let Mn D f1; : : : ; ng. If Z is a space and n 2,
D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj g for any distinct i; j 2 Mn . The set
let nij .Z/S
n .Z/ D fnij .Z/ W 1 i < j ng is called the n-diagonal of Z.
Fact 0. For any n 2 N; n 2 and any space Z, we have s.Z n / D s.Z n nn .Z//.
Proof. Since Z n nn .Z/ Z n , we have s.Z n nn .Z// s.Z n /. We prove the
inverse inequality by induction on n; to simplify the notation, let n D n .Z/ and
D 2 . If p1 W Z Z ! Z is the natural projection onto the first coordinate,
then p1 .Z 2 n/ D Z (we consider Z to be infinite), so s.Z 2 n/ s.Z/ because
continuous maps do not increase the spread. Since is homeomorphic to Z, we
have s./ D s.Z/ s.Z 2 n/. It is an easy exercise that a finite union of spaces
of spread also has spread , so if D s.Z 2 n/, it follows from Z 2 D
[ .Z 2 n/ that s.Z 2 / and hence s.Z 2 / D s.Z 2 n/.
Assume that we have proved, for all n m, that we have s.Z n / s.Z n nn /.
Let s.Z mC1 nmC1 / D ; denote by pm W Z mC1 ! Z m the natural projection onto
the face determined by the first m coordinates. Since pm .Z mC1 nmC1 / D Z m nm ,
we have s.Z m / s.Z m nm / by the induction hypothesis and the fact that
continuous maps do not increase the spread.
Observe also that ijmC1 .Z/ is homeomorphic to Z m1 ; since is
homeomorphic to Z, the space ijmC1 .Z/ is homeomorphic to Z m for any distinct
i; j 2 Mn and hence mC1 is a finite union of spaces homeomorphic to Z m .
As a consequence, s.mC1 / s.Z m / s.Z m nm / . Now it follows from
s.mC1 / and Z mC1 D mC1 [ .Z mC1 nmC1 / that s.Z mC1 / . As we
accomplished the required inductive step, we can conclude that s.Z n / D s.Z n nn /
for all n 2 N; n 2, so Fact 0 is proved. t
u
For notational convenience we consider that Jn D f0g [ I1 [

[ In where
Ik D .0; 1 fkg for each k 2 Mn . Assume that s.Cp .X; Jn // D and we have
a discrete D X n with jDj D C . By Fact 0 we can assume, without loss of
generality, that D X n nn .X /. Let fd W < C g be a faithful enumeration of
the set D where d D .x1 ; : : : ; xn / for each < C . For each < choose sets
U1 ; : : : ; Un such that
(1) Ui 2 .xi ; X / for each i 2 Mn ;
(2) if i; j 2 Mn and i j , then clX .Ui / \ clX .Uj / D ;;
(3) W \ D D fd g where W D U1

Un for each < C .
S
For all < C let V D fUi W i 2 Mn g and take a function fi 2 C.X; 0; 1 /
such that fi .xi / D 1 and fi j.X nUi / 0 for all i 2 Mn . Let g .x/ D 0 for
all x 2 X nV ; if x 2 V , then there is a unique i 2 Mn such that x 2 Ui , so let
g .x/ D .fi .x/; i /. This gives us a map g W X ! Jn for all < C .
Observe that the map k W 0; 1 ! Ik [ f0g Jn defined by k .0/ D 0 and
k .t/ D .t; k/ for all t 2 .0; 1 is continuous. Therefore the map k fk W X ! Jn

is continuous for all k 2 Mn . Now, to see that g is continuous, take any x 2 X ; the
family U D fU1 ; : : : ; Un g is discrete by (2), so there is U 2 .x; X/ such that U
intersects at most one element of U , say, Uk . It is evident that g jU D . k fk /jU ,
so g jU is a continuous map. Therefore we can apply Fact 1 of S.472 to conclude
that the mapping g is continuous, i.e., g 2 Cp .X; Jn / for each < C .
If ; < C and , then d W by (3), so there is k 2 Mn such

that xk Uk . This implies g .xk / D .1; k/ 2 Ik while g .xk / Ik because
g1 .Ik / Uk . This shows that g g and therefore the correspondence ! g

is a bijection between the sets D and E D fg W < C g Cp .X; Jn /.

For each < C , the set O D ff 2 Cp .X; Jn / W f .xk / 2 Ik for all k 2 Mn g is
open in Cp .X; Jn / and g 2 O for all < C . If ; < C and , then d

W by (3), so there is k 2 Mn such that xk Uk . This implies g .xk / Ik because
g1 .Ik / Uk ; as a consequence, g O which proves that O \ E D fg g.

Thus E is a discrete subspace of Cp .X; Jn / and jEj D jDj D C which contradicts

s.Cp .X; Jn // . This contradiction proves that s.X n / D s.Cp .X; Jn // for
each n 2 N.
Fact 1. The space J.n/ embeds in R2 for each n 2 N.

0 0 0 0
p any points z; z 2 R such that z D .x; y/ and z D .x ; y /, let
2
Proof. Given
0 0 0
d.z; z / D .x x / C .y y / . Then d is a metric which generates the usual
2 2
topology on R2 (see Problems 130 and 205 of [TFS]). Denote by the usual metric
on J.n/ (see Problem 222 of [TFS]). Let '.0/ D w D .0; 0/ 2 R2 and, for any
k 2 Mn and any x D .t; k/ 2 Ik , let '.x/ D .t; kt/ 2 R2 . Observe that for the map
' W J.n/ ! R2 , we have
(*) for any x; y 2 J.n/, we have d.'.x/; '.y// .n C 1/
.x; y/.
Indeed, if x D .t; k/ 2 Ik [f0g and y D .s; m/p2 Im [f0g for p some k m, then
d.'.x/; '.y// d.'.x/; w/ C d.w;p '.y// D t 1 C k 2 C p s 1 C m2 . It follows
from k n and m n that we have 1 C k 2 < n C 1 and 1 C m2 < n C 1, so
d.'.x/; '.y// .s C t/.n C 1/ D .n C 1/
.x; y/.pIf x D .t; k/ 2 Ik [ f0g and
y D .s; k/ 2 Ik [ f0g, then d.'.x/; '.y// D jt sj
1 C k 2 .n C 1/
jt sj D
.n C 1/
.x; y/.
An immediate consequence of . / is that the map ' W J.n/ ! R2 is continuous;
it is clear that J.n/ is compact and ' is injective, so ' is a homeomorphism between
J.n/ and Tn D '.J.n// R2 . Fact 1 is proved. t
u
Returning to our solution note that Cp .X; Jn / Cp .X; R2 / by Fact 1 and
Problem 089 of [TFS]. Thus s.Cp .X; Jn // s.Cp .X; R2 // D s.Cp .X / Cp .X //
because the spaces Cp .X; R2 / and Cp .X / Cp .X / are homeomorphic by Problem
112 of [TFS]. This proves that s.Cp .X; Jn // s.Cp .X / Cp .X //, so our solution
is complete.
that hd.X n / hl.Cp .X; Jn // hl.Cp .X / Cp .X // for any space X .
let nij .Z/S
Fact 0. For any n 2 N; n 2 and any space Z, we have hd.Z n / D
hd.Z n nn .Z//.
Proof. Since Z n nn .Z/ Z n , we have hd.Z n nn .Z// hd.Z n /. We prove the
D 2 . If p1 W Z Z ! Z is the natural projection onto the first coordinate, then
p1 .Z 2 n/ D Z (we consider Z to be infinite), so hd.Z 2 n/ hd.Z/ because
continuous maps do not increase the hereditary density. Since is homeomorphic
to Z, we have hd./ D hd.Z/ hd.Z 2 n/. It is an easy exercise that a finite
union of spaces of hereditary density also has hereditary density , so if
D hd.Z 2 n/, it follows from Z 2 D [ .Z 2 n/ that hd.Z 2 / and hence
hd.Z 2 / D hd.Z 2 n/.
Assume that we have proved, for all n m, that we have hd.Z n /
hd.Z n nn /. Let hd.Z mC1 nmC1 / D ; denote by pm W Z mC1 ! Z m the
natural projection onto the face determined by the first m coordinates. Since
pm .Z mC1 nmC1 / D Z m nm , we have hd.Z m / hd.Z m nm / by
the induction hypothesis and the fact that continuous maps do not increase the
hereditary density.
i; j 2 Mn , and hence mC1 is a finite union of spaces homeomorphic to Z m .
As a consequence, hd.mC1 / hd.Z m / hd.Z m nm / . Now it follows
from hd.mC1 / and Z mC1 D mC1 [ .Z mC1 nmC1 / that hd.Z mC1 / .
As we accomplished the required inductive step, we can conclude that hd.Z n / D
hd.Z n nn / for all n 2 N; n 2 so Fact 0 is proved. t
u

[ In where Ik D
.0; 1 fkg for each k 2 Mn . Assume that hl.Cp .X; Jn // D and hd.X n / > .
Then there is a subspace D D fd W < C g X n which is left-separated by
its indexation (see Problem 004 and Fact 2 of T.004); by Fact 0, we can assume,
without loss of generality, that D X n nn .X /. Let d D .x1 ; : : : ; xn / for each
< C . For every < choose sets U1 ; : : : ; Un such that:
(3) W \ D D ; where D D fd W < g and W D U1

Un for each
< C.
S
all x 2 X nV ; if x 2 V , then there is a unique i 2 Mn such that x 2 Ui so let

If < < C , then d W by (3), so there is k 2 Mn such that xk Uk .

This implies g .xk / D .1; k/ 2 Ik while g .xk / Ik because g1 .Ik / Uk .

This shows that g g and therefore the correspondence ! g is a bijection

between the sets D and E D fg W < C g Cp .X; Jn /.
For each < C the set O D ff 2 Cp .X; Jn / W f .xk / 2 Ik for all k 2 Mn g
is open in Cp .X; Jn / and g 2 O for all < C . If < < C , then d W

by (3), so there is k 2 Mn such that xk Uk . This implies g .xk / Ik because
g1 .Ik / Uk , so we have the following property:

(*) g O whenever < < C .

Take any < C ; given any < , we have < 0 for any 0 .
An immediate consequence of . / is that g0 O . Therefore O \ E E
where E D fg
W
< g. This shows that the set E is open in E for
each < C , and hence E is a right-separated (by its indexation) subspace of
Cp .X; Jn /. Since jEj D jDj D C , this contradicts hl.Cp .X; Jn // ; thus
hd.X n / D hl.Cp .X; Jn // for each n 2 N.
Finally note that Cp .X; Jn / Cp .X; R2 / by Problem 089 of [TFS] and Fact 1
of T.019. Thus hl.Cp .X; Jn // hl.Cp .X; R2 // D hl.Cp .X / Cp .X // because
the spaces Cp .X; R2 / and Cp .X / Cp .X / are homeomorphic by Problem 112 of
[TFS]. This proves that hl.Cp .X; Jn // hl.Cp .X / Cp .X //, so our solution is
complete.
that hl.X n / hd.Cp .X; Jn // hd.Cp .X / Cp .X // for any space X .
let nij .Z/S
Fact 0. For any n 2 N; n 2 and any space Z, we have hl.Z n / D
hl.Z n nn .Z//.
Proof. Since Z n nn .Z/ Z n , we have hl.Z n nn .Z// hl.Z n /. We prove the
D 2 . If p1 W Z Z ! Z is the natural projection onto the first coordinate,
then p1 .Z 2 n/ D Z (we consider Z to be infinite), so hl.Z 2 n/ hl.Z/
because continuous maps do not increase the hereditary Lindelf number. Since
is homeomorphic to Z, we have hl./ D hl.Z/ hl.Z 2 n/. It is an easy exercise
that a finite union of spaces with hereditary Lindelf number also has hereditary
Lindelf number , so if D hl.Z 2 n/, it follows from Z 2 D [ .Z 2 n/ that
hl.Z 2 / and hence hl.Z 2 / D hl.Z 2 n/.
Assume that we have proved, for all n m, that we have hl.Z n / hl.Z n nn /.
Let hl.Z mC1 nmC1 / D ; denote by pm W Z mC1 ! Z m the natural projection onto
the face determined by the first m coordinates. Since pm .Z mC1 nmC1 / D Z m nm ,
we have hl.Z m / hl.Z m nm / by the induction hypothesis and the fact that
continuous maps do not increase the hereditary density.
i; j 2 Mn , and hence mC1 is a finite union of spaces homeomorphic to Z m .
As a consequence, hl.mC1 / hl.Z m / hl.Z m nm / . Now it follows
from hl.mC1 / and Z mC1 D mC1 [ .Z mC1 nmC1 / that hl.Z mC1 / .
As we accomplished the required inductive step, we can conclude that hl.Z n / D
hl.Z n nn / for all n 2 N; n 2, so Fact 0 is proved. t
u

[ In where Ik D
.0; 1 fkg for each k 2 Mn . Assume that hd.Cp .X; Jn // D and hl.X n / > .
Then there is a subspace D D fd W < C g X n which is right-separated by
its indexation (see Problem 005 and Fact 2 of T.005); by Fact 0, we can assume,
without loss of generality, that D X n nn .X /. Let d D .x1 ; : : : ; xn / for each
< C . For every < choose sets U1 ; : : : ; Un such that
(3) W \ D D where D D fd W < C 1g and W D U1

Un for
each < C .
S
all x 2 X nV ; if x 2 V , then there is a unique i 2 Mn such that x 2 Ui so let

If < < C , then d W by (3), so there is k 2 Mn such that xk Uk .

We have g .xk / D .1; k/ 2 Ik while g .xk / Ik because g1 .Ik / Uk .

This shows that g g and therefore the correspondence ! g is a bijection

between the sets D and E D fg W < C g Cp .X; Jn /.
For each < C the set O D ff 2 Cp .X; Jn / W f .xk / 2 Ik for all k 2 Mn g
is open in Cp .X; Jn / and g 2 O for all < C . If < < C , then d W

by (3), so there is k 2 Mn such that xk Uk . This implies g .xk / Ik because
g1 .Ik / Uk , so we have the following property:
(*) g O whenever < < C .
Take any < C ; given any < , we have g O by . /. Therefore
O \ E D ; where E D fg W < g. This shows that the set E is closed in
E for each < C , and hence E is a left-separated (by its indexation) subspace
of Cp .X; Jn /. Since jEj D jDj D C , this contradicts hd.Cp .X; Jn // ; thus
hl.X n / D hd.Cp .X; Jn // for each n 2 N.
Finally note that Cp .X; Jn / Cp .X; R2 / by Problem 089 of [TFS] and Fact 1
of T.019. Thus hd.Cp .X; Jn // hd.Cp .X; R2 // D hd.Cp .X / Cp .X // because
the spaces Cp .X; R2 / and Cp .X / Cp .X / are homeomorphic by Problem 112 of
[TFS]. This proves that hd.Cp .X; Jn // hd.Cp .X / Cp .X //, so our solution is
complete.
T.022. For any space X prove that s.Cp .X / Cp .X // D s .Cp .X //.
Solution. It is evident that s.Cp .X /Cp .X // s .Cp .X //, so we must only prove
that s .Cp .X // s.Cp .X / Cp .X //. Let Xi be a homeomorphic copy of X for all
i 2 N. Then .Cp .X //n is homeomorphic to the space Cp .X1

Xn / for all n 2 N
(Problem 114 of [TFS]). Let Yn D X1

Xn ; then Ynk is a finite union of spaces
homeomorphic to X k . Since finite unions do not increase spread, we have s.Ynk /
s.X k / s .X / for all k 2 N. Consequently, s .Yn / s .X / for all n 2 N. Apply
Problem 016 to conclude that s..Cp .X //n / D s.Cp .Yn // s .Yn / s .X / for all
n 2 N, i.e., s .Cp .X // s .X /. Since s.X n / s.Cp .X / Cp .X // for all n 2 N
by Problem 019, we have s .X / s..Cp .X //2 /. Finally, s .Cp .X // s .X /
s.Cp .X / Cp .X //, so s .Cp .X // D s.Cp .X / Cp .X //.
T.023. For any space X prove that hl.Cp .X / Cp .X // D hl .Cp .X //.
Solution. It is evident that hl.Cp .X / Cp .X // hl .Cp .X //, so we must only
prove that hl .Cp .X // hl.Cp .X / Cp .X //. Let Xi be a homeomorphic copy
of X for all i 2 N. Then .Cp .X //n is homeomorphic to the space Cp .X1

Xn / for all n 2 N (see Problem 114 of [TFS]). Let Yn D X1

Xn ; then
Ynk is a finite union of spaces homeomorphic to X k . Since finite unions do not
increase the hereditary density, we have hd.Ynk / hd.X k / hd .X / for all
k 2 N. Consequently, hd .Yn / hd .X / for all n 2 N. Apply Problem 017 to
conclude that hl..Cp .X //n / D hl.Cp .Yn // hd .Yn / hd .X / for all n 2 N,
i.e., hl .Cp .X // hd .X /. Since hd.X n / hl.Cp .X / Cp .X // for all n 2
N by Problem 020, we have hd .X / hl..Cp .X //2 /. Finally, hl .Cp .X //
hd .X / hl.Cp .X / Cp .X // so hl .Cp .X // D hl.Cp .X / Cp .X //.
T.024. For any space X prove that hd.Cp .X / Cp .X // D hd .Cp .X //.
Solution. It is evident that hd.Cp .X / Cp .X // hd .Cp .X //, so we must only
prove that hd .Cp .X // hd.Cp .X /Cp .X //. Let Xi be a homeomorphic copy of
the space X for all i 2 N. Then .Cp .X //n is homeomorphic to the space Cp .X1

Xn / for all n 2 N (see Problem 114 of [TFS]). Let Yn D X1

Xn ;
then Ynk is a finite union of spaces homeomorphic to X k . Since finite unions do
not increase the hereditary Lindelf number, we have hl.Ynk / hl.X k / hl .X /
for all k 2 N. Consequently, hl .Yn / hl .X / for all n 2 N. Apply Problem
018 to conclude that hd..Cp .X //n / D hd.Cp .Yn // hl .Yn / hl .X / for all
n 2 N and hence hd .Cp .X // hl .X /. Since hl.X n / hd.Cp .X / Cp .X //
for all n 2 N by Problem 021, we conclude that hl .X / hd..Cp .X //2 /. Finally,
observe that the inequalities hd .Cp .X // hl .X / hd.Cp .X / Cp .X // show
that hd .Cp .X // D hd.Cp .X / Cp .X //.
T.025. Prove that s .X / D s .Cp .X // for any space X .
Solution. Let Xi be a homeomorphic copy of X for all i 2 N. Then .Cp .X //n is

homeomorphic to the space Cp .X1

Xn / for all n 2 N by Problem 114 of
[TFS]. Let Yn D X1

Xn ; then Ynk is a finite union of spaces homeomorphic
to X k . Since finite unions do not increase spread, we have s.Ynk / s.X k / s .X /
for all k 2 N. Consequently, s .Yn / s .X / for all n 2 N. Apply Problem 016
to conclude that s..Cp .X //n / D s.Cp .Yn // s .Yn / s .X / for all n 2 N;
this implies that s .Cp .X // s .X /. Since s.X n / s.Cp .X / Cp .X // for all
n 2 N by Problem 019, we have s .X / s..Cp .X //2 / s .Cp .X // so s .X / D
s .Cp .X //.
T.026. Prove that hl .X / D hd .Cp .X // for any space X .


Xn / for all n 2 N by Problem 114
of [TFS]. Consider the space Yn D X1

Xn ; then Ynk is a finite union
of spaces homeomorphic to X k . It is evident that finite unions do not increase
hereditary Lindelf number, so we have hl.Ynk / hl.X k / hl .X / for all
k 2 N. Consequently, hl .Yn / hl .X / for all n 2 N. We have hd..Cp .X //n / D
hd.Cp .Yn // hl .Yn / hl .X / for all n 2 N (see Problem 018); this implies
that hd .Cp .X // hl .X /. Since hl.X n / hd.Cp .X / Cp .X // for all n 2 N
by Problem 021, we have hl .X / hd..Cp .X //2 / hd .Cp .X // whence
hl .X / D hd .Cp .X //.
T.027. Prove that hd .X / D hl .Cp .X // for any space X .

Xn / for all n 2 N by Problem 114
of [TFS]. Consider the space Yn D X1

Xn ; then Ynk is a finite union
of spaces homeomorphic to X k . It is evident that finite unions do not increase
hereditary density, so we have hd.Ynk / hd.X k / hd .X / for all k 2 N.
Consequently, hd .Yn / hd .X / for all n 2 N. We have hl..Cp .X //n / D
hl.Cp .Yn // hd .Yn / hd .X / for all n 2 N (see Problem 017); this implies
that hl .Cp .X // hd .X /. Since hd.X n / hl.Cp .X / Cp .X // for all n 2 N
by Problem 020, we have hd .X / hl..Cp .X //2 / hl .Cp .X // whence
hd .X / D hl .Cp .X //.
T.028. For an infinite cardinal , suppose that s.Cp .X // D and .X / .
Prove that s .Cp .X // and hence s .Cp .X // D s.Cp .X //. In particular, if X
is a space with a G -diagonal, then s .Cp .X // D s.Cp .X // D s .X /.
Solution. For an arbitrary n 2 N, let Mn D f1; : : : ; ng; we will also need the set
Sn D f W is a bijection and W Mn ! Mn g. If Z is a space and n 2, let
nij .Z/ D Sfz D .z1 ; : : : ; zn / 2 Z n W zi D zj g for any distinct i; j 2 Mn . The set
n .Z/ D fnij .Z/ W 1 i < j ng is called the n-diagonal of the space Z.
If the space Z is clear, we write n instead of n .Z/.
Fact 1. Given a space Z and a natural number n 2, we have s.Z n nU /
s.Cp .Z// for any U 2 .n .Z/; Z n /.
Proof. Assume that s.Cp .Z// D ; if we are given a point z D .z1 ; : : : ; zn / 2 Z n ,

then supp.z/ D fz1 ; : : : ; zn g; let n D n .Z/. For any point z 2 Z n nn , consider
the set Pz D fz0 2 Z n nn W supp.z0 / D supp.z/g. It is evident that jPz j D n
for any z 2 Z n nn . It is also clear that either Pz \ Py D ; or Pz D Py for any
y; z 2 Z n nn . Therefore the family P D fPz W z 2 Z n nn g is disjoint; if we
choose an element zP 2 P for each P 2 P, we obtain a set C D fzP W P 2 Pg
such that C Z n nn and jC \ Pz j D 1 for every z 2 Z n nn .
Given any permutation 2 Sn , consider the map p W Z n ! Z n defined by

p .z/ D .z .1/ ; : : : ; z .n/ / for any z D .z1 ; : : : ; zn / 2 Z n . It is immediate that
each pT is a homeomorphism such that p .n / D n . As a consequence, the set
W D fp .U / W 2 Sn g is an open neighborhood of n such that W U and
Pz Z n nW for any z 2 Z n nW . Since Z n nU Z n nW , it suffices to prove that
s.Z n nW / . Suppose that there is a discrete D Z n nW of cardinality C .
Since Pz is finite for any z 2 Z n nn , the set R D fz 2 C W Pz \ D ;g has
cardinality C . For each z 2 R choose dz 2 Pz \ D; then D 0 D fdz W z 2 Rg is
again a discrete subspace of Z n nW of cardinality C such that
(*) jD 0 \ Pz j D 1 for each z 2 D 0 .
The set F D D 0 nD 0 is a closed subset of Z n and F Z n nW (the bar denotes
the closure in Z n ). Let T D Z n n.W [ F /; there exist E D 0 and m 2 N such that
jEj D C and jPd \ T j D m for every d 2 E. For each d D .d1 ; : : : ; dn / 2 E and
i n, choose a set Oid 2 .di ; Z/ such that
(1) Oid \ Ojd D ; if i j ;
(2) O d \ D 0 D fd g where O d D O1d

Ond ;
(3) if y D .dj1 ; : : : ; djn / 2 n , then Ojd1

Ojdn W ;
(4) if z D .dj1 ; : : : ; djn / 2 .Pd nfd g/ \ T , then .Ojd1

Ojdn / \ .E [ F / D ;.
For each d 2 E and any i 2 Mn take a function fid 2 C.Z; 0; 1 / such that
fi .di / D 1 and fid .ZnOid / f0g; let fd D f1d C

C fnd . If d D .d1 ; : : : ; dn / 2
d
E, then the set Ud D ff 2 Cp .Z/ W f .di / > 0 for all i 2 Mn g is open and fd 2 Ud .
Take any distinct a; d 2 E with a D .a1 ; : : : ; an / and d D .d1 ; : : : ; dn /. If fd 2 Ua ,
then for all i n, there is ji 2 Mn such that fjdi .ai / > 0 and hence ai 2 Ojdi . An
immediate consequence is that a 2 Ojd1

Ojdn . If there exist i; k n such that
i k and ji D jk , then y D .dj1 ; : : : ; djn / 2 n and a 2 Ojd1

Ojdn \ W by
(3) which contradicts E \ W D ;.
Thus ji jk if i k, i.e., the map W Mn ! Mn defined by .i / D ji
is a bijection; let D 1 . Given any z D .dk1 ; : : : ; dkn / 2 Pd \ T , the point
a.z/ D .a.k1 / ; : : : ; a.kn / / belongs to Pa \ T because a.z/ 2 .Okd1

Okdn /nW
and .Okd1

Okdn / \ F D ; by the property (4). This shows that the set A D
fag [ fa.z/ W z 2 Pd \ T g is contained in Pa \ T which contradicts the fact that
Pa \ T has m elements by the choice of E while jAj D m C 1 and A Pa \ T .
This contradiction shows that fd Ua whenever d a which implies that
Ua \ E D ffa g for each a 2 E and, in particular, fd fa for distinct d and a,
i.e., the correspondence d ! fd is a bijection. Therefore ffa W a 2 Eg is a discrete
subspace of Cp .Z/ of cardinality C which is a contradiction with s.Cp .Z// .
Thus s.Z n nU / and Fact 1 is proved. t
u
Returning to our solution, fix T n 2 N with n 2; take an arbitrary family U
.X X / such that jUj and U D D 2 .X /. Given distinct i; j 2 Mn , let
qij W X n ! X X be the natural projection onto the face defined by i and j , i.e.,
for any x D .x1 ; : : : ; xn / 2 X n we have qij .x/ D .xi ; xj / 2 X X . It is clear that
T
nij .X / D qij1 ./ and therefore nij .X / D Uij where Uij D fqij1 .U / W S
U 2 Ug.
If Bn D f.i; j / 2 Mn Mn W i < j g, then the family V D fU D fUij W
.i; j / 2 Bn g W Uij 2 Uij for all .i; j / 2 Bn g consists of open
T Ssubsets of X n and
V D n .X /. It is evident that jVj and X nn .X / D fX nV W V 2 Vg.
n n
For any set F 2 F S D fX nV

n
W V 2 Vg we have s.F / by Fact 1. It follows
from jF j and F D X n nn .X / that s.X n nn .X // (it is an easy exercise
to show that a union of -many spaces of spread each has spread ). Let
n D n .X /; we proved that:
(**) s.X n nn / for any n 2 N with n 2.
Now we can prove that s .X / by induction on the power of X . If n D 2,
then s.X n / by Problem 016. Assume that we proved that s.X m / ; then, for
n D mC1, we have X n D n [.X n nn /. Observe that nij is homeomorphic to
X n2 which in turn is homeomorphic to X n1 D X m because is homeomorphic
to X . As a consequence, n is a finite union of spaces homeomorphic to X m so
s.n / by the induction hypothesis. Since s.X n nn / by . /, we have
s.X n / , so our inductive step is accomplished. This proves that s .X / and
hence s .Cp .X // D s .X / (see Problem 025), so our solution is complete.
T.029. For an infinite cardinal , suppose that hl.Cp .X // D and .X / .
Prove that hl .Cp .X // and hence hl .Cp .X // D hl.Cp .X //. In particular, if
X is a space with a G -diagonal, then hl .Cp .X // D hl.Cp .X // D hd .X /.
let nij .Z/S
n .Z/ D fnij .Z/ W 1 i < j ng is called the n-diagonal of the space Z. If
the space Z is clear, we write n instead of n .Z/. Call a set U 2 .Z n / marked if
U D U1

Un where fUi W i 2 Mn g .Z/ and Ui \ Uj D ; for any distinct
i; j 2 Mn .
Fact 1. Given n 2 N, for any space Z and any marked set U Z n , we have
hd.U / hl.Cp .Z//.
Proof. Assume that hl.Cp .Z// D ; we have U D U1

Un where the family
fUi W i 2 Mn g consists of open subsets of Z and Ui \ Uj D ; for any distinct
i; j 2 Mn . If hd.U / > , then there exists a set P D fz W < C g U which
is left-separated by its indexation (see Problem 004 and Fact 2 of T.004). For any
< C , we have z D .z1 ; : : : ; zn / where zi 2 Ui for any i 2 Mn .
For each < C there exist O1 ; : : : ; On 2 .Z/ with the following
properties:
(1) zi 2 Oi for all i 2 Mn ;
(2) O \ P D ; where O D O1

On and P D fz W < g;
(3) Oi Ui for all i 2 Mn .
For any < C and i 2 Mn take a function fi 2 C.Z; 0; 1 / such that

D 1 and fi .ZnOi / f0g; let f D f1 C

C fn . We will prove that the
fi .zi /
set E D ff W < C g is right-separated by its indexation.
Let U D ff 2 Cp .Z/ W f .zi / > 0 for all i 2 Mn g. It is clear that U is an open
subset of Cp .Z/ and f 2 U for all < C . Assume that < and f 2 U .
j
Then, for any i 2 Mn , we have f .zi / > 0 and hence zi 2 O for some j 2 Mn .
j
However, zi 2 Ui and O \ Ui D ; for any j i . Therefore zi 2 Oi for all
i 2 Mn whence z 2 O which contradicts (2). This contradiction shows that we
have
(*) f U for any < < C and, in particular, the map z ! f is a bijection
between P and E.
Take any < C ; given any < , we have < 0 for any 0 . An
immediate consequence of . / is that f0 U . Therefore U \ E E where
E D ff
W
< g. This shows that the set E is open in E for each < C and
hence E is a right-separated (by its indexation) subspace of Cp .Z/. Since jEj D C
by . /, we have a contradiction with hl.Cp .Z// (see Problem 005). Fact 1 is
proved. t
u
Fact 2. Given a space Z and n 2, we have hd.Z n nU / hl.Cp .Z// for any
U 2 .n .Z/; Z n /.
Proof. Let hl.Cp .Z// D ; we have s.Z n nU / s.Cp .Z// hl.Cp .Z// D by
Fact 1 of T.028. If hd.Z n nU / > , then there exists a set D D fz W < C g
Z n nU which is left-separated by its indexation (see Problem 004 and Fact 2 of
T.004). For any < C there is a marked set U 2 .z ; Z n /. We have hd.U /
by Fact 1. Thus U D U \ D cannot have C -many elements, i.e., jU j . The
family U D fU W < C g is an open cover of D and s.D/ . Besides, the
S hl.D/
space D is left-separated, so we can apply Problem 007 to conclude that
s.D/ . In particular, there is U 0 U such that jU 0 j and U 0 D D.
Since each U 2 U 0 has cardinality , we have jDj which is a contradiction.
Therefore hd.Z n nU / and Fact 2 is proved. t
u
Returning to our solution, fix T n 2 N with n 2; take an arbitrary family U
.X X / such that jUj and U D D 2 .X /. Given distinct i; j 2 Mn , let
qij W X n ! X X be the natural projection onto the face defined by i and j , i.e.,
for any x D .x1 ; : : : ; xn / 2 X n , we have qij .x/
T D .xi ; xj / 2 X X . It is clear that
nij .X / D qij1 ./ and therefore nij .X / D Uij where Uij D fqij1 .U / W S U 2 Ug.
.i; j / 2 Bn g W Uij 2 Uij for all .i; j / 2 Bn g consists of open
T Ssubsets of X n and
V D n .X /. It is evident that jVj and X nn .X / D fX nV W V 2 Vg.
n n
For any set F 2 F D S fX nV WnV 2 Vg we have hd.F

n
/ by Fact 2. It follows
from jF j and F D X nn .X / that hd.X n nn .X // (it is an easy
exercise to show that a union of -many spaces of hereditary density each
has hereditary density ). Let n D n .X /; we proved that
(**) hd.X n nn / for any n 2 N with n 2.

Now we can prove that hd .X / by induction on the power of X . If
n D 2, then hd.X n / by Problem 017. Assume that we established that
hd.X m / ; then, for n D m C 1, we have X n D n [ .X n nn /. Observe
that the space nij is homeomorphic to X n2 which is homeomorphic to
X n1 D X m because is homeomorphic to X . As a consequence, n is a finite
union of spaces homeomorphic to X m so hd.n / by the induction hypothesis.
Since hd.X n nn / by . /, we have hd.X n / , so our inductive step is
accomplished. This proves that hd .X / and hence hl .Cp .X // D hd .X /
(see Problem 027), so our solution is complete.
T.030 (Velichkos theorem). Prove that hd .Cp .X // D hd.Cp .X // D hl .X /
for any space X .
let nij .Z/S
n .Z/ D fnij .Z/ W 1 i < j ng is called the n-diagonal of the space Z. If
the space Z is clear, we write n instead of n .Z/. Call a set U 2 .Z n / marked if
U D U1

Un where fUi W i 2 Mn g .Z/ and Ui \ Uj D ; for any distinct
i; j 2 Mn .
Fact 1. Given n 2 N, for any space Z and any marked set U Z n , we have
hl.U / hd.Cp .Z//.
Proof. Assume that hd.Cp .Z// D ; we have U D U1

Un where the family
fUi W i 2 Mn g consists of open subsets of Z and Ui \ Uj D ; for any distinct
i; j 2 Mn . If hl.U / > , then there exists a set P D fz W < C g U which
is right-separated by its indexation (see Problem 005 and Fact 2 of T.005). For any
< C , we have z D .z1 ; : : : ; zn / where zi 2 Ui for any i 2 Mn .
For each < C there exist O1 ; : : : ; On 2 .Z/ with the following
properties:

(3) Oi Ui for all i 2 Mn .
For any < C and i 2 Mn take a function fi 2 C.Z; 0; 1 / such that
D 1 and fi .ZnOi / f0g; let f D f1 C

C fn . We will prove that the
fi .zi /
set E D ff W < C g is left-separated by its indexation.
Let U D ff 2 Cp .Z/ W f .zi / > 0 for all i 2 Mn g. It is clear that U is an open
subset of Cp .Z/ and f 2 U for all < C . Assume that < and f 2 U .
j
Then, for any i 2 Mn , we have f .zi / > 0 and hence zi 2 O for some j 2 Mn .
j
However, zi 2 Ui and O \ Ui D ; for any j i . Therefore zi 2 Oi for all
i 2 Mn whence z 2 O which contradicts (2). This contradiction shows that we
have
(*) f U for any < < C and, in particular, the map z ! f is a bijection
between P and E.
Take any < C ; given any < , we have f U by . /. Therefore
U \ E D ; where E D ff
W
< g. This shows that the set E is closed in
E for each < C and hence E is a left-separated (by its indexation) subspace of
Cp .Z/. Since jEj D C by . /, we have a contradiction with hd.Cp .Z// (see
Problem 004). Fact 1 is proved. t
u
Fact 2. Given a space Z and n 2, we have hl.Z n nU / hd.Cp .Z// for any
U 2 .n .Z/; Z n /.
Proof. Let hd.Cp .Z// D ; we have s.Z n nU / s.Cp .Z// hd.Cp .Z// D
by Fact 1 of T.028. Observe that t.Cp .Z// hd.Cp .Z// D and therefore
l .Z/ by Problem 149 of [TFS]. The set Z n nU is closed in Z n so
l.Z n nU / . For every z 2 F D Z n nU there exists a marked set Uz 2 .z; Z n /.
The family U D fUz W z 2 FSg is an open cover of the space F , so there is
U0 U Ssuch that jU 0 j and U 0 F . We have hl.V / for each V 2 U 0
0
so hl. U / (it is an easy exercise that a union of -many spaces with
S number has hereditary Lindelf number ). Therefore
hereditary Lindelf
hl.Z n nU / hl. U 0 / and Fact 2 is proved. t
u
Returning to our solution, let hd.Cp .X // D ; observe that hl.X X / by
Problem 018. Let D 2 .X /; for any z 2 .X X /n, pick any Uz 2 .z; X X /
such that U z \ D ;. It follows immediately from hl.X 2 / that the open cover
U D fUz W z 2 .X X /ng of the S space .X X /n has a subcover U 0 such that
jU j . Therefore .X X /n D fU W U 2 U 0 g, i.e., .X X /n is a union of
0
-many closed sets and therefore .X / .

T n 2 N with n 2; take an arbitrary family U .X X / such that jUj
Fix
and U D . Given distinct i; j 2 Mn , let qij W X n ! X X be the natural
projection onto the face defined by i and j , i.e., for any x D .x1 ; : : : ; xn / 2 X n we
have qij .x/ D .xi ; xj / 2 X X . It is clear that nij .X / D qij1 ./ and therefore
T 1
nij .X / D Uij where Uij D fq Sij .U / W U 2 Ug. If Bn D f.i; j / 2 Mn Mn W i <
j g, then the family V D fU D fUijTW .i; j / 2 Bn g W Uij 2 Uij for all .i; j / 2 Bn g
consists of openSsubsets of X n and V D n .X /. It is evident that jVj and
X n nn .X / D fX n nV W V 2 Vg. For any set F 2 F D S fX nV nW V 2 Vg we
n
have hl.F / by Fact 2. It follows from jF j and F D X nn .X / that

hl.X n nn .X // . Let n D n .X /; we proved that
(**) hl.X n nn / for any n 2 N with n 2.
Now we can prove that hl .X / by induction on the power of X . If
n D 2, then hl.X n / by Problem 018. Assume that we established that
hl.X m / ; then, for n D m C 1, we have X n D n [ .X n nn /. Observe
that the space nij is homeomorphic to X n2 which is homeomorphic to
X n1 D X m because is homeomorphic to X . As a consequence, n is a finite
union of spaces homeomorphic to X m , so hl.n / by the induction hypothesis.
Since hl.X n nn / by . /, we have hl.X n / so our inductive step is

accomplished. This proves that hl .X / and hence hd .Cp .X // D hl .X /
(see Problem 026), so our solution is complete.
T.031. Prove that s.Cp .Cp .X /// D s .X / for any space X .
Solution. If X is empty, there is nothing to prove. If X ;, then we have
the equalities s .X / D s .Cp .X // D s .Cp .Cp .X /// by Problem 025. Apply
Problem 182 of [TFS] to conclude that there exists a space Y such that Cp .X /
is homeomorphic to Y R. It follows from Problem 177 of [TFS] that the space
Cp .Cp .X // is homeomorphic to .Cp .Cp .X ///! and therefore s .Cp .Cp .X /// D
s.Cp .Cp .X /// D s .X /.
T.032. Prove that hd.Cp .Cp .X /// D hd .X / for any space X .
the equalities hd .X / D hl .Cp .X // D hd .Cp .Cp .X /// by Problem 026
and Problem 027. Now apply Problem 030 to conclude that hd .Cp .Cp .X /// D
hd.Cp .Cp .X /// D hd .X /.
T.033. Prove that hl.Cp .Cp .X /// D hl .X / for any space X .
the equalities hl .X / D hd .Cp .X // D hl .Cp .Cp .X /// by Problem 026 and
Problem 027. Apply Problem 182 of [TFS] to conclude that there exists a space
Y such that Cp .X / is homeomorphic to Y R. It follows from Problem 177 of
[TFS] that the space Cp .Cp .X // is homeomorphic to .Cp .Cp .X ///! and therefore
hl .Cp .Cp .X /// D hl.Cp .Cp .X /// D hl .X /.
T.034. Prove that for a zero-dimensional space X , we have s .X / D s.Cp .X //.
let nij .Z/S D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj g for any distinct i; j 2 Mn . The set
Let n D n .X / for any n 2; assume that for some infinite cardinal , we
have s.Cp .X // D and s.X n / > for some n 2 N. Then n > 2 by Problem 016;
by Fact 0 of T.019 there exists a discrete faithfully indexed D D fd W < C g
X n nn . We have d D .x1 ; : : : ; xn / for each < C . Given any < C , choose
sets U1 ; : : : ; Un such that
(1) Ui is a clopen subset of X and xi 2 Ui for each i 2 Mn ;
(2) if i; j 2 Mn and i j , then Ui \ Uj D ;;
(3) W \ D D fd g where W D U1

Un for each < C .
S
For all < C let V D fUi W i 2 Mn g and take a function fi such that
fi .Ui / D fi g and fi j.X nUi / 0 for all i 2 Mn . It is evident that fi is a
continuous function for all i 2 Mn , so the function g D f1 C

C fn is also
continuous for any < C . Let O D ff 2 Cp .X / W jf .xi / g .xi /j < 1=3 for
all i 2 Mn g. It is clear that O is open in Cp .X / and g 2 O for all < C .

If , then d W by (3) and hence there is i 2 Mn such that xi Ui .
Therefore g .xi / 2 .f0g [ Mn /nfi g which implies jg .xi / g .xi /j 1 > 13
which shows in turn that g O . As a consequence, for the set E D fg W
< C g we have O \ E D fg g for each < C . Thus E is a discrete
subset of Cp .X / of cardinality C > D s.Cp .X // which is a contradiction.
This contradiction shows that s.X n / s.Cp .X // for all natural numbers n, i.e.,
s .X / s.Cp .X //. However, s.Cp .X // s .Cp .X // D s .X / by Problem 025,
so we have s.Cp .X // D s .X /.
T.035. Prove that hd .X / D hl.Cp .X // for any zero-dimensional space X .
let nij .Z/S
Let n D n .X / for any n 2; assume that for some infinite cardinal , we
have hl.Cp .X // D < hd.X n / for some n 2 N. Then n > 2 by Problem 017;
by Fact 0 of T.020 and Fact 2 of T.004, there exists a set D D fd W < C g
X n nn which is left-separated by its indexation. We have d D .x1 ; : : : ; xn / for
each < C . Given any < C , choose sets U1 ; : : : ; Un such that
(1) Ui is a clopen subset of X and xi 2 Ui for each i 2 Mn ;
(2) if i; j 2 Mn and i j , then Ui \ Uj D ;;
(3) W \ D fd W g where W D U1

Un for each < C .
S
For all < C let V D fUi W i 2 Mn g and take a function fi such that
fi .Ui / D fi g and fi j.X nUi / 0 for all i 2 Mn . It is evident that fi is a
continuous function for all i 2 Mn , so the function g D f1 C

C fn is also
continuous for any < C . Let O D ff 2 Cp .X / W jf .xi / g .xi /j < 1=3 for
all i 2 Mn g. It is clear that O is open in Cp .X / and g 2 O for all < C .

If < , then d W by (3), and hence there is i 2 Mn such that xi Ui .
Therefore g .xi / 2 .f0g [ Mn /nfi g which implies jg .xi / g .xi /j 1 > 13
which shows in turn that g O . As a consequence, for the set E D fg W <
C g we have O \ E fg W g for each < C . Thus E is a right-separated
(by its indexation) subset of Cp .X / of cardinality C > D hl.Cp .X // which is
a contradiction. This contradiction shows that hd.X n / hl.Cp .X // for all natural
n, i.e., hd .X / hl.Cp .X //. However, hl.Cp .X // hl .Cp .X // D hd .X / by
Problem 027, so we have hl.Cp .X // D hd .X /.
T.036. Prove that, under SA, the following conditions are equivalent:
(i) s.Cp .X // D !;
(ii) hl..Cp .X //! / D !;
(iii) hd..Cp .X //! / D !.
In particular, if SA holds, then hl.Cp .X // D ! implies hl..Cp .X //! / D ! and
s.Cp .X // D ! implies s..Cp .X //! / D !.
Solution. It is evident that (ii)H)(i). The axiom SA says that (iii)H)(ii), so we

only must prove that (i)H)(iii). Assume that s.Cp .X // D !.
Fact 1. Suppose that Y and Z are spaces such that s.Y Z/ for some infinite
cardinal . Then either hd.Y / or hl.Z/ .
Proof. Assume that hd.Y / > and hl.Z/ > . It follows from hd.Y / > that
there exists a left-separated subspace of Y of cardinality C (see Problem 004).
Apply Fact 2 of T.004 to find a subspace D D fy W < C g Y which is
left-separated by its indexation. Analogously, hl.Z/ > implies that there exists a
subspace E D fz W < C g Z which is right-separated by its indexation (see
Fact 2 of T.005).
Consider the set H D f.y ; z / W < C g Y Z; for any ordinal < C
the set D D fy W < g is closed in D, so there exists U 2 .y ; Y / such
that U \ D D ;. Analogously, the set fz W < C 1g is open in E, so the set
E D fz W < g is closed in E, and therefore there exists V 2 .z ; Z/ such
that V \ E D ;. As a consequence, W D U V 2 ..y ; z /; Y Z/ and
W \ H D f.y ; z /g for each < C . This shows that H is a discrete subspace
of Y Z; since jH j D C , this is a contradiction with s.Y Z/ . Thus it is
impossible that hd.Y / > and hl.Z/ > , so we have hd.Y / or hl.Z/ .
Fact 1 is proved. t
u
Observe that s.X X / ! by Problem 016. Hence hl.X / ! or hd.X / !
by Fact 1; if hd.X / !, then apply SA to conclude that hl.X / ! anyway.
Assume first that for every point x 2 X , there exists a set Ux 2 .x; X/ such
that hl.Cp .U x // !. We have hd.U x U x / ! by Problem 017 and therefore
hl.U x U x / ! S by SA. Since the space X is Lindelf,
S there exists a countable
A X such that fUx W x 2 Ag D X . The set W D fUx Ux W x 2 Ag is open
in X X and D f.x; x/ W x 2 X g W . Any countable union of hereditarily
Lindelf spaces is hereditarily Lindelf, so hl.W / D ! which shows that is a
G -set in W and hence in X X .
Now assume that there is a point x 2 X such that hl.CS p .U // > ! for any set
U 2 .x; X/. For the space Y D X nfxg, we have Y D fYn W n 2 !g where
the subspaces Yn can be chosen closed in X with Yn YnC1 for all n 2 !. For
each n 2 !, take any Vn 2 .x; X/ such that V n \ Yn D ;. The set V n [ Yn is
a closed subspace of the (Lindelf and hence normal) space X , so the restriction
maps Cp .X / continuously onto Cp .Yn [ V n /. Since Yn [ V n is homeomorphic
to Yn V n , the space Cp .Yn [ V n / is homeomorphic to Cp .Yn / Cp .V n /. Thus
s.Cp .Yn / Cp .V n // D !; since hl.Cp .V n // > !, we have hd.Cp .YSn // ! by
Fact 1. As a consequence, we obtain hl.Yn Yn / !. As Y Y D fYn Yn W
n 2 !g, we have hl.Y Y / D ! and therefore hl.X X / !. This implies again
that is a G -subset of X .
We proved that in both cases s.Cp .X // D ! implies, under SA, that .X / D !.
Apply Problem 028 to conclude that s .X / D !. As before, for any natural n, the
inequality s.X n X n / ! implies hl.X n / ! by Fact 1 and SA. This proves
that hl .X / D ! and hence hd..Cp .X //! / D hd .Cp .X // D hl .X / D ! (see
Problem 012), so (i)H)(iii) and our proof is complete.
T.037. Prove that the following statements are equivalent (remember that all spaces
are assumed to be Tychonoff):
(i) there is a space X with s.X / D ! and d.X / > !;
(ii) there is a space X with hl.X / D ! and d.X / > !;
(iii) there is a left-separated space X with s.X / D ! and jX j D !1 .
Solution. If X is a space with s.X / D ! < d.X /, then there is a left-separated
uncountable subspace Y X (see Problem 004). It is an easy exercise that no
uncountable left-separated space is separable, so d.Y / > !. Besides, Y has to be
hereditarily Lindelf by Problem 007 which proves that (i)H)(ii).
If X is as in (ii), then there is an uncountable left-separated Y X . Take any
Z Y with jZj D !1 . Then Z is a left-separated space with jZj D !1 and
s.Z/ D !; this shows that the implication (ii)H)(iii) also holds.
Finally, if X is as in (iii), then X cannot be separable being uncountable and left-
separated. Hence X also satisfies (i) and the implication (iii)H)(i) is established.
T.038 (-system lemma).Prove that for any regular uncountable cardinal , if U
is a family of finite sets with jUj D , then there exists a set F (called the -root
for U) and a family V U (called the -system for U) such that jVj D and
A \ B D F for any distinct A; B 2 V.
Solution. It follows from regularity of that there is n 2 ! and a subfamily U 0 U
such that jU 0 j D and jU j D n for any U 2 U 0 . Thus, it is sufficient to prove our
statement for U 0 . This shows that we can assume that there is n 2 ! such that
jU j D n for any U 2 U. Our proof will be by induction on n.
If n D 0, then all elements of U are empty, so our statement holds vacuously.
Assume that we have proved our Delta-lemma for all n < mSand take any family U
of cardinality such that jU j D m for all U 2 U. Let X D U; we must consider
two cases:
Case 1. jfU 2 U W x 2 U gj < for any x 2 X .
S
LetSV be a maximal disjoint subfamily of U; if jVj < , then j Vj < . If
Y D V, then the family U 0 D fU 2 U W U \ Y ;g has cardinality <
because is regular and each point of Y belongs to < elements of U. However, if
U \Y D ;, then the family V [fU g U is disjoint and strictly larger than V which
contradicts maximality of V. Therefore U \ Y ; for all U 2 U, i.e., U D U 0 and
hence jUj D jU 0 j < which is a contradiction with jUj D .
This contradiction shows that V is a disjoint subfamily of U of cardinality , so
we can let F D ; finishing our proof in this case.
Case 2. There is x 2 X such that jfU 2 U W x 2 U gj D .
Let U 0 D fU 2 U W x 2 U g; the family W D fU nfxg W U 2 U 0 g has cardinality
and jW j D m 1 for each W 2 W. Therefore we can apply the induction
hypothesis to find a set F 0 and a subfamily W 0 W such that jW 0 j D and
V \ W D F 0 for any distinct V; W 2 W 0 . Now, letting F D F 0 [ fxg and
V D fW [ fxg W W 2 W 0 g we obtain a family V U such that jVj D and

A \ B D F for any distinct A; B 2 V. Thus our statement holds in this case as
well, so our proof is complete.
T.039. Prove that, under CH, there exists a hereditarily Lindelf non-separable
dense subspace X of the space f0; 1g!1 . In particular, L-spaces exist under CH.
Solution. Denote by D the discrete two-point space f0; 1g. We will need the space
D fx 2 D!1 W jx 1 .1/j !g. It was proved in Fact 3 of S.307 that is dense
in D!1 , countably compact and non-compact while A is compact for any countable
A (the bar denotes the closure in ). An immediate consequence is that
is a dense countably compact non-separable subspace of D!1 . Fix an arbitrary base
B D fB W < !1 g in the space such that B ; for each < !1 .
Fact 1. Let Z be a space such that c.Z/ ! and w.Z/ !1 . If CH holds, then
there exists a family N of nowhere dense closed subspaces of Z such that jN j !1
and, for any nowhere dense F Z, there is N 2 N such that F N . We will say
that the family N is cofinal in the family of all nowhere dense subsets of Z.
Proof. Take any base U in the space Z with jUj !1 . Since CH holds, the family
S subfamilies of U has cardinality at most !1 . Therefore
of all countable S the family
N D fZn. V/ W V is a countable disjoint subfamily of U and V is dense in Zg
consists of closed nowhere dense subsets of Z and jN j !1 .
To see that N is as promised, take any nowhere dense F Z. Let V be a
maximal disjoint subfamily of U such that V \ F D ; for any V 2 V. Since Z has
S Souslin property, the family V is countable;
the S it follows from maximality of V that
V is dense in Z whence N D Zn. V/ 2 N . It is immediate that F N , so
Fact 1 is proved. t
u
Returning to our solution, apply Fact 1 to the space to find a cofinal family
N D fN W < !1 g of closed nowhere dense subsets of . The space has the
Baire property (see Problem 274(ii) of [TFS]) and hence so does B for all < !1
(see Problem 275 of [TFS]). The set N0 being nowhere dense, we can choose a
point x0 2 B0 nN0 . Assume that < !1 and we have chosen points fx W < g
so that
S
(*) x 2 B n. fN
W
g/
for all < . Since
S B has the Baire property, it cannot be covered by the first
category set N D fN W g, so we can choose a point x 2 B nN .
This shows that our inductive construction can go on giving us a subspace X D
fx W < !1 g such that . / holds for all < !1 . We claim that X is
the required space. Observe first that X is dense in and hence in D!1 because
X \ B ; for all < !1 . As a consequence, X cannot be separable because a
countable dense subset of X would be a countable dense subset of which is not
separable, a contradiction.
The space X is dense in and hence it has no isolated points. Now, if A is
a right-separated subspace of X , then A is scattered (see Problem 006) and hence
there is a discrete subspace D of X such that A D. It is an easy exercise that every

discrete subspace of a space without isolated points is nowhere dense in this space so
A D is nowhere dense in X and hence in . By Fact 1, there is < !1 such that
A N . By . /, we have X \ N fx W < g and therefore A fx W < g
which shows that A is countable. Applying Problem 005 we conclude that X is
hereditarily Lindelf. Therefore hd.X / d.X / !1 > hl.X / D ! s.X / and
our solution is complete.
T.040. Prove that, under CH, there exists a hereditarily separable non-Lindelf
dense subspace X of the space f0; 1g!1 . Thus, S -spaces exist under CH.
Solution. Denote by D the discrete two-point space f0; 1g. If f is a function, then
dom.f / is its domain. For any set A, let Fn.A/ be the set of all finite functions
from A to D, i.e., Fn.A/ D fs W s 2 DB for some finite B Ag. In this definition
the set B can be empty, i.e., the empty function is considered to be an element of
Fn.A/. Any s 2 Fn.A/ determines a standard open subset s of DA by the formula
s D ff 2 DA W f jdom.s/ D sg. It is a standard practice to identify functions and
their graphs so the fact that f jdom.s/ D s can be written as s f . We will use this
agreement here for the sake of brevity. Given any set A !1 , let A W D!1 ! DA
be the natural projection of D!1 onto the face DA . Call a subset Z of the space D!1
finally dense in D!1 if there is < !1 such that !1 n .Z/ is dense in D!1 n . A set
Z D!1 is called hereditarily finally dense (or HFD) if any infinite Y Z is
finally dense in D!1 .
Fact 1. Assume that U D fUn W n 2 !g is a family of infinite sets. Then there
exists a disjoint family V D fVn W n 2 !g such that Vn is infinite and Vn Un for
all n 2 !. We will say that the family V is a (disjoint) -net for U.
Proof. Let fnk W k 2 !g be an enumeration of ! in which each n 2 ! occurs
infinitely often (see Fact 3 of S.286). Take x0 2 Un0 arbitrarily; if we have chosen
x0 ; : : : ; xk , we can choose xkC1 2 UnkC1 nfx0 ; : : : ; xk g because the set UnkC1 is
infinite. This inductive construction gives us a set fxi W i 2 !g such that the
correspondence i 7! xi is an injection. If An D fi 2 ! W ni D ng, then the
family fAn W n 2 !g is disjoint and consists of infinite sets. Therefore the set
Vn D fxi W i 2 An g is infinite for each n 2 !; the family fVn W n 2 !g is clearly
disjoint, consists of infinite sets and Vn Un for each n 2 !. Fact 1 is proved. u t
Fact 2. Under CH there exists an HFD space Z D fz W < !1 g D!1 .
Proof. Apply CH to fix an enumeration fA W < !1 g of all countably infinite

subsets of !1 . For each < !1 , let D minf 2 !1 W A g; it is evident that
! for all < !1 . We will define, by induction on , the element z ./ for all
< !1 . To start the procedure, let z .n/ D 0 for all n 2 ! and < !1 . Suppose
that < !1 and we defined z .
/ for all < !1 and
< in such a way that
(i) z .
/ D 0 if
;
(ii) if <
< and
, then, for any s 2 Fn.
n / (including s D ;),
the sets f W 2 A ; s z and z .
/ D 1g and f W 2 A ; s z and
z .
/ D 0g are infinite.
Observe that is (i) trivially fulfilled for D ! while (ii) holds vacuously. Assume
first that < and < . Given any non-empty function s 2 Fn.n /, we
have dom.s/ D f
1 ; : : : ;
n g where
1 <

<
n < . If s 0 D sjf
1 ; : : : ;
n1 g,
then the induction hypothesis can be applied to the function s 0 and ordinals and
D
n to conclude that the sets f W 2 A ; s 0 z and z .
n / D 1g and
f W 2 A ; s 0 z and z .
n / D 0g are infinite. As a consequence, the set
f W 2 A ; s 0 z and z .
n / D s.
n /g is infinite and coincides with the set
C;s D f W 2 A and s z g. Therefore the family C D fC;s W < ; <
and s 2 Fn.n /g [ fA W D g is countable and consists of countable infinite
sets. Apply Fact 1 to find a -net N D fNk W k 2 !g for the family C. Find an
infinite Mk Nk such Sthat Nk nMk is also infinite for all k 2 !. Given any < ,
let z ./ D 1 if 2 fMk W k 2 !g and z ./ D 0 otherwise. Letting z ./ D 0
for all , we finish our inductive construction.
It is now trivial that (i) holds for all ; let us prove that (ii) is also fulfilled
for all
. Since the statement of (ii) holds and does not depend on when
< , it suffices to show that (ii) is true for

D . In case when D
0 <
we proved that the set C;s is infinite for any s 2 Fn.n /. By our construction,
there is k 2 ! such that Nk C;s . Observe that the set f 2 A W s z and
z ./ D 1g coincides with the set P D C;s \ f W z ./ D 1g and Mk P by
our construction. Since Mk is infinite, the set P is also infinite. Analogously, the set
f 2 A W s z and z ./ D 0g coincides with the set Q D C;s \ f W z ./ D 0g
and Nk nMk Q by our construction. Since Nk nMk is infinite, the set Q is also
infinite. This shows that (ii) holds for all < such that < .
If D , then the only element of Fn.n / is the empty function, so the set
f 2 A W s z and z ./ D 1g coincides with Q D A \ f W z ./ D 1g. There
is l 2 ! such that Nl A ; by our construction we have Ml Q and hence Q
is infinite. Analogously, the set f 2 A W s z and z ./ D 0g coincides with
P D A \ f W z ./ D 0g. By our construction we have Nl nMl P and hence P
is infinite.
We have finally shown that (i) and (ii) are fulfilled for all
and therefore our
inductive construction can be continued giving us a set Z D fz W < !1 g D!1
with the properties (i) and (ii). To see that Z is an HFD, take any infinite Y Z.
There exists < !1 such that fz W 2 A g Y . We claim that Y is dense
beyond . Indeed, it suffices to prove that for any s 2 Fn.!1 n /, there is y 2 Y
such that s y. If s D ;, there is nothing to prove. If dom.s/ D f
1 ; : : : ;
n g where
1 <

<
n , then we can apply (ii) to and
D
n C 1 > to see that there
exists 2 A such that z .
/ D 1 and s z . Therefore y D z 2 Y and s y.
Thus any infinite Y Z is finally dense, so Z is an HFD and Fact 2 is proved. u t
If x; y 2 D!1 and the set f < !1 W x./ y./g is countable, we will call the
points x and y similar.
Fact 3. Suppose that Z D fz W < !1 g D!1 is an HFD. If a point y 2 D!1 is
similar to z for each < !1 , then the set Y D fy W < !1 g is also an HFD.
Proof. Take any infinite Y 0 Y ; there is a countably infinite set A !1 such that
P D fy W 2 Ag Y 0 . The set Z 0 D fz W 2 Ag is finally dense, so we can
take < !1 such that Z 0 is dense beyond . Since each y is similar to z , there
is a countable B !1 such that !1 nB .y / D !1 nB .z / for all 2 A. Take any
< !1 such that [ B

; it is immediate that !1 n
.P / D !1 n
.Z 0 /. It follows
from density of !1 n .Z 0 / in D!1 n that !1 n
.Y 0 / !1 n
.P / D !1 n
.Z 0 / is
dense in D!1 n
. Therefore Y 0 is finally dense in D!1 and Fact 3 is proved. t
u
Fact 4. Any HFD space Z D fz W < !1 g D!1 is hereditarily separable.
Proof. If Z is not hereditarily separable, then there exists Y D fy W < !1 g Z
such that y fy W < g for all < !1 . Choose a finite function s 2 Fn.!1 /
such that
(*) y 2 s and s \ fy W < g D ;
for all < !1 . Apply Delta-lemma (see Problem 038) to find an uncountable A
!1 and a finite set F !1 such that dom.s / \ dom.s / D F for any distinct
; 2 A. Since the set DF is finite, there is an uncountable A0 A such that
y jF D y jF for all ; 2 A0 . Take any countably infinite set Y 0 fy W 2 A0 g.
Since Y is an HFD, there exists
< !1 such that Y 0 is dense beyond
. Increasing
if necessary, we can consider that f W y 2 Y 0 g [ F

.
The family S D fdom.s /nF W < !1 g is disjoint, so only countably many
elements of S can intersect the set
. As a consequence, there exists 0 2 A0
such that dom.s0 /nF !1 n
and < 0 whenever y 2 Y 0 . The set Y 0
being dense beyond
there is < !1 such that y 2 Y 0 and y j.dom.s0 /nF D
y0 jdom.s0 /nF . We saw already that y jF D y0 jF and therefore y jdom.s0 / D
y0 jdom.s0 / which shows that y 2 s0 contradicting . /. This contradiction
shows that Z is hereditarily separable, so Fact 4 is proved. t
u
Returning to our solution, take any HFD space Z D fz W < !1 g D!1
(Fact 2). For each < !1 , let y ./ D 0 for all and y ./ D z ./ for each
> . Then y is similar to z for each < !1 , so the set fy W < !1 g is also an
HFD (as a matter of fact, if we take the HFD space Z D fz W < !1 g constructed
in Fact 2, then y D z for all < !1 ).
Observe that the family B D fs W s 2 Fn.!1 /g is a base in D!1 and jBj D !1 .
Choose any enumeration fB W < !1 g of B in which every U 2 B occurs !1 times
(see Fact 3 of S.286). We have B D s for some s 2 Fn.!1 /; let D D dom.s /
for each < !1 . Let x ./ D y ./ for all 2 !1 nD and x ./ D s ./
for all 2 D . Since we changed each y at finitely many coordinates, the space
X D fx W < !1 g is still an HFD by Fact 3. The set X intersects each B because
x 2 B \ X ; this shows that X is dense in D!1 . Any infinite subset of an HFD is
trivially an HFD, so we can consider that the point u whose all coordinates are zeros
does not belong to X (for otherwise we just throw it away from X ).
Let W D fx 2 D!1 W x./ D 1g. Then W is an open subset of D!1 and the
family W D fW W < !1 g covers X because u X . However the cover W
does not have a countable subcover. Indeed, if A !1 is countable, then take any
2 !1 with < for all 2 A. The function s W fg ! D defined by s./ D 1

occurs !1 times in the enumeration fs W < !1 g, so there exists > such that
s D s. Observe that D D fg, so when we changed y to x we only affected
the thScoordinate of y . Therefore x ./ D y ./ D 0 for all 2 A and hence
x fW W 2 Ag. As a consequence, the cover W of the space X has no
countable subcover, so X is not Lindelf. Finally, X is hereditarily separable by
Fact 4, so our solution is complete.
T.041. Prove that, under CH, any sequential compact space has points of countable
character.
Solution. Given a space X and a set A X denote by Seq.A/ the set of all limits
S D A and, if we
of sequences lying in A. It is clear that A Seq.A/ A. Let S0 .A/
have sets fS .A/ W < g for some < !1 , let S .A/ D Seq. fA W < g/.
Observe that the family fS .A/ W < !1 g has the following properties:
(1) A S .A/ A for all < !1 ;
(2) S .A/ S .A/ if < < !1
S
and therefore the set S.A/ D fS .A/ W < !1 g also lies between A and the
closure of A.
Fact 1. A space X is sequential if and only if S.A/ D A for any A X .
Proof. Suppose that S.A/ D A for any A X . If the space X is not sequential,
then there is a non-closed A X such that Seq.A/ D A and hence S.A/ D A A,
a contradiction which proves sufficiency.
Now assume that X is a sequential space and take any A X . Observe that
Seq.S.A// D S.A/; indeed, if B is a convergent sequence from S.A/, then B
S .A/ for some < !1 , and hence the limit of B belongs to SC1 .A/ S.A/.
Since X is sequential, the set S.A/ has to be closed and hence S.A/ A S.A/ D
S.A/ and Fact 1 is proved. t
u
Fact 2. If X is a sequential space and jAj c for some A X , then jAj c.
Proof. Given any set B, let P! .B/ D fC B W jC j !g. Observe first that if
jBj c, then jP! .B/j c. Now, if B X then the set of all convergent sequences
lying in B is a subfamily of P! .B/; therefore jSeq.B/j c whenever jBj c.
We have jS0 .A/j D jAj c. Assume
S that < !1 and we have proved that
jS .A/j c for all
S < . Then j fS .A/ W < gj !
c D c and therefore
jS .A/j D jSeq. fS .A/ W < g/j c as well which proves that jS .A/j c
for all < !1 . Consequently, jS.A/j !1
c D c. Finally, apply Fact 1 to conclude
that jAj D jS.A/j c finishing the proof of Fact 2. t
u
Fact 3. Any sequential space has countable tightness.
Proof. Let X be a sequential space. It suffices to prove that for any non-closed
A X , there is a countable B A such that BnA ; (see Lemma of S.162).
Since X is sequential, there is a convergent sequence S A (taken without its limit)
such that S ! x 2 X nA. It is clear that the set B D S is countable and x 2 BnA,
so t.X / ! and Fact 3 is proved. t
u
Fact 4. Let X be a compact space of countable tightness. Then there exists a

countable set A X and a non-empty G -set H X such that H A.
Proof. Recall that a set fx W < !1 g X is called a free sequence (of length !1 )
if fx W < g \ fx W g D ; for any < !1 . Assume that for any countable
A X , no non-empty G -subset of X is contained in the closure of A. Choose a
point x0 2 X and let F0 D X . Assume that < !1 and we have chosen points
fx W < g and non-empty closed G -sets fF W < g such that
(1) x 2 F F
whenever
< < ;
(2) fx W <
g \ F
D ; for all
< .
T
The set F D fF W < g is a non-empty G -subset of X ; since the set
A D fx W < g is countable, we have H 0 D F nA ;. Evidently, H is a non-
empty G -subset of X , so we can find a closed non-empty G -subset F H (see
Fact 2 of S.328). Picking any point x 2 F we complete our inductive construction
of the set S D fx W < !1 g and the family fF W < !1 g with the properties (1)
and (2).
We claim that the set S is a free sequence in X . Indeed, if < !1 , then we
have fx W g F by (1); thus fx W g F D F and therefore
fx W < g \ fx W g fx W < g \ F D ; for any < !1 by (2).
As a consequence, the compact space X has an uncountable free sequence which
contradicts t.X / D ! (see Problem 328 of [TFS]). Fact 4 is proved. t
u
Returning to our solution, assume CH and take any compact sequential space X .
The space X has countable tightness by Fact 3 and therefore there exists a countable
A X and a non-empty G -set H X such that H A by Fact 4. We have
jH j jAj c by Fact 2. Apply again Fact 2 of S.328 to find a non-empty closed
G -set P H ; then jP j jH j c D !1 . If .x; P / !1 for any x 2 P ,
then we can apply Problem 330 of [TFS] to conclude that jP j 2!1 > !1 which
is a contradiction. Therefore .x; P / ! for some x 2 P . Since P is a G -set in
X and fxg is a G -set in P , the point x is a G -set in X (see Fact 2 of S.358). As
a consequence, .x; X / D .x; X / D ! (see Problem 327 of [TFS]), i.e., x is a
point of countable character in X and our solution is complete.
T.042. Prove that, under CH, there is a P -point in !n!.
Solution. Let ! D !n!; given an infinite set A !, let A D A \ ! (the
closure is taken in !). Then the family B D fA W A is an infinite subset of !g
is a base in ! (see Fact 2 of S.370). Every element of B is clopen in ! (Fact 1
of S.370) and, under CH, the cardinality
S 0 of B is equal to !1 . If U is an arbitrary
clopen subset of ! , then U D B for some B 0 B; since U is compact, we
can consider that B 0 is finite. Since there are at most !1 finite subfamilies of B, the
family C of all clopen subsets of ! also has cardinality !1 .
Fact 1. No non-empty open subset of ! is a union of !1 of nowhere dense
subsets of ! .
S
Proof. Assume that, on the contrary, we have U fF W < !1 g where F is a
nowhere dense subset of ! . Since the closure of a nowhere dense subset is nowhere
dense, we can assume that F is closed in ! for each < !1 . Since B is a base in
! , we can find U0 2 B such that U0 U nF0 . Suppose that < !1 and we have
sets fU W < g with the following properties:
(1) U 2 B and U \ F D ; for all < ;
(2) U U
whenever
< < .
T
By compactness of ! , the set P D fU W < g is closed and non-empty;
being a G -subset of ! it has a non-empty interior by Problem 370 of [TFS];
choose any U 2 B with U Int.P /nF . It is evident that (1) and (2) hold for all
and therefore our inductive construction can be continued giving us a family
fU W < !1 g with the propertiesT (1) and (2). The property (2) and compactness
of ! imply that the set Q D fU W < !1 g is non-empty. S If x 2 Q, then
x 2 U and x F for any < !1 by (1). Therefore x 2 U n fF W < !1 g; this
contradiction finishes the proof of Fact 1. t
u
Returning to our solution observe that the family U of all countable unions of
elements of C has cardinality !1! D c D !1 ; given any U 2 U, the set U nU
is nowhere dense in ! . SinceSjUj !1 , we can apply Fact 1 to conclude that
there exists a point x 2 ! n fU nU W U 2 Ug. We claim that x is a P -point
of the space ! . To prove it, take any family fVn W n 2 !g .x; ! /. Since B
is a base in ! , there exists a family fWn W n 2 !g B such that WnC1 Wn
and x T 2 Wn Vn for all n 2 !. It suffices to show that x 2 Int.W / where

W D fWn W n 2 S !g. Now, if this is not true, then x 2 U where U D ! nW
is open and U D f! nWn W n 2 !g. Observe that ! nWn 2 C for each n 2 !
and therefore U 2 U. As a consequence, x 2 U nU for some U 2 U which is
a contradiction with our choice of the point x. Thus x is a P -point of ! , so our
solution is complete.
T.043. Let X be a Luzin space. Prove that hl.X / D ! and hd.X / !1 .
Solution. Take any right-separated A X ; then A is scattered (see Problem
006) and hence it has a dense discrete subspace D. It is easy to see that every
discrete subspace is nowhere dense in a space without isolated points. Therefore
A clX .D/ is also nowhere dense in X and hence countable because X is a Luzin
space. We proved that every right-separated A X is countable and hence X is
hereditarily Lindelf by Problem 005.
To show that hd.X / !1 , we will need several observations. Take any Y X ;
if U D Int.Y /, then Y nU is nowhere dense in X and hence countable. The set
Y 0 D Y \ U is dense in U and hence it is dense-in-itself (unless it is empty, of
course). Any nowhere dense subset of Y 0 is, evidently, nowhere dense in X and
hence countable. This shows that Y 0 is a Luzin space (or an empty space). As a
consequence, any Y X is a union of a Luzin space Y 0 (which may be empty) and
a countable set. Therefore, to prove that d.Y / !1 for every Y X , it suffices to
establish that d.Z/ !1 for any Luzin space Z.
We will prove first that any non-empty open subset of Z contains a non-empty
open subset of density !1 . Take any U 2 .Z/; there is a left-separated L U
which is dense in U (see Problem 009). Let < be the well-order that left-separates L.
For any x 2 L, let Lx D fy 2 L W y < xg. It is an exercise for the reader that if
jLx j ! for all x 2 L, then jLj !1 . If jLj !1 , there is nothing to prove
because then d.U / !1 ; if jLj > !1 , then the set Lx is uncountable for some
x 2 L; let z D minfy 2 L W jLy j > !g. The set Lz has cardinality !1 and hence
it must be dense in some open set V U . It is evident that d.V / !1 and hence
we established that for every U 2 .Z/, there is V 2 .Z/ such that V U and
d.V / !1 .
Now let U be a maximal disjoint family of non-empty open subsets of Z of
S !1 . Since every non-empty open U Z contains an element of U, the
density
set U is dense in Z. We have c.Z/ hl.Z/ D !, so U is countable. It is an
easy of spaces of density !1 has density !1 , so
S exercise that a countable union S
d. U/ !1 . Any dense subset of U is also dense in Z, so d.Z/ !1 and our
T.044. Prove that if a Luzin space X is separable, then all closed subsets of X are
separable.
Solution. Fix any dense countable A X ; if F is a closed subspace of X , then
P D F nInt.F / is nowhere dense in X and hence countable. The set Q D A \
Int.F / is countable and dense in Int.F / whence the set P [ Q is a countable dense
subset of the space F .
T.045. Prove that no Luzin space can be condensed onto a compact space.
Solution. Denote by D the discrete two-point space f0; 1g; we let D0 D f;g. If f is
a function, then dom.f / is its domain. It is a standard practice to identify functions
and their graphs, so the fact that for functions f and g, we have f jdom.g/ D g
can be written as g f . We will use this agreement here for the sake of brevity.
As usual, any ordinal is identified with the set of its predecessors; in particular,
n D f0; : : : ; n 1g for any n 2 !. If s 2 Dk for some k 2 !, then, for any i 2 D,
we denote by s _ i the function t 2 DkC1 such that tjk D s and t.k/ D i .
Fact 1. If Z is a FrchetUrysohn space without isolated points, then there exists a
closed separable dense-in-itself subspace Y Z.
Proof. Since no point of Z is isolated, it follows from the FrchetUrysohn property
of Z that, for each a 2 Z there is a sequence Sa Znfag which converges to a.
Take any countably infinite
S A0 Z; if we have countable sets A0

An Z,
let AnC1 D An [ . fSa W a 2 An g/. It is clear that this construction gives us a
sequence fAn W n 2 !g of countable subsets of Z such that every S point a 2 An is
a limit of a sequence from AnC1 nfag. Consequently, the set A D fAn W n 2 !g
is countable and has no isolated points. Therefore the space Y D A is the promised
dense-in-itself separable closed subspace of Z. Fact 1 is proved. t
u
Fact 2. Let Z be a compact dense-in-itself space; given a family fFn W n 2 !g of

closed nowhere
S dense subspaces of Z, there exists an uncountable closed subspace
K Zn fFn W n 2 !g.
Proof. Since F0 is nowhere dense in Z, we can choose U; 2 .Z/ with
U ; \ F0 D ;. Suppose that n 1 and, for any k < n and every s 2 Dk , we
have a non-empty open subset Us of the space Z such that
(1) if s 2 Dk , then U s \ Fk D ;;
(2) if s; t 2 Dk and s t, then U s \ U t D ;;
(3) if s 2 Dk ; t 2 Dm and t s, then U s Ut .
Take any t 2 Dn1 ; since Ut 2 .Z/ and Fn is nowhere dense in Z, we can
choose V; W 2 .Z/ such that V [ W Ut ; V \ W D ; and .V [ W / \ Fn D ;.
Let Ut _ 0 D V and Ut _ 1 D W . After carrying out this procedure for all t 2 Dn1 we
will obtain a non-empty open set Us for every s 2 Dn . The property (1), evidently,
holds for k D n. Given distinct s; t 2 Dn , we have two cases:
(1) sj.n 1/ D tj.n 1/; then, by our construction, Us D V and Ut D W for some
V; W 2 .Z/ such that V \ W D ;. Therefore U s \ U t D V \ W D ;, so
the property (2) also holds for k D n.
(2) s 0 D sj.n 1/ t 0 D tj.n 1/; then U s \ U t U s 0 \ U t 0 D ; (the last
equality is valid by the property (2) for k D n 1), and hence (2) is fulfilled in
this case as well.
To check the property (3) observe that it suffices to show that it holds for k D n
and any m < n. By our construction we have U s 0 Us where s 0 D sj.n 1/. The
property (3) applied to k D n 1 and m k shows that U t U s 0 Us , and hence
(3) is also true when k D n. Thus our inductive construction can go on providing
us, for any n 2 !, a non-empty open set Us for each s 2 Dn in such a way that the
properties (1)(3)TareSfulfilled for all k 2 !.
The set K D f fU s W s 2 Dn g W n 2 !g is compact being the intersection of
closed subspaces ofT Z. If z 2 K, then there is a sequence fsn W n 2 !g such that
sn 2 Dn and z 2 fU sn W n 2 !g. It follows from (1) that U sn \S Fn D ; for all
n 2 ! and therefore z Fn for all n 2 !. This shows that K Zn fFn W n 2 !g.
to see that the set K is uncountable, observe that for any s 2 D! , the set
Finally, T
Ks D fU sjn W n 2 !g is non-empty because the family fU sjn W n 2 !g is
decreasing by (3) and consists of compact sets.
Given distinct s; t 2 D! , there is a number n 2 ! such that sjn tjn and
hence Ks \ Kt U sjn \ U t jn D ; by (2). As a consequence, if we choose a
point ys 2 Ks for all s 2 D! , then s 7! ys is an injection of D! to K and hence
jKj jD! j D c > !. Fact 2 is proved. t
u
Fact 3. If Z is a first countable dense-in-itself compact space, then there exists an
uncountable disjoint family K of closed uncountable subsets of Z.
Proof. Any first countable space is FrchetUrysohn, so we can apply Fact 1 to
find a separable closed dense-in-itself Y Z. Of course, it suffices to construct
the promised family in Y , so we can consider that Z is separable; fix any countable
dense D Z. The family ffd g W d 2 Dg is countable and consists of nowhere
dense subsets of Z. This makes it possible to apply Fact 2 to find an uncountable
compact subset K0 ZnD. It follows from density of D that K0 is nowhere dense
in Z.
Suppose that < !1 and we have a family fK W < g of uncountable
nowhere dense subsets of Z with the following properties:
(4) K ZnD for all < ;
(5) K \ K
D ; for any distinct ;
< .
The family F D ffd g W d 2 Dg [ fF W < g is countable and consists of
nowhere dense subsets of Z. ThereforeS we can apply Fact 2 to find an uncountable
dense-in-itself compact set K Zn F . It follows from K ZnD that K
is nowhere dense in Z, so our inductive construction can be continued, giving us a
family K D fK W < !1 g with properties (4) and (5). An immediate consequence
of (5) is that the family K is disjoint; since its consists of uncountable compact
subsets of Z, the proof of Fact 3 is complete. t
u
Now it is easy to finish our solution. Suppose that there is a condensation
f W L ! K of a Luzin space L onto a compact space K. It is immediate that
any continuous image of a hereditarily Lindelf space is hereditarily Lindelf, so
hl.K/ ! because hl.L/ ! (see Problem 043). Therefore K is perfect (see
Problem 001) and hence .K/ D .K/ D ! (see Problem 327 of [TFS]). If x
is an isolated point of K, then f 1 .x/ is an isolated point of L; this contradiction
shows that K is dense-in-itself. By Fact 3, there is an uncountable disjoint family
K of uncountable closed subsets of K. The family L D ff 1 .F / W F 2 Kg is
also disjoint and consists of uncountable closed subsets of L. Since L is a Luzin
space, every P 2 L has non-empty interior in L. The interiors of all elements of
L form a disjoint uncountable family of non-empty open subsets of L which is a
contradiction with c.L/ s.L/ hl.L/ !. This contradiction shows that no
Luzin space can condense onto a compact space.
T.046. Prove that, under CH, there is a Luzin second countable space as well as a
Luzin non-separable space.
Solution. Denote by D the discrete two-point space f0; 1g. We will also need the
space D fx 2 D!1 W jx 1 .1/j !g.
Fact 1. Let X be a dense-in-itself space with the Baire property such that c.X / !
and w.X / !1 . Then, under CH, there is a dense Luzin subspace in the space X .
Proof. Fix an arbitrary base B D fB W < !1 g in the space X such that B ;
for each < !1 (repetitions are allowed in this enumeration to cover the case when
X has a countable base). Since c.X / D ! and w.X / !1 , we can apply Fact 1 of
T.039 to conclude that there exists a family N D fN W < !1 g of closed nowhere
dense subsets of X which is cofinal in the family of all nowhere dense subsets of
X , i.e., for any nowhere dense subset N of the space X , there is < !1 such that
N N . Since the space X has the Baire property, so does B for all < !1
(see Problem 275 of [TFS]). The set N0 being nowhere dense, we can choose a
point x0 2 B0 nN0 . Assume that < !1 and we have chosen points fx W < g
so that
S
(*) x 2 B n. fN
W
g [ fx
W
< g/
for all < . Since the subspace
S B has the Baire property, it cannot be covered
by a first category set P D . fN W g/ [ fx W < g, so we can choose a
point x 2 B nP .
This shows that our inductive construction can go on giving us a subspace L D
fx W < !1 g X such that . / holds for all < !1 . We claim that L is a
Luzin space. It follows from . / that x x whenever and therefore L
is uncountable. Furthermore, the set L is dense in X because L \ B ; for all
< !1 ; thus the space L has no isolated points. Now, if A is a nowhere dense
subspace of L, then A is nowhere dense in X ; since N is cofinal in the family of all
nowhere dense subsets of X , there is < !1 such that A N . By . /, we have
L \ N fx W < g and therefore A fx W < g which shows that A is
countable. This proves that L is a Luzin space and finishes the proof of Fact 1. u t
It was proved in Fact 3 of S.307 that is dense in D!1 , countably compact
and non-compact, while A is compact for any countable A (the bar denotes
the closure in ). An immediate consequence is that is a dense countably
compact non-separable subspace of D!1 . Any countably compact space has the
Baire property (see Problem 274 of [TFS]); besides, c./ D ! and w./ D !1 ,
so, under CH, we can apply Fact 1 to conclude that there is dense Luzin subspace
L . The space L cannot be separable because any countable dense subset of L
would be dense in which is not separable. This shows that CH implies existence
of a non-separable Luzin space.
Finally, if X D R and CH holds, then X is an uncountable dense-in-itself Baire
space with c.X / D w.X / D !. Therefore we can apply Fact 1 again to conclude
that there is a dense Luzin subspace M of the space R. Hence M is an example of
a second countable Luzin space, so our solution is complete.
T.047. Prove that MA(!) holds in ZFC (and hence CH implies MA), while MA(c) is
false in ZFC.
Solution. Given any function f the set dom.f / is the domain of f . If f; g are
functions and f jdom.g/ D g, i.e., the function f extends g, we say that g f .
Take any non-empty partially ordered set .P; /; suppose that Dn is dense in P
for each n 2 !. Pick any d0 2 D0 ; assume that we have d0 : : : dn such that
di 2 Di for all i n. Since the set DnC1 is dense in P, we can choose dnC1 2 DnC1
such that dnC1 dn . This shows that our inductive construction can be continued
to obtain a set fdn W n 2 !g such that dnC1 dn and dn 2 Dn for all n 2 !.
Now let F D fp 2 P W there is n 2 ! such that dn pg. We leave to the reader
the trivial verification of the fact that F is a filter; since F \ Dn 3 dn , the filter
F intersects every Dn and hence we proved MA.!/ for any (not necessarily ccc)
partially ordered set. Since CH says that the only infinite cardinal smaller than c is
!, this proves that CH implies the MartinsSaxiom.
To show that MA.c/ is false, let P D fDB W B is a finite subset of !g; here
D D f0; 1g. Let f g if g f ; we omit the trivial verification of the fact that
is a partial order on P. The set P is countable and hence ccc. Given any h 2 D! , let
Dh D ff 2 P W there is n 2 ! such that n 2 dom.f / and f .n/ h.n/g. Observe
that the set Dh is dense in P for any h 2 D! . Indeed, for any g 2 P, we can take
any n 2 !ndom.g/ and define f 2 P by f jdom.g/ D g and f .n/ D 1 h.n/.
It is clear that f 2 Dh and f g.
For each n 2 !, let Dn D ff 2 P W n 2 dom.f /g. The set Dn is also dense in
P for every n 2 !. To see it, take any g 2 P; if n 2 dom.g/, then g 2 Dn . If not,
then define f 2 P by f jdom.g/ D g and f .n/ D 0. It is clear that f 2 Dn and
f g.
The family D D fDn W n 2 !g [ fDh W h 2 D! g consists of dense subsets of P
and jDj D c. We claim that there is no filter in P which intersects all elements of D.
To see this, assume that F is a filter in P with F \ D ; for all D 2 D. Since
F \ Dn ; for each n 2 !, there is fn 2 F such that n 2 dom.fn /. Note that we
have the following property of F :
(*) if f; g 2 F , then f .n/ D g.n/ for any n 2 dom.f / \ dom.g/;
because there exists h 2 F such that f; g h and hence f .n/ D h.n/ D g.n/. As a
consequence, we can define a function h 2 D! by h.n/ D fn .n/ for each n 2 !. An
immediate consequence of . / is that f h for any f 2 F . However, F \ Dh ;
and hence there is f 2 F such that f 2 Dh which means f .n/ h.n/ for some
n 2 dom.f / and therefore f 6 h; this contradiction shows that MA(c) is false.
T.048. Prove that MA./ is equivalent to MA./ restricted to ccc partially ordered
sets of cardinality , i.e., if MA./ is true for all ccc partial orders of cardinality
, then it is true for all ccc partial orders.
Solution. Assume that MA./ holds for all ccc partial orders of cardinality
and take a ccc partially ordered set .P; / of arbitrary cardinality. Fix any family
fD W < g of dense subsets of P. For each < define a function f W P ! P
as follows: given any p 2 P, there is q 2 D such that q p; let f .p/ D q.
Besides, for any compatible p; q 2 P, choose b.p; q/ 2 P such that b.p; q/ p
and b.p; q/ q (if p and q are incompatible, we do not choose anything for the pair
.p; q/). Thus b W C ! P where C D f.p; q/ 2 P P W p and q are compatibleg.
Take any p0 2 P and let Q0 D fp0 g; assume that m < ! and we have sets
Q0

Qm P such that
S
(*) jQn j for each n m and b..Qn Qn / \ C/ [ . ff .Qn / W < g/
QnC1
S
for any n < m. Since jQm j , the set Q D b..Qm Qm / \ C/ [ . ff .Qm / W
< g/ also has cardinality ; if we let QmC1 D Qm [ Q, the property . /
also holds for n D m C 1 and hence we can construct an increasing sequence
fQm W mS
2 !g of subsets of P such that . / holds for all n 2 !. We claim that the
set Q D fQm W m 2 !g has the following properties:
(1) E D D \ Q is dense in .Q; / for each < ;
(2) the set Q with the order induced from P is ccc.
Take any q 2 Q. Then q 2 Qn for some n 2 !; the property . / implies that
p D f .q/ 2 D \ Q and p q. This proves (1). To establish (2), take any
uncountable A Q. Since P is ccc, the set A cannot be an antichain in P. Thus
there are p; q 2 A which are compatible in P. There is n 2 ! for which p; q 2 Qn ;
since .p; q/ 2 C, the point r D b.p; q/ belongs to QnC1 and hence to Q. As a
consequence, we found r 2 Q such that r p and r q which shows that p and
q are compatible in Q, i.e., A is not an antichain in Q.
Thus we can apply our reduced form of Martins axiom to the set Q to find a
filter F 0 in Q such that F 0 \ E ; for all < . The set F 0 may not be a filter in
P; however the set F D fp 2 P W there is q 2 F 0 such that q pg is a filter in P
(check it, please) and F \ D ; for all < .
T.049. Let .P; / be a partially ordered set. Call a subset A P centered if for
any n 2 N and any p1 ; : : : ; pn 2 A, there is r 2 P such that r pi for all i n.
Assume MAC:CH and take any ccc partially ordered set P. Prove that for any
uncountable R P, there exists an uncountable centered Q R. In particular, all
elements of Q are pairwise compatible.
Solution. Passing to smaller uncountable set R if necessary, we can assume that
R D fr W < !1 g where r r if . Given S an arbitrary p 2 P, let
O.p/ D fr 2 P W r pg; if A P, then O.A/ D fO.p/ W p 2 Ag. We will
also need the sets R D fr W > g and T D fp 2 P W O.p/ \ O.R / ;g
for each < !1 . It is clear that T T if ; however, we have the following
property:
(*) there exists < !1 such that T D T for all .
If . / is not true, then there exist sets f
W
< !1 g !1 and fp
W
< !1 g
P for which
< whenever
< and p
2 T
nT
C1 for all
< !1 . Choose
s
2 S
D fr W
<
C1 g with O.p
/ \ O.s
/ ; for all
< !1 ; this
makes it possible to take q
2 O.p
/ \ O.s
/ for all
< !1 . Now, if
< , then
O.q
/ O.p
/ and hence O.q
/ \ O.R / D ;. Besides, S R and therefore
O.q / O.s / O.S / O.R / which shows that O.q
/ \ O.q / D ;, i.e., q
is incompatible with q . Thus fq

W
< !1 g is an uncountable antichain in P which
contradicts the ccc property of P and proves . /.
Fix < !1 for which . / is fulfilled and consider the partially ordered set Q D
O.R / with the order induced from P. We first show that Q is ccc. Indeed, if
A Q is uncountable, then A cannot be an antichain in P, so there are p; q 2 A
such that r 2 O.p/ \ O.q/ for some r 2 P. By definition of Q, there is p 0 2 R
such that p p 0 ; this shows that r p 0 and hence r 2 Q; consequently, p and q
are also compatible in Q, i.e., A is not an antichain in Q, so Q is also ccc.
The set D
D O.R
/ is dense in Q for all
; to see this take any r 2 Q. It
is evident that ; O.r/ \ Q D O.r/ \ O.R /, so we can apply . / for D
to conclude that O.r/ \ O.R

/ ; which shows that there is s 2 O.R
/ D D
with s r. Now applying MA(!1 ) to the ccc partially ordered set Q and the family
fD
W
< !1 g of its dense subsets, we obtain a filter F in Q such that F \ D
;
for all
< !1 . Finally, let Q D fr 2 R W p r for some p 2 F g. The set Q is
centered for if q1 ; : : : ; qn 2 Q, then there are s1 ; : : : ; sn 2 F such that si qi for
all i n. It is an easy consequence of the property (F1) of the definition of the filter
that there exists s 2 F such that s si for all i n. Therefore s qi for all i n
and hence Q is centered. To see that jQj D !1 observe that for any
< !1 , there
is r 2 F \ D
which implies existence of >
such that r 2 O.r /, i.e., r r
whence r 2 Q. As a consequence, the set f < !1 W r 2 Qg is cofinal in !1 and
hence jQj D !1 .
T.050. Assume MAC:CH. Let Xt Q be a space with c.Xt / D ! for every t 2 T .
Prove that c.X / D !, where X D fXt W t 2 T g.
Solution. Say that !1 is a precaliber of a space Z if for any uncountable family
U .Z/, we can find an uncountable centered U 0 U .
Fact 1. Let Z be a space with c.Z/ D !. If MAC:CH holds, then !1 is a
precaliber of the space Z.
Proof. Let P D .Z/ where U V if U V . It is clear that is a partial order
on P; it is immediate that U; V 2 P are compatible if and only if U \ V ;.
An immediate consequence is that P is ccc because c.Z/ D !. Observe also that
a set A P is centered in the sense of Problem 049 if and only if the family A
is centered as a family of subsets of Z. Therefore we can apply Problem 049 to
conclude that any uncountable family U .Z/ contains an uncountable centered
subfamily, i.e., !1 is a precaliber of Z. Fact 1 is proved. t
u
Fact 2. Given an infinite
Q cardinal , suppose that c.Zs / for Q
all s 2 S . Assume
additionally that c. s2A Zs / for each finite A S . Then c. s2S Zs / .
Q Q
Proof. Let Z D fZs W s 2 S g and ZA D fZs W s 2 Ag for each A S . If
c.Z/ > , then there is a disjoint family fU W < C g .Z/. Shrinking each
U if necessary,
Q Q assume that every U is a standard open subset of Z, i.e.,
we can
U D . s2A Us /. s2S nA Zs / where A is a finite subset of S and Us 2 .Zs /
for each < C and s 2 A . Since C is an uncountable regular cardinal, we can
apply Delta-lemma (see Problem 038) to find a set E C and a finite A S such
C
that jEj
Q D s and A \ A D A for any distinct ; 2 E. For any 2 E let
V D s2A U ; then V D fV W 2 Eg .ZA /. As c.ZA / and jVj > ,
there exist distinct ; 2 E Ssuch that V \ V ;. Take any y 2 V \ V and
define a function z W S ! s2S Zs as follows: z.s/ D y.s/ for all s 2 A; let
z.s/ 2 Us for all s 2 A nA and z.s/ 2 Us for all s 2 A nA. This definition is
consistent because A nA is disjoint from A nA. Finally, define z.s/ arbitrarily for
all s 2 S n.A [ A /. It is immediate that z 2 U \ U ; this contradiction finishes
the proof of Fact 2. t
u
Returning to our solution, apply Fact 1 to conclude that !1 is a precaliber of

Xt for all t 2 T . By Fact 2, it suffices to show that every finite product of spaces
with precaliber !1 has the Souslin property. Observe that it is evident that any space
with precaliber !1 has the Souslin property, so it is sufficient to show that if !1 is a
precaliber of Y and Z, then it is a precaliber of Y Z for then a trivial induction
shows that any finite product of spaces with precaliber !1 also has precaliber !1 and
hence the Souslin property.
So assume that !1 is a precaliber of Y and Z. Given any uncountable family
U .Y Z/ we can assume that U D fU W < !1 g and, for any < !1 , we
have U D V W for some V 2 .Y / and W 2 .Z/. Since !1 is a precaliber
of Y , there is an uncountable A !1 such that the family V D fV W 2 Ag is
centered; apply also the fact that !1 is a precaliber of Z to choose an uncountable
B A such that the family W D fW W 2 Bg is also centered. We claim that the
family U 0 D fU W 2 Bg is centered as well. To see this, take any U1 ; : : : ; Un 2
U 0 ; since V is centered, we can find a point y 2 V1 \

\ Vn . The family W being
centered, there is a point z 2 W1 \

\ Wn . Hence w D .y; z/ 2 U1 \

\ Un ,
so the family U 0 U is centered and uncountable. This shows that precaliber !1 is
preserved by finite products and finishes our solution.
S families A; B exp ! such that jAj ; jBj and < c, suppose

T.051. Given
that Bn. A0 / is infinite for every B 2 B and any finite family A0 A. Prove that
MA./ implies that there exists M ! such that BnM is infinite for any B 2 B
while AnM is finite for any A 2 A.
Solution. Consider the set P D f.K; C/ W K is a finite subset of ! and C is a finite
subfamily of Ag. We will introduce a partial order on the set P as follows: given
P; P 0 2 P with PSD .K; C/ and P 0 D .K 0 ; C 0 /, we let P P 0 if P 0 P; C 0 C
and .KnK 0 / \ . C 0 / D ;. It is straightforward that is indeed a partial order
on P. Besides, .P; / is ccc. To see this, observe that for any finite K !, if
P D .K; C/ 2 P and P 0 D .K; C 0 / 2 P, then P and P 0 are compatible because
Q D .K; C [ C 0 / is a common extension of P and P 0 , i.e., Q P and Q P 0 .
Now, if P 0 is an uncountable subset of P, then there are distinct P; P 0 2 P 0 such
that P D .K; C/ and P 0 D .K; C 0 / for some finite K ! (because there are only
countably may finite subsets of !). Therefore P and P 0 are compatible and hence
P 0 is not antichain.
Given any A 2 A let DA D fP D .K; C/ 2 P W A 2 Cg. The set DA is dense
in P for any A 2 A for if P D .K; C/ 2 P, then Q D .K; C [ fAg/ belongs to
DA and is an extension of P . Observe also that the set D.B; n/ D fP D .K; C/ 2
P W K \ B is not contained in n D f0; : : : ; n 1gg is dense
S in P for any B 2 B
and n 2 !. S Indeed, for any P D .K; C/ 2 P the set Bn. C/ is infinite, so there is
m 2 Bn. C/ with m n. It is immediate that Q D .K [ fmg; C/ is an extension
of P and Q 2 D.B; n/.
The family D D fDA W A 2 Ag [ fD.B; n/ W B 2 B and n 2 !g consists of
dense subsets of P and has cardinality , so MA() isSapplicable to obtain a filter
F P such that F \ D ; for any D 2 D. Let L D fK W .K; C/ 2 F for some
(finite) C Ag. We claim that
(1) the set L \ B is infinite for any B 2 B.

Indeed, if L \ B is finite for some B 2 B, then L \ B n for some n 2 !.
However, there is some F D .K; C/ 2 F \ D.B; n/ and therefore K L and
K \ B 6 n which is a contradiction with L \ B K \ B. We will also need the
following property:
(2) L \ A is finite for any A 2 A.
To show that the property (2) holds take any F D .K; C/ 2 F \ DA . Given any
F 0 D .K 0 ; C 0 / 2 F the elements F and F 0 are compatible because F is a filter.
Thus there is F 00 D .K 00 ; C 00 / 2SF with F 00 F and F 00 F 0 . This implies K 00
K; K 00 K 0 and .K 00 nK/ \ . C/ D ;. Since A 2 C, we have .K 00 nK/ \ A D ;
0 00
and thereforeS K 0 \ A K 0\ A K \ A. Since F 0 was taken arbitrarily, we have
L \ A D fK \ A W .K ; C / 2 F for some C 0 Ag K \ A and therefore
0
L \ A K \ A is a finite set, i.e., (2) is proved.

Finally, observe that it is an immediate consequence of (1) and (2) that the set
M D !nL is as promised, so our solution is complete.
T
T.052 (Booth lemma). Let C exp ! be a family such that jCj D < c and C 0 is
infinite for every finite C 0 C. Prove that MA./ implies that there exists an infinite
L ! such that LnC is finite for any C 2 C.
SA 0D f!nC
Solution. Let T 0 W C 2 Cg0 and B D f!g. Then, for any finite A0 A,
0
S A / D C where C D f!nA W A 2 A g is a finite subfamily of C.
we have !n.
Thus Bn. A0 / is infinite for any (and the unique) B 2 B. Therefore we can apply
Problem 051 to the families A and B to conclude that there is M ! such that
!nM is infinite and AnM is finite for any A 2 A. Consider the set L D !nM ;
we already saw that L is infinite. Now if C 2 C, then .!nC / 2 A, and hence
.!nC /n.!nL/ D LnC is finite, so C is as promised.
T.053. Let A exp ! be an almost disjoint family ( all elements of A are infinite
while A \ B is finite whenever A and B are distinct elements of A). Suppose that
is an infinite cardinal and jAj D < c. Prove that MA./ implies that A is not
maximal.
S
Solution. Assume first that the set D D !n. A0 / is finite
S for some finite A0 A.
0 0
Since A is infinite, there is A 2 AnA ; we have AnD A and hence the infinite
0 0 0
S fA0 \ A W A 2 A g of finite sets
set AnD is a union of a finite family which is a
contradiction. This shows that !n. A / is infinite for any finite A0 A, so if
B D f!g, then the families A and B satisfy the premises of Problem 051. Therefore
there exists M ! such that !nM is infinite and AnM is finite for any A 2 A.
Finally, if L D !nM , then L is an infinite subset of ! such that L \ A is finite for
any A 2 A. Hence A is not maximal because the family A [ fLg is almost disjoint
and is strictly larger than A.
T.054. Assume MAC:CH. Let X be a space such that .x; X / < c for some x 2 X .
Prove that for any countable A X with x 2 A, there exists a sequence fan gn2!
A such that an ! x.
Solution. If x 2 A, then our statement is trivially true, so let us assume that

x A. Then the set U \ A is infinite for any U 2 .x; X/. Consider the family
C DTfU \ A W U 2 Ug where U is some base at the point x with jUj < c. Note
that C 0 is infinite for any finite C 0 C because this set is also the intersection of
A with a neighborhood of x. Our idea is to identify A with ! and apply the Booth
lemma (see Problem 052). Formally, take any bijection ' W A ! !; the family
B DT f'.P / W P 2 Ag has cardinality < c and consists of infinite subsets of ! such
that B 0 is infinite for every finite B 0 B. Therefore we can apply Problem 052 to
the family B to find an infinite set M ! such that M nP is finite for any P 2 B.
Since ' is a bijection, we have the same property for the set L D ' 1 .M /, i.e., L is
infinite and LnC is finite for any C 2 C. Take any enumeration fan W n 2 !g of the
set L. Then fan W n 2 !g D L A and, for any W 2 .x; X/, there is U 2 U such
that x 2 U W . Thus C D U \ A 2 C, so LnC is finite. Since C W , the set
LnW is also finite and hence there is m 2 ! such that an 2 W for all n m. This
proves that an ! x.
T.055. Let X be a second countable space of cardinality < c. Prove that, under MA,
any subset of X is a G in X .
Solution. We need some easy facts about second countable spaces.
Fact 1. For any second countable space Z there exists a metric d on Z and a base
B D fUn W n 2 !g in the space Z such that .d / D .Z/ and diamd .Un / ! 0
when n ! 1.
Proof. We can consider that Z is a subset of I! (see Problem 209 of [TFS]); take
any metric on I! which generates the topology of I! . It is straightforward that the
metric as well as the restriction d of the metric to Z Z is totally
S bounded.
Therefore, for each n 2 N, there exists a finite set An Z such that fBd .a; n1 / W
a 2 An g D Z. Let B D fBd .a; n1 / W n 2 N and a 2 An g; if " > 0, then there is
m 2 N such that n1 < 3" for all n m. If B 0 D fBd .a; n1 / W n < m; a 2 An g,
then B 0 is a finite subfamily of B such that diamd .U / < " for any U 2 BnB 0 . This
shows that if we take any faithful enumeration fUn W n 2 !g of the family B, then
diamd .Un / ! 0 if n ! 1.
To prove that B is a base in Z, take any z 2 Z and any U 2 .z; Z/; there is
n 2 N such that Bd .z; n2 / U . Take any a 2 An such that z 2 Bd .a; n1 /; it is
immediate that Bd .a; n1 / Bd .z; n2 / U which proves that B is a base in Z, so
Fact 1 is proved. t
u
Now apply Fact 1 to find a metric d which generates the topology of X and a
base U D fUn W n 2 !g in X such that diamd .Un / ! 0 when n ! 1. It is evident
that U has the following property:
T T
(*) for any infinite U 0 U, either U 0 D ; or j U 0 j D 1.
For the cardinal D jX j < c choose any enumeration fx W < g of the set
X and take an arbitrary Y X . Let P D fn 2 ! W x 2 Un g for each < and
consider the families B D fP W x 2 Y g and A D fP W x 2 X nY g. It is clear that
A; B exp ! and S jAj ; jBj . Besides, if A0 is a finite subfamily of A and

P 2 B, then P n. A0 / cannot be finite; for otherwise, there 0
T is P 2 A such that
P D P \ P is infinite and hence the infinite intersection fUn W n 2 P g contains
the two-point set fx ; x g which is impossible by . / (observe that our enumeration
of X need not be faithful; however, the points x and x are distinct because x 2 Y
and x 2 X nY ).
Consequently, we can apply Problem 051 to the families A and B to find an
infinite set M ! such that BnM is infinite for any B 2 B and AnM is finite for
any A 2 A. For the set L D !nM , the setSA \ L is finite for any A 2 A and B \ L
is infinite for any B 2 B. Now let Wn D fUm W m 2 L and m ng for all n 2 !.
Given any < such that x 2 Y and any n 2 ! note that L \ P is infinite and
therefore there is m n such that m 2 P \ L whence
T x 2 Um Wn . This proves
that Y Wn for all n 2 ! and hence Y H D fWn W n 2 !g.
On the other hand no point of X nY can belong to H ; indeed, if x 2 X nY , then
L \ P is finite and hence there is n 2 ! such that m L \ P forTany m n. As
a consequence x Wn and hence x H . This proves that Y D fWn W n 2 !g,
so Y is a G -set in X and our solution is complete.
T.056. Prove that MA implies 2 D c for any infinite < c.
Solution. It is evident that 2 2! D c, so we must only prove that 2 c. Since
< c, we can take a set X R with jX j D . Fix any countable base B in the space
X ; since every open subset of X is a union of a subfamily of B, the number of all
open subsets of X does not exceed j exp Bj c. If G is the family of all G -subsets
of X , then jGj j.X/j! c! D c. Now apply Problem 055 to conclude that every
subset of X is a G -set in X , so we have 2 D jexp.X /j D jGj c and therefore
2 D c.
T.057. Let X be a second countable dense-in-itself space. Given a cardinal such
that 0 < < c, suppose that N is a S
nowhere dense subset of the space X for any
< . Prove that, under MA, the set fN W < g is of first category in X .
Solution. Take any base U D fUn W n 2 !g .X / of the space X and let
Bj D fi 2 ! W Ui Uj g for each j 2 !. Since X has no isolated points, the set
Bj is infinite for all j 2 !. Let B D fBj W j 2 !g; consider also the set A D fi 2
! W Ui \ N ;g for every < . The families B and A D fA W < g have
cardinality strictly less than c. If k; j 2 ! and A0 D fA1 ; : : : ; Ak g A, then the
set D D N1 [

[ Nk is nowhere dense in X , so there is W 2 .X / such that
W Uj nD. Since S X is dense-in-itself, Sthe set P D fi 2 ! W Ui W g is infinite
and P Bj n. A0 /. Therefore Bj n. A0 / is infinite for any finite A0 A and
any j 2 !. Consequently, we can apply Problem 051 to families A and B to obtain
a set M ! such that AnM is finite for every A 2 A and BnM is infinite for each
B 2 B. Hence, for L D !nM , the setS L \ A is finite for all A 2 A and L \ B is
infinite for each B 2 B. The set Wn D fUk W k 2 L and k ng is open and dense
in X for all n 2 !. Of course, we must only check density of Wn .
To do it, fix any j 2 !; since there are infinitely many k 2 L \ Bj , we can find
k n with k 2 L\Bj . By definitions of L; Wn and Bj we have Uk Uj \Wn . This
shows that Wn \ Uj ; for each j 2 ! and hence Wn is denseSin X . Consequently,
the set Fn D X nWn is nowhere dense in X , so the set F D fFn W n 2 !g is of
first category.
Take any < ; since the set L\A is finite, there is n 2 ! such that L\A n
and therefore Wn \ N D ;. An immediate consequence S is that N Fn F . The
ordinal was taken arbitrarily, so we proved that N D fN W < g F and
hence the set N is of first category in X .
T.058. Prove that Martins axiom is equivalent to the following statement: Given
a compact space X such thatSc.X / D !, for any family
of nowhere dense subsets
of X with j
j < c, we have
X .
Solution. The following statement:
S that c.X/ D !, for any family
of nowhere dense subsets
Given a compact space X such
of X with j
j < c, we have
X
will be called the topological version of Martins axiom. We must prove the
equivalence of Martins axiom and its topological version.
Take any compact space X with c.X / D ! and assume that
is a family of
nowhere dense subsets of X with j
j < c. Let P D .X /; introduce a partial order
on P by U V if and only if U V . It is clear that any U; V 2 P are incompatible
if and only if U \ V D ;. The Souslin property of X implies that any antichain in
P is countable being a disjoint family of non-empty open subsets of X . Hence P
is ccc. For any N 2
let DN D fU 2 P W U \ N D ;g. The set N is nowhere
dense, so for any V 2 P, there is U 2 P such that U V and U \ N D ; (we
used regularity of X here: remember that all our spaces are assumed Tychonoff if
the opposite is not stated explicitly). This proves that DN is a dense subset of P and
therefore the family D D fDN W N 2
g is a family of dense subsets of P such that
jDj < c. If MA holds, then there is a filter F in P such that F \ DN ; for all
N 2
. It is evident that
T F is a centered family of subsets of X , so by compactness
of X , we have P D fF W F 2 F g ;. Pick any x 2 P ; given any N 2
, there
is F 2 F \ DN and henceSF \ N D ;. Since x 2 F , this shows that x N for any
N 2
and therefore x
. Thus Martins axiom implies the topological version
of Martins axiom.
Fact 1. Let .P; / be an arbitrary partially ordered set. Then, for any p; q 2 P, the
set D.p; q/ D fr 2 P W r p and r qg [ fr 2 P W r ? qg [ fr 2 P W r ? pg is
dense in P.
Proof. Take any r 2 P; if there exists some s r which is incompatible either
with p or with q, then s 2 D.p; q/ and we are done. Hence we can assume that
each s r is compatible with both p and q. In particular, r is compatible with p,
so there is t 2 P with t r and t p. By our assumption t has to be compatible
with q, so there is s 2 P such that s t and s q. Since s t r and s t p,
we have s p, so s 2 D.p; q/ and Fact 1 is proved. t
u
Now assume that the topological version of Martins axiom is true and take any
cardinal < c; we must prove MA() for an arbitrary non-empty ccc partially
S set .P; /. For any p 2 P let Np D fq 2 P W q pg. It is immediate
ordered
that fNp W p 2 Pg D P; if r 2 Np \ Nq , then Nr Np \ Nq , so the family
fNp W p 2 Pg is a base of some topology on the set P (see Problem 006 of [TFS]).
Denote by P the space .P; /. Observe that P need not even be a T1 -space.
Let B D fU 2 D nfeg W U D Int.U /g (the bar denotes the closure in P and
the interior is taken in P as well). The elements of B will be called regular open
sets of P . Let us formulate some simple properties of regular open sets (we leave
their easy proofs to the reader):
(1) Int.W / 2 B for any W 2 ;
(2) if W 2 and W P , then P nW 2 B;
(3) if U; V 2 B and U \ V ;, then U \ V 2 B.
To apply our topological version of the Martins axiom, we are going to use the
family B for constructing a compact Hausdorff space X . This construction (called
the Stone space of a Boolean algebra) can be carried out in a very general situation;
however our concrete case is sufficient for our solution.
A set F B is called a filter if it has the following properties:
(4) U \ V 2 F for any U; V 2 F ;
(5) U 2 F and U V 2 B implies V 2 F .
A filter F B is called an ultrafilter on B if it is maximal with respect to
inclusion, i.e., if G B is a filter and F G, then F D G. An easy application of
the Zorns lemma shows that any filter on B is contained in an ultrafilter on B.
Fact 2. The following properties are equivalent for any F B:
(6) F is an ultrafilter on B;
(7) F is a maximal centered subfamily of B;
(8) F is centered and, for any U 2 B, either U 2 F or P nU 2 F .
Proof. Let us first show that any centered family C on B is contained in a filter
on B. Let ^C be the family of all finite intersections of the elements from C. It is
immediate that U \ V 2 ^C for any U; V 2 ^C. If C 0 D fU 2 B W U V for
some V 2 ^Cg, then C ^C C 0 and C 0 is a filter: we leave the trivial verification
of this fact to the reader.
Now, if F is an ultrafilter, then it is centered; this is proved by an easy induction
using (4). Assume that there is a strictly larger centered family C F . By the
observation above, there is a filter C 0 C. Since F is an ultrafilter, we have F D C 0
and therefore C D F D C 0 which is a contradiction showing that (6)H)(7).
Now, if F is a maximal centered family on B, then, again, take a filter F 0 F .
Since F 0 is also centered and F is maximal centered, we have F D F 0 , i.e., F is a
filter. Furthermore, no filter on B can be larger than F because any filter is centered
and F is maximal centered. As a consequence, F is an ultrafilter and we established
that (6) (7).
If F is a maximal centered family on B take any U 2 B. If U F , then, by

maximality of F , there are U1 ; : : : ; Un 2 F such that U \ U1 \

\ Un D ;.
Analogously, if V D P nU F , then we can choose sets T V 1 : : : ; Vk 2 F
T such that
V \ V1 : : : \ Vk D ;. Since F is centered, the set W D . i n Ui / \ . i k Vi / is
non-empty. It is clear that W \ .U [ V / D ; which contradicts the fact that U [ V
is dense in P . This proves that any maximal centered subfamily of B satisfies (8)
and hence (7)H)(8).
Finally, take any F as in (8). If F is not maximal, then there is U 2 B such
that U F and F 0 D F [ fU g is still centered. The property (8) says that V D
P nU 2 F and hence the sets U and V belong to the centered family F 0 . However,
U \ V D ;; this contradiction shows that (8)H)(7), so Fact 1 is proved. t
u
Returning to our solution let X D fF W F is an ultrafilter on Bg. Given any U 2
B, let OU D fx 2 X W U 2 xg. Let us check that the family O D fOU W U 2 Bg
generates a topology on X as a base. Since no xS2 X is empty, there is U 2 B with
U 2 x. Therefore x 2 OU which proves that O D X . Now, if x 2 OU \ OV ,
then U; V 2 x and hence W D U \ V 2 x. Thus x 2 OW OU \ OV , so we can
apply Problem 006 of [TFS] to see that O is a base of a topology on X . Next we
prove that
(9) .X; / is a Hausdorff compact space.
In the sequel we will write X instead of .X; /. By Problem 118 of [TFS], S to
prove compactness of X , it suffices S to show that for every U O such that UD
X , there is a finite U 0 U with U 0 D X . Fix a family B 0 B such that U D
fOU W U 2 B 0 g. If the cover UShas no finite subcover, then for any U1 ; : : : ; Un 2 B 0
there is x 2 X such that x i n OUi which is equivalent to Ui x for all i n.
T
Therefore P nU i 2 x for all i n (Fact 2) and hence fP nU i W i ng ;. This
shows that the family
D fP nU W U 2 B 0 g is centered; if x is any ultrafilter on B
with
x, then P nU 2 x for all U 2 S B 0 and therefore U Sx for all U 2 B 0 .
Thus x OU for all U 2 B whence x fOU W U 2 B 0 g D U D X which is
0
a contradiction. Therefore U has a finite subcover, so X is compact.

To see that X is Hausdorff, take distinct x; y 2 X . There is U 2 B such that U 2
x but U y. Thus V D P nU 2 y by Fact 2; it is immediate that x 2 OU ; y 2 OV
and OU \ OV D ; so X is Hausdorff and hence Tychonoff (see Problem 124 of
[TFS]).
Let us establish next that X has the Souslin property. It suffices to show that no
uncountable subfamily of O can be disjoint. Suppose that B 0 B is an uncountable
family. Observe that OU \ OV D OU \V , so the family fOU W U 2 B 0 g is disjoint if
and only if B 0 is disjoint. Now if B 0 is disjoint, then for each U 2 B 0 there is p.U / 2
P such that Np.U / U . It turns out that the family fNp.U / W U 2 B 0 g is disjoint,
i.e., the set fp.U / W U 2 B 0 g is an uncountable antichain. This contradiction with
ccc property of P shows that c.X / !.
Now assume that D is a dense subset of S P for all < . For each p 2 P,
let Up D Int.N p /. Consider the set W D fOUp W p 2 D g. We claim that W
is an open dense subspace of X . Since the openness is clear, let us prove that W is
dense in X . Given any U 2 B take any p 2 P with Np U . The set D is dense in

P, so there is q 2 D with q p. As a consequence, Nq Np U and therefore
Uq D Int.N q / IntU D U which shows that OUq OU \ W . We proved that
W intersects every element of O, so W is dense in X for all < .
The set F D X nW is nowhere S dense in X , so by the
T topological version of
Martins axiom, there is x 2 X n fF W < g D fW W < g. Recall
that x is an ultrafilter on B. We claim that the family C D fp 2 P W Up 2 xg is
centered (the term centered is used in the sense of Problem 049 here; we hope
that its meaning is always clear from the context). Indeed, if p1 ; : : : ; pn 2 F , then
Up1 ; : : : ; Upn 2 x and hence U D Up1 \

\ Upn ;. Observe that Npi is dense
in Upi and hence in U for every i n. Therefore Np1 \ U ;; if k < n and we
proved that Gk D Np1 \

\ Npk \ U ;, then Gk is a non-empty open subset of
U , so NpkC1 \ Gk D Np1 \

NpkC1 \ U ;. This shows that our inductive proof
can go on to finally establish that Gn D Np1 \

\Npn \U D Np1 \

\Npn ;.
If r 2 Np1 \

\ Npn , then r pi for all i n, so p1 ; : : : ; pn are compatible
and hence C is centered. Observe also that p 2 C and p q implies q 2 C because
Uq Up 2 x implies Uq 2 x, i.e., q 2 C. Note also that C \ D ; for each
< because, by definition of W , there exists p 2 D such that x 2 OUp , i.e.,
Up 2 x and therefore p 2 C \ D .
To sum up, we proved that the topological version of Martins axiom implies
that for any ccc partially ordered set P if D is a family of dense subsets of P with
jDj , then there is a centered C P such that C \ D ; for any D 2 D and C
is backwards closed, i.e., p 2 C and p q implies q 2 C.
Now take any ccc partially ordered set P of cardinality . Given any family
D of dense subsets of P with jDj , let D0 D D [ fD.p; q/ W p; q 2 Pg where
D.p; q/ D fr 2 P W r p and r qg [ fr 2 P W r ? qg [ fr 2 P W r ? pg for all
p; q 2 P. Each D.p; q/ is dense in P by Fact 2, so D0 is a family of dense subsets
of P and jD0 j . We showed that there must exist a centered backwards closed
set F such that F \ D ; for any D 2 D0 . We claim that F is a filter. Indeed,
take any p; q 2 F ; since F \ D.p; q/ ;, there is r 2 F \ D.p; q/. The family
F is centered, so it is impossible that r ? q or r ? p. Thus r p and r q which
shows that F is a filter such that F \ D ; for any D 2 D and hence MA()
is established for all ccc partially ordered sets of cardinality . As a consequence
MA() holds in any ccc partially ordered set by Problem 048. Since the cardinal
< c was chosen arbitrarily, we established that the Martins axiom holds in its full
generality whenever its topological version is true.
T.059. Show that, under MAC:CH, if s .X / D !, then hd.X ! / D hl.X ! / D !.
In particular, neither strong S -spaces nor strong L-spaces exist under MAC:CH.
Solution. Let us prove first that for any space Z, under MA+:CH, it follows from
s .Z/ D ! that hl.Z/ D !. If this is not true, then there exists a space Z such that
s .Z/ D ! and there is a right-separated R Z with jRj D !1 (see Problem 005).
Since s .R/ D ! and hl.R/ > !, we can assume that Z D R; furthermore, without
loss of generality, we can assume that Z D fx W < !1 g and Z is right-separated
by its indexation (see Fact 2 of T.005).
The set Z D fx W g is an open neighborhood of x , so we can choose

U 2 .x ; Z/ such that U Z for all < !1 . A subspace Y Z is called
special if x 2 Y implies that Y \ U is finite. It is clear that every special subspace
of Z is discrete.
Consider the set P DSfp W p is a finite subset of Zg. Given p; q 2 P, let p q
if p q and .pnq/ \ . x 2q U / D ;. We omit a straightforward verification that
.P; / is a partially ordered set. If p 2 P, let i .p/ D minf < !1 W x 2 pg
and iC .p/ D maxf < !1 W x 2 pg. Let us show that P is ccc. If there is an
uncountable antichain C P, then we can apply the Delta-lemma (see Problem
038) to find an uncountable C 0 C and a finite set r Z such that p \ q D r for
any distinct p; q 2 C 0 . The family C0 D fpnr W p 2 C 0 g is disjoint and consists of
finite sets, so, for any < !1 , the family fd 2 C0 W d \ ;g is countable. This
makes it possible to choose an uncountable C 00 C 0 with the following properties:
(1) i .pnr/ > iC .r/ for any p 2 C 00 ;
(2) C 00 D fp W < !1 g and iC .p nr/ < i .p nr/ for all < < !1 ;
(3) there is n 2 N such that jpnrj D n for all p 2 C 00 .
If p nr D fy1 ; : : : ; yn g (the order of the elements of p nr is fixed but taken
arbitrarily), then let z D .y1 ; : : : ; yn / 2 Z n for each < !1 . It follows from (2)
and the definition of the family fU
W
< !1 g that
S
(4) for any < < !1 , we have .p nr/ \ . x
2p U
/ D ;
S
because x
2p U
fx W iC .p /g. Thus, if < , then for incompatibility
S
of p and p it is necessary that .p nr/ \ W ; where W D x
2p nr U
.
Now consider the set O D fz D .z1 ; : : : ; zn / 2 Z n W zi 2 W for some i ng.
It is clear that O is an open subset of Z n ; our last observation shows that z 2 O
for each . On the other hand, it follows from (4) that zC1 O ; therefore
G D OC1 nO is an open subset of Z n such that G \.fzC1 W < !1 g/ D fzC1 g
for each < !1 . Consequently, the set fzC1 W < !1 g is an uncountable discrete
subspace of Z n which contradicts s.Z n / s .Z/ D !. This contradiction proves
that P is ccc.
For each < !1 let D D fp 2 P W iC .p/ > g; the set D is dense in P
for each < !1 . Indeed, if p 2 P, then let D maxfiC .p/ C 1; C 1g; then
q D p [ fx g p and q 2 D . If MAC:CH holds, then the family D D fD W
< !1 g has cardinality < c and consists of dense subsets of P. Choose any S filter
F P such that F \ D ; for all < !1 . We claim that the set Y D F
is special. To see this, take any x 2 Y and any p 2 F such that x 2 p. If q is
an arbitrary element of F , then q is compatible with p and hence there is r p
and r q. As a consequence, .qnp/ \ U .rnp/ \ U D ; by the definition of
the order . Therefore q \ U p \ U ; since this is true for any q 2 F , the set
Y \ U p \ U is finite and therefore Y is a special subset of Z. Finally, observe
that Y has to be uncountable because any countable set is contained in some Z
while F \ DC1 ; which shows that there is p 2 F with pnZ ; and hence
Y nZ ;. This final contradiction shows that hl.Z/ !.
Thus we have proved that, under MAC:CH, any space Z with s .Z/ D !
is hereditarily Lindelf. Besides, if s .Z/ D !, then s .Z n / D s .Z/ D ! for
any n 2 N and therefore hl .Z/ D !. Furthermore, s .Cp .Z// D s .Z/ D
! (see Problem 025) and hence hl .Cp .Z// D ! which shows that hd .Z/ D
hl .Cp .Z// D ! (see Problem 027). Thus, for Z D X , it follows from s .X / D !
and MAC:CH that hd .X / D hl .X / D !. Finally, apply Problem 011 and
Problem 012 to conclude that, under MAC:CH, we have hl.X ! / D hd.X ! / D !
for each space X such that s .X / D !.
T.060. Prove that, under MAC:CH, every compact space of countable spread is
Solution. Let Z be a compact space with s.Z/ D !. Observe first that t.Z/ D !
because any free sequence is a discrete subset of Z, so all free sequences in Z are
countable (see Problem 328 of [TFS]; note that Martins axiom is not needed for
this conclusion). Let us show first that there is a countable A Z such that A has
non-empty interior. Assuming that this is not true we are going to construct a family
fF W < !1 g with the following properties:
(1) F is a closed separable subset of Z (and hence Int.F / D ;);
(2) F F if < < !1 ;
(3) For each < !1 there is a countable A FC1 nF such that F A.
To start off, pick any x 2 Z and let F0 D fxg. If < !1 and we have a collection
fF W g with the properties (1)(3), fix a dense countable set B D fbn W n 2
!g F ; since F has S empty interior, every bn is in a closure of a countable set
An ZnF . If A D n2! An and FC1 D F [ A, then the conditions (1)(3) are
fulfilled for the family fF W C 1g. If is a limit ordinal and we have a family
S
fF W < g with (1)(3), let F D fF W < g. It is immediate that (1)(3)
are also fulfilled for the collection fF W g. Thus our inductive construction
can be continued S providing us a family fF W < !1 g with the properties (1)
(3). Let F D fF W < !1 g. It is an easy exercise to see that t.Z/ D !
implies that F is a closed subspace of Z and hence compact. Note also that c.F /
s.F / s.Z/ D !. Besides, (3) implies that F is nowhere dense in FC1 and
hence in F . Therefore the compact space F is a union of !1 -many (and hence < c-
many) nowhere dense subsets which is a contradiction with the topological version
of Martins axiom (see Problem 058).
This proves that, under Martins axiom, every compact space of countable spread
has a non-empty separable open set. Now, given any compact space Z with s.Z/ D
!, consider a maximal disjoint family U of non-empty separable openSsubsets of Z.
Observe that U is countable because c.Z/ s.Z/ D !. Besides, U hasS to be
dense in Z; for otherwise, there is W 2 .Z/ such that P D W Zn. U/.
Since P is a compact space of countable spread, it has a non-empty open separable
V P . The set U D V \ W is separable, non-empty and open in Z; it is evident
that U U and the family U [ fU g is disjoint which contradicts maximality of U.
This contradiction shows that U is dense in Z. Being a countable union of separable
S
S space U is also separable. If we take any countable A U which
subspaces, the
is dense in U, then A is dense in Z, so Z is separable. As a consequence, every
compact space of countable spread is separable under MAC:CH. Thus, if Z is
compact and s.Z/ D !, then all closed subsets of Z are separable under MAC:CH
because the previous statement is also true for all closed subspaces of Z.
Finally, take any Y Z; since Y is separable, we can take a countable A Y
with A D Y . It follows from t.Z/ D ! that, S for any a 2 A, there is a countable
Ba Y such that a 2 B a . The set B D fBa W a 2 Ag Y is countable and
B A D Y Y . Therefore B is a countable dense subset of Y and we proved
that every subspace of Z is separable. Thus MAC:CH implies that every compact
space of countable spread is hereditarily separable.
T.061. Prove that, under MAC:CH, every compact space of countable spread is
perfectly normal.
Solution. To show that every compact space of countable spread is hereditarily
Lindelf under MAC:CH, take any compact X with s.X / D !. If X is not
hereditarily Lindelf, then there is Y D fy W < !1 g X which is
right-separated by its indexation (see Fact 2 of T.005). The space Y is also a
counterexample to our statement, so to obtain a contradiction, there is no loss of
generality in assuming that Y D X .
The set Y D fy W g is an open neighborhood of y in Y , so we can
choose U 2 .y ; X / such that U \ Y Y for all < !1 . A subspace Z Y
is called special if y 2 Z implies that Z \ U is finite. It is clear that every special
subspace of Y is discrete. Given any A !1 , let YA D fy W 2 Ag.
Consider the set P D fp W p is a non-empty finite subset of Y g, and for any
n 2 N, let PnSD fp 2 P W jpj D ng. Given p; q 2 P, let p q if p q
and .pnq/ \ . y 2q U / D ;. We omit a straightforward verification that .P; /
is a partially ordered set. Given any p 2 P, let i .p/ D minf < !1 W y 2 pg
and iC .p/ D maxf < !1 W y 2 pg. To proceed by contradiction, assume that
P is ccc. The set D D fp 2 P W iC .p/ > g is dense in P for each < !1 .
Indeed, if p 2 P, let D maxfiC .p/ C 1; C 1g; then q D p [ fy g p and
q 2 D . Thus the family D D fD W < !1 g consists of dense subsets of P.
If MAC:CH holds, then D has cardinality < c, so we can choose S a filter F P
such that F \ D ; for all < !1 . We claim that the set Z D F is special.
To see this, take any y 2 Z and any p 2 F such that y 2 p. If q is an arbitrary
element of F , then q is compatible with p and hence there is r 2 F such that r p
and r q. As a consequence, .qnp/ \ U .rnp/ \ U D ; by the definition of
the order . Therefore q \ U p \ U ; since this is true for any q 2 F , the set
Z \ U p \ U is finite and therefore Z is a special subset of Y . Finally, observe
that Z has to be uncountable because any countable set is contained in some Y
while F \ DC1 ; which shows that there is p 2 F with pnY ; and hence
ZnY ;. Thus Z is an uncountable discrete subspace of X which contradicts
s.X / D !. This contradiction shows that P cannot be ccc.
T
Let R D Y nY for all < !1 ; we will also need the set R D fR W < !1 g.
By compactness of X , the set R is compact and non-empty. It is an immediate
consequence of the fact that Y is open in Y that R \ Y D ; for each < !1 and
therefore R \ Y D ;. Given p; q 2 P, we say that p < q if iC .p/ < i .q/.
Say that a family B of subsets of Y is adequate if it has the following
properties:
(4) every B 2 B is countable and B \ R D ;;
(5) there is n 2 N, a set fq W < !1 g Pn and a family fB W < !1 g B such
that q < q and B \ q ; whenever < < !1 .
We proved that P cannot be ccc; therefore there is an uncountable antichain
C P; applying the Delta-lemma (see Problem 038) we can find an uncountable
C 0 C and a finite set r Y such that p \ q D r for any distinct p; q 2 C 0 .
The family C0 D fpnr W p 2 C 0 g is disjoint and consists of finite sets, so, for any
< !1 , the family fd 2 C0 W d \ ;g is countable. This makes it possible to
choose an uncountable C 00 C 0 with the following properties:
(6) i .pnr/ > iC .r/ for any p 2 C 00 ;
(7) C 00 D fp W < !1 g and p nr < p nr for all < < !1 ;
(8) there is n 2 N such that jp nrj D n for all < !1 .
the family fU W < !1 g
It follows from the property (7) and the definition of S
that, for any ordinals < < !1 , we have .p nr/ \ . x
2p U
/ D ; because
S
x
2p U
fx W iC .p /g. Thus, if < , then for incompatibility of p
and p it is necessary that
S
(9) .p nr/ \ W ; where W D x
2p nr U
.
Now it follows from (6)(9) that the family D of all finite unions of the family
fU \ Y W < !1 g is adequate because we can let q D p nr and B D W for all
< !1 .
Say that a collection fB W < !1 g of subsets of Y is cofinally centered on a
set Z Y if for any uncountable T Z there is < !1 such that the family
fB \ T W g is centered.
Lemma. Suppose that B is an adequate family closed under finite unions, i.e., any
finite union of elements of B is an element of B. Then there exists an uncountable
A !1 and a collection fB W < !1 g B which is cofinally centered on the set
YA D fy W 2 Ag.
Proof. Fix the smallest n 2 N for which there exists a set fq W < !1 g Pn and a
family B 0 D fB W < !1 g B which satisfy (5). Let A D fi .q / W < !1 g and
suppose that B 0 is not cofinally centered on YA . Then there exists an uncountable
counterexample C A such that, for each < !1 , there is a finite set b !1 n
for which
T
(*) YC \ . fB W 2 b g/ D ;.
The set C will be still a counterexample if we make it smaller. The property . /

will still hold if we substitute each b with some b./ for which ./ > . Thus
we can make the necessary changes in the set C and in the collection fb W < !1 g
to show that we can assume, without loss of generality, that there is an increasing
enumeration f W < !1 g of the set C such that
(10) b < for each < !1 ;
(11) if, for each < !1 , we let r D q
for the ordinal
defined by i .q
/ D ,
then < < !1 implies r < b .
S
Consider the sets s D r nf g and C D fB
W
2 b g for each < !1 .
If < , then, by . /, there is some
2 b such that y B
. However B
\r
; by (5) and (11), so B
\ s ;. Since
2 b , we have B
C which
implies s \ C ;. An easy consequence of (5) for the set fq W < !1 g is
that s < s whenever < < !1 . Thus the property (5) is fulfilled for the set
fs W < !1 g Pn1 and the family fC W < !1 g B contradicting the
minimality of n and finishing the proof of our lemma. t
u
Returning to our solution, apply the lemma to our adequate family D to choose
an uncountable A !1 and a collection fD W < !1 g D which is cofinally
centered on YA . By definition of D,
Tthe set K D D is compact and disjoint from
R for every
S < ! 1 . Let H D > K for each < !1 and consider the set
H D <!1 H . Since t.X / D !, the set H is compact (observe that some of
the sets H can be empty). The set H being disjoint from R, the decreasing family
fR \ H W < !1 g of compact subsets of X has empty intersection. Therefore
R \ H D ; for some < !1 . For the set C D An, find < !1 such that
the family fD \ YC W > g is centered. Then the family fK \ YC W > g
is also centered and therefore the family of compact sets fK \ R W > g
is centered
T because YC D R . By compactness of the space X , there is a point
x 2 fK \ R W > g D H \ R H \ R D ; which is a contradiction.
This contradiction shows that all right-separated subspaces of X are countable and
hence X is perfectly normal (see Problem 001), so our solution is complete.
T.062. Suppose that MAC:CH hold. Prove that every compact space X with
s.X X / ! is metrizable.
Solution. If Z is a space, let Z D f.z; z/ W z 2SZg Z Z be the diagonal of Z.
Given a cover U of the space Z, let St.z; U/ D fU 2 U W z 2 U g for any z 2 Z. A
cover V of the space Z is a barycentric refinement of the cover U if, for any z 2 Z,
there is U 2 U such that St.z; V/ U . Any compact space Z is paracompact and
hence any open cover of Z has an open barycentric refinement (see Problem 230 of
[TFS]).
Fact 1. Let Z be a compact space such that D Z is a G -subset of Z Z. Then
Z is second countable and hence metrizable.
Proof. The diagonal is closed in Z Z; if it is a G -subset of Z Z, then, by

paracompactness of Z, there exists a sequence fOn W n 2 !g .Z Z/ and a
collection fUn W n 2 !g of open finite covers of Z such that
(1) TnC1 On for each n 2 !;
O
(2) SfOn W n 2 !g D ;
(3) fU U W U 2 Un g On for every n 2 !;
(4) the cover UnC1 is a barycentric refinement of Un for all n 2 !.
S S
The family S D fUn W n 2 !g is countable and S D Z, so there exists a
topology on the set Z generated by S as a subbase (see Problem 008 of [TFS]).
Let us prove that Y D .Z; / is a Hausdorff space. Take any distinct a; b 2 Z;
since c D .a; b/ , there is n 2 ! such that c On . We claim that the set
fa; bg is not contained in St.x; UnC1 / for any x 2 Z. Indeed, if there is x 2 Z
with fa; bg St.x; UnC1 /, then there is W 2 Un such that St.x; UnC1 / W and
therefore fa; bg W . Thus c 2 W W On which is a contradiction.
Now, take U; V 2 UnC1 such that a 2 U and b 2 V ; if z 2 U \ V , then fa; bg
St.z; UnC1 / which we proved not to be possible. Therefore U \ V D ;, so the space
Y is Hausdorff. The identity map i W Z ! Y is continuous because i 1 .W / D
W 2 .Z/ for any W 2 S (see Problem 009 of [TFS]). Being a continuous image
of a compact space Z, the space Y is a compact Hausdorff and hence Tychonoff
space (see Problem 124 of [TFS]). The subbase S of the space Y is countable; the
family of all finite intersections of the elements of S is a countable base of Y , so Y is
second countable and hence metrizable (see Problems 209 and 212 of [TFS]). As a
consequence, i condenses Z onto a metrizable space Y . Therefore Z is metrizable
by Problem 140 of [TFS]. Any metrizable compact space is second countable, so
Fact 1 is proved. t
u
Returning to our solution, observe that the inequality s.X X / ! implies
that, under MAC:CH, the space X X is perfectly normal (see Problem 061) and
hence its diagonal X is a G -set in X X . Now apply Fact 1 to conclude that X
is second countable and metrizable.
T.063. Prove that MAC:CH implies that there are no Luzin spaces.
S space X and consider

Solution. Suppose to the contrary that there exists a Luzin
the family U D fU 2 .X / W jU j !g; if X 0 D U, then U is an open
cover of X 0 . Since
S 0hl.X / ! (see Problem 043), there is a countable U 0 U
0
such that X D U . Consequently, X is countable. If some V 2 .X nX 0 / is
0
countable, then there is W 2 .X/ such that W \ .X nX 0 / D V . Therefore W SD

V [ .W \ X 0 / 2 .X/ is countable which is a contradiction because W 6 U.
Thus all non-empty open subsets of Y D X nX 0 are uncountable and, in particular,
Y has no isolated points. Furthermore, any nowhere dense subset A of Y is also
nowhere dense in X and hence jAj !. This proves that
(*) if there is a Luzin space, then there is a Luzin space whose all non-empty open
subsets are uncountable. We will call such spaces everywhere uncountable.
If A is any set, let Fin.A/ be the family of all finite subsets of A; if f is function,
then dom.f / is its domain. If U and V are families of subsets of a set X , then V is
properly inscribed in U if for any V 2 V there is U 2 U such that V is properly
contained in U , i.e., V U and V U .
Fact 1. Let M be a second countable space. Suppose that Z is an uncountable
subset of M . Then, under MAC:CH, there exists a disjoint family fF W < !1 g
of closed subsets of M such that jF \ Zj !1 for each < !1 .
Proof. Fix some countable base B of the space M which is closed under finite
unions, i.e., any finite union of elements of B belongs to B. It is easy to find a
disjoint family fZ W < !1 g exp.Z/ such that jZ j D !1 for all < !1 . Let P
be the set of pairs p D .s; a/ such that
(1) s and a are functions for which there is a finite set D.p/ !1 ! such that
dom.s/ D dom.a/ D D.p/;
(2) s W D.p/ ! B and a W D.p/ ! Fin.Z/;
(3) a.; n/ 2 Fin.Z / for each .; n/ 2 D.p/;
(4) a.; n/ s.; n/ for any .; n/ 2 D.p/.
Given any elements p D .s; a/ 2 P and q D .t; b/ 2 P, say that p q
if D.q/ D.p/ and p.; n/ q.; n/; b.; n/ a.; n/ for each .; n/ 2
D.q/. It is immediate that .P; / is a partially ordered set. Let us show that P
has ccc. Assume that S is an uncountable subset of P; by the Delta-lemma (see
Problem 038) there is an uncountable S 0 S and a finite set F !1 ! such
that D.p/ \ D.q/ D F for any distinct p; q 2 S 0 . Since B is countable, there are
only countably many functions from F to B. Thus there is an uncountable S 00 S 0
such that, for any p D .s; a/ 2 S 00 and q D .t; b/ 2 S 00 , we have sjF D tjF .
As a consequence, given distinct p D .s; a/ 2 S 00 and q D .t; b/ 2 S 00 , we can
define a function w W D.p/ [ D.q/ ! B by the conditions w.; n/ D s.; n/ for all
.; n/ 2 D.p/ and w.; n/ D t.; n/ for all .; n/ 2 D.q/nD.p/. Furthermore, let
c.; n/ D a.; n/ for all points .; n/ 2 D.p/nD.q/ and c.; n/ D b.; n/ for all
.; n/ 2 D.q/nD.p/; besides, let c.; n/ D a.; n/ [ b.; n/ for all .; n/ 2 F .
It is evident that the point r D .w; c/ 2 P is a common extension of p and q, i.e.,
r p and r q, so S is not an antichain which shows that P has ccc.
Given any < !1 and z 2 Z , let E.z; / D f.s; a/ 2 P W z 2 a.; n/ for some
n 2 !g. The set E.z; / is dense in P for all < !1 and z 2 Z . Indeed, if p D
.s; a/ 2 P, then take any n 2 ! such that .; n/ D.p/, any B 2 B with z 2 B
and let q D .t; b/ 2 P be defined by tjD.p/ D s; bjD.p/ D a; b.; n/ D fzg and
t.; n/ D B. It is evident that q p and q 2 E.z; /.
Now, for any distinct ordinals ; 2 !1 and any m; n 2 !, let H.; ; m; n/ D
fp D .s; a/ 2 P W s.; n/ \ s.; m/ D ;g. The set H.; ; m; n/ is also dense in P
for all possible 4-tuples .; ; m; n/. To show this, take any point p D .s; a/ 2 P.
If .; n/ D.p/, then choose any z 2 Z and any U 2 B \ .z; M / and define
p 0 D .s 0 ; a0 / 2 P by s 0 jD.p/ D s; a0 jD.p/ D a; s 0 .; n/ D U and s 0 .; n/ D fzg.
It is clear that p 0 p, so it suffices to find q 2 H.; ; m; n/ with q p 0 . An
identical reasoning for the case when .; m/ D.p/ shows that we can assume,
without loss of generality, that Q D f.; n/; .; m/g D.p/.
The finite sets a.; n/ Z and a.; m/ Z are disjoint because

Z \ Z D ;. Since the family B is closed under finite unions, we can find sets
U; V 2 B such that a.; n/ U; a.; m/ V and U \ V D ;. Now let
t.; n/ D U \ s.; n/; t.; m/ D s.; m/ \ V and t.
; k/ D s.
; k/ for all
.
; k/ 2 D.p/nQ. It is immediate that q D .t; a/ 2 H.; ; m; n/ and q p
which shows that H.; ; m; n/ is dense in P for all distinct ; 2 !1 and any
m; n 2 !. Thus the family
DDfE.z; / W < !1 ; z 2 Z g[fH.; ; m; n/ W ; 2 !1 ; m; n 2 ! and g
consists of dense subsets of P and has cardinality < c. Applying the Martins axiom
we can find a filter F P such that F \ D ; for any D 2 D. We will consider
for technical purposes that for any p D .s; a/ 2 P, we have s.; n/TD M and
a.; n/ D ; for all .; n/ 2 .!1 !/nD.p/.
S Consider the set F;n D fs.; n/ W
.s; a/ 2 F for some ag and let K D n2! F;n . It is clear that the set F;n is
closed in M for any .; n/ 2 !1 !. Observe first that Z K for each <
!1 ; indeed, if z 2 Z , then pick any p D .s; a/ 2 F \ E.z; /. There is some
n 2 ! such that z 2 p.; n/. Now take any q D .t; b/ 2 F ; there is r D .u; c/ 2
F such that r p and r q. In particular, u.; n/ s.; n/ \ t.; n/ while
c.; n/ a.; n/ [ b.; n/ and therefore z 2 c.; n/ u.; n/. As a consequence,
z 2 u.; n/ t.; n/ t.; n/ which shows that z 2 t.; n/ for all q D .t; b/ 2 F ,
i.e., z 2 F;n K ; the point z 2 Z was chosen arbitrarily, so Z K for all
< !1 .
Now take any distinct ordinals ; < !1 and any m; n 2 !. There is a point
p D .s; a/ 2 F \H.; ; m; n/; it is clear that F;n \F;m s.; n/\s.; m/ D ;.
An immediate consequence is that K \ K D ; for any distinct ; < !1 .
Since each K covers an uncountable set Z , there is n./ 2 ! such that F;n./ \
Z is uncountable. The sets F D F;n./ are as promised, so Fact 1 is proved. u t
Fact 2. Let X be a Luzin space. If M is a second countable space and MAC:CH
holds, then f .X / is countable for any continuous map f W X ! M .
We can consider that M D f .X /; assume that M is uncountable. If, in the
formulation of Fact 1, we let Z D M , then we obtain an uncountable disjoint
family U of uncountable closed subsets of M . The family V D ff 1 .P / W P 2 Ug
consists of disjoint closed uncountable subsets of X . The space X being Luzin, we
have IntX .V / ; for any V 2 V. Therefore fIntX .V / W V 2 Vg is an uncountable
disjoint family of non-empty open subsets of X which contradicts the fact that
c.X / hl.X / D ! (see Problem 043). Fact 2 is proved.
Fact 3. Let I be the space 0; 1 with the natural topology. No Luzin space can
be mapped continuously onto I . Note that no additional axioms are needed for the
proof of this fact while, under MAC:CH, it is an immediate consequence of Fact 2.
Proof. By Fact 3 of T.045, there exists a disjoint family fF W < !1 g of
uncountable closed subsets of I . If L is a Luzin space and f W L ! I is a
continuous onto map, then U D IntL .f 1 .F // ; for all < !1 because
f 1 .F / is an uncountable closed subset of L. Thus fU W < !1 g is an

uncountable disjoint family of non-empty open subsets of L which is a contradiction
with c.L/ hl.L/ D ! (see Problem 043). Fact 3 is proved. t
u
A space is zero-dimensional if it has a base which consists of clopen sets.
Fact 4. Let Z be an arbitrary space. If Z is not zero-dimensional, then it maps
continuously onto I . No additional axioms are needed for the proof of this fact.
Proof. Since Z is not zero-dimensional, there is a point z 2 Z and U 2 .z; Z/
such that U Z and there is no clopen set W for which z 2 W U . The space
Z being Tychonoff, there is a continuous function f W Z ! I such that f .z/ D 1
and f j.ZnU / 0. If t 2 I nf .Z/, then t 2 .0; 1/ and W D f 1 ..t; 1 / is an open
subset of Z with z 2 W U . Besides, ZnW D f 1 .0; t / D f 1 .0; t// is again
an open set, so W is clopen which is a contradiction. Thus f .Z/ D I and Fact 4 is
proved. t
u
Fact 5. Any Luzin space is zero-dimensional (no additional axioms are needed here
either).
Proof. If L is a Luzin space, then it cannot be continuously mapped onto I by
Fact 3. Therefore L is zero-dimensional by Fact 4, so Fact 5 is proved. u
t
Fact 6. Let X be an everywhere uncountable Luzin space. Suppose that we have a
family U D fU.n; k/ W n; k 2 !g of clopen subsets of X (some of which
S might be
empty) such that the family Un D fU.n; k/ W k 2 !g is disjoint and Un is dense
in X for all n 2 !. Then, under MAC:CH, the set
[ \
H D fInt. fU.n; h.n// W n 2 !g/ W h 2 ! ! g
is dense in X .
T
Proof. Let Eh D fU.n; h.n// W n 2 !g for each h 2 ! ! . Given any U 2 U, let
U W X ! f0; 1g be the characteristic function of U defined by U .U / D f1g and
U .X nU / D f0g (if U D ;, then U 0; if U D X , then U 1). The function
f D fU W U 2 Ug maps X onto a space Y f0; 1gU . Since U is countable,
the space Y is second countable, so jY j ! by Fact 2. Next observe that for every
h 2 ! ! , the set Eh is non-empty if and only if there is a point yh 2 Y such that
Eh D f 1 .yh /. Indeed, given, U.n; k/ 2 U, let yh .U.n; k// D 1 if and only if
k D h.n/; otherwise let yh .U / D 0. It is immediate that yh 2 Y if and only if
Eh ; and f 1 .yh / D Eh .
Now take any set W 2 .X /. Since we have only countably many of non-
empty elements of the family fEh W h 2 ! ! g and jW j > !, there is h 2 ! !
such that Eh \ W is uncountable. Since X is a Luzin space, there is V 2 .X /
such that V Eh \ W . Every U.n; k/ is also closed, so Eh is closed as well. Thus
V Eh \ W E h \W D Eh \W . Since V \W ;, we have W1 D V \W ;,
and hence W1 Eh \W which shows that H \W ;. Since the set W was chosen
arbitrarily, we have established that H is dense in X , so Fact 6 is proved. t
u
Returning to our solution, assume MAC:CH and suppose that there exists a
Luzin space. By . / there exists an everywhere uncountable Luzin space X . Let
T0 D fX g. Suppose that < !1 and we constructed families fT W < g of
clopen subsets of X with the following properties:
S
(5) T is countable, disjoint and T is dense in X for every < ;
(6) T is properly inscribed in T
whenever
< < .
If D 0 C 1, then for each U 2 T0 choose a proper clopen OU U (which is
possible because X is zero-dimensional (Fact 5) and everywhere uncountable (see
. /). Let T D fOU W U 2 T0 g [ fU nOU W U 2 T0 g. It is immediate that (5) and
(6) still hold for the collection fT W g.
If is a limit ordinal, then choose an increasing sequence fn W n 2 !g cofinal
in and choose an enumeration fU.n; k/ W k 2 !g for the family Tn for all n 2 !.
If some Tn is finite, then enumerate its elements as fU.n; k/ W k < mn g for some
mn 2 ! and let U.n; k/ S D ; for
Tall k mn . Now we can apply Fact 6 to conclude
that the open set H D fInt. fU.n; h.n// W n 2 !g/ W h 2 ! ! g is dense in X .
By zero-dimensionality of X (Fact S 5), there exists a disjoint family T .X/ of
clopen subsetsTof X such that T is dense in X and each U 2 T is contained in
the set Eh D fU.n; h.n// W n 2 !g for some h 2 ! ! .
The property (5) is, clearly, fulfilled for the collection fT W g. To see
that (6) also holds, we must prove it for D and
< . Take any U 2 T ;
there is n 2 ! such that
< n . There is h 2 ! ! such that U Eh . Therefore
U U.n; h.n// 2 Tn ; since Tn is properly inscribed in T
, there is W 2 T
such
that U.n; h.n// is properly included in W . Thus U is properly included in W , so (5)
and (6) hold for the collection fT W g. Therefore our inductive construction
can go on to give us the collection fT W < !1 g such that (5) and (6) are fulfilled
for all < !1 . S
Observe that the family T D fT W < !1 g has the following property:
(7) if U; V 2 T and U \ V ;, then either U V or V U and in both cases
the inclusion is proper.
Indeed, if U \ V ;, then U 2 T and V 2 T where . Assume, for
example, that < . Since T is properly inscribed in T , there is W 2 T such that
V W . If U W , then U \ W D ; because T is disjoint and hence U \ V D ;,
a contradiction. Thus U D W and hence V U and the inclusion is proper by (6).
Given U; V 2 T , let U V iff U V . Then is a partial order on T and .T; /

is ccc because c.X / D ! and T .X/. S The family T of non-empty elements of

T is uncountable; for otherwise, T fT WS < g for some < !1 . However,
there are non-empty elements S of T because T is dense in X , and by (6), no
element of T can belong to fT W < g which is a contradiction. It is evident
that .T ; / also has ccc, so, under MAC:CH, we can apply Problem 049 to find
an uncountable S T such that all elements of S are pairwise compatible. By (7),
this means U V or V U for any U; V 2 S .
Take any U 2 S . If U 2 T , then, for any < there is at most one element of
T which contains U . Therefore
(8) for any U 2 S the family PU D fV 2 S W U V g is countable.
Choose any U0 2 S ; suppose that we have fU W < g for some S < !1 such
that U is properly contained in U
if
< < . Since P D fPU W < g
is countable, there is U 2 S nP . By (7), the set U is properly included in U for
each < . As a consequence, we can construct a collection fU W < !1 g such
that < < !1 implies U U and the last inclusion is proper. If V D U0 nU
for all < !1 , then fV W S< !1 g .X /. SinceSX is hereditarily Lindelf, there
is < !1 such that H D fV W < g D G D fV W < !1 g. Since V V
for all < , we have H V D U0 nU . Since UC1 is properly included in U ,
there is x 2 U nUC1 . Thus x 2 U0 nUC1 D VC1 and therefore x 2 G. However,
H U0 nU so x H . This contradiction shows that the existence of our family T
is contradictory, so our solution is complete.
T
T.064. Let Cn !1 be a club for any natural n. Prove that fCn W n 2 !g is a club.
T
Solution. It is evident that C D n2! Cn is closed in !1 ; to see that it is
unbounded, fix any < !1 . Let fnk W k 2 !g be an enumeration of ! in which every
m 2 ! occurs infinitely many times. Since Cn0 is unbounded, there is 0 2 Cn0 such
that < 0 . Suppose that we have 0 ; : : : ; k such that < 0 <

< k < !1
and i 2 Cni for all i k. Since CnkC1 is unbounded, there is kC1 2 CnkC1 with
k < kC1 . This proves that our inductive construction can continue to provide an
increasing sequence fi W i 2 !g !1 such that < 0 and i 2 Cni for all i 2 !.
Let D supfi W i 2 !g. Then the sequence S D fi W i 2 !g converges
to . Furthermore, for every m 2 ! the set Sm D fi W ni D mg is an infinite
subsequence of S and therefore Sm ! as well. Since Sm Cm and Cm is closed,
the ordinal belongs to Cm for all m 2 !. Thus 2 C and > , so C is
uncountable.
T.065. Prove that
(i) every stationary subset of !1 is uncountable;
(ii) not all uncountable subsets of !1 are stationary;
(iii) if A !1 contains a stationary set, then A is stationary;
(iv) S of !1 with a club is stationary;
Any intersection of a stationary subset
(v) if An !1 for each n 2 ! and fAn W n 2 !g is stationary, then An is
stationary for some n 2 !.
Solution. If a set A !1 is countable then A for some < !1 . However, the
set C D !1 n is a club such that A \ C D ;. Hence A is not stationary and (i) is
proved.
Let C be the set of all limit ordinals of !1 . Topologically, C is the set of non-
isolated points of !1 , so C is closed in !1 . Besides, C cannot be countable because
otherwise the infinite (and even uncountable) set !1 n is closed and discrete in
!1 for some < !1 which is a contradiction with countable compactness of !1
(see Problem 314 of [TFS]). Thus C is a club; an immediate consequence is that

A D !1 nC is not stationary. Since the uncountable set f C 1 W < !g is contained
in A, the set A is uncountable and non-stationary so (ii) is proved.
If A A0 and the set A0 is stationary, then, for any club C !1 , we have
A \ C A0 \ C ;, so A is stationary and (iii) is proved.
Assume that A !1 is stationary and C !1 is a club. Given any club D !1 ,
the set C \ D is also a club (see Problem 064) and therefore A \ .C \ D/ D
.A \ C / \ D ;. Thus the set A \ C intersects any club in !1 , so A \ C is
stationary and (iv) is proved.
S To settle (v), assume that the set An is not stationary for all n 2 ! and let A D
fAn W nT2 !g. There is a club Cn !1 such that An \ Cn D ; for all n 2 !. The
set C D fCn W n 2 !g is a club by Problem 064 while A \ C D ; which is a
contradiction. Therefore some An must be stationary, so our solution is complete.
T.066. Let A be a stationary subset of !1 . Prove that there exists a disjoint family of
stationary subsets fA W < !1 g such that A A for each 2 !1 .
Solution. For any countable ordinal > 0 choose an onto map f W ! ! .
If A.; n/ D f 2 A W f .n/ D g for all > 0 and n 2 !, then the family
An D fA.; n/ W > 0g is disjoint for every n 2 !. Indeed, if 0 and
0 0
2 A.; n/ \ SA. ; n/, then f .n/ D D which is a contradiction. It is easy
S check that fA.; n/ W n 2 !g D .!1 n. C 1// \ A for every > 0. Therefore
to
fA.; n/ W n 2 !g is a stationary set for each > 0 (see Problem 065) which
implies by Problem 065 that A.; n / is stationary for some n 2 !. There exists
m 2 ! such that the set P D f > 0 W n D mg is uncountable. Therefore, for
all ordinals 2 P , the set A.; n / D A.; m/ belongs to Am . As a consequence,
the disjoint family Am has uncountably many stationary elements and hence we can
choose an uncountable subfamily A0 Am such that every A 2 A0 is stationary.
Choosing any enumeration fA W < !1 g of the family A0 we obtain the promised
disjoint uncountable family of stationary subsets of A.
T.067 (Fodors lemma; also called pressing-down lemma). Let A !1 be a
stationary subset of !1 . Suppose that f W A ! !1 is a map such that f ./ < for
any 2 A. Prove that there is 0 2 !1 such that the set f 1 .0 / is stationary (and,
in particular, jf 1 .0 /j D !1 ).
Solution. Assume that f 1 ./ is not stationary (maybe empty) for any < !1 .
Then we have a club C !1 suchTthat f 1 ./ \ C D ; for each < !1 .
Consider the set C D f < !1 W 2 < C g. We claim that C is a club.
If 2 C nC , then there is an increasing sequence fn W n 2 !g C which
converges to . Fix any < ; there is m 2 ! such that < m . Since n 2 C ,
we have n 2 C for all n m. The set C is closed in !1 so D lim n 2 C .
The ordinal was chosen arbitrarily, so 2 C for all < , i.e., 2 C . This
contradiction shows that C is a closed subset of !1 . T
To see that C is uncountable, fix any 2 !1 nf0g. The set D0 D fC W < g
is a club by Problem 064, so there is 0 > such that 0 2 D0 . Let 1 D
T
and assume that we have ordinals 1 < 0 <

< n such T that i 2 fC W
< i 1 g for all i 2 f0; : : : ; ng. Since the set DnC1 D fC W < n g is
a club by Problem 064, there is nC1 2 DnC1 such that n < nC1 . Thus our
inductive construction can be carried out T for all n 2 ! to construct an increasing
sequence fn W n 2 !g such that n 2 fC W < n1 g for all n 2 !. Let
D supfn W n 2 !g; given any T
< , there is m 2 ! such that
< m . By the
definition of n , we have n 2 fC W < n1 g C
for all n m C 1. Thus
the sequence S D fn W n m C 1g is contained in C
; the set C
being closed,
we have D lim S 2 C
. It turns out that 2 C
for each
< so 2 C . Since
> , we proved that C is a club.
Since the set A is stationary, there exists 2 C \ A; if D f ./ < , then
2 f 1 ./ \ C and hence 2 C by the definition of the set C . This proves that
2 f 1 ./ \ C while f 1 ./ \ C D ; by the choice of the set C . The obtained
contradiction shows that our assumption is false and hence there is 0 < !1 such
that f 1 .0 / is stationary.
T.068. Given < !1 , let S D ff 2 ! W f is an injectiong. In the set S D
S
<!1 S , consider the following partial order: f g if and only if f g.
Observe that .S; / is a tree with all its chains countable. Prove that there exists
an !1 -sequence fs W < !1 g S such that s 2 SSand s j s for all
< < !1 . Deduce from this fact that the subtree T D <!1 fs 2 S W s s g
of the tree S is an Aronszajn tree. Hence Aronszajn codings and trees exist in ZFC.
Solution. It is evident that is a partial order on S . Given any f 2 ! , observe
that the set Lf D fg 2 S W g < f g coincides with the set ff j W < g.
It is easy to see that the map i W Lf ! defined by i.f j/ D , is an order
isomorphism, so Lf is well-ordered being isomorphic to . This proves that S is
a tree. If ft W < !1 g is a chain in S , then t j D t whenever < < !1 .
Let t./ D tC1 ./ for all < !1 . It is straightforward that t W !1 ! ! and
tj D t for all < !1 . An immediate consequence is that t is an injection which
is a contradiction. Therefore
(*) S is a tree with no uncountable chains.
If f is a function, then dom.f / is its domain. Suppose that we have a set of
functions ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D fj j.dom.fi / \
dom.fj // for any i; j 2 I . Then we can define a function f with dom.f / D
S
i 2I dom.fi / as follows: given any x 2 dom.f /, find any i 2 I with x 2 dom.fi /
and let f .x/ D fi .x/. It is easy to check that the value of f at x does not depend on
the Schoice of i , so we have consistently defined a function f which will be denoted
by ffi W i 2 I g (this makes sense if we identify each function with its graph).
We are going to construct the promised !1 -sequence fs W < !1 g by transfinite
induction. Start off with s0 D ;. Assume that
< !1 and we have a set fs W <
g
with the following properties:
(1) s 2 S and s j s whenever < <
;
(2) !ns .f W < g/ is infinite for every <
.
Let ran.s / D s .f W < g/ for all <

. If
D C 1 for some ordinal ,
then pick any n 2 !nran.s /; the function s
, defined by the conditions s
j D s
and s
./ D n, maps
to !. It is clearly injective and satisfies (1) and (2).
If
is a limit ordinal, fix any increasing sequence f
n W n 2 !g which converges
to
. Let t0 D s
0 ; assume that we have functions ti 2 S
i for all i n such that
(3) ti C1 j
i D ti for all i D 0; : : : ; n 1.
(4) ti s
i for all i D 0; : : : ; n.
0 0
Define an auxiliary function tnC1 2 !
nC1 by the conditions tnC1 j
n D tn and
0 0
tnC1 j.
nC1n
n / D s
nC1 j.
nC1 n
n /. The function tnC1 might fail to be injective.
However, we have s
nC1 j
n s
n tn and hence there is a finite set P
n such
0 0
that s
nC1 j.
n nP / D tn j.
n nP / D tnC1 j.
n nP /. Thus tnC1 j.
nC1nP / is injective
because it coincides with the injective function s
nC1 . Recall that the function
0 0
tnC1 j
n D tn is also injective, so the injectivity of tnC1 can fail only in case when the
0
set Q D tnC1 .P /\s
nC1 .
nC1 n
n / is non-empty. If this is the case, then there exists
a finite set Q0
nC1 n
n such that tnC1 0
.Q0 / D tn .Q/ and hence tnC1 0
j.
nC1 nQ0 /
is an injection which coincides with tn on
n . Since the set !nran.s
nC1 / is infinite,
we can define a function tnC1 as follows: tnC1 j.
nC1 nQ0 / D tnC1 0
j.
nC1 nQ0 / and
0 0
tnC1 jQ W Q ! !nran.s
nC1 / is defined arbitrarily with the only condition that
tnC1 jQ0 be injective. It is immediate that tnC1 j
n D tn and tnC1 s
nC1 , so our
inductive construction can be continued to obtain a sequence fti W i 2 !g with the
properties (3) and (4) true for all i 2 !.
Observe
S that it follows from the fact that each ti is injective that the function
t D fti W i 2 !g is injective as well. Besides,
(5) for each <
, we have tj s
because there is some
i > and hence tj D .tj
i /j s
i j s by (1). It
is possible that (2) does not hold for the function t; to correct this, note that the set
R D t.f
i W i 2 !g/ is infinite because of injectivity of t and hence we can find
an infinite (and faithfully indexed) set R0 D fi W i 2 !g R such that RnR0 is
also infinite. Finally, let W D
nf
i W i 2 !g and define s
by s
jW D tjW and
s
.
i / D i for all i 2 !. The injectivity of s
is evident; the property (2) holds for
D
because !ns
.f W <
g/ contains an infinite set RnR0 . Finally, (1) is true
because for each <
the function s
j can be obtained from tj by changing the
values of tj on the finite set f
i W i 2 !g \ . Therefore s
j tj s by (5).
This finishes our inductive construction and shows that there exists a collection
fs W S < !1 g with the properties (1) and (2). Observe, furthermore, that the set
T D <!1 fs 2 S W s s g is a subtree of S because given any < !1 and any
s 2 T D fs 2 S W s s g we have sj s j s for any < [see (1)]. It is
evident that the th level of T is the set T . All elements of T are obtained from
s by changing its value at some finite set of points. Since there are only countably
many finite subsets of , the set T is countable, and hence T is a tree without
uncountable chains [because even the tree S T does not have uncountable chains
by . /] with all of its levels countable. Therefore T is an Aronszajn tree, so our
T.069. Observe that Jensens axiom implies CH; prove that it is equivalent to any of
the following statements:
(i) for every < !1 , there exists a countable family A exp such that, for
any A !1 , the set f W A \ 2 A g is stationary;
(ii) for any < !1 , there is a set B such that, for any B !1 !1 , the
set f W B \ . / D B g is stationary;
(iii) for any < !1 , there is a function f W ! such that, for any f W !1 ! !1 ,
the set f W f j D f g is stationary.
(iv) for any < !1 , there is a function g W ! !1 such that, for any
g W !1 ! !1 , the set f W gj D g g is stationary.
(v) there exists a set S of cardinality !1 and a set of functions fh W < !1 g
such that, h W ! S for all < !1 and, for any h W !1 ! S , the set
f W hj D h g is stationary;
(vi) for any set T of cardinality !1 , there exists a family fk W < !1 g of functions
such that k W ! T for all < !1 and, for any function k W !1 ! T , the set
f W kj D k g is stationary.
Solution. Let fA W < !1 g be a collection of sets which witnesses the Jensens
axiom. Given any P !, there is 2 !1 n! such that P \ D P \ ! D P D A .
Letting f ./ D P , we obtain a surjection f W W ! exp.!/ for some W !1 .
Therefore c D j exp !j jW j !1 and hence c D !1 if the Jensens axiom holds,
i.e., the Jensens axiom implies CH.
Given any set Q, a set P !1 and a function f W P ! Q, viewing every ordinal
2 P as a set f
W
< g may cause confusion when we take images under f .
So, our agreement will be as follows: the expression f ./ means the element of Q
which is the respective image of the point ; this element can also be considered a
subset of Q if Q !1 , i.e., f ./ could be either a point of Q or the set f
W
<
f ./g if Q !1 . However, if we want the image of the set !1 , then we use
square brackets, i.e., the set f .f
W
< g/ is denoted by f .
continuous map r W !1 ! !1 such that < < !1
Fact 1. Suppose that we have a S
implies r./ r./. If Q D f r./ W < !1 g !1 !1 , then, for any
surjective map s W !1 ! Q, the set C D f < !1 W s D r./g is a club.
Proof. To see that C is closed in !1 , take any increasing sequence fn W n 2 !g
C which converges to an ordinal . We have sn D n r.Sn / for each n 2 !.
Since the mapping
S r is continuous and non-decreasing, we have n2! r.n / D r./
and hence fn r.n / W n 2 !g D r./. Furthermore,
[ [ [
s D s. fn W n 2 !g/ D fsn W n 2 !g D fn r.n / W n 2 !g D r./;
which shows that 2 C , i.e., the set C is closed in !1 .

To show that C is unbounded, fix any < !1 and let 0 D . Since s0 is
countable, there exists 0 2 !1 n0 such that s0 0 r.0 / (the fact that r is
non-decreasing was used here again).
Assume that n 2 ! and we have sequences of ordinals fi W i ng and fi W

i ng with the following properties:
(1) D 0 and i i i C1 for any i D 0; : : : ; n 1.
(2) si i r.i / si C1 for all i D 0; : : : ; n 1.
The set n r.n / is countable and the map s surjective, so there is an ordinal
nC1 2 !1 nn such that n r.n / snC1 . Once more use the fact that r is non-
decreasing to find an ordinal nC1 2 !1 nnC1 such that snC1 nC1 r.nC1 /.
It is evident that (1) and (2) are now fulfilled for all i n and hence we can carry
out our inductive construction to obtain sequences fi W i 2 !g and fi W i 2 !g
with the properties (1) and (2) for all i 2 !. Let
D lim n D lim n . Then
[ [ [
s
D s fn W n 2 !g D fsn W n 2 !g fn r.n / W n 2 !g D
r.
/:
We used (2) to assert the inclusion; as before, the last equality holds because r is
continuous and non-decreasing. On the other hand
!
[ [ [
s
D s nC1 D fsnC1 W n 2 !g fn r.n / W n 2 !g D
r.
/;
n2!
which proves that s

D
r.
/. Thus, for any < !1 we found
with
2 C . Hence C is a club and Fact 1 is proved. t

u
Fact 2. Given any f W !1 ! !1 , the set Cf D f < !1 W f g is a club.
Proof. Assume that fn W n 2 !g Cf is an increasing sequence with D lim n .
Then
!
[ [ [
f D f n D f n n D ;
n2! n2! n2!
which shows that 2 Cf and hence Cf is closed in !1 .

To see that Cf is unbounded, take any < !1 and let 0 D . Suppose that we
have D 0 <

< n such that f i i C1 for all i < n. Since f n is a
countable subset of !1 , there is nC1 > n such that f n nC1 . Thus we can
inductively construct a sequence fi W i 2 !g !1 such that f i i C1 for all
i 2 !. If D lim i , then
!
[ [ [
f D f i D f i i C1 D ;
i 2! i 2! i 2!
which shows that 2 Cf . Thus, for any < !1 there is > with 2 Cf ; hence
Cf is a club, so Fact 2 is proved. t
u
Returning to our solution, assume that } holds; fix a collection fA W < !1 g

which witnesses this and let A D fA g for all < !1 . It is immediate that the
!1 -sequence fA W < !1 g satisfies (i) so } H)(i).
Now assume that (i) is true and fix any bijection s W !1 ! ! !1 . Given any set
A !1 , let A D s.A/; if B ! !1 , then B 0 D s 1 .B/. For each < !1 , let
B D fA W A 2 A g. We claim that
(3) for any B ! !1 , the set PB D f < !1 W B \ .! / 2 B g is stationary.
To prove (3) note first that the set C D f < !1 W s D ! g is a club by
Fact 1 applied to the map s and the map r W !1 ! !1 defined by r./ D ! for all
2 !1 . By (i) the set H0 D f W B 0 \ 2 A g is stationary and therefore so is
the set H D H0 \ C (see Problem 065). If 2 H , then s D ! and hence
B \ .! / D .B 0 \ / 2 B because B 0 \ 2 A . As a consequence, H PB ,
so PB is stationary and (3) is proved.
Let fBk W k 2 !g be an enumeration of B for all < !1 (repetitions are allowed
and there is no loss of generality to assume that B ; for all < !1 ). We will
k
need the sets B;n D f < !1 W .n; / 2 Bk g for all k; n 2 ! and < !1 . Consider
the collection Cn D fB;nn
W < !1 g. We claim that there exists n 2 ! for which the
!1 -sequence Cn witnesses }.
Indeed, if this is not the case, then for any n 2 ! there exists a set Bn !1
S the set f W Bn \ D B;n g is non-stationary. Now, looking at
n
such that the set
B D ffng Bn W n 2 !g ! !1 , we can see that B \ .! / D Bn implies
B \ .fng / D B;n n
and therefore Bn \ D B;n n
. This shows that the set Pn D
f W B \ .! / D B g is non-stationary.S
n
Any countable union of non-stationary
sets is non-stationary by Problem 065, so n2! Pn D f W B \ .! / 2 B g
is non-stationary which contradicts (3) and completes the proof of (i)H) }. Thus
(i) }.
Take any bijection s W !1 ! !1 !1 . If r W !1 ! !1 is defined by r./ D for
all 2 !1 , then Fact 1 can be applied to r and s to conclude that the set C D f 2
!1 W s D g is a club. Assume that a collection fA W < !1 g witnesses }
and let B D s.A / for all 2 C ; if 2 !1 nC , then let B D ;. Given any set
B !1 !1 , the set Q D f W s 1 .B/ \ D A g is stationary and hence so is the
set Q0 D Q \ C . For any 2 Q0 , we have
B D s.A / D s.s 1 .B/ \ / D B \ s./ D B \ . /
which shows that the set f W B \ . / D B g is stationary because it contains

the stationary set Q0 . Thus (ii) holds, i.e., we proved that } H)(ii).
Now assume that (ii) is true and fix an !1 -sequence fB W < !1 g which
witnesses this. Let p W !1 !1 ! !1 be the natural projection to the first factor;
we claim that the !1 -sequence fp.B / W < !1 g witnesses }. It is evident that
A D p.B / for all < !1 . Given any set A !1 , consider the set B D AA.
The set R D f W B \ . / D B g is stationary; for any 2 R we have
A D p.B / D p.B \ . // D A \ and hence the set R0 D f W A \ D A g
is stationary because R R0 . This settles (ii)H) } and hence } (ii).
To see that (iii)H)(iv), let g D f for each < !1 . The implication (iv)H)(v)
becomes evident if we let S D !1 and h D g for all < !1 .
Assume that the property (v) holds and take any set T of cardinality !1 ; fix any
bijection s W S ! T and let k D s h for all < !1 . Given any function
k W !1 ! T , let h D s 1 k; since (v) holds, the set W D f W hj D h g is
stationary. If 2 W and < , then k./ D s.h.// D s.h .// D k ./,
i.e., kj D k , so the set W 0 D f W kj D k g is stationary because W 0 W .
This proves that (v)H)(vi). Since the implication (vi)H)(v) is evident, we have
(v) (vi). It is also immediate that (vi) is stronger than (iv); as a consequence,
(iv) (v) (vi).
Now assume that (iv) holds and fix an !1 -sequence fg W < !1 g that witnesses
it. Let D D f W g g. For any 2 D, let f D g ; if 2 !1 nD, then let
f ./ D 0 for all < . We claim that the set ff W < !1 g witnesses (iii). To
see this, take any function f W !1 ! !1 and apply Fact 2 to conclude that the set
C D f W f g is a club.
The set E D f < !1 W f j D g g is stationary by our assumption, so the
set C 0 D E \ C is also stationary. If 2 C 0 , then f j D g D f , so the set
U D f < !1 W f j D f g is stationary because C 0 U . This shows that (iii)
holds and proves (iv)H)(iii). Consequently, (iii) (iv) (v) (vi).
We prove next that (ii)H)(iii). If f W X ! Y is a function, then the set f D
f.x; f .x// W x 2 X g X Y is its graph. It is easy to see that two functions
f; g W X ! Y coincide if and only if f D g . Fix an !1 -sequence fB W < !1 g
as in (ii) and let D D f < !1 W there is a function p W ! such that B D p g.
For each 2 D, let f D p ; if 2 !1 nD, then let f ./ D 0 for all < .
The !1 -sequence ff W < !1 g witnesses (iii). Indeed, if f W !1 ! !1 , then
f !1 !1 , and therefore the set E D f W f \ . / D B g is stationary.
Given any 2 E, the set f \ . / is a graph of some function on and hence
B D f \ . / is also a graph of some function on . Therefore, by our choice
of D, we have E D and hence f \ . / D p D f for each 2 E. It is
an easy exercise to check that f \ . / D f is equivalent to f j D f , so
the set E 0 D f W f j D f g is stationary because E E 0 . Thus (iii) holds, so we
proved the implication (ii)H)(iii).
Finally, assume that the set ff W < !1 g witnesses (iii) and let W D f
for all < !1 . The collection W D fW W < !1 g is almost what we need to
prove }. To see it, take any non-empty set W !1 and fix some
2 W . Define a
function f W !1 ! !1 by f ./ D if 2 W and f ./ D
for all 2 !1 nW .
Since (iii) holds, the set B D f W f j D f g is stationary and hence B 0 D Bn
is
also stationary. If 2 B 0 , then f ./ D for any 2 W \ and f ./ D
for
any 2 nW . Consequently, W D f D .W \ / [ f
g D W \ because
2 W \ . Hence, for stationary-many s, we have W \ D W , i.e.,

(4) for any non-empty W !1 , the set f W W \ D W g is stationary.
Therefore the only set not captured by the collection W is the empty set because
W ; for all ordinals < !1 . To correct this situation, observe that the set
E D f < !1 W W D f0gg is stationary [just let W D f0g and apply (4)]. Apply
Problem 066 to find disjoint stationary sets E0 ; E1 E such that E D E0 [ E1 .

Now let A D W for all 2 !1 nE0 and A D ; for all 2 E0 . Then the empty
set is captured by the collection A D fA W < !1 g because if A D ;, then the
set f W A \ D A g is stationary being at least as large as E0 . To see that we still
capture all non-empty sets by the family A, take any non-empty A !1 . The set
D D f > 0 W A \ D W g is stationary. If A f0g, then there is 2 Anf0g and
hence the set f W A \ D A g is stationary because it contains the stationary set
Dn. If, on the other hand, A D f0g, then the set f W A \ D A g is stationary
because it contains the stationary set E1 . Thus the set f W A\ D A g is stationary
for all sets A !1 . Therefore (iii)H) } and this was the last implication we needed
to show that } (i) (ii) (iii) (iv) (v) (vi), so our
T.070. Prove that if Jensens axiom holds, then there is a Souslin tree.
S
Solution. All through this solution we assume }. Let T D f! W < !1 g;
denote by T the set ! for all < !1 . Given f; g 2 T , let f g iff f g.
It is an easy exercise that is a partial order on T . Any subset of T will be also
considered a partially ordered set with the order induced from T . For any g 2 T ,
we denote by ht.g/ the unique ordinal such that g 2 T . Observe that the set
Lg D ff 2 T W f < gg can also be represented as ff j W < D ht.g/g. It is
immediate that the correspondence 7! f j is an isomorphism of onto Lg , so
Lg is well-ordered for any g 2 T , and hence T is a tree. If f 2 T; ht.f / D and
n 2 !, then g D f _ n is the function defined by gj D f and g./ D n. It is clear
that ht.g/ D C 1.
S
theS choice of i , so we have consistently defined a function f which will be denoted
Given any set Q, a set P !1 and a function f W P ! Q, viewing every ordinal
2 P as a set f
W
< g may cause confusion when we take images under f .
So, our agreement will be as follows: the expression f ./ means the element of Q
which is the respective image of the point ; this element can also be considered a
subset of Q if Q !1 , i.e., f ./ could be either a point of Q or the set f
W
<
f ./g if Q !1 . However, if we want the image of the set !1 , then we use
square brackets, i.e., the set f .f
W
< g/ is denoted by f .
Observe that jT j D j! j D c D !1 because we have CH under the Jensens
axiom (see Problem 069). Thus there exists a collection ff W < !1 g such that
f W ! T for each < !1 and, given any function f W !1 ! T , the set
f W f j D f g is stationary (see Problem 069).
We are going to construct a subtree S of the tree T by induction on < !1 .
This construction will give us a set S T , and the resulting tree will be the set
S
S D <!1 S . To start off, let S0 D T0 D ;. Suppose that < !1 and we have
constructed a collection fS W < g with the following properties:
(1) S T and jS j ! for all < ;
(2) if s 2 S and C 1 < , then s _ n 2 SC1 for all n 2 !;
(3) if <
< and s 2 S , then there is t 2 S
such that s < t;
(4) if <
< and t 2 S
, then tj 2 S ;
(5) if
< Sis a limit ordinal and the set A D f

is a maximal antichain in
S.
/ D <
S , then, for any t 2 S
there is s 2 A such that s < t.
If D
C 1 is a successor ordinal, then let S D fs _ n W s 2 S
and n 2 !g.
The properties (1), (2), (4) and (5) clearly hold for all . To see that (3) also
holds, take any s 2 S for some < . Since (3) is true for all < , there is
u 2 S
such that s < u. Therefore t D u_ 0 2 S and s < t. Thus all properties
(1)(5) hold for all .
If is a limit ordinal, choose an increasing sequence fn W n 2 !g with
lim n D ; we must consider two cases:
(1) The set f is not a maximal antichain in S./. For each s 2 S./, there
is m 2 ! such that ht.s/ < m and therefore there is s0 2 Sm with s < s0
[see (3)]. Using the property (3), we can construct, by an evident induction, a
sequence fsn W n 2 !g SS./ such that sn 2 SmCn and sn < snC1 for all
n 2 !. Letting w.s/ D n2! sn we obtain a function w.s/ 2 T such that
s < w.s/. If S D fw.s/ W s 2 S./g, then the conditions (1)(5) are satisfied
for all .
(2) If A D f is a maximal antichain in S./, then take any s 2 S./ and
observe that there is ts 2 A such that s ts or ts s because otherwise s A
and A [ fsg is an antichain in S./ which contradicts the maximality of A.
There exists m 2 ! such that ht.s [ ts / m ; apply (3) to find s0 2 Sm
such that s [ ts s0 . As before, we can construct by induction a sequence
fsn W n 2 !g SS./ such that sn 2 SmCn and sn < snC1 for all n 2 !.
Letting w.s/ D n2! sn we obtain a function w.s/ 2 T such that s < w.s/
and ts < w.s/. If S D fw.s/ W s 2 S./g, then the conditions (1)(5) are
satisfied for all . Indeed, (1)(4) are evident by our construction. As to
(5) it must only be checked for the set A D f S./. Since every t 2 S
is w.u/ for some u 2 S./, we have tu < w.s/ D t, so, if s D tu , then s 2 A
and s < t which shows that (5) is also fulfilled for all .
Therefore
S our inductive construction can be continued until we obtain a set
S D fS W < !1 g. It follows from (4) that Ls S for any s 2 S and
hence S is a tree. It is also immediate from our construction that all levels of S are
countable, so S is an !1 -tree. All levels of S are non-empty by (3), so jS j D !1 .
Let us show that all antichains of S are countable.
Suppose that U S is an uncountable antichain. It is an easy exercise that every
antichain is contained in a maximal antichain, so we can assume, without loss of
generality, that U is a maximal antichain in S . Take any bijection f W !1 ! U .
Next observe that the set
(6) C D f < !1 W f S./ and f is a maximal antichain in S./g is

a club.
To see this, assume that fn W n 2 !g !1 is anSincreasing sequence
S such that
n 2 C for all n 2 !. If D lim n , then f D n2! f n n2! S.n / D
S./ so f S./.
Now, if f is not a maximal antichain in S./, then there is s 2 S./nf
such that fsg [ f is still an antichain. However, s 2 S.n / for some n 2 !,
and hence fsg [ f n is an antichain in S.n / strictly larger than f n which
contradicts the maximality of f n in S.n /. Therefore C is a closed subset of !1 .
To see that C is unbounded, fix any < !1 and let 0 D . There is 0 < !1
such that 0 < 0 and f 0 S.0 /. Assume that we have increasing sequences
fi W i ng and fi W i ng such that
(7) i < i < i C1 for all i n 1;
(8) if i n 1 and s 2 S.i /, then some ts 2 f i C1 is compatible with s;
(9) f i S.i / for all i n.
Since the set U is a maximal antichain, for every function s 2 S.n / there is ts 2
U which is compatible with s. Choose any ordinal nC1 > n for which f nC1
fts W s 2 S.n /g. Since the set f nC1 is countable, there is nC1 > nC1 such
that f nC1 S.nC1 /. It is clear that (7)(9) now hold for fi W i n C 1g
and fi W i n C 1g, and hence our inductive construction can go ahead giving us
increasing sequences fi W i 2 !g !1 and fi W i 2 !g !1 such that (7)(9)
take place for all i 2 !. We claim that D lim n D lim n 2 C , i.e., f S./
and the set f is a maximal antichain in S./.
For the first assertion observe that
!
[ [ [
f D f n D f n S.n / D S./
n2! n2! n2!
[the inclusion takes place by (9)] and hence f S./. For the second statement
assume that s 2 S./nf is incompatible with all elements of f . Choose n 2
! such that s 2 S.n /. By (7) there is ts 2 f nC1 f which is compatible
with s. It is clear that ts 2 f and hence s is compatible with an element of f .
This contradiction shows that f is a maximal antichain in S./ and hence, for
any < !1 , there is > with 2 C . Therefore C is a club, so we finished the
proof of (6).
Recalling that the collection ff W < !1 g witnesses the Jensens axiom, we
convince ourselves that the set E D f W f j D f g must be stationary and
therefore so is the set H D E \ C . Take any 2 H ; by definition of C , the
set A D f D f is a maximal antichain in S./.
Since the set A is countable, we can pick s 2 U nA such that ht.s/ . Then
t D sj 2 S by (4); apply (5) to conclude that there is u 2 f with u < t. We
have u < t < s for distinct u; s 2 U which is a contradiction. Thus S has ccc.
To finish our proof that S is a Souslin tree, assume that A is an uncountable chain
in S . Passing to an appropriate uncountable subset of A if necessary, we can assume
that AS D fs W < !1 g where s < s whenever < . It is immediate that
s D <!1 s is well-defined and hence there exists a function s W !1 ! ! such
that sj 2 S for all < !1 .
Let t D .sj/_ n where n D s./C1 for all < !1 . It is clear that t 2 SC1
and t ? sj. C 1/ for all < !1 ; we claim that B D ft W < !1 g is an
antichain. Assume that < ; the only possibility for t to be compatible with t is
to satisfy the condition t < t because C 1 D ht.t / < ht.t / D C 1. However,
sj. C 1/ sj < t and therefore sj. C 1/ < t . It turns out that the incompatible
elements t and sj. C 1/ are less than t while Lt must be well-ordered because
S is a tree. This contradiction shows that B is an uncountable antichain in S which
in turn contradicts the ccc property of S we established above. Hence S is a Souslin
tree and our solution is complete.
T.071. Prove that MAC:CH implies there are no Souslin trees.
Solution. Assume that MAC:CH holds and .S; / is a Souslin tree. Since S is
uncountable and has ccc, we can apply Problem 049 to conclude that there is an
uncountable set C S such that every s; t 2 C are compatible. However, in any
tree any compatible elements are comparable. Thus s t or t s for any s; t 2 C ,
i.e., C is an uncountable chain in S which is a contradiction with the fact that S is
a Souslin tree.
T.072. Given two topologies and on the same set X , say that is weaker than
if . If and , then is said to be strictly weaker than . If .X; /
is a linearly ordered set and Y X , then Y is the order considered only on the
points of Y . Let .L; / be any linearly ordered space:
(i) Prove that for any M L, the topology .M / on M generated by the order
M is weaker than the topology M L
of the subspace on M induced from L.
L
(ii) Show that .M / can be strictly weaker than M even if M is a dense or a
clopen subspace of L.
(iii) Assume that M is order dense in L, i.e., for any a; b 2 L, if a < b, then a
p < q b for some p; q 2 M . Prove that M is dense in L and .M / D M L
.
(iv) Prove that there exists a compact linearly ordered space .K; / such that, for
some order dense N K, there exists an order isomorphism between .N; N /
and .L; /. In particular, .N / D NK , the space N is dense in K and .L; /
is order isomorphic to .N; N /.
Solution. Given any points a; b 2 L, we let .a; b/ D fs 2 L W a < s < bg. Besides,
.a; !/ D fs 2 L W a < sg and . ; a/ D fs 2 L W s < ag. If B D f.a; b/ W a; b 2 L
and a < bg [ f.a; !/ W a 2 Lg [ f. ; a/ W a 2 Lg, then B is a base of the
space L.
(i) For any p; q 2 M such that p < q, we let .p; q/M D .p; q/\M . Analogously,
.p; !/M D .p; !/ \ M and . ; p/ D . ; p/ \ M for any p 2 M .
Observe that .p; q/M D fx 2 M W p <M x <M qg, i.e., .p; q/M is the
interval in .M; M / determined by the order M and points p; q. Analogously,

.p; !/M D fx 2 M W p <M xg and . ; p/M D fx 2 M W x <M pg and
therefore the family
BM D f.p; q/M W p; q 2 M and p < qg[f.p; !/M W p 2 M g[f. ; p/M W p 2 M g
is a base of the space .M; .M //. By our definition, for any U 2 BM there is
V 2 B such that U D V \ M . Therefore every element of BM belongs to the
topology M L
. Since BM is a base of .M /, every W 2 .M / is a union of
a subfamily of BM . Since M L
is a topology, every union of a subfamily of BM
also belongs to M . Consequently, .M / M
L L
and hence we proved (i).
(ii) Let L D f n W n 2 Ng [ f n W n 2 Ng; the order on L is induced from the order
1 1
in R. The set M D f n1 W n 2 Ng [ f1g is clopen in L (prove it please!) and

the point s D 1 is isolated in M if M has the subspace topology induced from
L. Indeed, fsg D . 12 ; !/, so s is even isolated in L and hence in M . However,
s is not isolated in .M; .M //. To see this, observe that the base at s in M is
given by the sets .t; !/M where t runs over M nfsg. Now, if t 2 M nfsg, then
t D n1 for some n 2 N and therefore .t; !/M f m1 W m > ng is an infinite
set. Thus every neighborhood of s in .M; .M // is infinite, so s is not isolated
in .M; .M //. Hence it is possible that .M / M L
for a clopen M .
To show that the same is possible for a dense set M L, consider the space
L D f n1 W n 2 Ng [ f0g [ f1g with the order induced from R. It is evident
that 0 2 f n1 W n 2 Ng, so the set M D Lnf0g is dense in L. As before, the
point s D 1 is isolated in .M; M L
// because s is isolated in L. However, s is
not isolated in M : we proved it in the previous case considering the same M .
This shows that it is possible that .M / M L
for a dense M L, so (ii) is
established.
(iii) Assume that M is order dense in L. To see that M is dense in L it suffices to
show that M \ B ; for any non-empty B 2 B. We have three possibilities
for the set B:
(1) If B D . ; a/ for some a 2 L, pick any s 2 B; then s < a, so there
are p; q 2 M such that s p < q a. Therefore p 2 . ; a/ \ M , so
M \ B ; in this case.
(2) If B D .a; !/ for some a 2 L, then again take any s 2 B and observe
that a < s implies that there are p; q 2 M with a p < q s and hence
q 2 .a; !/ \ M which shows that M \ B ; in this case as well.
(3) If B D .a; b/ for some a; b 2 L such that a < b, pick any s 2 .a; b/.
Again, there are p; q 2 M with a p < q s which implies q 2
.a; b/ \ M and therefore M \ B ;, so M is dense in L.
L
As to M D .M /, it suffices to show that M
L
.M /. Take any U 2 M
L
and V 2 .L/ such that U D V \ M . Given any s 2 U there exists an interval

I 2 B such that s 2 I V . We must consider the same three cases:
(1) I D .a; b/ for some a; b 2 L with a < s < b. By order density of M ,

there are p; q; r; t 2 M such that a p < q s and s r < t b.
Therefore s 2 .p; t/M U .
(2) I D .a; !/ for some a 2 L with a < s. By order density of M , there are
p; q 2 M such that a p < q s. Consequently, s 2 .p; !/M U .
(3) I D . ; a/ for some a 2 L with s < a. By order density of M , there are
p; q 2 M such that s p < q a. Consequently, s 2 . ; q/M U .
Thus we proved that for any x 2 U , there is I 2 BM such that s 2 I U .
Therefore BM is a base for ML L
, so M .M / and (iii) is proved.
(iv) Given a subset x L, call x a gap ray if it satisfies the following conditions:
(1) x has no maximal element, i.e., there is no s 2 x such that t s for all
t 2 x;
(2) s 2 x and t s implies t 2 x;
(3) there is no minimal element in Lnx.
Call a set x L a point ray if x D Ls D ft 2 L W t < sg for some s 2 L.
A set x L is a left ray if it satisfies (2). It is evident that any point ray is a
left ray.
Let K D fx L W x is a gap ray or a point rayg; note that we have
(4) any left ray x without a maximal element is in K,
because if there is a minimal element s in the set .Lnx/, then x D Ls . If
x; y 2 K, then let x y if x y. It is evident that is a partial order on K.
We will also need the following property of left rays:
(5) If x L is a left ray and s 2 L, then x Ls if and only if s x.
Indeed, if t 2 x and t s, then s 2 x, so s x implies t < s for all t 2 x, i.e.,
x Ls . If, on the other hand, x Ls , then t < s for any t 2 x and hence s x.
This proves (5).
It is immediate from the definition that if x L is a left ray and s 2 x, then
Ls x. On the other hand, if Ls x and x Ls , then s 2 x; for otherwise,
x Ls by (5) and hence x D Ls . Thus we proved that
(6) for any left ray x L if s 2 x, then Ls x; besides if Ls x, then Ls x
if and only if s 2 x.
To establish that is also a linear order, take any x; y 2 K. If x D y, then
there is nothing to prove. If x y, then xny ; or ynx ;. If xny ;,
then take any s 2 xny and apply (5) and (6) to conclude that y Ls x and
hence y x. Analogously, if ynx ;, then x y which proves that is a linear
order on K. If a; b 2 K, then .a; b/K D fx 2 K W a x bg; analogously,
.a; !/K D fx 2 K W a xg and . ; b/K D fx 2 K W x bg. The family
B.K/ D f.a; b/K W a; b 2 K and a bg[f.a; !/K W a 2 Kg[f. ; b/K W b 2 Kg
is a base in the space K.

Let us show that .K; / is compact. TakeS any closed non-empty F K; observe
that any x 2 F is a subset of L so z D fx W x 2 F g is also a subset of L. It is
immediate that z is a left ray. We prove first that z 2 K. If z has no maximal element
then z 2 K by (4). If we have a maximal element t 2 z then t 2 x for some x 2 F
and hence z Lt [ ftg x, i.e., z x and therefore z D x because z is a union of
all elements of F which implies x z. Thus z D x 2 F so z 2 K.
To prove that z 2 F assume that this is not true; since F is closed, there is
B 2 B.K/ such that z 2 B KnF . We have three possibilities for the set B.
Assume that B D .a; b/K for some a; b 2 K with a z b; since no element
of F can be greater than or equal to z, we have z 2 .a; !/K KnF . The case
B D . ; a/K is impossible because F . ; a/K for any a z and hence
. ; a/K \ F ;. Thus, in all possible cases, there is a 2 K such that a z and
.a; !/K \ F D ;. Since z a, there is s 2 zna; it follows the definition of z that
there is x 2 F such that s 2 x and therefore s 2 xna. It follows from (5) and (6)
that a Ls x, so a x. Besides, s 2 xna implies a x and therefore a x
which shows that x 2 .a; !/K \ F D ;, a contradiction. Consequently, z 2 F is
the maximal element of F . T
To show that F also has a minimal element, let w D fx W x 2 F g. It is an easy
exercise that any intersection of left rays is a left ray. It is also evident that w x
for any x 2 F , so if w 2 F , then w is the minimal element of F . However, it is not
even clear whether w 2 K.
Assume for a moment that w 2 K. If w F , then there is B 2 B.K/ such
that w 2 B KnF . There are three possibilities for the set B. Note first that B
cannot be .a; !/K for any a 2 K because w 2 B implies a w; since w x for
any x 2 F , we have a x for any x 2 F , i.e., F .a; !/K which contradicts
the equality .a; !/K \ F D ;. Now, if B D .a; b/K for some a; b 2 K, then
.w; b/K \ F D ;; if B D . ; b/K , then since there is no x 2 F with x w, we
also have .w; b/K \ F D ;.
To sum up, if w 2 KnF , then there is b 2 K such that .w; b/K \ F D ;. Since
w b, there is t 2 bnw; by definition of w, there is x 2 F such that t x. Apply
(5) and (6) to conclude that x Lt b. Since t 2 b, we have Lt b and therefore
x Lt b, so x 2 .w; b/K \ F which is a contradiction.
Thus we proved that
(7) if w 2 K, then w 2 F .
Now assume that w K. By (4), the ray w has a maximal element s. The point
u D Ls 2 K does not belong to F because s u. Observe also that u x for
all x 2 F and hence u x for each x 2 F . The set F being closed in K there
is B 2 B.K/ such that u 2 B KnF . Again, we have three possibilities for B.
If B D .a; b/K , then .u; b/K \ F D ;. If B D .a; !/K , then a u x for every
x 2 F which shows that F B, a contradiction with B \ F D ;. If B D . ; b/K ,
then it follows from . ; u/K \ F D ; that .u; b/K \ F D ;. As a result we showed
that w K implies .u; b/K \ F D ; for some b 2 K with u b.
Now u b implies u b and u b. Since wnu D fsg and w K cannot
coincide with b 2 K, there exists t 2 b \ .Lnw/. By definition of w, there is
x 2 F such that t x. Apply (5) and (6) to conclude that x Lt b. Since

t 2 b, we have Lt b and therefore x Lt b, so x 2 .u; b/K \ F which
is a contradiction. This contradiction shows that w 2 K and hence w 2 F by (7).
We have finally proved that F has both the smallest and the largest elements. The
closed non-empty set F K was chosen arbitrarily, so K is compact by Problem
305 of [TFS].
Given any s 2 L, let '.s/ D Ls for all s 2 L; this defines a map ' W L ! K.
If s; t 2 L and s < t, then Ls Lt , i.e., '.s/ '.t/. Besides, s 2 Lt nLs , so
'.s/ '.t/ and hence '.s/ '.t/. This shows that ' W L ! '.L/ is an order
isomorphism and hence L is order isomorphic to the set N D '.L/ K. Observe
first that N is precisely the set of point rays in K. Let us prove that N is order dense
in K. Take any a; b 2 K with a b. If both a and b are point rays, then we can
take p D a and q D b to obtain p; q 2 N such that a p q b.
Now if b D Ls for some s 2 L, then a Ls and it follows from a Ls that
there is some t 2 Ls na. Apply (5) to conclude that a Lt and hence a Lt . Now,
if p D Lt and q D Ls D b, then a p < q b, so the order density of N is
established in this case as well.
Finally, assume that b is a gap ray. Since a b, we can choose some s 2 bna.
No gap ray has a maximal element, so there is t 2 b with s < t. Now, if p D Ls and
q D Lt , then a p q b (we used (5) and (6) again) which finishes the proof
of order density of N in K. Now apply (iii) to conclude that N is dense in K and
NK coincides with the topology generated by the order N . We have now proved
everything promised in (iv), so our solution is complete.
T.073. Prove that, under }, there exists a linearly ordered hereditarily Lindelf
non-separable compact space.
Solution. We will first prove some facts about linearly ordered spaces. Let .L; /
be a linearly ordered space. Given any points a; b 2 L, such that a < b, we let
.a; b/ D fs 2 L W a < s < bg. If a > b, it is convenient to consider that .a; b/ D
.b; a/. Of course, .a; b/ D ; if a D b. Furthermore, a; b D .a; b/ [ fa; bg for any
a; b 2 L; besides, .a; !/ D fs 2 L W a < sg and . ; a/ D fs 2 L W s < ag. By
definition, the family BL D f.a; b/ W a; b 2 Lg [ f.a; !/ W a 2 Lg [ f. ; a/ W
a 2 Lg is a base of the space L. The elements of BL are called open intervals in L.
We consider the whole space L to be an open interval as well.
Fact 1. Let .K; / be a compact linearly ordered space. Then every non-empty
open U K is a disjoint union of open intervals.
Proof. Since K is compact, there exist m D min K and M D max K (see Problem
305 of [TFS]). Say that a set P K is convex if, for any x; y 2 P , we have
x; y P . Note that every open interval is a convex set. Besides,
T S
(1) if a family C consists of convex sets and C ;, then C is also a
convex set.
S
Indeed, take Tany x; y 2 C D C. There are A; B 2 C such that x 2 A and
y 2 B. Since C ;, we can take some z 2 A \ B. Then x; z A and
z; y B, so x; y x; z [ z; y A [ B C which shows that x; y C
for any x; y 2 C , i.e., C is a convex set.
Given x; y 2 U , say that x y if x; y U . It is immediate that is an
equivalence relation on U . Let Ex D fy 2 U W y xg be the equivalence class
of x for all x 2 U . The set Ex is convex because y; z 2 Ex implies y; z
y; x [ x; z Ex . Besides, Ex is open in K; indeed, if y 2 Ex , then there is
B 2 BK such that y 2 B U . Since B is a convex set, we have B Ey ; since
Ex D Ey , we have B Ex , and hence every y 2 Ex is contained in Ex together
with an open set, i.e., Ex is open for every x 2 U . Since K is compact, there exist
Mx D max E x and mx D min E x (see Problem 305 of [TFS]). If mx < z < Mx ,
then the set B D .z; !/ is open and contains Mx 2 E x , so there is b 2 Ex with
z < b. Analogously, there is a 2 Ex with a < z. Therefore z 2 a; b Ex and
hence z 2 Ex . This proves that .mx ; Mx / Ex . If mx Ex and Mx Ex , then
Ex D .mx ; Mx / is an open interval.
Assume now that Mx Ex and mx 2 Ex . If mx D m, then Ex D . ; Mx /
is an open interval. If m < mx , then the set m; mx / is closed in K being the
intersection of a closed set m; mx with the closed set KnEx . If nx D maxm; mx /
(we can take the respective maximum by Problem 305 of [TFS]), then Ex D
.nx ; Mx / is an open interval. This proves that Ex is an open interval if mx 2 Ex
and Mx Ex . Analogously, Ex is an open interval if mx Ex and Mx 2 Ex .
Finally, assume that Ex D mx ; Mx . We still have some subcases here.
(1) if mx D m and Mx D M , then Ex D K is an open interval;
(2) if mx D m and Mx < M , then the set .Mx ; M has a minimal element Nx and
hence Ex D . ; Nx / is an open interval;
(3) if mx > m and Mx D M , then the set m; mx / has a maximal element nx , so
Ex D .nx ; !/ is an open interval;
(4) if mx > m and Mx < M , then the set .Mx ; M has a minimal element Nx , so
Ex D .nx ; Nx / is an open interval.
This proves that Ex is an open interval for any x 2 U . It is clear that U is a
disjoint union of the sets Ex , so Fact 1 is proved. u
t
Fact 2. Let .L; / be a linearly ordered space. If c.L/ !, then L is hereditarily
Lindelf.
Proof. There exists a compact linearly ordered space .K; / such that L is order
isomorphic to a dense N K such that the topology of the subspace on N coincides
with the topology generated by the restriction of the order to N (see Problem
072). Since N is homeomorphic to L, we have ! D c.L/ D c.N / D c.K/, so it
suffices to prove our Fact for K because if K is hereditarily Lindelf, then so is N
and hence L.
Observe that s.K/ D c.K/ D ! (see Fact 2 of S.304). Note also that t.K/ D !
because any free sequence is a discrete subset of K, so all free sequences in K are
countable (see Problem 328 of [TFS]). Therefore .K/ D t.K/ D ! (see Problem
303 of [TFS]). For each x 2 K fix a local base fUnx W n 2 !g BK at the point
x
x such that UnC1 Unx for all n 2 !. Take any points a; b 2 K such that a < b
and observe that the set Fn D .a; b/n.Una [ Unb / D a; b n.Una [ Unb / is closed for
each n 2 !. Given any x S 2 .a; b/, there is n 2 ! such that x Unb [ Una and hence
x 2 Fn . This shows that n2! Fn D .a; b/, so we proved that every open interval
.a; b/ is an F -set in K. S
Now if Gn D . ; b/nUnb , then the set Gn is closed in K and n2! Gn D . ; b/
and therefore . ; b/ is an F -set for any b 2 K. Analogously, .a; !/ is an F -set
for any a 2 K so we proved that
(2) every open interval is an F -set in K.
Now take any non-empty
S open U K; there is a family U of disjoint open
intervals such that U D U (see Fact 1). Since c.K/ D !, the family U is
countable. Any countable union of F -sets is an F -set, so U is an F -set by (2).
It turns out that every open subset of K is an F -set, so K is perfectly normal.
Applying Problem 001 we conclude that K is hereditarily Lindelf, so Fact 2 is
proved. t
u
We will need to introduce some notation for working with trees. Given a tree
.S; / and any s 2 S , the set L.s; S / D ft 2 S W t < sg is well-ordered and hence
isomorphic to an ordinal ; let ht.s; S / D . Let S./ D fs 2 S W ht.s; S / D g;
call S./ the -th level of S . If s 2 S , then H.s; S / D ft 2 S W s tg. Another
important observation is that given any s 2 S with ht.s; S / D , for any <
there is a unique t D s./ 2 S./ with t s.
Fact 3. Assume that there exists a Souslin tree. Then there exists a Souslin tree S
(3) S.0/ is a singleton;
(4) given any < < !1 and any s 2 S./, there is t 2 S./ such that s < t;
(5) for any < !1 any s 2 S./ and any n 2 N, there is > such that
jft 2 S./ W s < tgj n.
Proof. Take any Souslin tree T and consider the set U D ft 2 T W jH.t; T /j D !1 g.
Observe that s t implies H.s; T / H.t; T / and hence jH.t; T /j D !1 implies
jH.s; T /j D !1 for any s < t which shows that L.t; T / U for any t 2 U and
therefore U is a subtree of T . Of course, we havent even proved that U is non-
empty. We will show more, namely, that jU j D !1 and hence U is also a Souslin
tree and, besides, jH.s; U /j D !1 for any s 2 U . S
To prove that jU j D !1 , note
S that for each < !1 , the set T0 D T n. < T .//
is uncountable and T0 D fH.s; T / W s 2 T ./g. Therefore H.s; T / is
uncountable for some s 2 T ./ and hence s 2 U . It turns out that T ./ \ U ;
for each < !1 so jU j D !1 . Furthermore, if s 2 U , then for each S> ht.s; T /,
the uncountable set H.s; T / \ T0 is covered by the countable union fH.t; T / W
t 2 T ./g which shows that H.t; T / \ H.s; T / \ T0 is uncountable for some
t 2 T ./ and hence t 2 U . Besides, H.t; T / \ H.s; T / ; implies that

t 2 H.s; T / \ U D H.s; U /. Thus, for any > ht.s; T /, we found an element
of H.s; U / in T ./ which shows that jH.s; U /j D !1 .
Now, pick an arbitrary a 2 U and let S D H.a; U /; we claim that S is a tree with
the required properties. It is clear that S.0/ D fag so (3) is fulfilled. To see that (4) is
true, fix any < !1 , any > and any s 2 S./. The set H.s; S / D H.s; U / is
uncountable and hence there is u 2 H.s; S / such that ht.u; S / > . Since L.u; S /
is isomorphic to some
> , there is t < u such that t 2 S./. We have s < u
and t < u; since L.u; S / is well-ordered, the elements t and s are comparable and
hence s < t. This settles (4).
To prove the property (5) by induction on n, we start off with n D 2; since
the uncountable set H.s; S / cannot be a chain, there exist a; b 2 H.s; S / with
H.a; S / \ H.b; S / D ;. If ht.a; S / D ht.b; S /, then we are done. If not, then
assume that
D ht.a; S / < ht.b; S /. Then b 0 D b.
/ a and a; b 0 2 S.
/ \
H.s; S / which proves (5) for n D 2.
Now, if the property (5) is true for n D k, then fix any
> and distinct
t1 ; : : : ; tk 2 S.
/ \ H.s; S /. Applying (5) for n D 2 to tk , we can find >
and distinct a; b 2 S./ \ H.tk ; S /. The property (4) guarantees existence of

s1 ; : : : ; sk1 2 S./ such that si .
/ D ti for all i k 1. It is clear that
s1 ; : : : ; sk1 ; a; b 2 H.s; S / are distinct and belong to S./, so (5) holds for
n D k C 1. Therefore (5) holds for any n 2 and Fact 3 is proved. t
u
Fact 4. If there is a Souslin tree, then there exists a linearly ordered hereditarily
Lindelf non-separable compact space.
Proof. Apply Fact 3 to fix a Souslin tree S with the properties (3)(5) and fix any
well-order on S . Denote by L the set of all maximal chains of S (with respect
to inclusion). Given any x 2 L let .x/ D supfht.s; S / W s 2 xg. Observe that for
any x 2 L and any s 2 x with ht.s; S / D , there is t 2 S. C 1/ \ H.s; S /
by (4). Therefore the chain x cannot have a maximal element, and hence .x/ is
a limit ordinal for any x 2 L. Besides, if x 2 L and s 2 x, then L.s; S / x,
so the set x \ S./ is a singleton for any < .x/. Let x be the element of
S./ such that x \ S./ D fx g. If we have distinct points x; y 2 L, then let
d.x; y/ D minf W x y g.
We define an order on L as follows: if x; y 2 L are Distinct, then x < y if
and only if xd.x; y/ yd.x; y/ . It is evident that any two elements of L are
comparable. It is immediate that x < y and y < x cannot hold at the same time,
so x y and y x implies x D y. To check the transitivity, assume that x y
and y z for some x; y; z 2 L. The proof is trivial if there is some equality among
x; y; z, so assume that x < y and y < z; let D d.x; y/ and D d.y; z/:
(1) If < , then y z while x D y and hence x z . Since
x
D y
D z
for all
< , we have d.x; z/ D which shows that
x < z.
(2) If D , then x y z and therefore x z by transitivity of

. Since x
D y
D z
for all
< , we have d.x; z/ D which shows
that x < z.
(3) If < , then y D z which implies that x y D z . Since
x
D y
D z
for all
< , we have d.x; z/ D which shows that
x < z.
Thus is a linear order on L. Considering that we endow L with the topology
generated by , let us show that c.L/ D !. Assume that there is an uncountable
disjoint family U of non-empty open subsets of L. Since BL is a base in L, we can
assume that U consists of elements of BL . Furthermore, given a; b 2 L, one of the
sets . ; a/ and . ; b/ is contained in the other which shows that U can contain at
most one set . ; a/. Analogously, U can contain at most one element .b; !/, so
throwing away at most two elements from U, we can consider that U D f.a ; b / W
< !1 g where the intervals .a ; b / are non-empty and disjoint. There will be no
loss of generality to assume that a < b for all < !1 . Choose any x 2 .a ; b /
for each < !1 . Observe that if D d.a ; b /, then x D a D b for
every < . Consider also the ordinals D d.x ; a / and D d.x ; b /. Since
a b , the point x must be distinct from either a or b .
Thus, minf ; g D ; let D maxf ; g; observe that any chain x 2 L with
x 2 x coincides with x on all levels under . Since the relation a < x < b
is determined by the points x ; x and the fact that x D a D b
for all < , we have
(6) x 2 .a ; b / for any x 2 L such that x 2 x.
Now, if p D x , then A D fp W < !1 g is an antichain in S . To see
this, assume that and p 2 H.p ; S /. It is easy to see that there exists a
maximal chain x 3 p . As a consequence, p 2 x and therefore we can apply
(6) to conclude that x 2 .a ; b / \ .a ; b / which is a contradiction. Hence A is an
uncountable antichain in S ; since S is s Souslin tree, this gives another contradiction
which shows that c.L/ D !.
Our next step is to prove that L is not separable. To check this, assume that X
is a countable subset of L. Since every x 2 X is a countable chain of S , there
is < !1 such that .x/ < for any x 2 X . We have S./ ; by (4), so
pick any s 2 S./. Apply (5) to find > and distinct t; u; v 2 S./ such
that ft; u; vg H.s; S /. There exist a; b; c 2 L with t 2 a; u 2 b and v 2 c.
Since the situation is symmetric, we can assume, without loss of generality, that
a b c, i.e., b 2 .a; c/ and hence the interval .a; c/ is non-empty. However, for
every x 2 .a; c/ \ X , we have x
D a
D c
D s.
/ for all
< which
shows that .x/ which contradicts the choice of . Therefore X \ .a; c/ D ;
and hence X is not dense in L. This proves that L is not separable.
Finally, let .K; v/ be a compact linearly ordered space such that .L; / is
isomorphic to an order dense N K. Of course, N is also homeomorphic to L
so c.N / D !. Since N is dense in K by Problem 072, we have c.K/ D ! and
hence K is hereditarily Lindelf by Fact 2. To see that K is not separable observe
that K is perfectly normal by Problem 001 and hence t.K/ .K/ D .K/ D !
(see Problem 327 of [TFS]). Now, if a countable X K is dense in K, then X N
and hence,Sfor any x 2 X , we can choose a countable Ax N with x 2 Ax . The
set A D fAx W x 2 X g N is countable and x 2 A for all x 2 X . Therefore
X A and hence N K D X A which shows that a countable set A is dense in
N contradicting the fact that N is homeomorphic to a non-separable space L. This
contradiction shows that K is a non-separable hereditarily Lindelf compact space,
so Fact 4 is proved. t
u
Now it is very easy to solve our problem. If Jensens axiom holds, then there
exists a Souslin tree by Problem 070. Applying Fact 4 we conclude that there exists a
non-separable hereditarily Lindelf linearly ordered compact space, so our solution
is complete.
T.074. Prove that a linearly ordered compact L-space exists if and only if there
exists a Souslin tree.
Solution. We have proved in Fact 4 of T.073 that if there is a Souslin tree, then
there exists a compact linearly ordered hereditarily Lindelf non-separable space.
Thus we only have to establish necessity constructing a Souslin tree from a linearly
ordered hereditarily Lindelf non-separable compact space.
Let .L; / be any linearly ordered space. Say that the order is dense in L if
for any a; b 2 L for which a < b, there is c 2 L such that a < c < b. Given any
a; b 2 L, such that a < b we let .a; b/L D fs 2 L W a < s < bg. If a > b, it
is convenient to consider that .a; b/L D .b; a/L . Of course, .a; b/L D ; if a D b.
Furthermore, a; b L D .a; b/L [ fa; bg for any a; b 2 L; besides, .a; !/L D fs 2
L W a < sg and . ; a/L D fs 2 L W s < ag. By definition, the family
BL D f.a; b/L W a; b 2 Lg [ f.a; !/L W a 2 Lg [ f. ; a/L W a 2 Lg
is a base of the space L. The elements of BL are called open intervals in L. We

consider the whole space L to be an open interval as well. We will also need the
family CL D f.a; b/L W a; b 2 Lg; the elements of CL are called proper intervals.
A linearly ordered space L is called a Souslin line if c.L/ D ! < d.L/. Given any
linearly ordered space .L; / and any A L, a point a 2 A is extreme for A if a is
either maximal or a minimal point of A. By A we denote the order restricted to
the points of A. The sets .a; !/L are called right rays and the sets . ; b/L are the
left rays of L.
Fact 1. Suppose that there is a Souslin line. Then there exists a Souslin line .S; /
(i) S has neither minimal nor maximal element;
(ii) the order is dense in S ;
(iii) no U 2 .S / is separable.
Proof. Take any Souslin line .X; /. Given x; y 2 X , let x y if the interval
.x; y/X is separable or empty. For technical reasons we will consider empty spaces
separable too. Observe that
(1) an interval .x; y/X is separable if and only if so is x; y X .
Indeed, sufficiency holds in (1) because .x; y/X is open in x; y X and necessity
is true because adding the set fa; bg to a countable dense subset of .x; y/X we obtain
a countable dense subset of x; y X .
It is immediate from the definition that for any x; y 2 X , we have x x and
x y is equivalent to y x. To see that is transitive, assume that x y
and y z. Then .x; z/X is an open subset of the separable space x; y X [ y; z X ,
so .x; z/X is separable and hence is an equivalence relation. For any x 2 X
consider the equivalence class Ex D fy 2 X W y xg of the point x. The family
L D fEx W x 2 X g of all equivalence classes for is crucial for constructing our
promised Souslin line. Observe that
(2) every Ex 2 L is a convex set, i.e., y; z 2 Ex implies y; z X Ex .
Indeed, the space y; z X is separable because y x z; if we take an arbitrary
t 2 y; z X , then .y; t/X is separable being an open subset of a separable space
y; z X ; therefore t y x which shows that any t 2 y; z X belongs to Ex , i.e.,
y; z X Ex so (2) is proved.
It turns out that any element of Ex order represents it in the following sense:
(3) Assume that Ex ; Ey 2 L and Ex Ey ; take any z 2 Ex and t 2 Ey . Then
z t if and only if z0 t 0 for any z0 2 Ex and t 0 2 Ey .
Sufficiency is evident. As to necessity, assume that t 0 z0 for some z0 2 Ex and
t 2 Ey . Then u z0 for all u 2 Ey ; for otherwise, z0 2 t 0 ; u X for some u 2 Ey and
0
hence z0 2 Ey by (2), a contradiction. In particular, t z0 and therefore t 2 z; z0 X

whence t 2 Ex by convexity of Ex [see (2)]; this contradiction shows that (3) is
true.
Now we can introduce the following linear order on L: given distinct
points a D Ex 2 L and b D Ey 2 L, let a < b if and only if
x y. The property (3) shows that this definition is consistent and gives
us a linear orderSon L. To be completely rigorous, we should choose a set
W X such that fEw W w 2 W g D X and u w implies Eu \ Ew D ; for any
u; w 2 W . The property (3) implies that for any u; w 2 W , we have Eu < Ew if and
only if u w, so .L; / is isomorphic to the set .W; W /. Our construction of L
guarantees that
(4) the order is dense in L.
Indeed, if .a; b/L D ; for some a; b 2 L with a < b, take any x 2 a and y 2 b.
Since the interval .x; y/X is an open subset of a separable space a [ b (here we
consider a [ b to be a subspace of X ), the set .x; y/X is separable (maybe empty)
and hence x y which implies Ex D Ey and contradicts a D Ex b D Ey . This
proves (4). Our next observation is that
(5) the set Ex is separable for every x 2 X .

Let U be a maximal disjoint family of non-empty intervals .u; v/X such that u; v 2
Ex . The family U is countable because c.X / D !. For any U D .u; v/X 2 U
we have u x v and hence the space .u; v/X is separable; apply (1) to fix a
S dense set DU u; v X with fu; vg DU . Consider the countable set
countable
D D fDU W U 2 Ug [ N where N is the (possibly empty) set of extreme points
S
Sshow it, assume that z 2 Ex n U and pick any
of Ex . The set D is dense in Ex ; to
W 2 BX with z 2 W and W \ . U/ D ;. Observe that z cannot be an extreme
point of Ex , so there are r; s 2 Ex such that r < z < s. The set .r; s/X \ W
has to be a proper interval, say .u; v/X . It is evident that u; v 2 Ex ; the maximality
of U implies that there is .p; q/X 2 U such that .p; q/X \ .u; v/X ;. The set
D being dense in .p; q/X , we have D \ .u; v/X \ .p; q/X ; which contradicts
D \ .u; v/X D ;. This shows that the set D is dense in Ex so (5) is proved. For
any a 2 L considered as a subspace of X , apply (5) to choose a dense countable
Pa a. Now let us prove that for the linearly ordered space .L; /, we have the
following property:
(6) no non-empty open subspace of L is separable.
It suffices to show that no non-empty U 2 BL is separable. If U D . ; a/L for
some a 2 L, then take any b 2 U . If U is separable, then .b; a/L is also separable
being a non-empty open subset of . ; a/L [see (4)]. Analogously, if some non-
empty interval .a; !/L is separable, then for any b 2 .a; !/L the interval .a; b/L
is also separable. Thus to prove (6) it suffices to show that no interval .a; b/L can
be separable for distinct a; b 2 L. The family fEx W x 2 X g is disjoint; since
c.X / D !, the set M D fa 2 L W IntX .a/ ;g is countable.
Now assume that (6) is false and choose a; b 2 L with a < b such that some
countable Q .a; b/L is dense in .a; b/L . Pick any z 2 a and t 2 b; we claim
that the interval
S .z; t/X is separable. Let M 0 D M \ a; b L ; it is evident that the
set D D fPc W c 2 Q [ M 0 g [ Pa [ Pb is countable. Assume that w 2 .z; t/X
and w clX .D/; take any B 2 BX such that w 2 B and B \ D D ;. It is clear
that B 0 D B \ .z; t/ 2 BX and w 2 B 0 X nD. Besides, B 0 D .s; u/X for some
s; u 2 z; t X . If s u, then c D Es D Eu 2 M 0 and hence .s; u/X \ D Pc ;.
This contradiction shows that c D Es Eu D d and therefore .c; d /L is a non-
empty open subset of .a; b/L [see (4)]. By density of Q in .a; b/L , there is e 2
Q \ .c; d /L and hence D \ .s; u/X D \ e Pe . This contradiction shows that
w 2 clX .D/; the point w 2 .z; t/X was chosen arbitrarily, so .z; t/X clX .D/ and
therefore .z; t/X is separable whence z t which again gives a contradiction with
a D Ez b D Et . This finishes the proof of (6).
Our next step is to prove that c.L/ D !. It suffices to show that any disjoint
subfamily U .L/ of the family BL is countable. Assuming the contrary pick
any uncountable U 2 .L/ \ BL and observe that any two distinct non-empty left
rays have non-empty intersection as well as any two distinct non-empty right rays
of L. This shows that throwing away at most two elements from U, we will have a
disjoint uncountable family of proper intervals of L. Thus there will be no loss of
generality to assume that U D f.a ; b /L W < !1 g. It follows from (4) that taking
a smaller interval inside each .a ; b /L if necessary, we can assume that the family
fa ; b L W < !1 g is disjoint as well.
Choose x 2 a and y 2 b for all < !1 . Observe that the interval
.x ; y /X is non-empty for each < !1 . Indeed, the property (4) implies that
V D .a ; b /L ;; if c 2 V , then v 2 .x ; y /X for any v 2 c by (3). Therefore
the family V D f.x ; y /X W < !1 g is uncountable and consists of non-empty
intervals of X . Besides, V is disjoint because and u 2 .x ; y /X \ .x ; y /X
implies Eu 2 a ; b L \ a ; b L D ; which is a contradiction. Thus V .X /
is disjoint which contradicts c.X / D !. This proves that c.L/ D !.
Finally, let S D LnE where E is the (possibly empty) set of the extreme points
of L. Considering the set S with the order induced from L, we will write instead
of S . It follows from (4) that S is order dense in L in the sense of Problem 072
and hence the topology of S as a subspace of L is generated by the order (see
Problem 072). Note first that S is a non-empty open subspace of L and therefore
c.S / D !. It is an immediate consequence of (6) that no non-empty open subset
of S is separable, so S is a Souslin line with the property (iii). Now, if a 2 S ,
then a E and hence there are b; c 2 L such that b < a < c. Apply (4) to find
b 0 ; c 0 2 L such that b < b 0 < a < c 0 < c; it is evident that b 0 ; c 0 2 S and hence a
cannot be an extreme point of S . Therefore (i) is true for S . To see that is dense
in S , take any a; b 2 S with a < b; by (4) there is c 2 L with a < c < b. It is clear
that c cannot be an extreme point of L, so c 2 S and hence (ii) holds in S as well.
Fact 1 is proved. t
u
Returning to our solution, assume that there exists a linearly ordered hereditarily
Lindelf non-separable compact space. Any such space is a Souslin line, so we can
apply Fact 1 to fix a Souslin line .S; / with the properties (i)(iii). It is an easy
consequence of (i) that CS is a base in S . We omit the index S when using intervals
in S , i.e., .a; b/ is .a; b/S and a; b D a; b S for any a; b 2 S .
ST0 be a maximal disjoint family of non-empty proper intervals of S . It is clear
Let
that T0 is dense in S . Assume that < !1 and we have families fT W < g
T is a disjoint family of non-empty proper intervals of S for any < ;
(7) S
(8) T is dense in S for all < ;
(9) if <
< , then, for any V 2 T and any W 2 T
, we have either V \W D ;
or W V and V nW ;.
Assume first that D C 1. For each U 2 T denote by PU a maximal
disjoint family of non-empty proper intervals .a; b/ such that a; b 2 U . Observe
that .a; b/ a; b and therefore U n.a; b/ fa; bgSis a non-empty set for every
.a; b/ 2 PU . The maximality of PUS implies that PU is dense in U for any
U 2 T . Therefore the family T D fPU W U 2 T g has a union which is dense
in S , i.e., (8) holds for T . The property (7) is evidently true for D . To check
(9) we can assume, without loss of generality, that
D . If V 2 T ; W 2 T and
V \ W ;, then, by our construction, there is U 2 T
such that W 2 PU . Since
V \ U V \ W ;, we have U V ; since W U , we have W V . Besides,

V nW U nW ;, so V nW ; and hence (9) also S holds.
Now assume that is a limit ordinal. Let T0 D fT W < g and consider the
family U D fU 2 CS W U ;, and for any V 2 T0 , we have either U \ V D ;
or U V and V nU ;g. From the definition of U, it is not even clear whether
U ;. However, we will prove that
S
(10) if W is a maximal disjoint subfamily of U, then W is dense in S .
To prove (10) let us first show that
(*) for any non-empty interval I D .p; q/, there is U D .a; b/ 2 U such that
a; b I .
Since the family T0 is countable, there is a countable set P S such that all
endpoints of the intervals from T0 belong to P . Since the set I is not separable by
(iii), there is a non-empty interval .a; b/ I such that a; b I and P \a; b D ;.
To see that U D .a; b/ 2 U take any V D .c; d / 2 T0 such that c < d and
V \ U ;. The choice of U implies c; d .a; b/; if c > b or d < a, then
U \ V D ;, a contradiction. Since c a; b , we have c < a. Analogously,
d a; b implies d > b and hence a; b V together with H D V nU ;
because fa; bg H . This proves that U 2 U and therefore S . / holds.
Now it is easy to establish (10). Assume that G D W is not dense in S and
take any non-empty interval I D .p; q/ such that I \ G D ;. We proved that there
exists U D .a; b/ 2 U such that U I . Therefore the family W [ fU g U is still
disjoint which contradicts maximality of W. This settles (10).
To finish our inductive step, let T be any maximal disjoint subfamily of U. We
have jT j ! because c.S / D !. The property (10) together with the method of
construction of the family T imply that the properties (7)(9) hold for all ,
so our inductive procedure can be continued to obtain a collection fT W < !1 g
with the properties
S (7)(9) fulfilled for all < !1 .
Let T D fT W < !1 g; given any U; V 2 T , let U V if V U . It is
evident that is a partial order on T . We are going to show that .T; / is a Souslin
tree.
To first establish that T is a tree, take any U 2 T ; then U 2 T for some < !1 .
If V 2 T for some > , then it is impossible that V U by (9); the property (7)
implies that no V 2 T nfU g is contained in U . Therefore if V U , then V 2 T
for some < . By (7) the set LU D fV 2 T W V U g has at most one element in
each T . Furthermore, the property (8) implies that for each < , there is V 2 T
with V \ U ; and hence V U . This shows that LU \ T D fV g for some
V 2 T . As a consequence, the correspondence V ! is an isomorphism of LU
onto which proves that LU is well-ordered and hence T is a tree.
The properties (8) and (9) show that T is uncountable because each T is non-
empty and T \ T D ; if < by (9). If A T is an antichain, then U \ V D ;
for any distinct U; V 2 A by (7) and (9). Thus A is countable because c.S / D !.
Finally, if C is an uncountable chain, then we can consider that C is maximal and
hence C D fU W < !1 g where fU g D C \ T for each < !1 . By (9) the

family fU nU C1 W < !1 g is uncountable and consists of non-empty disjoint
open subsets of S which again contradicts c.S / D !. Hence .T; / is a Souslin tree
and our solution is complete.
T.075. Let L be a non-separable linearly ordered space such that c.L/ !. Prove
that c.L L/ > !. In particular, if X is a perfectly normal non-separable linearly
ordered compact space, then c.X / D ! but c.X X / > !.
Solution. Given a linearly ordered space .M; / and any a; b 2 M , such that a <
b, we let .a; b/M D fs 2 M W a < s < bg; besides, .a; !/M D fs 2 M W a < sg
and . ; a/M D fs 2 M W s < ag. By definition of a linearly ordered space, the
family
BM D f.a; b/M W a; b 2 M g [ f.a; !/M W a 2 M g [ f. ; a/M W a 2 M g
is a base of the space M.

Fact 1. Let M be an arbitrary linearly ordered space. If U 2 .M / and U has
no isolated points, then there are a; b 2 U such that a < b, the interval .a; b/M is
non-empty and .a; b/M U .
Proof. Pick any s 2 U ; since BM is a base in M , there is V 2 BM such that s 2
V U . Therefore there exists a non-empty V 2 BM with V U . If V D .a; b/M
for some a; b 2 U , then there is nothing to prove. If V D . ; c/M for some c 2 M ,
then V is infinite because it has no isolated points. This makes it possible to choose
a; b; d 2 V such that a < d < b. In particular, a; b 2 V U; .a; b/M V U
and .a; b/M ; because d 2 .a; b/M . Analogously, if V D .c; !/M , then V
has to be infinite because it has no isolated points. This makes it possible to choose
a; b; d 2 V such that a < d < b. We have again a; b 2 V U; .a; b/M V U
and .a; b/M ; because d 2 .a; b/M ; thus Fact 1 is proved. t
u
Returning to our solution, denote by D the set of all isolated points of L. We
omit the index L in the name of the base in L and its intervals, i.e., we use B instead
of BL and .a; b/ instead of .a; b/L for any a; b 2 L.
Since fd g is a non-empty open set for each d 2 D and the family ffd g W d 2 Dg
is disjoint, we have jDj ! because c.L/ D !. The space L is not separable, so
U D LnD is not separable and has no isolated points. Apply Fact 1 to choose a
non-empty interval I0 D .a0 ; c0 / U . Since I0 contains no isolated points, it is
infinite, so there is b0 2 .a0 ; c0 / such that both intervals .a0 ; b0 / and .b0 ; c0 / are
non-empty.
Assume that < !1 and we have constructed points a ; b ; c 2 L with the
following properties:
(1) a < b < c for all < ;
(2) .a ; b / ; and .b ; c / ; for each < ;
(3) .a ; c / \ fb W < g D ; for all < .
The set B D fb W < g is countable, so W D U nB is a non-empty open

set without isolated points. We can apply Fact 1 again to find a non-empty interval
I D .a ; c / W ; since I contains no isolated points, it is infinite, so there
is b 2 .a ; c / such that both intervals .a ; b / and .b ; c / are non-empty. It is
evident that the properties (1)(3) hold for all , so our inductive construction
can be continued to obtain a collection fa ; b ; c W < !1 g of points of L such
that the properties (1)(3) are fulfilled for all < !1 .
It is clear that U D .a ; b / .b ; c / is an open subset of L L. The property
(2) implies U ; for all < !1 . If < , then b .a ; c / so we have two
cases.
(1) b a . Then .a ; b / \ .a ; b / D ; and hence U \ U D ;.
(2) c b , Then .b ; c / \ .b ; c / D ; and again U \ U D ;.
This shows that the family fU W < !1 g .L L/ is disjoint and
hence c.L L/ > !. Finally observe that if X is a linearly ordered non-separable
perfectly normal compact space, then X is hereditarily Lindelf by Problem 001. As
a consequence, c.X / hl.X / D ! and hence the result proved for L is applicable
to X . Thus c.X X / > !, so our solution is complete.
T.076. Let X be a linearly ordered hereditarily Lindelf non-separable compact
space. Prove that Cp .X / is not Lindelf.
Solution. Let be a linear order which generates the topology of X . Given a; b 2
X , such that a < b, let .a; b/ D fx 2 X W a < x < bg and a; b D .a; b/ [ fa; bg.
We will also need the intervals a; b/ D fx 2 X W a x < bg and .a; b D fx 2
X W a < x bg. Assume that Cp .X / is Lindelf; then t.X / D ! by Problem 189
of [TFS]. Denote by D the set of all isolated points of X . Since fd g is a non-empty
open set for each d 2 D and the family ffd g W d 2 Dg is disjoint, we have jDj !
because c.X / D !. The space X is not separable, so U D LnD is not separable
and has no isolated points.
Recall that a family F exp.X / is called point-countable if fF 2 F W x 2 F g
is countable for every x 2 X . Any compact space of countable tightness has a point-
countable -base (see Problem 332 of [TFS]), so fix a point-countable -base C in
the space X . It is evident that the family C 0 D fC 2 C W C U g is a -base in the
space U . Of course, C 0 is point-countable; besides,
(1) U has no countable -base,
because otherwise U is separable which is false. Observe that each C 2 C 0 is a
non-empty open set without isolated points. This makes it possible to apply Fact 1
of T.075 to choose a non-empty interval WC D .aC ; bC / C for each C 2 C 0 . It is
evident that W D fWC W C 2 C 0 g is also a point-countable -base of the space U .
For every set W D .a; b/ 2 W take any function fW 2 Cp .X / such that fW W
X ! 0; 1 ; fW .x/ D 0 for all x a and fW .x/ D 1 for all x b. This choice
is possible because X is a normal space and the sets F D fx 2 X W x ag and
G D fx 2 X W b xg are closed and disjoint.
Consider the set P D ffW W W 2 Wg; we claim that

(*) for any f 2 Cp .X / there is Of 2 .f; Cp .X // such that the set Of \ P is
countable.
Assume first that there exists x 2 X such that f .x/ f0; 1g. Then the set
Of D fg 2 Cp .X / W g.x/ 1 and g.x/ 0g is open in Cp .X / and f 2 Of .
Now, if fW 2 Of for some W D .a; b/ 2 W, then fW .x/ 2 .0; 1/ and hence
x 2 .a; b/. Since there are only countably many W D .a; b/ 2 W such that x 2 W ,
the set P \ Of is countable.
Now assume that f .X / f0; 1g. Since X is compact, it has a minimal element
m and a maximal element M . It is evident that fW .m/ D 0 and fW .M / D 1 for
every W 2 W. If f P , then there is Of 2 .f; Cp .X // such that Of \ P D
;, so there is nothing to prove. Therefore we can assume that f 2 P and hence
f .m/ D 0; f .M / D 1. Thus the set H D f 1 .1/ is closed and non-empty, so it
has a minimal element p 2 H . Observe that m; p/ D f 1 .0/ \ m; p is a closed
subset of X and therefore it has a maximal element q (see Problem 305 of [TFS]),
so the interval .q; p/ is empty. Consider the set Of D fg 2 Cp .X / W g.q/ < 1 and
g.p/ > 0g. It is immediate that Of 2 .f; Cp .X //.
Now, if fW 2 Of for some W D .a; b/ 2 W, then fW .p/ > 0 and hence
p > a; furthermore, fW .q/ < 1 implies q < b. It is impossible that b < p because
then b 2 .q; p/ D ;, a contradiction. Thus p b, i.e., p 2 .a; b . Analogously, if
q < a, then a 2 .q; p/ D ; which is a contradiction. Therefore a q and hence
q 2 a; b/. Observe also that the equalities p D b and q D a cannot hold at the
same time because otherwise ; D .q; p/ D .a; b/ ; which is a contradiction.
Thus p 2 .a; b/ or q 2 .a; b/. The family W being point-countable, there are only
countably many W D .a; b/ such that fp; qg \ W ;. This proves that Of \ P is
countable in this case as well so . / is proved.
The family U D fOf W f 2 Cp .X /g is an open cover Sof the Lindelf space
Cp .X /. If U 0 U is a countable subcover of U, then P D fP \ Of W Of 2 U 0 g
is countable because Of \ P is countable for each Of 2 U 0 (see . /).
Now observe that for any W D .a; b/ 2 W and W 0 D .a0 ; b 0 / 2 W, if
fW D fW 0 , then W D .a; b/ D fW1 ..0; 1// D fW10 ..0; 1// D .a0 ; b 0 / D W 0 .
As a consequence, W D fW W fW 2 P g is a countable -base of U which
contradicts (1). Thus Cp .X / cannot be Lindelf, so our solution is complete.
T.077. Suppose that s .X / D !. Prove that X condenses onto a hereditarily
separable space.
Solution. Given any n 2 N, it follows from s.X n X n / ! that hd.X n / D ! or
hl.X n / D ! (see Problem 014). If hd.X n / D ! for some n, then there is nothing to
prove because the space X itself is hereditarily separable. Therefore we can assume
that hl .X / D ! and hence hd .Cp .X // D !. In particular, d.Cp .X // D !
and hence X condenses onto a second countable space (see Problem 174 of [TFS])
which is, of course, hereditarily separable.
T.078. Suppose that Cp .X / has countable spread. Prove that it can be condensed
onto a hereditarily separable space.
Solution. For an arbitrary n 2 N, let Mn D f1; : : : ; ng; we will also need the set
Sn D f W is a bijection and W Mn ! Mn g. If Z is a space and n 2, let
nij .Z/ DSfz D .z1 ; : : : ; zn / 2 Z n W zi D zj g for any distinct i; j 2 Mn . The set
Fact 1. If Z is a left-separated space, then Z n is left-separated for any n 2 N.
Proof. Let us prove first that Z Z is left-separated. Take any well-order < on Z
which left-separates Z; a well-order on Z Z which left-separates Z Z will be
defined as follows: given x D .a; b/ 2 Z Z and y D .c; d / 2 Z Z, let x y
if max.a; b/ < max.c; d /; if max.a; b/ D max.c; d /, then x y if a < c; finally,
if max.a; b/ D max.c; d / and a D c, then x y if c < d .
We omit a routine verification of the fact that is a linear order on Z Z. Let
us prove that well-orders Z Z. For any x D .a; b/ 2 Z Z, let m.x/ D
max.a; b/; p1 .x/ D a and p2 .x/ D b. Observe that we only use symbols min
and max for the minimum and maximum with respect to the order <. Given any
non-empty set A Z Z, let m0 D minfm.x/ W x 2 Ag and A0 D fx 2
A W m.x/ D m0 g; it is clear that A0 is a non-empty set. Therefore, the element
r0 D minfp1 .x/ W x 2 A0 g is well-defined and hence the set A1 D fx 2 A0 W
p1 .x/ D r0 g is non-empty. Thus we have the element r1 D minfp2 .x/ W x 2 A1 g
and the set A2 D fx 2 A1 W p2 .x/ D r1 g is non-empty. It is immediate that the set
A2 can have at most one point and this point is the -minimal element of A.
To see that the order left-separates Z Z, take any x D .a; b/ 2 Z Z and let
A D fc 2 Z W c < ag and B D fc 2 Z W c < bg. Since the order < left-separates
Z, the sets A and B are closed in Z. We have three cases:
(1) a D b; then A D B and Px D fz 2 Z Z W z xg D .A A/ [ .fag A/ [
.A fag/ which shows that Px is closed being a union of three closed sets;
(2) a < b; then the set Px D .B B/ [ .A fbg/ is closed being a union of two
closed sets;
(3) b < a; then the set Px D .A A/ [ .A fag/ [ .fag B/ is closed being a
union of three closed sets.
Therefore the well-order left-separates Z Z, i.e., we proved our Fact for
k
n D 2. As a consequence, the space Z 2 is left-separated for each k 2 N. It is clear
that a subspace of a left-separated space is left-separated; since for any n 2 N, the
k
space Z n is a subspace of some Z 2 , Fact 1 is proved. t
u
Fact 2. For any space Z if 2 .Z/ is a G -set in Z Z, then n .Z/ is a G -set in
Z n for any n 2 N and any infinite cardinal .
T
Proof. Take an arbitrary family U .Z Z/ such that jUj and U D
D 2 .Z/. Given distinct i; j 2 Mn , let qij W Z n ! Z Z be the natural
projection onto the face defined by i and j , i.e., for any z D .z1 ; : : : ; zn / 2 Z n we
have qij .z/ D .zi ; zj / 2 Z Z. It is clear that nij .Z/ D qij1 ./ and therefore
T
nij .Z/ D Uij where Uij D fqij1 .U S / W U 2 Ug. If Bn D f.i; j / 2 Mn Mn W
i < j g, then the family V D fU D fUij W T .i; j / 2 Bn g W Uij 2 Uij for all
.i; j / 2 Bn g consists of open subsets of Z n and V D n .Z/. Since it is evident
that jVj , Fact 2 is proved. t
u
Fact 3. Given any space Y, a subspace Z Y and a natural number n 2, we have
s.Z n nU / s.Cp .Y // for any U 2 .n .Z/; Z n /.
Proof. This proof is a slight modification of the proof of Fact 1 of T.028. Assume
that s.Cp .Y // D ; if we are given a point z D .z1 ; : : : ; zn / 2 Z n , then supp.z/ D
fz1 ; : : : ; zn g; let n D n .Z/. For any point z 2 Z n nn , consider the set Pz D
fz0 2 Z n nn W supp.z0 / D supp.z/g. It is evident that jPz j D n for any z 2 Z n nn .
It is also clear that either Pz \ Py D ; or Pz D Py for any y; z 2 Z n nn . Therefore
the family P D fPz W z 2 Z n nn g is disjoint; if we choose an element zP 2 P
for each P 2 P, we obtain a set C D fzP W P 2 Pg such that C Z n nn and
jC \ Pz j D 1 for every z 2 Z n nn .
Given any permutation 2 Sn , consider the map p W Z n ! Z n defined by
p .z/ D .z .1/ ; : : : ; z .n/ / for any z D .z1 ; : : : ; zn / 2 Z n . It is immediate that each
p is a homeomorphism such that p .n / D n . As a consequence, the set W D
T
fp .U / W 2 Sn g is an open (in Z) neighborhood of n such that W U and
Pz Z n nW for any z 2 Z n nW . Since Z n nU Z n nW , it suffices to prove that
s.Z n nW / . Suppose that there is a discrete D Z n nW of cardinality C .
Since Pz is finite for any z 2 Z n nn , the set R D fz 2 C W Pz \ D ;g has
cardinality C . For each z 2 R, choose dz 2 Pz \ D; then D 0 D fdz W z 2 Rg is
again a discrete subspace of Z n nW of cardinality C such that
(*) jD 0 \ Pz j D 1 for each z 2 D 0 .
The set F D D 0 nD 0 is a closed subset of Z n and F Z n nW (the bar denotes
the closure in Z n ). Let T D Z n n.W [ F /; there exist E D 0 and m 2 N such that
jEj D C and jPd \ T j D m for every d 2 E. For each d D .d1 ; : : : ; dn / 2 E and
i n, choose a set Oid 2 .di ; Z/ such that
(1) Oid \ Ojd D ; if i j ;
(2) O d \ D 0 D fd g where O d D O1d

Ond ;
(3) if y D .dj1 ; : : : ; djn / 2 n , then Ojd1

Ojdn W .
(4) if z D .dj1 ; : : : ; djn / 2 .Pd nfd g/ \ T , then .Ojd1

Ojdn / \ .E [ F / D ;.
For each d 2 E and any i 2 Mn , we have di clY .ZnOid / and hence there
exists a function fid 2 C.Y; 0; 1 / such that fid .di / D 1 and fid .ZnOid / f0g;
let fd D f1d C

C fnd . If d D .d1 ; : : : ; dn / 2 E, then the set Ud D ff 2 Cp .Y / W
f .di / > 0 for all i 2 Mn g is open and fd 2 Ud . Take any distinct a; d 2 E with
a D .a1 ; : : : ; an / and d D .d1 ; : : : ; dn /. If fd 2 Ua , then, for all i n, there is
ji 2 Mn such that fjdi .ai / > 0 and hence ai 2 Ojdi . An immediate consequence is
that a 2 Ojd1

Ojdn . If there exist i; k n such that i k and ji D jk , then
y D .dj1 ; : : : ; djn / 2 n and a 2 Ojd1

Ojdn \ W by (3) which contradicts
E \ W D ;.
Thus ji jk if i k, i.e., the map W Mn ! Mn defined by .i / D ji ,

is a bijection; let D 1 . Given any z D .dk1 ; : : : ; dkn / 2 Pd \ T , the point
a.z/ D .a.k1 / ; : : : ; a.kn / / belongs to Pa \ T because a.z/ 2 .Okd1

Okdn /nW
and .Okd1

Okdn / \ F D ; by the property (4). This shows that the set A D
fag [ fa.z/ W z 2 Pd \ T g is contained in Pa \ T which contradicts the fact that
Pa \ T has m elements by the choice of E while jAj D m C 1 and A Pa \ T .
This contradiction shows that fd Ua whenever d a which implies that
Ua \ E D ffa g for each a 2 E and, in particular, fd fa for distinct d and a,
i.e., the correspondence d ! fd is a bijection. Therefore ffa W a 2 Eg is a discrete
subspace of Cp .Y / of cardinality C which is a contradiction with s.Cp .Y // .
Thus s.Z n nU / and Fact 3 is proved. t
u
Fact 4. Given any space Y such that s.Cp .Y // for some infinite cardinal ,
there is a dense Z Y such that hl .Z/ .
Proof. There exists a dense left-separated Z Y (see Problem 009). We have
s.Y Y / (see Problem 016) and hence s.Z Z/ . Since Z Z is
also left-separated by Fact 1, we have hl.Z Z/ by Problem 007 and, in
particular, .Z/ . Apply Fact 2 to conclude that n .Z/ is a G -set in Z n for
any n 2 N; n 2.
Fix any n 2 N; n > 2; since n .Z/ is a G -set in n
T the space Z , there exists
a family V .n .Z/; Z / such that jVj and V D n .Z/. For any set
n
F 2 FSD fZ n nV W V 2 Vg we have s.F / by Fact 3. It follows from jF j

that s. F / (it is an easy exercise to show that a union of -manyS spaces
of spread each has spread ). It is evident that Z n nn .Z/ D F , so we
have s.Z n nn .Z// . Applying Fact 0 of T.019, we convince ourselves that
s.Z n / D s.Z n nn .Z// . Besides, the space Z n is left-separated by Fact 1, so
we can apply Problem 007 again to conclude that hl.Z n / . The number n 2 N
was chosen arbitrarily, so we showed that hl .Z/ . Fact 4 is proved. t
u
Returning to our solution, apply Fact 4 to find a dense Z X with hl .Z/ !.
Since Z is dense in X , the space Cp .X / condenses onto a space Y Cp .Z/. We
have hd.Y / hd .Cp .Z// D hl .Z/ D ! (see Problem 026) and therefore
Cp .X / condenses onto a hereditarily separable space Y . Thus our solution is
complete.
T.079. Prove that, under Jensens axiom, there is a space X of countable spread
which does not condense onto a hereditarily separable space.
Solution. It was proved in Problem 073 that, under Jensens axiom, there exists a
hereditarily Lindelf non-separable compact space X . We have s.X / hl.X / D !
while every condensation of X is a homeomorphism because X is compact. Hence
X is a space of countable spread which cannot be condensed even onto a separable
space.
T.080. For an arbitrary space X , let Y be a second countable space such that the
space Cp .X; Y / is dense in Y X . Fix any base B .Y / in the space Y ; an open
set U Cp .X; Y / is called B-standard (or standard with respect to X; Y and B)
if there exist n 2 N, points x1 ; : : : ; xn 2 X and sets O1 ; : : : ; On 2 B such that

U D ff 2 Cp .X; Y / W f .xi / 2 Oi for all i D 1; : : : ; ng. Prove that Cp .X; Y / is
perfectly normal if and only if any open subset of Cp .X; Y / is a union of countably
many B-standard open subsets of Cp .X; Y /.
Solution. For technical purposes we will use an even more general concept of a
standard set with respect to a base. Given spaces Z; T , a base C .T / of the
space T and a set P Cp .Z; T /, call a set U 2 .P / standard in P with respect to
Z; T and C if there is a C-standard open subset U 0 of the space Cp .Z; T / such that
U D U 0 \ P . According to the definition, a set U 0 2 .Cp .Z; T // is C-standard
in Cp .Z; T / if there exist n 2 N; z1 ; : : : ; zn 2 Z and O1 ; : : : ; On 2 C such that
U 0 D z1 ; : : : ; zn I O1 ; : : : ; On .Z; T; C/ D ff 2 Cp .Z; T / W f .zi / 2 Oi for all
i D 1; : : : ; ng. Given any A Z, let A W Cp .Z; T / ! Cp .A; T / be the restriction
map, i.e., A .f / D f jA for all f 2 Cp .Z; T /. Observe that A is continuous being
the restriction of the natural projection pA W T Z ! T A of the product space T Z
onto its face T A (see Problem 107 of [TFS]). A set F Z is called a zero-set if
there is f 2 Cp .Z/ such that F D f 1 .0/. A set U Z is a cozero-set if ZnU is
a zero-set.
Fact 1. Given arbitrary spaces Z and T and any base C .T / of the space T ,
take any non-empty set A Z and assume that V CA D A .Cp .Z; T //
Cp .A; T / is a standard open subset of CA (with respect to A; T and C). Then U D
A1 .V / is a C-standard open subset of Cp .Z; T /.
Proof. By definition, there exist z1 ; : : : ; zn 2 A and O1 ; : : : ; On 2 C such that
V D V 0 \ CA where V 0 D z1 ; : : : ; zn I O1 ; : : : ; On .A; T; C/. It suffices to show
that U D A1 .V / D W D z1 ; : : : ; zn I O1 ; : : : ; On .Z; T; C/. Indeed, if we take
any f 2 W , then f .zi / 2 Oi for each i n and hence A .f /.zi / D f .zi / 2 Oi
for all i n which shows that A .f / 2 V , i.e., f 2 A1 .V /. This proves that
W A1 .V /. On the other hand, if f 2 A1 .V /, then f .zi / D A .f /.zi / 2 Oi
for each i n and therefore f 2 W . Thus U D W and Fact 1 is proved. t
u
Fact 2. A space Z is perfectly normal if and only if every closed subset of Z is a
zero-set or, equivalently, if every open subset of Z is a cozero-set.
Proof. The statements with zero-sets and with cozero-sets are clearly equivalent, so
necessity follows immediately from Fact 1 of S.358. Now assume that every closed
F Z is a zero-set. Every zero-set is a G -set: this is an easy exercise ,so Z is
perfect and we only have to prove normality of Z. Given disjoint closed F; G Z
jf .z/j
fix f; g 2 C.Z/ such that F D f 1 .0/ and G D g 1 .0/. If h.z/ D jf .z/jCjg.z/j for
all z 2 Z, then h is a continuous function on Z for which we have h.F / f0g and
h.G/ f1g, so Z is normal and Fact 2 is proved. t
u
Fact 3. Given any space Z and a second countable space T take any base C
.T / in the space T . If any U 2 .Cp .Z; T // is a countable union of C-standard
subsets of Cp .Z; T /, then Cp .Z; T / is perfectly normal. Observe that we are stating
a stronger fact than sufficiency in Problem 080 because we do not assume that
Cp .Z; T / is dense in T Z .
Proof. S D fzI O .Z; T; C/ W z 2 Z and O 2 Cg. It is evident that every C-standard

open subset of Cp .Z; T / is a finite intersection of elements of S. Any O 2 C is
a cozero-set because any second countable space is perfectly normal (see Fact 2).
Let ' W T ! R be a continuous function such that F D T nO D ' 1 .0/. For any
point z 2 Z the mapping fzg W Cp .Z; T / ! T is continuous being the restriction
to Cp .Z; T / of the natural projection of T Z to its zth factor. Furthermore
1 1
zI O .Z; T; C/ D fzg .O/ D fzg .' 1 .Rnf0g// D .' fzg /1 .Rnf0g/;
which shows that zI O .Z; T; C/ is a cozero-set for any z 2 Z and O 2 C. Observe

that any finite intersection of cozero-sets is a cozero-set because any finite union of
zero-sets is a zero-set (see Fact 1 of S.499). As a consequence, any C-standard open
subset of Cp .Z; T / is a cozero-set. By our assumption every open set is a countable
union of C-standard sets which we proved to be cozero-sets of Cp .Z; T /. Any
countable union of cozero-sets is a cozero-set because any countable intersection
of zero-sets is a zero-set (see Fact 1 of S.499). Therefore every open subset of
Cp .Z; T / is a cozero-set, so we can apply Fact 2 to conclude that Cp .Z; T / is
perfectly normal. Fact 3 is proved. t
u
Returning to our solution observe that if every open U Cp .X; Y / is a countable
union of B-standard open subsets of Cp .X; Y /, then Fact 3 is applicable to conclude
that Cp .X; Y / is perfectly normal and hence we proved sufficiency.
Now assume that the space Cp .X; Y / is perfectly normal. Given any non-empty
open set U Cp .X; Y /, there exists a continuous function ' W Cp .X; Y / ! R
such that F D Cp .X; Y /nU D ' 1 .0/ (see Fact 2). Since R is second countable
and Cp .X; Y / is a dense subspace of the product Y X of second countable spaces,
we can apply Problem 299 of [TFS] to find a countable set A X and a continuous
function W CA D A .Cp .X; Y // ! R such that ' D A . If G D 1 .0/, then
F D A1 .G/ and hence U D A1 .V / where V D CA nG. The set V is open in
the second countable space CA ; since the standard open subsets of CA (with respect
to A; Y and B) form a base in CA , the set V is a union of these standard open sets.
Any second countable space is hereditarily Lindelf, so there is a countable
S family
V of standard subsets of CA (with respect to A; Y and B) such that V D V .
If U D fA1 .V / W V 2 Vg, Sthen U consists of B-standard subsets of Cp .X; Y / by
Fact 1; since jUj ! and U D U , we represented any open U Cp .X; Y / as a
countable union of B-standard open subsets of Cp .X; Y /. This proves necessity and
completes our solution.
T.081. Suppose that Cp .X / is perfectly normal. Prove that all closed subsets of X
X are separable.
Solution. Given arbitrary spaces Z and T take any base C in the space T ; say that
a set U 2 .Cp .Z; T // is C-standard in the space Cp .Z; T / if there exist n 2 N,
distinct points z1 ; : : : ; zn 2 Z and sets O1 ; : : : ; On 2 C such that
U D z1 ; : : : ; zn I O1 ; : : : ; On .Z; T; C/
D ff 2 Cp .Z; T / W f .zi / 2 Oi for all i D 1; : : : ; ng:
GivenSany n 2 N, let Ji D .0; 1 fi g for each i 2 Mn D f1; : : : ; ng and

Hn D fJi W i 2 Mn g [ fg. For any x; y 2 Hn ; x D .t; k/; y D .s; l/,
let .x; y/ D jt sj if k D l. If k l, then .x; y/ D t C s. Let .x; / D
t; .; y/ D s and .; / D 0. The space Hn with the topology generated by
the metric is called the Kowalsky hedgehog with n spines. It is evident that for
every number i 2 Mn , there exists a homeomorphism hi W 0; 1 ! Ji [ fg such
that hi .0/ D . Observe that hi is a continuous map considered as a function from
0; 1 to the space Hn Ji [ fg. Observe that the family Bn D fU 2 .Hn / W
either 2 U or U Ji for some i ng is a base in the space Hn . Given any space
Z, let n .Z/ D fz D .z1 ; : : : ; zn / W there are distinct i; j n with zi D zj g.
Fact 1. Let Z be an arbitrary space; take any distinct points z1 ; : : : ; zk 2 Z and a
set Ui 2 .zi ; Z/ for each i k such that U i \ U j D ; if i j . Suppose also that
n 2 N and we are given points tS 1 ; : : : ; tk 2 Hn nfg. Then there exists a function
f 2 Cp .Z; Hn / such that f .Zn fUi W i kg/ fg, and, for every i k, we
have f .zi / D ti and f .Ui / Jq.i / [ fg where q.i / 2 Mn is the unique natural
number for which ti 2 Jq.i / .
Proof. The Tychonoff property of Z implies that there exists gi 2 Cp .Z; 0; 1 /
such that gi .ZnUi / f0g and gi .zi / D 1 for all i k. Let ri D h1 q.i / .ti / for all
i k; for the function fi D ri
gi 2 Cp .Z; 0; 1 / we have fi .ZnUi / f0g and
fi .zi / D ri for each i S k. Finally define a function f W X ! Hn as follows:
f .z/ D if z 2 Zn. fUi W i ng/; if i n and z 2 Ui , then let f .z/ D
hq.i / .fi .z//. S
Observe first that by our definition of f , we have f .Zn fUi W i ng/ fg
and f .zi / D hq.i / .fi .zi // D hq.i / .ri / D ti for all i k. Besides, f jUi D .hq.i /
fi /jUi which shows that f .Ui / D hq.i / .fi .Ui // hq.i / .0; 1 / D Jq.i / [ fg. To
see that f is continuous take any z 2 Z; since the closures of Ui s are disjoint, there
is G 2 .x; X/ such that G intersects at most one of the sets U1 ; : : : ; Uk , say Ui .
Then f jG D .hq.i / fi /jG is a continuous function, so f is continuous by Fact 1
of S.472. Fact 1 is proved. t
u
Fact 2. For any n 2 N and any space Z the set Cp .Z; Hn / is dense in HnZ .
Proof. The space HnZ is homeomorphic to the space Cp .Z 0 ; Hn / where Z 0 is the set
Z considered with the discrete topology. The family D D .Hn / is, evidently, a
base in the space Hn . The D-standard subsets of the space Cp .Z 0 ; Hn / form a base
in Cp .Z 0 ; Hn /, so it suffices to show that Cp .Z; Hn / intersects every D-standard
subset of Cp .Z 0 ; Hn /.
To establish this, take any D-standard subset U of the space Cp .Z 0 ; Hn /. By
definition, there exist distinct points z1 ; : : : ; zk 2 Z and sets O1 ; : : : ; Ok 2 .Hn /
such that U D ff 2 Cp .Z 0 ; Hn / W f .zi / 2 Oi for all i kg. Since Hn nfg is dense
in Hn , we have Oi nfg ;; pick ti 2 Oi nfg for all i k. Then ti 2 Hn nfg for
each i k and hence we can apply Fact 1 to conclude that there is f 2 Cp .Z; Hn /
such that f .zi / D ti 2 Oi for all i k. As a consequence f 2 U \ Cp .Z; Hn /, so
Cp .Z; Hn / is dense in Cp .Z 0 ; Hn / D HnZ and Fact 2 is proved. t
u
Fact 3. Given any space Z, let F be a non-empty closed subspace of Z n nn .Z/.
Let UFz D z1 ; : : : ; zn I J1 ; : : : ; Jn .Z; Hn ; Bn / for every z D .z1 ; : : : ; znS / 2 F.
Assume that W is a Bn -standard subset of Cp .Z; Hn /, such that W fUFz W
z 2 F g. Then W D z1 ; : : : ; zn ; y1 ; : : : ; yk I O1 ; : : : ; On ; G1 ; : : : ; Gk .Z; Hn ; Bn /
for some point z D .z1 ; : : : ; zn / 2 F and Oi Ji for all i D 1; : : : ; n.
Proof. Since W is Bn -standard, there exist m 2 N, distinct points x1 ; : : : ; xm 2 Z
and sets O1 ; : : : ; Om 2 Bn such that W D x1 ; : : : ; xm I O1 ; : : : ; Om .Z; Hn ; Bn /.
If 2 Oi , then let q.i / D 0; if Oi , then there is a unique q.i / 2 Mn such
that Oi Jq.i / . Choose Vi 2 .xi ; Z/ for all i m in such a way that the family
fV i W i mg is disjoint. We claim that
(1) for any l 2 Mn , there is i m such that q.i / D l.
Indeed, if this is not the case, then let N D fi 2 Mm W q.i / 0g and choose any
ti 2 OS i for all i 2 N . By Fact 1 there exists a function f 2 Cp .Z; Hn / such that
f .Zn fVi W i 2 N g/ fg; f .xS i / D ti and f .Vi / Jq.i / [ fg for all i 2 N .
Observe that if q.i / D 0, then xi fVi W i 2 N g, so f .xi / D 2 Oi as well and
hence f 2 W . Since q.i / l for all i m, we have f .Z/ \ Jl D ;. However
there is z D .z1 ; : : : ; zn / 2 F such that f 2 UFz . The definition of UFz requires that
f .zl / 2 Jl , a contradiction which shows that (1) is true.
It follows from (1) that changing the order of the points x1 ; : : : ; xm and the sets
O1 ; : : : ; Om if necessary we can assume that there are k1 ; : : : ; kn 2 N such that
W D z11 ; : : : ; z1k1 ; : : : ; zn1 ; : : : ; znkn ; y1 ; : : : ; yp ;
O11 ; : : : ; Ok11 ; : : : ; O1n ; : : : ; Oknn ; Q1 ; : : : ; Qp .Z; Hn ; Bn /;
where 2 Qi for all i D 1; : : : ; p and Oil Jl for l D 1; : : : ; n and i D 1; : : : ; kl .

For every n-tuple j D .j1 ; : : : ; jn / such that 1 ji ki , let zj D .z1j1 ; : : : ; znjn /. If
some zj is in F , then everything is proved. If not, then we can choose Qil 2 .zli ; Z/
for every l 2 Mn and i 2 Mkl such S that the family Q D fclZ .Qi / W l 2 Mn ; i 2
l
Mkl g is disjoint, y1 ; : : : ; yp Q and for every n-tuple j D .j1 ; : : : ; jn / with

1 ji ki the set Qj D Qj11

Qjnn is disjoint from F . Fact 1 shows that
S
Sg 2
there exists a function Cp .Z; Hn / such that g.zli / 2 Oil ; g 1 .Jl / fQil W
i 2 Mkl g and g.X n fQil W l 2 Mn ; i 2 Mkl g/ fg.
There is z DS.z1 : : : ; zn / 2 F with g 2 UFz . Then g.zl / 2 Jl for each l 2 Mn and
therefore zl 2 fQil W i 2 Mkl g i.e., zl 2 Qjl l for some jl 2 Mkl . Now we have a
contradiction for the n-tuple j D .j1 ; : : : ; jn / because z 2 Qj \ F D ;, so Fact 3
is proved. t
u
Fact 4. If n 1 and the space Cp .Z; Hn / is perfectly normal, then all closed
subspaces of Z n are separable.
Proof. For the sake of brevity, call a space T a CS -space if all closed subspaces
of T are separable. We will prove our fact by induction on n. For the case when
n D 1, take any non-empty closed F Z; for each z 2 F , consider the set
S
Uz D ff 2 Cp .Z; H1 / W f .z/ 2 J1 g. The set UF D fUz W z 2 F g is open
in Cp .Z; H1 /, so we can apply Problem 080 to conclude that S there is a countable
family V of B1 -standard subsets of Cp .Z; H1 / such that V D UF . For each
V 2 V we have V D z1 ; : : : zk I O1 ; : : : ; Ok .Z; H1 ; B1 / where
S zi 2 Z and Oi 2 B1
for all i 2 Mk ; let supp.V / D fz1 ; : : : ; zk g. The set A D fsupp.V / W V 2 Vg is
countable and hence so is the set B D A \ F . It suffices to show that B D F .
Assuming the contrary we can find y 2 F nB and a function f 2 Cp .Z; H1 /
such that f .B/ fg and f .y/ 2 J1 . It is clear that f 2 Uy UF , so there
is V D z1 ; : : : zk I O1 ; : : : ; Ok .Z; H1 ; B1 / 2 V such that f 2 V . Apply Fact 3 to
conclude that there is z D zi 2 supp.V / for which z 2 F and Oi J1 . However,
zi 2 A \ F D B and f .zi / D by our choice of f , so f .zi / Oi , a contradiction
with f 2 V . Therefore B is dense in F and hence F is separable. This proves that
Z is a CS -space, so the case when n D 1 is settled.
Assume that n > 1 and our fact is proved for all m < n; let n D n .Z/.
Observe that Hm is a subspace of Hn for all m < n and hence Cp .Z; Hm / is
perfectly normal being a subspace of Cp .Z; Hn /. The inductive hypothesis implies
that X m is a CS -space for all m < n. The space nij D fz D .z1 ; : : : ; zn / 2 Z n W
zi D zj g is homeomorphic to Z Mn nfi;j g 2 which S n is homeomorphic to Z
n1
because 2 is homeomorphic to Z. Since n D fi;j W i; j 2 Mn ; i j g, the

space n is a finite union of spaces homeomorphic to X n1 . It is immediate that a
finite union of CS -spaces is a CS -space, so the induction hypothesis implies that
n is a CS -space. Therefore it suffices to show that Z n nn is a CS -space. We
consider that the base Bn is fixed in Hn .
Let F be a closed set in Z n nn . For any z D .z1 ; : : : ; zn / 2 F consider
S the set
Uz D ff 2 Cp .Z; Hn / W f .zi / 2 Ji for all i 2 Mn g. The set UF D fUz W z 2 F g
is open in Cp .Z; Hn /; Fact 2 makes it possible to apply Problem 080 to conclude
that there S family V of Bn -standard subsets of Cp .Z; Hn / such that
S exists a countable
UF D V. Let A D f.supp.V //n W V 2 Vg; it is evident that A is countable, so
it suffices to prove that the countable set B D A \ F is dense in F .
If it is not so, then take any y D .y1 ; : : : ; yn / 2 F nB; for each i 2 Mn there
exists Wi 2 .yi ; Z/ such that the family fW i W i 2 Mn g is disjoint and W \ B D ;
where W D W1

Wn . Apply Fact 1 to Sfind f 2 Cp .Z; Hn / with f .yi / 2 Ji ,
f 1 .Ji / Wi for all i 2 Mn and f .Zn fWi W i 2 Mn g/ D fg. There exists
V 2 V such that f 2 V . Now Fact 3 guarantees existence of distinct z1 ; : : : ; zn 2
supp.V / with z D .z1 ; : : : ; zn / 2 F and f .zi / 2 Ji for all i 2 Mn . But then zi 2 Wi
and therefore z 2 W ; since z 2 .supp.V //n A \ F D B, we have z 2 W \ B
which is a contradiction. Fact 4 is proved. t
u
Returning to our solution, let h.t/ D h1 .t/ for all t 2 0; 1 ; if t 2 1; 0/, then let
h.t/ D h2 . t/. It is an easy exercise that the map h W I ! H2 is a homeomorphism
and hence H2 is homeomorphic to I. As a consequence, the space Cp .X; H2 / is
homeomorphic to a subspace of Cp .X /. Thus perfect normality of Cp .X / implies
perfect normality of Cp .X; H2 /, so X X is a CS -space by Fact 4 and hence our
T.082. Let X be a compact space with Cp .X / perfectly normal. Prove that X X

is hereditarily separable.
Solution. For the sake of brevity we say that a space Z is a CS -space if all closed
subspaces of Z are separable. Our solution will be based on two simple statements.
Fact 1. If Z is a CS -space, then s.Z/ D !.
Proof. Indeed, if D Z is uncountable and discrete, then F D D is closed in Z,
so there is a countable dense A F . However, D is open in F , so for any d 2 DnA
(which exists because A is countable and D is not), the set U D fd g is non-empty,
open in F and U \ A D ;; this contradiction completes the proof of Fact 1. t
u
Fact 2. If K is a compact CS -space, then K is hereditarily separable.
Proof. Observe first that any free sequence in K is a discrete subset of K; we have
s.K/ D ! by Fact 1, so all free sequences in K are countable and hence t.K/ D !
(see Problem 328 of [TFS]). Take any Y K; since Y is separable, we can take a
countable A Y with A D Y . It follows from t.K/ D ! S that for any a 2 A, there
is a countable Ba Y such that a 2 B a . The set B D fBa W a 2 Ag Y is
countable and B A D Y Y . Therefore B is a countable dense subset of Y and
we proved that every subspace of K is separable. Thus K is hereditarily separable
and Fact 2 is proved. t
u
Returning to our solution, observe that all closed subspaces of X X are
separable by Problem 081, so we can apply Fact 2 to conclude that X X is
T.083. Let X be a compact space with Cp .X / perfectly normal. Prove that under
MAC:CH, the space X is metrizable.
Solution. Perfect normality of Cp .X / implies that all closed subspaces of X X are
separable (see Problem 081). Therefore s.X X / D ! by Fact 1 of T.082. Under
Martins axiom and the negation of CH the countability of the spread of X X
implies metrizability of X (see Problem 062).
T.084. Prove that the following properties are equivalent for any space X :
(i) Cp .X / Cp .X / is perfectly normal;
(ii) .Cp .X //n is perfectly normal for any natural n 1;
(iii) .Cp .X //! is perfectly normal;
(iv) Cp .X; Y / is perfectly normal for any second countable space Y ;
(v) for every n 2 N, all closed subsets of X n are separable.
subspaces of Z are separable. Given arbitrary spaces Z and T , take any base C
in the space T which is invariant under finite intersections, i.e., the intersection of
finitely many elements of C is still an element of C. A set U 2 .Cp .Z; T // is called
C-standard in the space Cp .Z; T / if there exist n 2 N, distinct points z1 ; : : : ; zn 2 Z
and sets O1 ; : : : ; On 2 C such that U D z1 ; : : : ; zn I O1 ; : : : ; On D ff 2
Cp .Z; T / W f .zi / 2 Oi for all i D 1; : : : ; ng. Observe that if we omit the condition
saying the points z1 ; : : : ; zn in the definition of U are distinct, then we obtain the
same concept because if zi D zj then considering Oi \ Oj instead of Oi for the
point zi and omitting zj and Oj we obtain the same set U . After finitely many of
such transformations we obtain the same set U written as y1 ; : : : ; yk I W1 ; : : : ; Wk
where y1 ; : : : ; yk are distinct and W1 ; : : : ; Wk 2 C. That is why, in this solution, we
do not assume that the points z1 ; : : : ; zn in the definition of U are distinct.
It is evident that (iii)H)(ii)H)(i). Since the space .Cp .X //! is homeomorphic
to Cp .X; R! /, we have (iv)H)(iii).
Now assume that (v) holds; take any second countable space Y and any countable
base B .Y / of the space Y which is invariant under finite intersections. Fix any
U 2 .Cp .X; Y //; given any n 2 N and any B D .B1 ; : : : ; Bn / 2 B n , the set
P .B; n/ D fx D .x1 ; : : : ; xn / 2 X n W W .x; B/ D x1 ; : : : ; xn I B1 ; : : : ; Bn U g
is closed in X n . Indeed, if y D .y1 ; : : : ; yn / 2 X n nP .B; n/, then there is a function

f 2 W .y; B/ such that f U . The set Of D f 1 .B1 /

f 1 .Bn / is open
in X n and y 2 Of . If z D .z1 ; : : : ; zn / 2 Of , then f .zi / 2 Bi for all i n and
hence f 2 W .z; B/; since f U , we proved that W .z; B/ is not contained in U
and hence z P .B; n/. The point z 2 Of was chosen arbitrarily and therefore
Of \ P .B; n/ D ; which proves that P .B; n/ is closed in X n . Since (v) is fulfilled
for X , we can choose, for all n 2 N and all B 2 B n , a countable set Q.B; n/
P .B; n/ which is dense in P .B; n/. The family W D fW .x; B/ W x 2 Q.B; n/
for some n 2 N and B 2 B n g is countable and consists of B-standard subsets of
Cp .X; Y /.
If g 2 U , then, by the definition of the topology of pointwise convergence, there
are n 2 N; x D .x1 ; : : : ; xn / 2 X n and B D .B1 ; : : : ; Bn / 2 B n such that g 2
W .x; B/ U ; it follows from the inclusion W .x; B/ U that x 2 P .B; n/. The
set V D g 1 .B1 /

g 1 .Bn / is an open neighborhood of x in X n ; since Q.B; n/
is dense in P .B; n/, there exists y 2 Q.B; n/\V . It Sis clear that g 2 W .y; B/ U
which shows that W .y; S B/ 2 W and hence g 2 W. The function g 2 U was
chosen arbitrarily, so W D U and we proved that every U 2 .Cp .X; Y // is a
countable union of B-standard subsets of Cp .X; Y /. Applying Fact 3 of T.080 we
conclude that Cp .X; Y / is perfectly normal and hence (v)H)(iv).
Finally, assume that Cp .X / Cp .X / is perfectly normal. Since .Cp .X //2 is
homeomorphic to Cp .X; R2 /, the space Cp .X; R2 / is also perfectly normal. For
every n 2 N the space Hn is homeomorphic to a subset of R2 by Fact 1 of T.019
and hence Cp .X; Hn / is homeomorphic to a subspace of Cp .X; R2 /. Therefore
Cp .X; Hn / is perfectly normal for each n 2 N, so we can apply Fact 4 of T.081
to conclude that X n is a CS -space for all n 2 N, i.e., the property (v) holds. We
proved that (i)H)(v) so our solution is complete.
T.085. Prove that for any compact space X , the space Cp .X / Cp .X / is perfectly
normal if and only if .Cp .X //! is hereditarily Lindelf.
subspaces of Z are separable. If Cp .X / Cp .X / is perfectly normal, then X n is a
CS -space for all n 2 N by Problem 084. Since X is compact, we can apply Fact 2
of T.082 to conclude that hd.X n / D ! for all n 2 N. Therefore hl .Cp .X // D !
by Problem 027 and hence hl..Cp .X //! / D ! by Problem 011.
T.086. Prove that, under SA, if Cp .X / is perfectly normal, then .Cp .X //! is
Solution. If Cp .X / is perfectly normal, then ext.Cp .X // D ! by Problem 295 of
[TFS] (in fact, only normality suffices for this). If D Cp .X / is an uncountable
Ssubspace, then D is open in F D D; perfect normality of F implies that
discrete
D D fDn W n 2 !g where each Dn is closed in F and hence in Cp .X /. As a
consequence, some Dn is a closed uncountable subspace of Cp .X /, a contradiction
with ext.Cp .X // D !. This proves that s.Cp .X // D !. Now, if SA holds, we can
apply Problem 036 to conclude that .Cp .X //! is hereditarily Lindelf.
T.087. Let X be a space with a G -diagonal. Prove that Cp .X / is perfectly normal
if and only if .Cp .X //! is perfectly normal.
subspaces of Z are separable. For any n 2 N, let Mn D f1; : : : ; ng; denote by Sn the
family of all bijections from Mn to Mn . By id W Mn ! Mn , we denote the identity
map. If Z is a space and n 2 N, then n .Z/ D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj for
some distinct i; j 2 Mn g. If 2 Sn , then we have a map 'Z W Z n ! Z n defined
by 'Z .z/ D .z .1/ ; : : : ; z .n/ / for any z D .z1 ; : : : ; zn / 2 Z n . It is clear that 'Z is
a homeomorphism of Z n onto Z n such that 'Z .Z n nn .Z// D Z n nn .Z/. Recall
that given a space Z and A Z, the pseudocharacter T .A; Z/ of A in Z is the
minimal cardinality of a family U .Z/ such that U D A.
Call an open set U Cp .Z; 0; 1 / standard in the space Cp .Z; 0; 1 / if there
exist k 2 N, distinct points z1 ; : : : ; zk 2 Z and sets O1 ; : : : ; Ok 2 .0; 1 / such
that U D z1 ; : : : ; zk I O1 ; : : : ; Ok D ff 2 Cp .Z; 0; 1 / W f .xi / 2 Oi for all
i 2 Mk g; let supp.U / D fz1 ; : : : ; zk g.
Fact 1. If Cp .Z/ is perfectly normal, then s.Cp .Z// D !. In particular, if Z is
perfectly normal and has a G -diagonal, then s .Z/ D !.
Proof. We have ext.Cp .Z// D ! by Problem 295 of [TFS] (in fact, only normality
suffices for this). If D Cp .Z/ is an uncountable discrete subspace, then D is
S
open in F D D; perfect normality of F implies that D D fDn W n 2 !g where
each Dn is closed in F and hence in Cp .Z/. As a consequence, some Dn is a closed
uncountable discrete subspace of Cp .Z/; this contradiction with ext.Cp .X // D !
shows that s.Cp .Z// D !. If Z has a G -diagonal, then s .Z/ D s .Cp .Z// D
s.Cp .Z// D ! by Problem 028 so Fact 1 is proved. t
u
Fact 2. Assume that Cp .Z/ is perfectly normal. If n 2 N and a set P Z n nn .Z/
is closed in Z n , then P is separable.
Proof. We will first prove a weaker fact, namely, that

(*) if F is a closed subspace of Z n such that F Z n nn .Z/ and 'Z .F / \ F D ;
for any 2 Sn nfidg, then F is separable.
It is clear that the space Cp .Z; 0; 1 / is also perfectly normal. For an arbitrary
point z D .z1 ; : : : ; zn / 2 F , let Uz S
D ff 2 Cp .Z; 0; 1 / W f .zi / > 0 for all
i 2 Mn g. We claim that the set UF D fUz W z 2 F g has the following property:
(1) if W D z1 ; : : : ; zp I Q1 ; : : : ; Qp is an arbitrary standard subset of the space
Cp .Z; 0; 1 / such that W UF , then there are numbers i1 ; : : : ; in 2 Mp such
that .zi1 ; : : : ; zin / 2 F and Oik .0; 1 for all k 2 Mn .
Suppose that W D z1 ; : : : ; zp I Q1 ; : : : ; Qp is a standard subset of Cp .Z; 0; 1 /
for which W UF and (1) does not hold. Denote by f0 the function which is
identically zero on Z; then f0 UF because otherwise f0 2 Ut for some point
t D .t1 ; : : : ; tn / 2 F and hence f0 .ti / > 0 for all i 2 Mn which is a contradiction.
However, if 0 2 Qi for all i 2 Mp , then f0 2 W which is impossible because
W UF . This shows that we can assume without loss of generality that there is
some r 2 Mp such that Q1 ; : : : ; Qr .0; 1 and 0 2 Qi for all i 2 Mp nMr .
Since the condition (1) is not satisfied for the set W , we have .zi1 ; : : : ; zin / F
for any fzi1 ; : : : ; zin g fz1 ; : : : ; zr g. The set F is closed in Z n , so we can choose
Gi 2 .zi ; Z/ for all i 2 Mp in such a way that the family fGi W i 2 Mp g is disjoint
and
.Gi1

Gin / \ F D ; if fi1 ; : : : ; in g Mr :
It is easy to see that S there exists f 2 Cp .0; 1 / such that f .zi / 2 Qi for all
i 2 Mr and f .Zn fGi W i 2 Mr g/ f0g. It is clear that f 2 W , so there
is t D .t1 ; : : : ; tn / 2 F such that f 2 Ut . By definition
S of the set Ut , we have
f .ti / > 0 for all i 2 Mn ; therefore ft1 ; : : : ; tn g fGi W i 2 Mr g which implies
that for each k 2 Mn , there is ik 2 Mr for which tk 2 Gik . As a consequence
t 2 .Gi1

Gin / \ F contradicting the choice of the sets Gi . Thus (1) is true.
Now apply Problem 080 to conclude that thereS is a family fWk W k 2 !g of
standardS subsets of C p .Z; 0; 1 / such that U F D fWk W k 2 !g. Since the set
B D fsupp.Wk / W k 2 !g is countable, so is the set A D B n . Thus it suffices to
show that A \ F is dense in F .
Assume that this is not true and fix any point y D .y1 ; : : : ; yn / 2 F nA \ F .
It follows from the hypothesis in . / that f'Z .y/ W 2 Sn g \ A \ F D ; and
therefore there exist disjoint sets H1 ; : : : ; Hn 2 .Z/ such that yi 2 Hi for all
i 2 Mn and .Hj1

Hjn / \ .A \ F / D ; whenever S fj1 ; : : : ; jn g Mn .
Choose any function g 2 Cp .Z; 0; 1 / such that g.Zn fHi W i 2 Mn g/ f0g and
g.yi / D 1 for all i 2 Mn . Then g 2 Uy UF and therefore there exists k 2 ! for
which g 2 Wk . The property (1) implies that Wk D z1 ; : : : ; zp I Q1 ; : : : ; Qp where
z D .zi1 ; : : : ; zin / 2 F for some i1 ; : : : ; in 2 Mp and Qik .0; 1 for all k 2 Mn .
Given any k 2 Mn , we have g.zik / > 0, so there exists jk 2 Mn such that zik 2 Hjk .
Consequently, z 2 .Hj1

Hjn / \ .A \ F /, and this contradiction completes
the proof of . /.
Now let us prove first that P is locally separable, i.e.,
(2) for each z D .z1 ; : : : ; zn / 2 P , there is a separable Wz 2 .z; P /.
Since all points z1 ; : : : ; zn are distinct, we can choose a set Vi 2 .zi ; Z/ for
every i 2 Mn in such a way that V i \ V j D ; if i j . Observe that we have
V D V1

Vn 2 .z; Z n / and 'Z .V / \ V D ; for each 2 Sn nfidg. Therefore
the set V \ P satisfies the conditions given in the hypothesis of . / and hence . /
can be applied to conclude that V \ P is separable. The set Wz D V \ P is open in
P and z 2 Wz . Besides, Wz is open in a separable smaller subspace V \ P , so Wz is
also separable which shows that we proved (2).
Let W D fWz W z 2 P g; then W is an open cover of P with separable subspaces
of P . Observe also that s.Cp .Z// D ! by Fact 1 and apply Fact 1 of T.028 to
conclude that s.P / D ! as well. By Fact 1 of T.007 we can find S a countable W 0
W and a discrete (and hence countable) set D P such that W 0 [ D D P .
A countable union of separable
S S spaces is separable, so there is a countable A
W 0 which is dense in W 0 . Therefore A [ D is a countable dense subset of P ,
so P is separable and Fact 2 is proved. t
u
Fact 3. Given an infinite cardinal , if Z is a space with .2 .Z/; Z/ , then
.n .Z/; Z n / for any n 3. In particular, if Z has a G -diagonal, then n .Z/
is a G -subset of Z n for all n 2.
Proof. Fix nT2 N with n 3; take an arbitrary family U .Z Z/ such that
jUj and U D 2 .Z/. Given distinct i; j 2 Mn , let qij W Z n ! Z Z be the
natural projection onto the face defined by i and j , i.e., for any z D .z1 ; : : : ; zn / 2
Z n we have qij .z/ D .zi ; zj / 2 Z Z. It is an immediate consequence from the
definition of qij that nij .Z/ D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj g D qij1 .2 .Z//
T
and therefore nij .Z/ D Uij where Uij D fqij1S .U / W U 2 Ug. If Bn D f.i; j / 2
Mn Mn W i < j g then the family V D fU D fUij T W .i; j / 2 Bn g W Uij 2 Uij
for all .i; j / 2 Bn g consists of open subsets of Z n and V D n .Z/. Since it is
evident that jVj , Fact 3 is proved. t
u
Returning to our solution, observe that we must only prove that perfect normality
of Cp .X / implies perfect normality of .Cp .X //! . By Problem 084 it suffices to
show that if Cp .X / is perfectly normal, then X n is a CS -space for any n 2 N. By
Problem 081 this is true for n D 2. Assume that n > 2 and we proved CS -property
of X m for all m < n; let n D n .X /. The space nij D fx D .x1 ; : : : ; xn / 2 X n W
xi D xj g is homeomorphic to X Mn nfi;j g 2 which
S nis homeomorphic to X
n1
because 2 is homeomorphic to X . Since n D fij W i; j 2 Mn ; i j g, the

space n is a finite union of spaces homeomorphic to X n1 . It is immediate that a
finite union of CS -spaces is a CS -space, so the induction hypothesis implies that
n is a CS -space. Therefore it suffices to show that X n nn is a CS -space.
Take any closed F X n nn . Fact 3 implies

T that n is a G -subset of X n ;
let U .X / be a countable family with U D n . For each U 2 U the set
n
FU DS F nU is separable being a closed subset of X n (see Fact 2). It is evident that

F D fFU W U 2 Ug, so F is a countable union of its separable subspaces. Thus
F is separable so our solution is complete.
T.088. Prove that, under MA+:CH, all closed subspaces of Cp .X / are separable
if and only if .Cp .X //! is hereditarily separable.
subspaces of Z are separable. It suffices to show that, under MAC:CH, if Cp .X / is
a CS -space, then .Cp .X //! is hereditarily separable. The space Cp .X / is separable
being a closed subspace of itself; therefore i w.X / D ! and hence .X / D !.
Another observation we need is that s .X / D ! by Fact 1 of T.087. Now, since
MAC:CH holds, we have hl .X / D ! by Problem 059. As a consequence, we
can apply Problems 026 and 012 to conclude that hd..Cp .X //! / D !.
T.089. Prove that, under CH, there exists a subspace X of f0; 1g!1 such that for all
n 2 N, all closed subsets of X n are separable and X is not hereditarily separable.
Therefore, Cp .X / is a perfectly normal non-Lindelf space.
Solution. If Z is a space, then A Z is called a meager subset of Z if A is of
first category in Z, i.e., A is a countable union of nowhere dense subsets of Z. For
any n 2 N let n .Z/ D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj for some distinct
i; j 2 Mn D f1; : : : ; ngg. Of course, 1 .Z/ D ;. Let D be the set f0; 1g endowed
with the discrete topology. A family N of nowhere dense subsets of Z is called
cofinal in Z if, for every nowhere dense P Z, there is N 2 N such that P N .
Say that a set P Z is somewhere dense in Z if Int.P / ;.
Fact 1. Suppose that Z is any space and Y is a space with w.Y / D !. Given a set
A Z Y , let Az D fy 2 Y W .z; y/ 2 Ag. Assume that P Z Y is a nowhere
dense (meager) set; then there is a meager set M.P / Z such that Pz is a nowhere
dense (meager) set in Y for any z 2 ZnM.P /.
Proof. Fix any -base B D fBn W n 2 !g in the space Y . Assume first that P
S dense in Z Y and let An D fz 2 Z W P z Bn g for all n 2 !. If
is nowhere
A D fAn W n 2 !g and z 2 ZnA, then P z contains no Bn in its closure and
hence Pz is nowhere dense in Y . If some An is not nowhere dense, then there is
some U 2 .Z/ with clZ .An / U . Then U Bn clZY .P /; to see it, take
any point p D .u; b/ 2 U Bn . Given any basic neighborhood W D G H 2
.p; Z Y /, there is z 2 G \ An because U clZ .An /. Since clY .Pz / Bn ,
we have clZY .fzg Pz / fzg Bn . Thus .z; b/ is in the closure of fzg Pz , so
.Z H / \ .fzg Pz / ; and hence there is t 2 Pz \ H . It is immediate that
.z; t/ 2 .G H / \ P which shows that p 2 clZY .P /. As a consequence, a non-
empty open set U Bn is in the closure of P which is a contradiction. Thus our fact
is proved for a nowhere dense set P .
S
Now, if P D n2! Pn where each Pn is nowhere dense, then there is a meager
set M.Pn / in Z such that S.Pn /z is a nowhere dense set for each z 2 ZnM.Pn /.
It is clear that M.P / D S fM.Pn / W n 2 !g is a meager subset of Z, and if
z 2 ZnM.P /, then Pz D f.Pn /z W n 2 !g is a meager set because so is .Pn /z for
every n 2 !. Fact 1 is proved. t
u
Fact 2. Suppose that Y is a product of second countable spaces. If we have a space
Z such that c.Z Y / D !, then for any nowhere dense (meager) set P Z Y
there is a meager set M.P / Z such that Pz D fy 2 Y W .z; y/ 2 P g is a nowhere
dense (meager) set in Y for any z 2 ZnM.P /.
Proof.QAssume first that P is a nowhere dense subspace of ZY . By our hypothesis
Y D fYt W t 2Q T g where Yt is second countable for any t 2 T . Given any set
S T , let YS D fYt W t 2 S g and let qS W Y ! YS and S W Z Y ! Z YS
be the respective natural projections.
Call an open subset W of the product Z QY a standard subset of Z Y if
W D W0 U for some W0 2 .Z/ and U D t 2T Ut such that Ut 2 .Yt / for all
t 2 T and Ut D Yt for all but finitely many ts; let supp.W / D ft 2 T W Ut Yt g.
It is evident that the family B of all standard subsets of Z Y is a base in Z Y .
Let U be a maximal disjoint subfamily of Sthe collection fB 2 B W B \ P D ;g.
Since P is nowhere dense, the set G D U is dense inSZ Y . Besides, U is
countable because c.Z Y / D !. Therefore the set S D fsupp.W / W W 2 Ug
is also countable. Observe that the set Q D S .P / is nowhere dense in Z YS .
Indeed, S1 .S .W // D W for every W 2 U and therefore S .W / \ Q D ; for
each W 2 U which in turn implies Q \ S .G/ D ;. The map S is open and
continuous (see Problem 107 of [TFS]), so S .G/ is an open dense subset of Z Y
contained in .Z YS /nQ which shows that Q is nowhere dense in Z YS . Since
the space YS is second countable, we can apply Fact 1 to conclude that there is a
meager set M.P / Z such that Qz is nowhere dense in YS for any z 2 ZnM.P /.
It is an easy exercise that any inverse image of a nowhere dense set under an open
map is a nowhere dense set, so qS1 .Qz / is nowhere dense in Y . On the other hand,
Pz qS1 .Qz / for each z 2 Z which shows that Pz is nowhere dense in Y for any
z 2 ZnM.P /, soS our fact is proved for a nowhere dense P .
Now, if P D n2! Pn where each Pn is nowhere dense, then there is a meager
set M.Pn / in Z such that S .Pn /z is a nowhere dense set for each z 2 ZnM.Pn /.
It is clear that M.P / D S fM.Pn / W n 2 !g is a meager subset of Z, and if
z 2 ZnM.P /, then Pz D f.Pn /z W n 2 !g is a meager set because so is .Pn /z for
every n 2 !. Fact 2 is proved. t
u
Fact 3. Suppose that Z D fz W < !1 g is a faithfully indexed separable space
(recall that faithfully indexed means z z whenever ). Given any n 2 N
and a point z D .z1 ; : : : ; zn / 2 Z n , let max.z/ D maxf1 ; : : : ; n g and min.z/ D
minf1 ; : : : ; n g. An !1 -sequence P D fp W < !1 g Z n is called increasing
if < implies max.p / < min.p /. Assume that for any n 2 N, any increasing
!1 -sequence P Z n nn .Z/ is somewhere dense in Z n . Then Z n is a CS -space
for all n 2 N.
Proof. For each ordinal < !1 , let Z D fz W < g. We will prove this
fact by induction on n. If n D 1, then n .Z/ D ; and every uncountable subset
of Z contains an !1 -sequence which is increasing. Thus every uncountable set
is somewhere dense and therefore every nowhere dense subset of Z is countable
(observe that we cannot say that Z is Luzin because it might have isolated points).
Given any closed F Z, the set U D Int.F / has a countable dense set A because
Z is separable. The set F nU is nowhere dense and hence countable; this shows that
.F nU /[A is a countable set which is dense in F . This proves that Z is a CS -space.
Now assume that n > 2 and we proved CS -property of Z m for all m < n; let
n D n .Z/. For each i 2 Mn we denote by pi W Z n ! Z the natural projection
of Z n onto its i th factor. The space nij D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj g
is homeomorphic to Z Mn nfi;j g 2 which
S is homeomorphic to Z
n1
because 2
is homeomorphic to Z. Since n D fnij W i; j 2 Mn ; i j g, the space
n is a finite union of spaces homeomorphic to Z n1 . It is immediate that a finite
union of CS -spaces is a CS -space, so the induction hypothesis implies that n is a
CS -space.
Take any closed F Z n and let U D fU 2 .Z n / S W U P for some
increasing !1 -sequence P F nn g. It is clear that G D U is an open set in
Z n contained in F . Since Z is separable, so is Z n and hence there is a countable
A G with G A. The set F nG is closed in Z n and there is no increasing
!1 -sequence P .F nG/nn because otherwise W D Int.P / ; and hence
W G, a contradiction. Thus there exists < !1 such that .F nG/nn H D
S
fpi1 .Z / W i 2 Mn g which shows that F nG H [ n . It is evident that H is
a countable union of spaces homeomorphic to Z n1 ; we observed already that n
is a finite union since of spaces homeomorphic to Z n1 . Since any countable union
of CS -spaces is a CS -space, this proves that H [ n is a CS -space, so the closed
subspace F nG of the space H [ n has a dense countable subset B. It is clear that
A [ B is a countable dense subset of F , so Fact 3 is proved. t
u
Recall that D fx 2 D!1 W jx 1 .1/j !g is a countably compact dense
subspace of the space D!1 ; besides, is not separable because the closure of any
countable subset of is a metrizable compact space (see Fact 3 of S.307). Take any
base W D fW W 2 !1 n!g ./ of the space . Any countably compact
space has the Baire property (see Problem 274 of [TFS]) and hence every open
subset of has the Baire property (see Problem 275 of [TFS]). By density of in
D!1 any nowhere dense subspace of D!1 which is contained in is nowhere dense
in . This shows that if some W is meager in D!1 , then it is meager in itself which
is a contradiction with the Baire property of W . Thus W is not a meager subset of
D!1 for each < !1 .
Denote by Sn the set of all bijections from Mn onto itself. Given any 2 Sn
and any z D .z1 ; : : : ; zn / 2 .D!1 /n , let T .z/ D .z .1/ ; : : : ; z .n/ /. It is clear that
T W .D!1 /n ! .D!1 /n is a homeomorphism. Call a subset A .D!1 /n symmetric
if T .A/ D A for any 2 Sn .
From now on we assume that CH holds. Observe that for each k 2 N, there exists
a family Nk D fNk W < !1 g with the following Properties:
(1) Nk is a closed symmetric nowhere dense subset of .D!1 /k for each k 2 N and
< !1 ;
(2) Nk is cofinal in the family of all nowhere dense subsets of .D!1 /k for each
k 2 N.
Indeed, for any natural k, there is a cofinal family Nk0 of cardinality !1 of
nowhere dense S closed subspaces of .D!1 /k by Fact 1 of T.039. Given any N 2 Nk0 ,
0
the set N D fT .N / W 2 Sk g is symmetric, closed and nowhere dense in
.D!1 /k while N N 0 . Thus the family Nk D fN 0 W N 2 Nk0 g is still cofinal and
consists of closed symmetric nowhere dense subsets of .D!1 /k ; it is evident that Nk
has cardinality !1 . Taking any enumeration ofSNk we obtain the promised family
Nk with the properties (1) and (2). If Mk D fNk W < g for any k 2 N and
< !1 , then let Mk D fMk W 2 !1 n!g for each k 2 N. It is immediate that the
families Mk have the following properties:
(3) Mk consists of symmetric meager subsets of .D!1 /k and Mk Mk whenever
! < < !1 ;
(4) if M is a meager subset of .D!1 /k , then there exists D .M; k/ 2 !1 n! such
that M Mk .
Take any n 2 N and any natural k > n. For any z D .z1 ; : : : ; zn / 2 .D!1 /k ,
let pnk .z/ D .z1 ; : : : ; zn / and qnk .z/ D .znC1 ; : : : ; zk /. Observe that if we consider
.D!1 /k to be the product .D!1 /n .D!1 /kn , then pnk and qnk are the natural
projections onto the respective factors. In particular, we can apply Fact 2 to the
product .D!1 /n .D!1 /kn and the set Mk to find a meager set R.k; n; / in the
space .D!1 /n such that
(5) the set Mk x D qnk ..pnk /1 .x/ \ Mk / is meager in the space .D!1 /kn for
every x 2 .D!1 /n nR.k; n; /.
Observe that substituting R.k; n; / with a larger set we still have (5), so we can
assume, without loss of generality, that R.k; n; / R.k; n; / whenever < .
For technical reasons, it is convenient to let R.1; 1; / D M1 for all 2 !1 n!.
The space D!1 is separable (see Problem 108 of [TFS]), so we can fix a countable
faithfully indexed dense subset fxn W n 2 !g of D!1 . Choose a point x! 2 D!1
arbitrarily and let D for all !. Assume that !1 > > ! and we have
chosen points fx W < g D!1 and ordinals f W < g !1 such that
(6) for each < ;
(7) if ! < , then for any n 2 N and any infinite ordinals
< 1 <

<
n < we have > .M
k .x1 ; : : : ; xn / ; k n/ for all k > n;
(8) for any number n 2 N and any infinite ordinals
< 1 <

< n < if
x D .x1 ; : : : ; xn /, then M
k x is a meager set in .D!1 /kn for any k > n;
(9) given any natural number l, we have .x1 ; : : : ; xl / M
l for any infinite
ordinals
< 1 <

< l < .
To construct the point x and the ordinal , let D supf.M

k x ; k n/ W
n; k 2 N; k > n; x D .x1 ; : : : ; xn / for S some infinite
< 1 <

< n < g.
Take any > maxf; g; the set S D fR.k; 1; / W k 2 Ng is meager, so
there exists a point x 2 W n.S [ fx W < g/ (this choice of x is possible
because W is not a meager subset of D!1 ). We claim that the sets fx W g
and f W g still satisfy (6)(9). The properties (6) and (7) hold for by
our choice of . To prove (8) we can assume that n D . If n D 1, then for
any k 2 N we have xn D x R.k; 1;
/ because
< < 0 which implies
R.k; 1;
/ R.k; 1; 0 / S while we have chosen x outside of S . Therefore
M
k x is a meager set in .D!1 /k1 by the definition of R.k; 1;
/.
Now, fix any n > 1 and k > n; let y D .x1 ; : : : ; xn1 /. By the property (8) for
the ordinals
< 1 <

< n1 , the set M
k y is meager in .D!1 /knC1 ; besides,
M
k y has been involved in the definition of and hence D .M
k y ; k
n C 1/ < . By the definition of the function , we have M
k y MknC1 .
Observe also that x R.k n C 1; 1; / and hence x R.k n C 1; 1; /
which shows that MknC1x is of first category in .D!1 /kn . Finally note that
M
k x D .M
k y /x MknC1 x is a meager set because so is MknC1 x .
This shows that (8) is true.
To show that the property (9) holds assume that
< 1 <

< l < and
x D .x1 ; : : : ; xl / 2 M
l . By the induction hypothesis, we can assume that l D .
If l D 1, then M
l D R.1; 1;
/ R.1; 1; / because >
. By our choice of
x we have x R.1; 1; / and hence x R.1; 1;
/ which is a contradiction
showing that (9) is true if l D 1.
Now assume that l > 1 and consider the point y D .x1 ; : : : ; xl1 /. It is evident
that xl 2 M
l y and hence M
l y ;. Applying the property (7) for the numbers
k D l; n D l 1 and the ordinals
< 1 <

< l , we conclude that l >
D .M
l y ; 1/ and hence M
l y M1 . Since < l , the point xl was chosen
outside of M1 which is a contradiction because xl 2 M
l y M1 .
This proves that we have the properties (6)(9) for the sets fx W g and
f W g and hence our inductive construction can be continued, giving us sets
fx W < !1 g and f W < !1 g with the properties (6)(9) for all 2 !1 n!.
Let X D fx W < !1 g; we claim that the space X is as promised. The space X
is separable because the countable set A D fxn W n 2 !g is dense in D!1 and hence
in X .
Now fix any n 2 N and assume that S D fp W < !1 g X n nn .X / is an
increasing !1 -sequence. If S is nowhere dense in X n , then it is nowhere dense in
.D!1 /n . Apply (4) to find < !1 n! such that S Mn . However, the sequence S
is increasing, and hence there exists > such that p D .x1 ; : : : ; xn / where
i > for all i 2 Mn . Since p 2 X n nn .X /, all coordinates of p are distinct,
and hence there exists a bijection 2 Sn such that p D T .p / D .x10 ; : : : ; xn0 /
where 10 <

< n0 . We have < 10 <

< n0 , so the property (9) is applicable
to the point p to conclude that p Mn . The set Mn being symmetric, we have
p D T 1 .p/ T 1 .Mn / D Mn which is a contradiction with p 2 S Mn .
This shows that no increasing sequence of X n nn .X / is nowhere dense in X n and

therefore we can apply Fact 3 to conclude that X n is a CS -space for all n 2 N.
Observe finally that X 0 D X \ X is a dense subspace of because it
intersects W for each 2 !1 n!; since is not separable (see Fact 3 of S.307), the
space X 0 cannot be separable, so X is not hereditarily separable and our solution is
complete.
T.090. Prove that a compact space X is metrizable if and only if X 3 is hereditarily
normal.
Solution. If X is metrizable, then X 3 is metrizable as well, so X 3 is hereditarily
normal because so is every metrizable space. This proves necessity.
Fact 1. Given any spaces Y and Z, assume that Z has a countable non-closed
subset and Y has a closed subspace which is not a G -set in Y . Then Y Z is
not hereditarily normal.
Proof. Assume that Y Z is hereditarily normal; fix a set B0 D fzn W n 2 !g Z
for which there is a point z 2 B 0 nB0 , and a closed F Y which is not a G -
set in Y . If B D B0 [ fzg, then the space Y B has to be hereditarily normal
and, in particular, the space T D .Y B/n.F fzg/ has to be normal. It is easy
to see that the sets P D .Y nF / fzg and Q D F B0 are closed in T and
disjoint. Since T is normal, there are U; V 2 .T / such that P U; Q V and
U \ V D ;. The sets U and V are also open in Y B because T is open in Y B.
Therefore the set Wn D fy 2 Y W .y; zn / 2 V g is open in Y for each n; it is also
T that F Wn for all n 2 !. As F is not a G -set in Y , we can take a point
immediate
y 2 . fWn W n 2 !g/nF . Then R D fyg B0 V and therefore t D .y; z/ 2 R.
The set U 3 t being open in Y B we have U \ R ;; since R V we have
U \ V ;, this contradiction completes the proof of Fact 1. t
u
Fact 2. Any infinite compact space has a non-closed countable subspace.
Proof. Let K be an infinite compact space; take any countably infinite A K. If
A is not closed, then we are done; if it is, then A is a countable compact space. The
space A cannot be discrete because no infinite discrete space is compact. Thus there
is a non-isolated point a in the space A. Therefore Anfag is a non-closed countable
subset of K and Fact 2 is proved. t
u
Returning to our solution observe that X 3 is homeomorphic to X 2 X , so we
can apply Fact 1 to the spaces Y D X 2 and Z D X to conclude that all closed
subsets of X 2 must be G -sets in X 2 because Z D X has a non-closed countable
set by Fact 2. In particular, the diagonal D f.x; x/ W x 2 X g is a G -subset in X 2 .
Thus we can apply Fact 1 of T.062 to conclude that X is metrizable. This proves
sufficiency and makes our solution complete.
T.091. Prove that w.X / D .X / for any infinite compact space X . Deduce from
this fact that a compact space X is metrizable if and only if the diagonal of X is a
G -subspace of X X .
Solution. Let D f.x; x/ W x 2 X g be the diagonal of X . Suppose that w.X /

and fix a base B in the space X with jBj . The family P D f.U; V / W U; V 2 B
and U \ V D ;g has cardinality at most
D . If .U; V / 2 P, then F .U; V / D
U V is a closed subset of X X such that F .U; V / \ D ;, i.e., F .U; V /
.X X /n. If .x; y/ 2 .X X /n, then x y, so there exist sets U 0 ; V 0 2 .X/
such that x 2 U 0 ; y 2 V 0 and U 0 \ V 0 D ;. Pick U; V 2 B for which x 2 U
U 0 and y 2 V V 0 ; then S .U; V / 2 P and .x; y/ 2 F .U; V /. This proves the
equality .X X /n D fF .U; V / W .U; V / 2 Pg; since jPj , we showed
that .X X /n is an F -subset of X X and hence is a G -subset of X X
which implies that .X / . Thus .X / w.X /; observe that we did not need
compactness of X to prove this inequality.
To establish that w.X / .X / assume that .XT/ D and choose a family
H 2 .; X X / such that jHj and D H. For any H 2 H take
a finite cover UH of the space X such that U U H for any U 2 UH ; by
paracompactness of X we can find a finite open barycentric refinement VH of the
cover UH . S S
The family S D fVH W H 2 Hg has cardinality and S D X , so there
exists a topology on the set X generated by S as a subbase (see Problem 008
of [TFS]). Let us prove that Y D .X; / is a Hausdorff space. Take any distinct
a; b 2 X ; since c D .a; b/ , there is H 2 H such that c H . We claim that
the set fa; bg is not contained in St.x; VH / for any x 2 X . Indeed, if there is x 2 X
with fa; bg St.x; VH /, then we can find a set W 2 UH such that St.x; VH / W
and therefore fa; bg W . Thus c 2 W W H which is a contradiction.
Now, take any sets U; V 2 VH such that a 2 U and b 2 V ; if z 2 U \ V ,
then fa; bg St.z; VH / which we proved not to be possible. Therefore U \ V D ;
and hence the space Y is Hausdorff. The identity map i W X ! Y is continuous
because i 1 .W / D W 2 .X/ for any W 2 S (see Problem 009 of [TFS]). Being
a continuous image of a compact space X , the space Y is a compact Hausdorff and
hence Tychonoff space (see Problem 124 of [TFS]). The subbase S of the space Y
has cardinality at most ; the family of all finite intersections of the elements of S
is a base of Y of cardinality so w.Y / . As a consequence, i condenses X
onto a space Y with w.Y / . This condensation has to be a homeomorphism (see
Problem 123 of [TFS]), so w.X / and hence we proved that w.X / .X /, i.e.,
w.X / D .X /.
Finally, observe that a compact space X is metrizable if and only if it is second
countable (see Problems 209 and 212 of [TFS]) which, in turn, happens if and only
if .X / D !, i.e., the diagonal of X is a G -subset of X X .
T.092. Prove that a countably compact space X is metrizable if and only if every
subspace Y X with jY j !1 is metrizable.
Solution. Given an arbitrary space Z, a set Y Z and a family A exp.Z/, let
AjY D fA \ Y W A 2 Ag; besides, Az D fA 2 A W z 2 Ag for any z 2 Z.
Fact 1. Suppose that Z is a space and A Z. Then .p; A/ D .p; A/ for any
point p 2 A.
Proof. Since A A, we have .p; A/ .p; A/. Now assume that .p; A/ D
and take any local base B of the point p in the space A such that jBj . For each
U 2 B fix OU 2 .A/ such that OU \ A D U . Then B 0 D fOU W U 2 Bg is a
family of open neighborhoods of p in the space A and jB 0 j . To see that B 0 is a
local base at p in the space A, take any W 2 .p; A/. There exist V 2 .p; A/ and
U 2 B such that clA .V / W and U V \ A; we have
OU clA .OU / D clA .U / clA .V \ A/ clA .V / W
and hence OU W which proves that B 0 is a local base at the point p in the space
A whence .p; A/ jB 0 j . Consequently, .p; A/ .p; A/ which implies
.p; A/ D .p; A/, so Fact 1 is proved. t
u
Fact 2. Let Z be any space. Assume that w.Y / ! for every Y Z with
jY j !1 . Then w.Z/ D !.
Proof. If there is a left-separated Y Z with jY j D !1 , then w.Y / D ! and hence
hd.Y / D ! which implies that there are no uncountable left-separated subspaces
in Y by Problem 004. This contradiction shows that Z is hereditarily separable (see
Problem 004). Take any dense countable A Z; for any z 2 Z, the set A [ fzg is
countable, so .z; A [ fzg/ D ! because w.A [ fzg/ D ! by our hypothesis. Now
apply Fact 1 to conclude that

.z; Z/ D z; fzg [ A D .z; fzg [ A/ D !
for any z 2 Z, i.e., .Z/ D !. Fix a countable local base

S Bz at the point z in the
space Z for all z 2 Z. Given any set A Z, let B.A/ D fBz W z 2 Ag.
Take any z0 2 Z and let Y0 D fz0 g; suppose that < !1 and we have sets
fY W < g with the following properties:
(1) Y Z and jY j ! for all < ;
(2) Y Y
whenever <S
< ;
(3) for any
< if Y
0 D fY W <
g, then B.Y
0 /jY
is not a base in Y
.
S
The family B 0 D B. fY W < g/ is countable, so it cannot be a base for the
space Z. Therefore there exists a point z 2 Z and a set U 2 .z ; Z/ such that
V nU ; for any V S 2 B 0 z . Choose a point zV 2 V nU for each V 2 B 0 z
and let Y D fz g [ . fY W < g/ [ fzV W V 2 B 0 z g. It is immediate that
the properties (1) and (2) hold for the family fY W g. The property (3) also
takes place because the point z and the set U \ Y 2 .z ; Y / witness that B 0 jY
is not a base in Y . Thus our inductive construction can be continued to obtain a
familySfY W < !1 g for which (1)(3) are satisfied for all < !1 . The set
Y D fY W < !1 g has cardinality at most !1 so w.Y / D !.
It is evident that B.Y /jY is a base in Y , so we can apply claim of S.088 to
conclude that there is a countable V B.Y / such that VjY is a base in Y . The
property (2) implies that there is < !1 such that V B.Y / and hence B.Y /jY
is a base in Y . An immediate consequence is that B.Y /jYC1 is a base in YC1

which contradicts (3) and finishes the proof of Fact 2. u
t
Given a space Z and a set Y Z, call a family U .Z/ an external base of Y
in Z if for any y 2 Y and any U 2 .y; Z/ there is V 2 U such that y 2 V U .
Fact 3. Given any space Z and Y Z suppose that we have a collection U .Z/
such that jUj and U is not an external base of Y in Z. Then there exists a set
A Z such that jAj and there is a point z 2 A \ Y which witnesses that UjA
is not a base in A.
Proof. Since U is not an external base of Y in Z, there exists z 2 Y and W 2
.z; Z/ such that there is no U 2 U such that z 2 U W . For each U 2 Uz pick
any zU 2 U nW and let A D fzg [ fzU W U 2 Uz g. It is clear that jAj jUj ;
evidently, the point z 2 Y \ A and its neighborhood W \ A 2 .z; A/ witness the
fact that UjA is not a base in A, so Fact 3 is proved. t
u
Returning to our solution, observe that if X is metrizable, then every Y X
is metrizable so necessity is clear. Now assume that X is not metrizable while any
Y X is metrizable if jY j !1 . Then w.X / !1 , so Fact 2 is applicable to
conclude that there is E X such that jEj !1 and w.E/ !1 . Take any z 2 E;
since fzg [ E is metrizable, we have .z; fzg [ E/ D ! and hence .z; E/ D !
by Fact 1. The point z 2 E was chosen arbitrarily so .E/ D !; besides, w.E/
w.E/ !1 . It is evident that E is a non-metrizable countably compact space such
that any subspace Y E is metrizable whenever jY j !1 . This shows that for
obtaining a contradiction, we can assume without loss of generality that X D E
and hence .X / D !.
Fix a countable local base Ux at the S point x for each x 2 X . Given any set
A X , we will need the family U.A/ D fUx W x 2 Ag. Given a family A of open
sets of a space Z we will say that a point z 2 Z witnesses that A is not a base in Z
if there exists U 2 .z; Z/ such that there is no V 2 A for which z 2 V U .
Take a point y0 2 X and let Y0 D fy0 g. Assume that < !1 and we have a
family fY W < g of subsets of X with the following properties:
(4) jY j ! for all < ;
(5) if < 0 < , then Y Y0 ; S
(6) if < is a limit ordinal, then Y D fY
W
< g; S
(7) if DS
C 1 < , then for any finite V U.Y
/ such that V X , we have
Y n. V/ ;;
(8) if is a limit ordinal with C 1 < , then there is a point z 2 Y \ YC1 which
witnesses that U.Y /jYC1 is not a base in YC1 .
S
If is a limit ordinal, then let Y D fY W < g; it is immediate that
we still have the properties (4)(8) for all . If D
C 1 where
is a
non-limit S ordinal, consider the family W D fV W V S is a finite subfamily of U.Y
/
such that V X g. Choose a point a.V/ 2 X n. V/ for any V 2 W and let
Y D Y
[ fa.V/ W V 2 Wg; it is evident that the properties (4)(8) now hold for all
.
Now, if D
C 1 where
is a limit ordinal, let us prove first that U.Y
/
is not an external base for Y
. Indeed, if U.Y
/ is an external base for Y
, then
Y
X ; for otherwise, w.X / D ! which is a contradiction. Fix any q 2 X nY
and consider the family A D fU 2 U.Y

/ W q U g. Since U.Y
/ is an external
base of Y
, the family A is a countable open cover of Y
; the latter space being
S
countably compact, there is a finite V A such that Y
V. Since
is a limit
ordinal, the property (6) implies that there exists <
such that V U.Y /. Now
S S
V U.Y /; Y S Y
Y
V and V X which shows that there S is a
point x 2 YC1 n. V/ by (7) which is a contradiction with YC1 Y
V.
Now that we established that U.Y
/ is not an external base for Y
, apply Fact 3
to find a countable set Q Z and a point z 2 Y
\ Q such that z witnesses that
U.Y
/jQ is not a base in Q.SNow consider the family W D fV W S V is a finite
subfamily of U.Y
/ such that V X g. Choose a point a.V/ 2 X n. V/ for any
V 2 W and let Y D Q [ Y
[ fa.V/ W V 2 Wg; it is evident that the properties
(4)(7) now hold for all . As to (8), it holds because z 2 Y
\ Y
C1 and z
witnesses that U.Y
/jY
C1 is not a base in Y
C1 due to the fact that z does it for the
family U.Y
/jQ and Q Y
C1 .
Therefore our inductive construction can be continued to obtain a collection fY W
< !S 1 g for which the properties (4)(8) hold for all ordinals < !1 . The space
Y D fY W < !1 g is metrizable because jY j !1 . Take any -discrete base
B in the space Y (see Problem 221 of [TFS]) and fix, for each V 2 B, a set OV 2
.Z/ such that OV \ Y D V . Let C D fOV W V 2 Bg and observe that C D
fU 2 C W U \ Y ;g is countable for all because Y is countable and each
y 2 Y belongs to at most countably many elements of C. Let U D U.Y /; given
an ordinal < !1 , call a pair .B; B 0 / 2 C C adequate if there is U 2 U for
which B \ Y U \ Y B 0 \ Y . For each < !1 and for each adequate pair
.B; B 0 / 2 C C , choose WB;B 0 2 U such that B \ Y WB;B 0 \ Y B 0 \ Y and
let E D fWB;B 0 2 U W .B; B 0 / 2 C C is an adequate pairg. Note that jE j !
and hence there exists ./ > such that E U.Y./ /.
Take any 0 < !1 and inductively let nC1 D .n / for all n 2 !. For the
sequence fn W n 2 !g !1 we have
(9) n < nC1 and En U.YnC1 / for all n 2 !.
Observe that D supfn W n 2 !g is a limit ordinal; the property (8) shows that
there exists z 2 Y \ YC1 and W 2 .z; X / such that
(10) U \ Y 6 W \ Y for every U 2 U.Y /z .
Observe that CjY and UjY are bases in the space Y and therefore there exist
B; B 0 2 C and U 2 U such that z 2 B \ Y U \ Y B 0 \ Y W \ Y . We
have z 2 Y , so B \ Y ; and B 0 \ Y ;. It follows from (5) and (6) that
there exists n 2 ! such that B \ Yn ; and B 0 \ Yn ; whence B; B 0 2 Cn .
Since the pair .B; B 0 / 2 Cn Cn is adequate, we have at our disposal the set
WB;B 0 2 En for which z 2 WB;B 0 \ Y B 0 \ Y W \ Y . Apply the property (9)
to conclude that WB;B 0 2 En U.YnC1 / U.Y / and therefore WB;B 0 2 U.Y /
which is a contradiction with the property (10). This contradiction shows that Y is
a non-metrizable subspace of X of cardinality !1 , so we obtained a contradiction
with our hypothesis. Thus our assumption of non-metrizability of X is false, so X
is metrizable and our solution is complete.
T.093. Give an example of a non-metrizable pseudocompact space P such that every
Y P with jY j !1 is metrizable.
Solution. Let D 2!1 and take any set T of cardinality S . It is easy to find a
disjoint family fT W < g exp.T S / such that fT W < g D T and
jT j D for each < . If Q D fIS W S is a countable subset of T g, then
jQj D jT j!
c D .2!1 /!
2! D 2!1
2! D 2!1 , so we can fix an enumeration
fq W < g of the set Q. For every < , let S be the unique countable subset
of T such that q 2 IS .
For any < define x 2 IT as follows: x .t/ D q .t/ if t 2 S ; if
t 2 T nS , then x .t/ D 1 and x .t/ D 0 for any t 2 T n.T [ S /. The space
P D fx W < g is as promised.
To prove it observe first that P fills all countable faces of I , i.e., for any
countable S T and any q 2 IS , there is p 2 P such that pjS D q. Indeed,
q D q and S D S for some < , so x jS D q D q by the definition of
x . This shows that the set P is dense in IT and hence pseudocompact by Fact 2
of S.433. Another easy observation is that P is not compact because P is dense in
IT and P IT due to the fact that the point u 2 IT defined by u.t/ D 1 for all
t 2 T does not belong to P . As a consequence, P is not metrizable because any
metrizable pseudocompact space is compact.
Finally, if a set A hasScardinality !1 , then jAj !1 < 2!1 D , and
S any < , the set fS W 2 Ag cannot cover the set T . Take any
hence, for
t 2 T n. fS W 2 Anfgg/; then x .t/ D 1 while x .t/ D 0 for all 2 Anfg.
Therefore x cannot be an accumulation point for the set fx W 2 Ag. Since any
subset of cardinality !1 in P can be written as fx W 2 Ag for some A
with jAj !1 , we proved that if Y P and jY j !1 then Y has no accumulation
points in P , i.e., Y is closed and discrete in P . Thus P is not metrizable while every
Y P with jY j !1 is metrizable being a discrete space. We finally proved that
P has all the required properties, so our solution is complete.
T.094. Let X be a non-metrizable compact space. Prove that there exists a
continuous map of X onto a (non-metrizable compact) space of weight !1 .
Solution. The space Cp .X / cannot be separable by Problem 213 of [TFS]. This
implies that hd.Cp .X // d.Cp .X // !1 and hence there is a left-separated
P Cp .X / with jP j D !1 . For each x 2 X let ex .f / D f .x/ for all f 2 P . We
have a map e defined by e.x/ D ex . Then e W X ! Cp .P / and e is a continuous
map (see Problem 166 of [TFS]); let Y D e.X /. Since w.Cp .P // D jP j D !1 (see
Problem 169 of [TFS]), we have w.Y / w.Cp .P // D !1 so w.Y / !1 .
We also have the mapping e W Cp .Y / ! Cp .X / defined by e .f / D f e
for any f 2 Y . The map e is an embedding (see Problem 163 of [TFS]) and
it is easy to see that P e .Cp .Y //. Thus P is an uncountable left-separated

subspace of e .Cp .Y //. Since e .Cp .Y // is homeomorphic to Cp .Y /, the latter
has an uncountable left-separated subspace and therefore w.Y / D nw.Y / D
nw.Cp .Y // hd.Cp .Y // !1 (we applied Problem 172 of [TFS], Problem
004 and Fact 4 of S.307). Therefore w.Y / !1 ; we have proved already that
w.Y / !1 , so Y is a continuous image of the space X such that w.Y / D !1 .
Of course, Y is compact and non-metrizable (see Problems 119 and 212 of [TFS]).
T.095. Let X be a perfectly normal compact space. Prove that d.X / !1 .
Solution. For every non-empty
T closed set F X there exists a countable family
OF .F; X/ such that OF D F ; this is possible because X is perfectly normal.
Pick any x0 2 X and let Y0 D fx0 g. Suppose that < !1 and we have chosen sets
fY W < g with the following properties:
(1) Y X and Y is countable for all < ;
(2) Y Y
if <
< ;
S
(3) let F D Y for all < ; Sif
< ; U
D S fOF W <
g and V is a finite
subfamily of U
such that V X then Y
n. V/ ;.
S
Let U S D fOF W < g and consider the collection W D fV S U W V is
finite and
S V X g. For each V 2 W, take any point x.V/ 2 X n V and let
Y D fY W < g [ fx.V/ W V 2 Wg. It is evident that the sets fY W g
still satisfy (1)(3), so our inductive construction can be continued to give us a
collection fY W < !1 g for which (1)(3)
S are fulfilled for all < !1 . Since all
Y s are countable, for the set Y D fY W < !1 g, we have jY j !1 , so it
suffices to prove that Y is dense in X .
S Observe that X is first countable (see Problem
S 327 of [TFS]) S and therefore Y D
fA W A Y and A is countableg D fY W < !1 g D fF W < !1 g.
If Y X , then fix any point q 2 X nY . For each y 2 Y there is < !1 such
that y 2 F and hence there is Oy 2 OF such that y 2 Oy and q Oy . As a
consequence, the family O D fOy W y 2 Y g is an open cover of the compact space
Y ; letSV O be a finite subcover
S of Y . There is < !1 such that V U ; besides,
V D V X because q V. The property (3) shows that YC1 nV ; which
is a contradiction with YC1 Y Y V . Therefore Y is dense in X ; since
jY j !1 , we have d.X / !1 and our solution is complete.
T.096. Let Y be a subspace of a perfectly normal compact space X . Prove that
nw.Y / D w.Y /.
Solution. Since nw.Y / w.Y / for any space Y , it suffices to establish the
inequality w.Y / nw.Y /; the case of nw.Y / < ! is trivial, so we assume that
D nw.Y / !. Take any network N in the space Y with jN j D and consider
the family F D fN W N 2 N g. We have .F; X / D .F; X / D ! for every
F 2 F (see Problem 327Sof [TFS]), so we can choose a countable outer base OF
for each F 2 F . If B 0 D fOF W F 2 F g, then jB 0 j
! D , so it is sufficient
to show that B D fU \ Y W U 2 B 0 g is a base in Y .
Take any point y 2 Y and W 2 .y; Y /; there exists a set V 2 .y; X/ such
that V \ Y D W . By regularity of X there is U 2 .y; X/ with U V . Since N
is a network in Y , we can find N 2 N such that y 2 N U \ Y . Consequently,
F D N 2 F and F U V ; the family OF being an outer base of F , there is
O 2 OF such that F O V . It is clear that we have y 2 F \ Y O \ Y
V \ Y D W . Since W 0 D O \ Y 2 B, we proved that for any y 2 Y and any
W 2 .y; Y / there is W 0 2 B such that y 2 W 0 W . Thus B is a base in Y and
hence w.Y / jBj D nw.Y /.
T.097. Prove that, under CH, there exists a strictly < -increasing !1 -sequence S D
ff W < !1 g P which is < -cofinal in ! ! .
Solution. It will be easy to construct the promised !1 -sequence after we prove the
following fact.
Fact 1. Let A be a countable subset of ! ! . Then there exists f 2 P such that
g < f for any g 2 A.
Phave A D fgn W n 2 !g (this enumeration of A can have repetitions). Let

Proof. We
f .n/ D ni;j D0 gi .j / C n C 1 for each n 2 !. It is immediate that f .n/ < f .m/
if n < m so f 2 P. Given any n 2 !, observe that gn .m/ < f .m/ for any m n,
so gn < f and Fact 1 is proved. t
u
Returning to our solution apply CH to choose some enumeration fg W < !1 g
of the set ! ! . Apply Fact 1 to the set fg0 g to find a function f0 2 P such that
g0 < f0 . Assume that < !1 and we have functions ff W < g P with the
(1) f
< f if
< < .
(2) g
< f whenever
< .
For the countable set A D fg W g [ ff W < g apply Fact 1 to
find a function f 2 P such that g < f for each g 2 A. It is clear that the set
ff W g still has properties (1) and (2), so our inductive construction can go on
to give us an !1 -sequence S D ff W < !1 g P for which the properties (1) and
(2) hold for every < !1 . It follows from (1) that S is strictly < -increasing. To
prove cofinality of S take any g 2 ! ! ; then g D g for some < !1 and hence
g D g < f 2 S by (2). Therefore S is cofinal in ! ! , so our solution is complete.
T.098. Assuming CH, prove that there exists a space Y with the following proper-
ties:
(i) jY j D !1 and Y is scattered;
(ii) Y is locally compact and locally countable, i.e., every point of Y has a
countable neighborhood (and hence Y is not Lindelf);
(iii) Y k is hereditarily separable for every k 2 N.
In particular, strong S -spaces exist under CH.
Solution. For an arbitrary n 2 N, let Mn D f1; : : : ; ng and denote by Sn the set of

all bijections from Mn onto itself. Given a space Z and a natural number n 2,
let nij .Z/ D fz DS.z1 ; : : : ; zn / 2 Z n W zi D zj g for any distinct i; j 2 Mn .
The set n .Z/ D fnij .Z/ W 1 i < j ng is called the n-diagonal of
Z. It is convenient to consider that 1 .Z/ D ;. Given any 2 Sn and any z D
.z1 ; : : : ; zn / 2 Z n , let T .z/ D .z .1/ ; : : : ; z .n/ /. It is clear that T W Z n ! Z n is a
homeomorphism and T .Z n nn .Z// D Z n nn .Z/.
Suppose that Z is a space which is faithfully indexed by the ordinals from !1 ,
i.e., Z D fz W < !1 g (faithfully indexed means z z whenever ).
Given any n 2 N and z D .z1 ; : : : ; zn / 2 Z n , let max.z/ D maxf1 ; : : : ; n g
and min.z/ D minf1 ; : : : ; n g. A point z D .z1 ; : : : ; zn / 2 Z n nn .Z/ is called
ordered if 1 <

< n . Call two distinct points p; q 2 Z n nn .Z/ comparable if
we have either max.p/ < min.q/ or max.q/ < min.p/. A set P Z n nn .Z/ is
called adequate if each p 2 P is ordered and every two distinct elements of P are
comparable.
Fact 1. Assume that we have a faithfully indexed space Z D fz W < !1 g.
Suppose that for any n 2 N, there is no uncountable adequate discrete subspace of
Z n nn .Z/. Then s.Z n / D ! for all n 2 N.
Proof. We will prove this by induction on n. For each z D z 2 Z, let
.z/ D .
If n D 1, then n .Z/ D ; and every uncountable D Z contains an !1 -sequence
P D fp W < !1 g such that
.p / <
.p / for any < < !1 . It is evident that
.p/ D max.p/ D min.p/ for any p 2 P , so the set P is adequate and hence not
discrete by our hypothesis. Consequently, the subspace D cannot be discrete either.
Thus Z has no uncountable discrete subspaces, i.e., s.Z/ D !.
Assume that our fact is proved for all m < n; we have s.Z n / D s.Z n nn .Z//
(see Fact 0 of T.019), so it suffices to establish that s.Z n nn .Z// D !. Given any
i 2 Mn , the map qi W Z n ! Z is the natural projection of Z n onto its i th factor and
Z D fz W < g for every < !1 .
Suppose that n > 1 and D is an uncountable discrete subspace of Z n nn .Z/.
It is clear that for any z 2 D there exists .z/ 2 Sn such that T .z/ .z/ is an ordered
point. Since Sn is finite, we can choose an uncountable D 0 D and 2 Sn such
that .z/ D for all z 2 D 0 and therefore the point T .z/ is ordered for every
z 2 D 0 . The mapping T W Z n nn .Z/ ! Z n nn .Z/ being a homeomorphism, the
subspace E D T .D 0 / is also discrete and consists of ordered points of Z n nn .Z/.
Since every subspace of E is discrete, it cannot contain an adequate uncountable
set.
It is an easy exercise to see that a subset A Z n nn .Z/ of ordered points of
Z nn .Z/ contains no adequate uncountable set if and only if the set M.A/ D
n
fmin.z/ W z 2 Ag is bounded in !1 . Therefore M.E/ is bounded in !1 and

hence there isS < !1 such that min.z/ < for any z 2 E. As a consequence,
E Q D fqi1 .Z / W i 2 Mn g. It is evident that qi1 .Z / is a countable
union of spaces homeomorphic to Z n1 for each i 2 Mn and therefore Q is also
a countable union of spaces homeomorphic to Z n1 . Since s.Z n1 / D ! by the
induction hypothesis, we have s.Q/ D ! (it is an easy exercise that a countable

union of spaces of countable spread has countable spread) which is a contradiction
with the fact that E is an uncountable discrete subspace of Q. Thus s.Z n / D !;
this completes our inductive step and shows that s.Z n / D ! for all n 2 N. Fact 1 is
proved. t
u
Fact 2. Let Z be any space (no axioms of separation are assumed). If, for any z 2
Z, there exists U 2 .z; Z/ such that U is compact and Hausdorff, then Z is a
Tychonoff space.
Proof. Recall that any compact Hausdorff space is Tychonoff by Problem 124 of
[TFS]; take an arbitrary point z 2 Z and any closed F Z with z F . There
exists U 2 .z; Z/ such that U is compact and Hausdorff. It is clear that G D
.U nU / [ .F \ U / is a closed subset of U such that z G. Therefore there is
a continuous function g W U ! 0; 1 such that g.z/ D 1 and g.G/ D f0g. Let
f .y/ D 0 for all y 2 ZnU and f .y/ D g.y/ for any y 2 U . It is immediate that
f W Z ! 0; 1 ; besides, f .z/ D g.z/ D 1 and f .F / f0g, so we must only check
that f is continuous at each y 2 Z. Fix an arbitrary " > 0.
If y 2 ZnU , then the set V D ZnU is a neighborhood of the point y such that
f .V / D f0g . "; "/ and hence f is continuous at y. If y 2 U , then there is
W 2 .y; U / such that g.W / .g.y/ "; g.y/ C "/. The set W 0 D W \ U
is open in U and hence in Z, so we have W 0 2 .y; Z/ and f .W / D g.W /
.g.y/ "; g.y/ C "/ D .f .y/ "; f .y/ C "/, so f is continuous at the point y.
Finally, if y 2 U nU , then f .y/ D 0. Choose O 2 .y; U / such that f .O/ D
g.O/ . "; "/; there is O 0 2 .Z/ such that O 0 \ U D O. It is clear that
O 0 2 .y; Z/ and it follows from f .O 0 nU / D f0g that f .O 0 / . "; "/, so f
is continuous at y. Fact 2 is proved. t
u
Returning to our solution, apply CH to fix a strictly < -increasing !1 -sequence
F D ff W < !1 g P which is cofinal in ! ! (see Problem 097). If f; g 2 ! ! are
distinct, then .f; g/ D minfn 2 ! W f .n/ g.n/g. It is convenient to consider
that .f; f / D ! for every f 2 ! ! . Observe that given any function f 2 ! ! and
any n 2 !, the set Un .f / D fg 2 ! ! W .f; g/ ng is clopen in ! ! and the family
Uf D fUn .f / W n 2 !g is a local base at f in the space ! ! . It is easy to check that
the function and the collection fUf W f 2 ! ! g have the following properties:
(0) .f; h/ minf.f; g/; .g; h/g for any functions f; g; h 2 ! ! ; if, addition-
ally, .f; g/ .g; h/, then .f; h/ D minf.f; g/; .g; h/g.
(1) for any n 2 ! and any f 2 ! ! if g 2 Un .f /, then Un .g/ Un .f /;
(2) UnC1 .f / Un .f / for any f 2 ! ! and n 2 !.
Choose an !1 -sequence fs W < !1 g with the following properties:
(3) s W ! ! is an injection for each < !1 ;
(4) s j s whenever < < !1 .
The existence of such an !1 -sequence (called an Aronszajn coding) was proved in
Problem 068. Let G D f < W s ./ f ..f ; f //g for each < !1 . If we
have some f D f 2 F , then let

.f / D and H.f / D ff W 2 G g. Recall
that each ordinal can be identified with the set of its predecessors and, in particular,
n D f0; : : : ; n 1g for any n 2 N. If f 2 ! n and g 2 ! ! , then f g says that
gjn D f . If A ! ! , say that A converges to a point f 2 ! ! if the set AnU
is finite for any U 2 .f; ! ! /. This terminology covers the usual convergence of a
sequence to a point as well as the case when A is finite. Thus it is worth remembering
that a finite subset of ! ! converges to any point of ! ! . We must first establish the
following fundamental property of the function H W F ! exp.F /.
(5) Given any f 2 F the set H.f / converges to f in the topology of ! ! or,
equivalently, the set fg 2 H.f / W .g; f / < ng is finite for any n 2 !.
To see that (5) is true assume the contrary; then we can find n 2 N and an
infinite set A H.f /nUn .f /. It is clear that .g; f / < n for any g 2 A, so for
r D f .0/ C

C f .n/ C 1 and D
.f /, we have s .
.g// f ..f; g// < r
for any g 2 A by definition of H.f /. However, s W ! ! is injective as well as
the function
W A ! . It turns out that s
W A ! ! is injective while s .
.A//
is finite, a contradiction. The property (5) is proved.
In what follows we consider the space F with its indexation given above and
fixed. This indexation is considered for applying the notions introduced before
Fact 1. Our promised space Y will have F as the underlying set; the topology of
Y will be constructed later using the function H . The following property of H will
guarantee the hereditary separability of all finite powers of the space Y :
(6) For any k 2 N and any uncountable adequate set P F k there exists a point
p D .p1 ; : : : ; pk / 2 P such that W .p/ D fq D .q1 ; : : : ; qk / 2 P W qi 2 H.pi /
for all i 2 Mk g is an infinite set.
The proof is not easy at all and will be done in several steps. Given any p D
.p1 ; : : : ; pk / 2 .! ! /k , let i .p/ D pi for all i 2 Mk . Since every subset of an
adequate set is an adequate set, we have
(7) for every uncountable P 0 P , the set fi .p/ W p 2 P 0 g is < -cofinal in ! !
for every i 2 Mk ;
Since P is an uncountable subspace of the second countable space .! ! /k , there
exists an uncountable PQ P which has no isolated points in the topology induced
from .! ! /k (see Fact 1 of S.151). Thus there is no loss of generality to assume
that PQ D P , i.e., P is dense-in-itself considered as a subspace of ! ! . Choose a
countable set D P which is dense in P ; it is clear that D does not have isolated
points as well. Fix any < !1 such that max.d / < for each d 2 D and let
P1 D fp 2 P W min.p/ > g; it is clear that P1 is uncountable. We are going to
prove first that
. / for every number l 2 N and any uncountable P 0 P1 , there exist distinct
points d1 ; : : : ; dl 2 D and p 2 P 0 such that i .dj / 2 H.i .p// for all i 2 Mk
and j 2 Ml .
Observe that for any p D .p1 ; : : : ; pk / 2 P 0 , we have p1 <

< pk and
hence there exists m.p/ 2 N such that pi .n/ < pj .n/ for any n m.p/ and i < j .
There is an uncountable P2 P 0 and m0 2 N such that m.p/ D m0 for all p 2 P2 .
Since for any p D .p1 ; : : : ; pk / 2 P2 we have f < p1 , there is n.p/ 2 N such
that f .n/ < p1 .n/ for all n n.p/. Choose an uncountable P3 P2 such that
there is m00 2 N for which n.p/ D m00 for all p 2 P3 . Let m D m0 C m00 . Since the
set fi .p/jm W p 2 P3 g is countable for each i 2 Mk , we can find an uncountable
P31 P3 and t1 2 ! m such that 1 .p/jm D t1 for all p 2 P31 . If h 1 and we have
sets P31

P3h , we can choose an uncountable P3hC1 P3h such that there is
thC1 2 ! m for which hC1 .p/jm D thC1 for all p 2 P3hC1 . Making k successive
choices as described above we obtain an uncountable set P4 D P3k P3 such that
there are functions fti W i 2 Mk g ! m for which i .p/jm D ti for all p 2 P4 and
i 2 Mk .
Observe also that the set fs
.i .p// j W p 2 P4 g is countable for every i 2 Mk :
this easily follows from (4). As in the previous paragraph we can refine k times
the set P4 to obtain an uncountable P5 P4 and functions ui W ! ! such that
s
.i .p// j D ui for all i 2 Mk and p 2 P5 . To sum up, we found an uncountable
P5 P 0 , a number m 2 N, functions t1 ; : : : ; tk 2 ! m and functions u1 ; : : : ; uk 2 !
such that
(8) for any p D .p1 ; : : : ; pk / 2 P5 , we have ti pi for all i 2 Mk ;
(9) for any p D .p1 ; : : : ; pk / 2 P5 and any i; j 2 Mk such that i < j , we have
pi .n/ < pj .n/ for every n m;
(10) for any p D .p1 ; : : : ; pk / 2 P5 , we have f .n/ < p1 .n/ for any n m;
(11) for any p D .p1 ; : : : ; pk / 2 P5 , we have ui s
.pi / for every i 2 Mk .
Given an indexed set A D fnu W u 2 U g ! say that A converges to infinity if
U is infinite and the set fu 2 U W nu hg is finite for any h 2 !. Observe that the
set U has to be countable; besides, if U 0 U is infinite, then the set fnu W u 2 U 0 g
also converges to infinity.
The set B.n/ D f1 .p/.n/ W p 2 P5 g cannot be finite for all n 2 !; for
otherwise, the function h defined by h.n/ D sup B.n/ for all n 2 ! is a < -
upper bound for the set f1 .p/ W p 2 P5 g, a contradiction with the property (7).
Let l1 D minfn 2 N W the set B.n/ is infiniteg. Observe that l1 m and the set
f1 .p/jl1 W p 2 P5 g is finite, so there is an infinite R1 P5 and w1 2 ! l1 such
that the set f1 .p/.l1 / W p 2 R1 g converges to infinity and 1 .p/jl1 D w1 for all
p 2 R1 .
Assume that 1 j < k and we have infinite sets R1 ; : : : ; Rj , natural numbers
l1 ; : : : ; lj and functions w1 ; : : : ; wj with the following properties:
(12) P5 R1

Rj ;
(13) li m and wi 2 ! li for all i 2 Mj ;
(14) i .p/jli D wi for all i 2 Mj ;
(15) the set fi .p/.li / W p 2 Ri g converges to infinity for all i 2 Mj .
The set fj C1 .p/.lj / W p 2 Rj g cannot be bounded in ! because, for each p 2

Rj , we have j C1 .p/.lj / > j .p/.lj / by (9) and (13). Therefore we can define
the number lj C1 D minfn 2 N W the set fj C1 .p/.n/ W p 2 Rj g is unboundedg.
Observe that lj C1 m by (8). The set fj C1 .p/jlj C1 W p 2 Rj g being finite, there
is an infinite Rj C1 Rj and wj C1 2 ! lj C1 such that the set fj C1 .p/.lj C1 / W p 2
Rj C1 g converges to infinity and j C1 .p/jlj C1 D wj C1 for all p 2 Rj C1 . It is clear
that the sets R1 ; : : : ; Rj C1 , the numbers l1 ; : : : ; lj C1 and functions w1 ; : : : ; wj C1
still satisfy (12)(15).
Thus our inductive construction can go on to give us infinite sets R1 ; : : : ; Rk ,
natural numbers l1 ; : : : ; lk and functions w1 ; : : : ; wk which satisfy (12)(15) for all
i 2 Mk . The set P6 D Rk P is infinite and we have
(16) for any p D .p1 ; : : : ; pk / 2 P6 we have pi jli D wi for all i 2 Mk ;
(17) for any n 2 ! there exists p D .p1 ; : : : ; pk / 2 P6 such that pi .li / > n for all
i 2 Mk .
The property (16) is evident and (17) can be easily deduced from the fact that the
set fi .p/.li / W p 2 P6 g converges to infinity for all i 2 Mk .
Since D is dense in P and has no isolated points (in the topology of .! ! /k ),
there exist distinct points d1 ; : : : ; dl 2 D such that wi i .dj / for all i 2 Mk
and j 2 Ml . Since i .dj / < f for all i 2 Mk , there is r 2 N such that r
P
m C kiD1 li C 1 and i .dj /.n/ < f .n/ for all n r and j 2 Ml . Apply (17) to
conclude that there is p D .p1 ; : : : ; pk / 2 P6 such that
pi .li / > maxfui .

.i .dj /// W j 2 Ml g C f .r/ for all i 2 Mk :
Observe that we have pi jli D wi D i .dj /jli for all i 2 Mk and j 2 Ml while
i .dj /.li / < i .dj /.r/ < f .r/ < pi .li / which shows that .i .dj /; pi / D li
and therefore
s
.pi / .
.i .dj /// D u
.pi / .
.i .dj /// < pi .li / D pi ..pi ; i .dj ///
for all i 2 Mk and j 2 Ml which shows that i .dj / 2 H.pi / for all i 2 Mk and
j 2 Ml . This completes the proof of . /.
Now assume that for any p 2 P1 , the set W .p/ is finite. Then there is l 2 N and
an uncountable P 0 P1 such that jW .p/j < l for all p 2 P 0 . However, . / says
that there exist distinct d1 ; : : : ; dl 2 D and p 2 P 0 such that di 2 W .p/ for all
i 2 Ml ; this contradiction shows that there exists p 2 P1 P such that W .p/ is
infinite. The property (6) is proved.
Let us define recursively a set C.f / such that f 2 C.f / fg 2 F W
.g/
.f /g for any f 2 F . If f D f0 , let C.f / D ff g. Assume that f D f and we

have defined C.g/ for all g 2 F D ff W < g. Let C.f / D ff g [ J.f / where
J.f / D fp 2 F W there exists q 2 H.f / such that p 2 C.q/ and .p; q/ >
.p; r/ for any r 2 .ff g [ H.f //nfqgg. Observe that
(18) H.f / C.f / for any f 2 F

because for any p 2 H.f / we can let q D p; then p 2 C.q/ and .p; q/ D ! >
.p; r/ for any r 2 .H.f / [ ff g/nfqg.
Now let Cn .f / D C.f / \ Un .f / for each n 2 ! and f 2 F . It turns out that
(19) for any n 2 ! and f 2 F , if g 2 Cn .f /, then Cm .g/ Cn .f / for some
m 2 !.
We will prove (19) by induction on
.f /; if f D f0 , then (19) is evidently true.
Assume that
.f / D < !1 and (19) is proved for all g 2 F . We must consider
three cases.
Case 1: g D f . then m D n shows that (19) is true.
Case 2: g 2 H.f /. Then g is an isolated point of Hf D ff g [ H.f / by (5).
Therefore there is k 2 N such that Uk .g/ \ Hf D fgg and hence .p; g/ < k
for any p 2 Hf nfgg. If m D k C n, then Cm .g/ Cn .f /. To prove it, take any
h 2 Cm .g/; we have .h; g/ m > .g; p/ for any p 2 Hf nfgg. Applying (0)
we conclude that .h; p/ D minf.h; g/; .g; p/g D .g; p/ < m .h; g/
for every p 2 Hf nfgg which shows that h 2 C.f /. The property (0) implies
that .h; f / minf.h; g/; .g; f /g D .g; f / n. Therefore h 2 Cn .f /
and we proved (19) for this case.
Case 3: g H.f /; then there exists a function h 2 H.f / such that g 2 C.h/
and we have ! > k D .g; h/ > .g; p/ for any p 2 Hf nfhg. Consequently,
.h; p/ D minf.h; g/; .g; p/g D .g; p/ < k for all p 2 Hf nfhg. This
shows that Uk .h/ \ Hf D fhg; since for any g 0 2 Uk .h/ we have Uk .g 0 /
Uk .h/, we obtain the equality .Hf nfhg/ \ Uk .g 0 / D ;. Thus, if g 0 2 Ck .h/,
then .g 0 ; h/ k > .g 0 ; p/ for any p 2 Hf nfhg and hence g 0 2 C.f /. As
a consequence, Ck .h/ C.f /. Besides, n .g; f / < .g; h/ D k whence
.h; f / n which shows that h 2 Un .f / and therefore Un .h/ Un .f /, so
Ck .h/ C.f / \ Un .h/ C.f / \ Un .f / D Cn .f /. Since g 2 Ck .h/ and

.h/ <
.f /, we can apply the induction hypothesis to h to conclude that there
exists m 2 ! such that Cm .g/ Ck .h/ Cn .f / which shows that the proof of
(19) is complete.
The property (19) implies that the family fCn .f / W f 2 F and n 2 !g generates
a topology on F as a base (see Problem 006 of [TFS]); let Y D .F; /. It is
an easy exercise that the topology 0 inherited by F from ! ! is contained in .
An immediate consequence is that Y is Hausdorff. Let us prove that
(20) C.f / is a compact subspace of Y for any f 2 F .
We prove this again by induction on
.f /. If f D f0 , then C.f / is a singleton
and hence compact. Assume that < !1 and we proved (20) for all g 2 F with
.g/ < ; let f D f . Since C.f / is countable, it suffices to show that it is

countably compact, i.e., every infinite set D C.f / has an accumulation point in
C.f / (see Problem 132 of [TFS]).
Assume first that for some n 2 !, the set Dn D fd 2 D W .d; f / D ng is

infinite. By the definition of the set C.f /, for any d 2 Dn , there is fd 2 H.f / such
that d 2 C.fd / and .d; fd / > .d; h/ for all h 2 Hf nffd g and, in particular,
.d; fd / > .d; f / D n for all d 2 Dn . Applying the property (0) to the functions
d; fd ; f , we obtain the equality .fd ; f / D minf.fd ; d /; .d; f /g D n for all
d 2 Dn . The property (5) implies that the set ffd W d 2 Dn g is finite, so there is
g 2 H.f / such that .g; f / D n and the set N D fd 2 Dn W fd D gg is infinite.
Let m D minf.d; g/ W d 2 N g; pick any d 2 N with .d; g/ D m.
It is evident that N Cm .g/. Observe that .g; h/ D minf.g; d /; .d; h/g D
.d; h/ for every h 2 Hf nfgg (we used (0) again). Since .d; h/ < .d; g/ D m,
we have .g; h/ < m for any h 2 Hf nfgg. Now take any g 0 2 Cm .g/;
we have .g 0 ; g/ m. On the other hand .g; h/ < m .g 0 ; g/ for all
h 2 Hf nfgg, and hence we can apply (0) once more to conclude that .g 0 ; h/ D
minf.g 0 ; g/; .g; h/g D .g; h/ < m .g0 ; g/. Thus .g 0 ; h/ < .g 0 ; g/
for any h 2 Hf nfgg so g 0 2 C.f /. The point g 0 2 Cm .g/ was taken arbitrarily so
Cm .g/ C.f /. Therefore N is an infinite subset of Cm .g/; since
.g/ <
.f /, we
can apply the induction hypothesis and conclude that the set N has an accumulation
point p in C.g/. But N Cm .g/ and Cm .g/ is closed in C.g/, so p 2 Cm .g/ and
hence p 2 C.f /, i.e., p is an accumulation point for D in C.f /. Finally, if the set
Dn is finite for each n 2 !, then every Cn .f / contains infinitely many points of D;
hence f is an accumulation point of D and our proof of (20) is complete.
Since C.f / is a neighborhood of f in Y , the space Y is locally compact and
hence Tychonoff by Fact 2. Let us show that s .Y / D !. Observe that we have
an enumeration on Y because the enumerated set F is the underlying set of Y .
Thus it suffices to show that any adequate discrete E Y k is countable for any
k 2 N (see Fact 1). Assume the contrary and take any uncountable discrete adequate
E Y k . For each p D .p1 ; : : : ; pk / 2 E, there exists n D n.p/ 2 N such that
Vp \ E D fpg where Vp D Cn .p1 /

Cn .pk /. There is an uncountable E 0 E
and n 2 N such that n.p/ D n for all p 2 E 0 . The property (6) implies that
there is p D .p1 ; : : : ; pk / 2 E 0 and an infinite Q E 0 such that p Q and
i .q/ 2 H.pi / for any i 2 Mk and any q 2 Q. Since the set E 0 is adequate, we
have i .q/ i .q 0 / for any distinct q; q 0 2 Q and any i 2 Mk . The infinite set
fi .q/ W q 2 Qg converges to pi (in the topology of ! ! ) for each i 2 Mk , so there
is q D .p1 ; : : : ; pk / 2 Q such that qi 2 Cn .pi / for every i 2 Mk and therefore
q 2 .Vp \ E/nfpg which is a contradiction. Thus s .Y / D !.
Next observe that Y is scattered; to see it suffices to show that every closed
Z Y has an isolated point. Take any z 2 Z; the set K D C.z/ \ Z is a compact
open neighborhood of z in Z. Since K is countable, it has an isolated point w. It
is clear that w is isolated in Z, so we established that Y is scattered and hence
not hereditarily Lindelf (see Problems 005 and 006). As a consequence, Y n is not
hereditarily Lindelf for all n 2 N. It follows from s.Y n Y n / D ! that Y n is
either hereditarily separable or hereditarily Lindelf by Problem 014. Since Y n is
not hereditarily Lindelf, we have hd.Y n / D ! for all n 2 N, i.e., hd .Y / D !.
Given any f 2 Y , the set C.f / is a countable neighborhood of f , so Y is locally
countable. Thus, we checked everything promised in (i)(iii) for Y so our solution
is complete.
T.099. Prove that, under CH, there exists a scattered compact space X which is not
first countable (and hence not metrizable) while X ! is hereditarily separable and
hence .Cp .X //! is hereditarily Lindelf. This implies that under CH, there exist
strong L-spaces and strong compact S -spaces. Observe that, under MAC:CH, any
compact space X is metrizable whenever .Cp .X //! is hereditarily Lindelf.
Solution. Let X be the one-point compactification of the space Y constructed in
Problem 098. The space X is well-defined because Y is locally compact. It is an
easy exercise that adding one point to a scattered space gives a scattered space so X
is scattered. Let a be the unique point of X nY . If .a; X / D !, then the space Y D
X nfag is -compact and hence Lindelf which contradicts Problem 098(ii). This
proves that .a; X / > ! and hence X is not first countable.
Fact 1. Let Z be any space; assume that z 2 Z and hd .Znfzg/ for some
infinite cardinal . Then hd .Z/ . In particular, if all finite powers of a space
are hereditarily separable, then adding one point to the space does not destroy this
property.
Proof. Let Mn D f1; : : : ; ng; we will prove by induction on n that hd.Z n /
for all n 2 N. Observe first that the space Z D .Znfzg/ [ fzg is a union of two
spaces of hereditary density . It is evident that even a countable union of spaces
of hereditary density has hereditary density so the case of n D 1 is clear.
Now assume that we proved that hd.Z k / for all k < n. Given any i 2 Mn ,
let pi SW Z n ! Z be the natural projection of Z n onto its i th factor. Consider the set
Q D fpi1 .z/ W i 2 Mn g; it is easy to see that Z n D .Znfzg/n [ Q. The space Q
is a finite union of spaces homeomorphic to Z n1 , so hd.Q/ by the induction
hypothesis and the above observation about countable unions of spaces of whose
hereditary density does not exceed . Since hd..Znfzg/n / by our hypothesis,
we represented the space Z n as a finite union of spaces of hereditary density .
Thus hd.Z n / for all n 2 N and hence Fact 1 is proved. t
u
Returning to our solution apply Fact 1 to conclude that hd .X / D ! and hence
hl..Cp .X //! / D hl .Cp .X // D hd .X / D ! (we applied Problems 011 and
027). Thus under CH we have a compact space X with all promised properties. If
MAC:CH holds and we have a compact space X such that hl..Cp .X //! / D !,
then s.X X / hd .X / D hl .Cp .X // D ! and hence we can apply
Problem 062 to conclude that X is metrizable.
T.100. Prove that, under CH, there exists a non-normal X such that the space
.Cp .X //! is hereditarily Lindelf.
Solution. Let K be a non-metrizable compact space with hd .K/ D !. The
existence of such a space under CH is proved in Problem 099. Apply Problem 090
to conclude that K 3 is not hereditarily normal and hence there is X K 3 such that
X is not normal. It is evident that hd .X / hd .K 3 / D hd .K/ D ! and hence
hl..Cp .X //! / D hl .Cp .X // D hd .X / D ! (we applied Problems 011 and
027). Thus X is a non-normal space for which .Cp .X //! is hereditarily Lindelf.
T.101. Prove that any metrizable space is strongly monolithic.

Solution. Let M be a metrizable space. If A M is an infinite set and jAj D ,
then A is also a metrizable space and therefore w.A/ D d.A/ by Problem 214
of [TFS]. Thus M is -monolithic for any infinite cardinal .
T.102. Prove that if X is a metrizable space with w.X / c, then X condenses onto
a second countable space.
Solution. We will need the following fact.
Fact 1. Let Z be an infinite space with w.Z/ D . Then, for any base B for the
space Z, there is B 0 B such that jB 0 j and B 0 is a base in Z. In other words,
any base of a space contains a base of minimal cardinality.
Proof. Take any base C in Z such that jCj D . Call a pair D .U; V / 2 C C
admissible if there is B 2 B such that U B V . If a pair D .U; V / is
Admissible, then fix some B D B./ 2 B such that U B V . The family B 0 D
fB./ W is an admissible pairg B has cardinality at most jC Cj D jCj D .
The family B 0 is a base in Z; indeed, if x 2 W 2 .Z/, then there is V 2 C such
that x 2 V W because C is a base in Z. By the same reason, there is B 0 2 B
with x 2 B 0 V . Analogously, we can find U 2 C such that x 2 U B 0 . Since
U B 0 V , the pair D .U; V / is admissible and therefore B D B./ 2 B 0 and
x 2 B V W . Therefore B 0 is a base in Z and Fact 1 is proved. t
u
Returning to our solution fix any -discrete base B in the space X (see
Problem 221 of [TFS]). Fact 1 S shows that we can assume, without loss of generality,
that jBj c. We have B D fBn W n 2 !g where each family Bn is discrete.
Since jBn j c, for each n 2 !, there exists an injection 'n W Bn ! T where
T D f.a; b/ 2 R R W a2 C b 2 D 1g is the unit circumference centered at
the origin z0 D .0; 0/ of R R. Fix any n 2 !; for any U 2 Bn , the set
IU D f.ta; tb/ 2 R2 W t 2 0; 1 and .a; b/ D 'n .U /g is homeomorphic to the
interval 0; 1 R and hence there exists a continuous map fU W X ! IU such that
X nU D fU1 .z0 / (we used Fact 1 of S.358 and perfect normality of any metrizable
S
space). Let gn W X ! R R be defined as follows: gn .x/ D z0 if x Bn ; if
x 2 U 2 Bn , then gn .x/ D fU .x/.
To see that gn is continuous take any x 2 X . Since the family Bn is discrete,
there is W 2 .x; X/ such that W intersects at most one element of Bn , say U . It is
evident that gn jW D fU jW is a continuous map, so we can apply Fact 1 of S.472
to see that gn is continuous for all n 2 !. The diagonal product g D fgn W n 2 !g
maps X onto a second countable space Y .R R/! . Thus it suffices to show that
g is an injection. Take any distinct x; y 2 X ; there is n 2 ! and U 2 Bn such that
x 2 U and y U . Therefore gn .y/ D fU .y/ D z0 while gn .x/ D fU .x/ z0
because x X nU . As a consequence, g.x/.n/ D gn .x/ gn .y/ D g.y/.n/ and
hence g.x/ g.y/, i.e., g W X ! Y is a condensation.
T.103. Suppose that A is a proper closed subset of a metric space .X; d / and let
d.x; A/ D inffd.x; a/ W a 2 Ag for any x 2 X . Prove that there exists a family
fUs ; as W s 2 S g (called a Dugundji system for X nA) such that
(i) Us X nA and as 2 A for any s 2 S ;
(ii) fUs W s 2 S g is an open locally finite (in X nA) cover of X nA;
(iii) d.x; as / 2d.x; A/ for any s 2 S and any x 2 Us .
Solution. A family A exp.X / is called inscribed in a family B exp.X / if,
for any A 2 A, there in B 2 B such that A B. As usual, if x 2 X and r > 0,
then B.x; r/ D fy 2 X W d.x; y/ < rg is the open ball of radius r centered at x.
It is evident that r.x/ D d.x; A/ > 0 for any x 2 X nA; it is also immediate
that B.x; r.x// X nA for any x 2 X nA. Thus the family V D fB.x; 14 r.x// W
x 2 X nAg is an open cover of X nA. Since any metrizable space is paracompact
(Problem 218 of [TFS]), there is a locally finite (in X nA) cover U D fUs W s 2 S g
of X nA inscribed in V. Therefore for each s 2 S there is xs 2 X nA such that
Us B.xs ; 14 r.xs //. By definition of d.xs ; A/, there is as 2 A such that d.xs ; as /
5
4 d.xs ; A/.
It is straightforward that the family W D fUs ; as W s 2 S g satisfies the conditions
(i)(ii). To check that W also has (iii), take an arbitrary s 2 S and pick any x 2 Us .
Let r D d.xs ; A/; it follows from x 2 B.xs ; 14 r/ that d.x; A/ 34 r (and hence
3
2
r 2d.x; A/) for otherwise there is a 2 A with d.x; a/ < 34 r, so we have
1 3
r D d.xs ; A/ d.xs ; a/ d.xs ; x/ C d.x; a/ < r C r D r;
4 4
a contradiction. On the other hand, d.x; as / d.x; xs /Cd.xs ; as / 14 r C 45 r D 32 r,

which shows that d.x; as / 32 r 2d.x; A/ so the family W is as promised.
T.104. Let A be a closed subspace of a metrizable space X and suppose that f W
A ! L is a continuous map of A into a locally convex linear topological space L.
Prove that there exists a continuous map F W X ! L such that F jA D f and
F .X / conv.f .A//.
Solution. Let us first develop some technique of handling linear topological spaces
and continuous maps into them. If M is a linear topological space, x 2 M and
P M , then x C P D fx C y W y 2 P g. Given any t 2 R and P M , let
tP D fty W y 2 P g; if P1 ; : : : ; Pn M , then
P1 C

C Pn D fx1 C

C xn W xi 2 Pi for all i D 1; : : : ; ng:
Assume that Z is a space and fbs W s 2 S g is a family of real-valued functions on

Z
P such that S.z/ D fs 2 S W bs .z/ 0g is finite for P
any z 2 Z. Then we denote by
b
s2S s the function b W Z ! R defined by b.z/ D s2S.z/ bs .z/ for each z 2 Z.
Fact 1. Given a linear space M and P M we have :

(1) P conv.P /;
(2) conv.conv.P // D conv.P /;
(3) the set P is convex if and only if conv.P / P and, in particular, conv.P / is a
convex set for any P M .
Proof. Given any x 2 P we have x D 1
x 2 conv.P / which proves (1). Therefore
conv.P / conv.conv.P //. Now assume that u 2 conv.conv.P // and hence there
exist s1 ; : : : ; sk 2 0; 1 with s1 C

C sk D 1 such that u D s1 x1 C

C sk xk
for some x1 ; : : : ; xk 2 conv.P /. For each i k there are ti1 ; : : : ; ti ni 2 0; 1 with
ti1 C

C ti ni D 1 and xi D ti1 yi1 C

C ti ni yi ni for some yi1 ; : : : ; yi ni 2 P .
P P
Therefore u D kiD1 . nj iD1 si tij yij /; it is clear that si tij 2 0; 1 for all i k and
P P P P Pk
j ni . The equality kiD1 . nj iD1 si tij / D kiD1 si . nj iD1 tij / D i D1 si D 1
implies u 2 conv.P / so conv.conv.P // conv.P / and (2) is proved.
Assume that conv.P / P ; then conv.P / D P by the property (1); observe that
tx C .1 t/y 2 conv.P / D P for any x; y 2 P and t 2 0; 1 , so P has to be
convex and we proved sufficiency in (3).
Now, if P is convex, then take any u D t1 x1 C

C tn xn 2 conv.P / such that
t1 ; : : : ; tn 2 0; 1 ; x1 ; : : : ; xn 2 P and t1 C

C tn D 1; we will prove that u 2 P
by induction on n. If n D 1, then t1 D 1 and hence u D x1 2 P .
Assume that n > 1, and,P for any k < n and any s1 ; : : : ; sk 2 0; 1 such that
s1 C

C sk D 1, we have kiD1 si yi 2 P for any y1 ; : : : ; yk 2 P . Observe that
u D .1 tn /z C tn xn where z D 1t t1
x1 C

C 1t
tn1
xn1 2 P by the induction
Pn1 ti n n
hypothesis because i D1 1tn D 1. Therefore u 2 P because P is convex. Since
the point u 2 conv.P / was taken arbitrarily, we established that conv.P / P , so
we settled (3) and hence Fact 1 is proved. t
u
Fact 2. Assume that M is a linear topological space; denote by 0 the zero vector of
M . Given a space Z, a map f W Z ! M is continuous at a point z 2 Z if and only
if for any V 2 .0; M / there is U 2 .z; Z/ such that f .y/ f .z/ 2 V for any
y 2 U.
Proof. If v 2 M , then the map Sv W M ! M defined by Sv .u/ D u C v for all
u 2 M is a homeomorphism (see Fact 1 of S.496) and Sv is the inverse map for Sv .
Assume first that f is continuous at the point z; given any V 2 .0; M / let W D
f .z/ C V D Sf .z/ .V /. Then W 2 .f .z/; M / being W a homeomorphic image of
V under the map Sf .z/ . By continuity of f at the point z there is U 2 .z; Z/ such
that f .U / W ; now, for any y 2 U we have f .y/ 2 W and therefore
f .y/ f .z/ 2 W f .z/ D Sf .z/ .W / D V;
which settles necessity.

Now assume that for any set V 2 .0; M / there is U 2 .z; Z/ such that f .y/
f .z/ 2 V for any y 2 U . Given any W 2 .f .z/; M / observe that we have
V D W f .z/ 2 .0; M /, so there is U 2 .z; Z/ such that f .y/ f .z/ 2 V

for all y 2 U . Consequently, f .y/ 2 f .z/ C V D W for any y 2 U and hence
f .U / W , i.e., f is continuous at the point z. Fact 2 is proved. t
u
Fact 3. Take a space Z and a linear topological space M . Then
(4) if f; g W Z ! M are continuous maps, then the map h D f C g W Z ! M
defined by h.z/ D f .z/ C g.z/ for each z 2 Z is continuous;
(5) if f W Z ! M is a continuous map, then the map m D b
f W Z ! M defined
by m.z/ D b.z/
f .z/ for each z 2 Z is continuous for any b 2 C.Z/;
(6) if f1 ; : : : ; fn W Z ! M are continuous maps and b1 ; : : : ; bn 2 C.Z/, then the
map u D b1 f1 C

C bn fn W Z ! M defined by u.z/ D b1 .z/f1 .z/ C

C
bn .z/fn .z/ for each z 2 Z is continuous.
Proof. Define a map W M M ! M by .p; q/ D p C q for any .p; q/ 2
M M ; let W R M ! M be the map defined by .t; p/ D t
p for each
.t; p/ 2 R M . The maps and are continuous by the definition of the topology
of a linear topological space. If h0 .x/ D .f .x/; g.x// and m0 .x/ D .b.x/; f .x//
for every x 2 X , then the maps h0 D ff; gg W X ! M M and m0 D fb; f g W
X ! R M are also continuous (see Introductory Part of V1.5). Therefore the
maps h D h0 and m D m0 are continuous so we proved (4) and (5). Finally,
(6) is obtained from (4) and (5) by an evident induction. Fact 3 is proved. t
u
Fact 4. Let Z be a paracompact space. Then, for any open cover U D fUs W s 2 S g
of the space Z, there exists a closed locally finite cover fFs W s 2 S g of the space Z
such that Fs Us for all s 2 S .
Proof. By paracompactness of Z, there exists a closed locally finite refinement F
of theS cover U. For each F 2 F there is s.F / 2 S such that F Us.F / . Let
Fs D fF 2 F W s.F / D sg; since every locally finite family is closure-preserving
(Fact 2 of S.221), the set Fs is closed for any s 2 S . If z 2 Z, then z 2 F for some
F 2 F and hence z 2 Fs.F / which shows that F 0 is a cover of Z. It is evident that
Fs Us for each s 2 S ; let us prove that the family F 0 D fFs W s 2 S g is locally
finite.
Take any point z 2 Z and choose a set W 2 .z; Z/ such that the family G D
fF 2 F W W \ F ;g is finite. The set S 0 D fs.F / W F 2 Gg is also finite, and
given any s 2 S nS 0 and any F 2 F with s.F / D s, we have W \ F D ; and
therefore W \ Fs D ;. This shows that W intersects only finitely many elements of
F 0 , so the family F 0 is locally finite and Fact 4 is proved. t
u
Returning to our solution, choose a metric d on X with .d / D .X/ and a
Dugundji system fUs ; as W s 2 S g for X nA (see Problem 103). Apply Fact 4 to find
a locally finite (in X nA) family fFs W s 2 S g such that Fs is closed in X nA and
Fs Us for each s 2 S . By normality of the space X nA, there exists a continuous
function cs W X nA ! 0; 1 such that cs .Fs / f1g and cs ..X nA/nUs / f0g
for every s 2 S . Since the family fUs W s 2 S g is locally finite in X nA, the set
S.x/ D fs 2 S W cs .x/ 0g is finite for every x 2 X nA and hence the function
P
c D s2S cs is well-defined. Given any x 2 X nA, there is W 2 .x; X nA/ such
that S 0 D fs 2 S W W \ UP 0
s ;g is finite; it is evident that S.y/ S for any
y 2 W and hence cjW D . s2S 0 cs /jW is continuous. Thus we can apply Fact 1
of S.472 to conclude that c W X nA ! R is continuous.
Note that for any x 2 X nA, there is s 2 S such that x 2 Fs and hence cs .x/ D 1;
this shows that c.x/ 1 > 0 for any x 2 X nA, so the function bs D ccs is well-
defined. It is immediate that
P
(7) . s2S bs /.x/ D 1 for any x 2 X nA.
P
Now, let F .x/ D f .x/ for any x 2 A and F .x/ D s2S bs .x/f .as / for each
x 2 X nA. Observe that F .x/ makes sense for any x 2 X nA because bs .x/ D 0
for any s 2 S nS.x/. We have FPjA D f by our definition P of F ; besides, for any
x 2 X nA, the point F .x/ D P
s2S b s .x/f .as / D s2S.x/ bs .x/f .as / belongs
to conv.f .A// due to the fact that s2S.x/ bs .x/ D 1 and bs .x/ 2 0; 1 for each
s 2 S.x/. This shows that F .X / conv.f .A//, so we only must prove that F is
continuous. Given any p 2 L let p W X nA ! L be the constant function defined
by p, i.e., p .x/ D p for every x 2 X nA.
Observe first that
0
P set S S , the mapping FS W X nA ! L defined by the equality
(8) for any finite 0
FS 0 .x/ D s2S 0 bs .x/f .as / for each x 2 X nA is continuous

because it is obtained from constant (and hence continuous) maps ff .as / W s 2 S 0 g
by operations described in Fact 3. Given any x 2 X nA, there is W 2 .x; X nA/
such that the set T .x/ D fs 2 S W W \ Us ;g is finite.
P As a consequence, S.y/
T .x/ for every y 2 W which shows that F jW D . s2T .x/ bs f .as / /jW D FT .x/ jW
is continuous because so is the function FT .x/ by the property (7). Therefore F is
continuous at the point x.
Now take any point a 2 A and any W 2 .0; L/. Since L is locally convex,
there is a convex V 2 .0; L/ such that V W . Since f is continuous on A, we
can apply Fact 2 to find > 0 such that a0 2 A and d.a; a0 / < imply f .a0 /
f .a/ 2 V . Let U D B.a; 3 / D fa0 2 X W d.a; a0 / < 3 g. If x 2 U \ A, then
F .x/ F .a/ D f .x/ f .a/ 2 V . Now, if x 2 U nA, then d.x; A/ d.x; a/ < 3 .
Besides, if x 2 Us , then d.as ; a/ d.as ; x/ C d.x; a/ 2d.x; A/ C d.x; a/ <
3 C 3 D (we used the fact that fUs ; as W s 2 S g is a Dugundji system and therefore
2
d.x; as / 2d.x; A/ because x 2 Us ). As a consequence, f .as / f .a/ 2 V for

any s 2 S such that x 2 Us . Let S 0 D fs 2 S W x 2 Us g; then S.x/ S 0 and we
have
X X X
F .x/ F .a/ D bs .x/f .as / f .a/ D bs .x/.f .as / f .a// 2 bs .x/V V I
s2S 0 s2S 0 s2S 0
P
we used the equality s2S 0 bs .x/ D 1, the convexity of V and Fact 1. Thus we
found U 2 .a; X/ such that F .x/ F .a/ 2 V W for any x 2 U , so F is
continuous at the point a by Fact 2. Therefore F has all required properties, so our
T.105. Prove that every metrizable non-separable space can be mapped continu-
ously onto a metrizable space of weight !1 .
Solution. Our solution will be based on the fact that there are sufficiently many
metrizable locally convex spaces.
Fact 1. Given a space X , let d.f; g/ D supfjf .x/ g.x/j W x 2 X g for any
functions f; g 2 C .X /. Then d is a metric on C .X / and M D .C .X /; d / is a
locally convex metrizable linear topological space.
Proof. It is evident that C .X / is a linear space if we consider it with the usual
operations of summing functions and multiplying them by a real number. The
function d is indeed a metric on C .X /: this was proved in Problem 248 of [TFS].
To see that M is locally convex, it suffices to show that the ball Bd .f; "/ D fg 2
C .X / W d.f; g/ < "g is convex for any f 2 C .X / and " > 0. Take any g; h 2
Bd .f; "/ and t 2 0; 1 ; there is > 0 such that maxfd.f; g/; d.f; h/g < < ".
Let p D tg C .1 t/h; then, for any x 2 X , we have jf .x/ p.x/j D jt.f .x/
g.x// C .1 t/.f .x/ h.x/j td.f; g/ C .1 t/d.f; h/ < t C .1 t/ D , so
d.f; p/ < " and therefore p 2 Bd .f; "/. Thus Bd .f; "/ is convex and hence L
is locally convex.
Let s W M M ! M be defined by s.f; g/ D f C g for any f; g 2 M ; define a
function m W R M ! M by m.t; f / D t
f for any t 2 R and f 2 M . To finish
the proof of our fact we must show that s and m are continuous.
Take arbitrary f; g 2 M and any W 2 .f C g; M /; there is " > 0 such that
Bd .f C g; "/ W . The set U D Bd .f; 2" / Bd .g; 2" / is an open neighborhood
of .f; g/ in M M . If .f1 ; g1 / 2 U , then d.f1 ; f / < 2" and d.g1 ; g/ < 2" , so
we can choose < 2" such that maxfd.f; f1 /; d.g; g1 /g < . Given any x 2 X
we have j.f1 C g1 /.x/ .f C g/.x/j jf .x/ f1 .x/j C jg.x/ g1 .x/j
d.f; f1 / C d.g; g1 / < 2. Thus d.f1 C g1 ; f C g/ 2 < " and hence f1 C
g1 D s.f1 ; g1 / 2 Bd .f C g; "/; the point .f1 ; g1 / 2 U was taken arbitrarily,
so s.U / Bd .f C g; "/ W which proves continuity of s at the point .f; g/.
Therefore s W M M ! M is continuous.
To prove continuity of m take any .t; f / 2 R M and any W 2 .tf; M /; there
is " > 0 such that Bd .tf; "/ W . Since f is a bounded function, there is r > 0
such that jf .x/j < r for any x 2 X . Take any > 0 such that < minf1; 3.rC1/ "
g
and jtj < 3 . The set U D .t ; t C / Bd .f; / is an open neighborhood of
"
.t; f / in the space R M . Take any .t1 ; f1 / 2 U ; then d.f; f1 / < , and hence, for
any x 2 X , we have jf1 .x/ f .x/j < whence jf1 .x/j < jf .x/jC < r C1. Thus
jf1 .x/j < r C 1 for any x 2 X . Furthermore, jt1 f1 .x/ tf .x/j D j.t1 t/f1 .x/ C
t.f1 .x/ f .x//j jt1 tjjf1 .x/jCjtjjf1 .x/ f .x/j .r C1/Cjtj < 3" C 3" D 2" 3 .
As a consequence, d.t1 f1 ; tf / 2" 3
< " so m.t 1 ; f1 / D t1 f1 2 Bd .tf; "/ W .
The point .t1 ; f1 / 2 U was chosen arbitrarily, so we proved that m.U / W and
hence m is continuous at the point .t; f /. Fact 1 is proved. t
u
Fact 2. For any infinite cardinal , there exists a metrizable locally convex space of
weight .
Proof. Recall that D./ is a discrete space of cardinality ; let Fin.D.// be the
family of all finite subsets of D./. Consider the space L D ff 2 C.D.// W
D./nf 1 .0/ is finiteg. In other words, L consists of functions on D./ which take
nonzero values at only finitely many points of D./. It is evident that L is a linear
subspace of a metrizable locally convex space C .D.// (see Fact 1). Therefore
L is a locally convex metrizable space. Let P D ff 2 L W f .x/ 2 Q for any
x 2 D./g. Given a finite A D./ the set PA D ff 2 P W f .D./nA/ D f0gg
is countable because theSmap f ! f jA maps PA injectively onto QA which is
countable. Thus jP j D j fPA W A 2 Fin.D.//gj !
jFin.D.//j D
! D .
Furthermore, the set P is dense in L; to see this take any f 2 L and any " > 0.
The set S D f 1 .Rnf0g/ is finite, so we can choose a rational number qs such that
jf .s/ qs j < 2" for each s 2 S . Let g.s/ D qs for each s 2 S ; if t 2 D./nS , then
let g.t/ D 0. Then g 2 P and jg.x/ f .x/j < 2" for every x 2 D./. Therefore
d.f; g/ 2" < ", so Bd .f; "/ \ P ; which proves that P is dense in L. As a
consequence, w.L/ D d.L/ jP j .
On the other hand, for any x 2 D./, let fx .x/ D 1 and fx .y/ D 0 for all y 2
D./nfxg. The subspace D D ffx W x 2 D./g of the space L is discrete because
Bd .fx ; 1/ \ D D ffx g for each x 2 D./. Therefore w.L/ s.L/ jDj D
which shows that w.L/ D and Fact 2 is proved. t
u
Returning to our solution, take any non-separable metric space X . Then
ext.X / D w.X / D d.X / !1 (see Problem 214 of [TFS]) and hence there
exists a closed discrete D X with jDj D !1 . Apply Fact 2 to take a metrizable
locally convex space L with w.L/ D !1 . Then ext.L/ D !1 , so we can find a
discrete E L. Let f W D ! E be any bijection. Then f is a continuous map
from D to L because D is discrete. By Problem 104, there exists a continuous
map F W X ! L such that F jD D f . If Y D F .X /, then E Y and hence
w.Y / s.Y / jEj D !1 so w.Y / D !1 . Thus F maps X continuously onto a
metric space Y with w.Y / D !1 , so our solution is complete.
T.106. Prove that a metrizable space is !-stable if and only if it is separable.
Solution. Let X be a metrizable space with w.X / > !. There exists a continuous
onto map f W X ! Y of the space X onto a metrizable space Y with w.Y / D !1
(see Problem 105). Since w.Y / c, there exists a condensation g W Y ! M such
that w.M / D ! (see Problem 102). If X is stable, then w.Y / D nw.Y / D ! which
is a contradiction. Thus no metrizable space of uncountable weight is stable, i.e.,
stability of a metrizable space implies its separability.
Now, if X is a metrizable separable space, then w.X / D ! and hence nw.Y / !
whenever Y is a continuous image of X . Thus X is !-stable.
T.107. Prove that any -product of spaces with a countable network is monolithic.
In particular, any -product of second countable spaces is monolithic.
Let Xt be a space with a countable network for each t 2 T . Fix any point
Solution. Q
a 2 X D fXt W t 2 T g; given any xQ2 X , let supp.x/ D ft 2 T W x.t/ a.t/g.
For any A T , let pA W X ! XA D fXt W t 2 Ag be the natural projection onto
the face XA . We must prove that the space D .X; a/ D fx 2 X W jsupp.t/j

!g is -monolithic for any infinite cardinal . S
Take any Y .X; a/ such that jY j and let S D fsupp.y/ W y 2 Y g; it
is clear that jS j . We have the equality y.t/ D a.t/ for any t 2 T nS and hence
Y F D fpT nS .a/g XS . It is an easy exercise that F is a closed subspace of X
and nw.F / (observe that F is homeomorphic to XS and use the fact that the
product of -many spaces with a countable network has the network weight ).
Therefore cl .Y / clX .Y / F and nw.cl .Y // nw.F / which proves
that .X; a/ is -monolithic.
T.108. Prove that any space X is -stable for any nw.X /. In particular, any
space with a countable network is stable.
Solution. If D nw.X / and Y is a continuous image of X , then nw.Y / . Thus
we have nw.Y / (the condensations of Y are irrelevant) which proves that
X is -stable.
T.109. Prove that any product of spaces with countable network is stable. In
particular, any product of second countable spaces is stable.
Solution. We will need the following general factorization theorem.
Fact 1.Q Suppose that Xt is a space such that nw.Xt / D ! for everyQt 2 T and let
X D fXt W t 2 T g; given any A T , the map pA W X ! XA D fXt W t 2 Ag
is the natural projection of X onto the face XA defined by pA .x/ D xjA for every
x 2 X . Suppose that Y is a dense subspace of the space X and f W Y ! M is
a continuous map of Y onto a space M such that w.M / D !. Then there
is a set S T and a continuous map g W pS .Y / ! M such that jS j and
f D g .pS jY /.
Proof. We will also need projections between faces of X , namely, if A B T ,
then we have a natural projection pAB W XB ! XA defined by pAB .x/ D xjA for any
x 2 XB . The map pAB is open, continuous and pAB PB D pA for all A B T .
Since w.M / , there is an embedding of M into R , so we can assume,
without loss of generality, that M R . For every < , the map W R ! R is
the natural projection of R onto its th factor; let q D jM . Given any < , we
can apply Problem 299 of [TFS] to the map f D q f W Y ! R to find a countable
set S T and a continuous
S map g W pS .Y / ! R such that f D g .pS jY /.
We claim that S D fS W < g is as promised.
Indeed, given z 2 pS .Y /, let g.z/./ D g .zjS / for each < ; this defines a
map g W pS .Y / ! R . It is immediate that g D fg .pSS jpS .Y // W < g and
hence the map g is continuous (see the introductory part of Sect. 1.4). Given any
y 2 Y , we have
f .y/./ D q .f .y// D f .y/ D g .pS .y// D g .pSS .pS .y/// D g.pS .y//./
for every < which shows that f .y/ D g.pS .y// and hence g.pS .y// 2 M for
each y 2 Y . As a consequence, g W pS .Y / ! M and f D g .pS jY /, so Fact 1 is
proved. t
u
Qthat nw.Xt / D !
Returning to our solution, fix any infinite cardinal ; suppose
for all t 2 T and take any continuous onto map f W X D t 2T Xt ! Y for
which there is a condensation g W Y ! M such that w.M / . If h D g f ,
then h W X ! M , so we can apply Fact Q 1 to find S T such that jS j , and
there is a continuous map b W XS D t 2S Xt ! M with b pS D h. For the
map d D g 1 b, we have d W XS ! Y and d pS D f . Observe also that for
any U 2 .Y /, the set d 1 .U / D pS .f 1 .U // is open in XS because f 1 .U / is
open in X and the map pS is open (see Problem 107 of [TFS]). Consequently, d is
a continuous map and hence nw.Y / nw.XS / (see Problem 157 of [TFS] and
observe that any product of -many spaces with a countable network has network
weight ). Therefore X is -stable and our solution is complete.
T.110. Prove that any -product of spaces with countable network is stable. In
particular, any -product of second countable spaces is stable.
Q
Solution. Given a product X D t 2T Xt , we know that natural projections of X
onto its faces are open maps (see Problem 107 of [TFS]). It turns out that the same
is true for certain subspaces of X .
Q
Fact 1. Let Xt be a space for each t 2 T ; given a point x 2 X D fXt W t 2 T g,
let .X; x/ D fyQ2 X W jft 2 T W x.t/ y.t/gj < !g. The natural projection
pS W X ! XS D t 2S Xt is defined by pS .x/ D xjS for any x 2 X . Suppose that
Y X and .X; x/ Y for any x 2 Y . Then the map pS jY W Y ! pS .Y / is open
for any S T .
Proof. The map pS is continuous (see Problem 107 of [TFS]), so p D pS jY is
also continuous. It is clear that p is surjective, so we only must show that p.U / is
open in Z D pS .Y / for any U 2 .Y /. It suffices to find Qa base B in Y such that
p.V / 2 .Z/ for any V 2 B. Let C D fU X W U D t 2T Ut W Ut 2 .Xt / for
all t 2 T and the set supp.U / D ft 2 T W Ut Xt g is finiteg. The family C is a base
of X (see Q Problem 101 of [TFS]), so B D fU \ Y W U 2 Cg is a base Q in Y . Take
any U D t 2T Ut 2 C and let V D U \ Y . Observe that the set US D t 2S Ut is
open in XS , so it suffices to show that pS .V / D US \ Z. Of course, there is no loss
of generality to assume that U ;.
Take any z 2 US \ Z and fix a point y 2 Y such that pS .y/ D z. We have
y.t/ D z.t/ 2 Ut for any t 2 S . Choose at 2 Ut for any t 2 supp.U /nS and define
u 2 X as follows: ujS D z; u.t/ D at for any t 2 supp.U /nS and u.t/ D y.t/ for
any t 2 T n.S [ supp.U //. Then the set A D ft 2 T W u.t/ y.t/g is finite, so
u 2 .X; y/ Y . Besides, u.t/ 2 Ut for all t 2 T and therefore u 2 U \ Y D V .
It is clear that S .u/ D z and hence z 2 pS .V /. This proves that US \ Z pS .V /.
Now, if y 2 V , then y.t/ 2 Ut for all t 2 T and hence pS .y/.t/ D y.t/ 2 Ut
for all t 2 S . Consequently, pS .y/ 2 US \ Z so pS .V / US \ Z. We proved that
pS .V / D US \ Z is an open subset of Z so Fact 1 is proved. t
u
Q to our solution assume that nw.Xt / D ! for every t 2 T and take any
Returning
a 2 X D t 2T Xt ; we must prove that the space
D .X; a/ D fx 2 X W jft 2 T W x.t/ a.t/gj !g
is -stable for any infinite cardinal . It is an easy exercise that is dense in X and
.X; x/ for any x 2 . Take any continuous surjective map f W ! Y such
that there is a condensation g W Y ! M of Y onto a space M with w.M / . We
can apply Fact 1 of T.109 to the map h D g f to find S T such that jS j
and there exists a continuous map b W pS ./ ! M such that h D b .pS j/.
For the map d D g 1 b we have d W pS ./ ! Y and d .pS j/ D f . Given
any U 2 .Y / the set d 1 .U / D pS .f 1 .U // is open in S D pS ./ because
f 1 .U / is open in and pS j is an open map by Fact 1. Consequently, d is a
continuous map.
Since S XS , we have nw.S / nw.XS / (it is an easy exercise to see
that a product of -many spaces with countable network has network weight ).
Since Y is a continuous image of S (under the mapping d ), we have nw.Y /
nw.S / (see Problem 157 of [TFS]) which proves that is -stable, so our
T.111. Prove that any -product of spaces with countable network is stable. In
particular, any -product of second countable spaces is stable.
Solution. Assume
Q that we have nw.Xt / D ! for every t 2 T . Given any point
x 2 X D t 2T Xt , let .X; x/ D fy 2 X W jft 2 T W y.t/ x.t/gj < !g and
fix any point a 2 X ; we must prove that the space D .X; a/ is -stable
Q for any
infinite cardinal . Given any S T the map pS W X ! XS D t 2S Xt is the
natural projection defined by pS .x/ D xjS for any x 2 X . It is an easy exercise
that is dense in X and .X; x/ for any x 2 .
Take any continuous surjective map f W ! Y such that there is a condensation
g W Y ! M of Y onto a space M with w.M / . We can apply Fact 1 of T.109 to
the map h D g f to find S T such that jS j and there exists a continuous
map b W pS ./ ! M such that h D b .pS j/.
For the map d D g 1 b we have d W pS ./ ! Y and d .pS j/ D f . Given
any U 2 .Y / the set d 1 .U / D pS .f 1 .U // is open in S D pS ./ because
f 1 .U / is open in and pS j is an open map by Fact 1 of T.110. Consequently,
d is a continuous map. Since S XS , we have nw.S / nw.XS / (it is an
easy exercise to see that a product of -many spaces with countable network has
network weight ). Since Y is a continuous image of S (under the mapping d ),
we have nw.Y / nw.S / (see Problem 157 of [TFS]) which proves that is
-stable so our solution is complete.
T.112. Prove that any -product of Lindelf P -spaces is !-stable.
Solution. Our solution needs some insight into the products and -products of
Lindelf P -spaces.
Fact 1. If X and Y are Lindelf P -spaces, then X Y is a Lindelf P -space.

Therefore any finite product of Lindelf P -spaces is a Lindelf P -space.
Proof. LetT P be a G -set in X Y ; fix a family O D fOn W n 2 !g .X Y / such
that P D O. If P D ;, then there is nothing to prove; if not, take an arbitrary
z D .x; y/ 2 P . It is easy to construct families fUn W n 2 !g .x; X/ and
fVn W n 2 !g .y; Y / such T that Un Vn On for every n T 2 !. Since X and Y are
P -spaces, we have U D fUn W n 2 !g 2 .X/ and V D fVn W n 2 !g 2 .Y /,
so W D U V 2 .z; X Y / and W P . Therefore every z 2 P is in the interior
of P , i.e., P is open in X Y . Consequently, X Y is a P -space.
To see that X Y is Lindelf, take any open cover W of the space X Y ; let
pX W X Y ! X and pY W X Y ! Y be the respective natural projections. Given
any set A X Y let AX D pX .A/ and AY D pY .A/. We can assume that the
elements of W belong to the standard base of the space X Y , i.e., W D WX WY
for any W 2 W. For each x 2 X the space Yx S D fxg Y is homeomorphic
T to Y , so
there is a countable Wx W such that Yx Wx . Let Ux D fWX W W 2 Wx g
for every x 2 X . The set Ux is open in X because X is a P -space;
S by the Lindelf
property of X , S there is a countable A X such that X D fUx W x 2 Ag. The
family W 0 D fWx W x 2 Ag W is countable, so it suffices to show that
S
W0 D X Y .
S Take any z D .a; b/ 2 X Y ; there is x 2 X such that a 2 Ux . Since Yx
Wx , we can find W 2 Wx such that .x; b/ 2 W and hence b 2 WY . By the
definition of Ux , we have WX Ux 3 a so a 2 WX and therefore .a; b/ 2 W 2 W 0 .
The point z 2 X Y was chosen arbitrarily, so W 0 is a countable subcover of W
and hence X Y is Lindelf.
Finally, the same result for any finite number of Lindelf P -spaces can be
obtained from what we proved for two factors by a trivial induction so Fact 1 is
proved. t
u
S
Fact 2. Suppose that X D fXn W n 2 !g where, for each n 2 !, the space Xn is
-stable for some infinite cardinal . Then X is -stable.
Proof. Take any continuous onto map f W X ! Y such that there is a condensation
g W Y ! M of the space Y onto some space M with w.M / . The space
Yn D f .Xn / is a continuous image of Xn and gjYn is a condensation of Yn onto a
S for every n 2 !. Consequently, nw.Yn / for any n 2 !;
space of weight
since Y D fYn W n 2 !g, we have nw.Y / , so X is -stable and Fact 2 is
proved. t
u
Fact 3. Every countable -product of Lindelf P -spaces is !-stable.
Proof. Let us check first that
. / any Lindelf P -space is !-stable.
To prove . / assume that Z is a Lindelf P -space and take any continuous onto
map f W Z ! Y such that there is a condensation g W Y ! M of Y onto some
space M such that w.M / D !. Every point of Y is a G -set in Y because so is
every point of M . Thus f 1 .y/ is a G -set in Z and hence an open subset of Z

for any y 2 Y . The family U D ff 1 .y/ W y 2 Y g is a disjoint S open cover of
Z; the space Z being Lindelf there is a countable U 0 U with U 0 D Z. This
shows that there is a countable A Y such that f 1 .A/ D Z which implies that
Y D f .Z/ D A is countable. Therefore nw.Y / jY j D ! so Z is !-stable and
. / is proved.
Q that Zn is a Lindelf P -space for every n 2 ! and choose any point
Now assume
a 2 Z D fZn W n 2 !g. Given any z 2 Z let supp.z/ D fn 2 ! W z.n/ a.n/g.
We must prove that the space .Z; a/ D fz 2 Z W jsupp.z/j < !g is !-stable.
Consider the set n .Z; a/ D fz
S2 Z W z.m/ D a.m/ for all m ng for every n 2 !.
It is evident that .Z; a/ D fn .Z; a/ W n 2 !g. It is also straightforward that
n .Z; a/ is homeomorphic to Z0

Zn1 so n .Z; a/ is a Lindelf P -space by
Fact 1. The property . / shows that n .Z; a/ is !-stable for every n 2 ! and hence
.Z; a/ is !-stable by Fact 2. Fact 3 is proved. t
u
Fact 4. Every -product of Lindelf P -spaces is Lindelf.
Proof.QAssume that Xt is a Lindelf P -space for every t 2 T ; take any point a 2
X D fXt W t 2 T g and let supp.x/ D ft 2 T W x.t/ a.t/g for every x 2 X . We
must prove that theSspace D .X; a/ D fx 2 X W jsupp.x/j < !g is Lindelf.
Observe that D fn W n 2 !g where n D fx 2 W jsupp.x/j ng for every
n 2 !. Therefore it suffices to show that n is Lindelf for every n 2 !. We will
do this by induction on n. If n D 0, then n D fag is a Lindelf space; assume that
n > 0 and we proved that n1 is Lindelf. Q
For any A T the map pA W X ! XA D t 2A Xt is the natural projection of
X onto its face XQ A defined by pA .x/ D xjA for each x 2 X . Call a set U X
standard if U D t 2T Ut where Ut 2 .Xt / for each t 2 T and the set supp.U / D
ft 2 T W Ut Xt g is finite. The family B of all standard sets is a base in X (see
Problem 101 of [TFS]), so toSprove that n is Lindelf, it suffices to show that
S for
every U B such that n U there is a countable U 0 U such that n SU 0 .
By the induction
S hypothesis, there is a countable V U such that n1 V.
The set A D fsupp.V / W V 2 Vg is countable; we claim that
S
. / supp.x/ A for any x 2 n n. V/.
S
Indeed, if . / is not true, then t0 2 supp.x/nA for some x 2 n n. V/. Let
y.t/ D x.t/ for all t 2 T nft0 g and y.t0 / D a.t0 /. Then y 2 n1 and hence y 2 V
for some V 2 V. We have supp.V / A and therefore pA1 ..pA .V // D V ; since
pA .x/ D pA .y/ and y 2 V , we have x 2 V which is a contradiction showing that
. / is true. S
It follows from . / that P D n n. V/ Q D n \ .XA fpT nA .a/g/.
It is straightforward that the mapping pA jQ is a homeomorphism of Q onto the
S D fy 2 XA W jft 2 T W y.t/ a.t/gj ng. Furthermore
space n .XA ; pA .a//
n .XA ; pA .a// D fRB W B A and jBj ng where RB D XB fpAnB .a/g
is homeomorphic to XB for any B A. The space RB is Lindelf being a finite
product of Lindelf P -spaces (see Fact 1). Any countable union of Lindelf spaces
is a Lindelf space, so n .XA ; pA .a// is Lindelf and hence Q is also Lindelf

being homeomorphic to n .XA ; pA .a//. Since P is a closed subset of n , it is also
closed in Q
Sand hence Lindelf. Thus we can choose a countable W U such
that PS W. It is clear that U 0 D V [ W is a countable subfamily of U and
n U 0 . TheS inductive step being carried out, we proved that every n is Lindelf
and hence D n is also Lindelf so Fact 4 is proved. t
u
We are finally ready to present our solution. Assume
Q that Xt is a Lindelf P -
space for every t 2 T and take any point a 2 X D fXt W t 2 T g. Let supp.x/ D
ft 2 T W x.t/ a.t/g for every x 2 X . We must prove that the space D
.X; a/ D fx 2 X W jsupp.x/j < !g is !-stable, so take any map f W ! Y
such that there is a condensation g W Y ! M of Y onto some space M with
w.M / D !. We know that is Lindelf (see Fact 4), so apply Problem 298 of
[TFS] to the map h D g f W ! M to obtain a countable S T and a
continuous map b W pS ./ ! M such that b pS D h. For the map d D g 1 b
we have d W pS ./ ! Y and f D d .pS j/. Given any U 2 .Y / the set
d 1 .U / D pS .f 1 .U // is open in S D pS ./ because f 1 .U / is open in
and pS j is an open map by Fact 1 of T.110. Therefore the map d is continuous.
We must also note that pS ./ is a countable -product of Lindelf P -spaces and
therefore pS ./ is !-stable by Fact 3. Since Y is a continuous image (under the
map d ) of an !-stable space S , we have nw.Y / ! because Y condenses onto a
second countable space M . This shows that is !-stable and hence our solution is
complete.
T.113. Prove that for any cardinal , if X is -monolithic, then any Y X is
-monolithic.
Solution. Take any A Y with jAj . Then nw.clY .A// nw.clX .A// .
Thus nw.clY .A// for any A Y with jAj so Y is -monolithic.
T.114. Let be an infinite Q cardinal. Suppose that X is a -monolithic space for
each < . Prove that fX W < g is also -monolithic. In particular, any
finite product of -monolithic spaces is -monolithic.
Q
Solution. For the space X D fX W < g the map p W X ! X is the natural
projection for every < . Take any A X Q with jAj ; then A D p .A/ X
and jA j . It is straightforward that A < A .
The space X being -monolithic, we have nw.A / for each < . We
Q -many spaces of
leave it as an easy exercise for the reader that any product of
network weight Q has network weight . Therefore nw. < A / whence
nw.A/ nw. < A / which shows that X is -monolithic.
T.115. Prove that if a space X is covered with a locally finite family of closed
monolithic subspaces, then X is monolithic.
Solution. Let F be a locally finite closed cover of X such that every F 2 F is
monolithic. Fix an infinite cardinal and any set A X with jAj . Each x 2 X
belongs to finitely many elements of F , so there is F 0 F such that jF 0 j
S 0
and A F . The family A D fA \ F W F 2 F 0 g is locally finite and hence
S S
closure-preserving (see Fact 2 of S.221), so A D A D fA \ F W F 2 F 0 g.
We have nw.A \ F / for any F 2 F 0 because A \ F F and F is
monolithic. Since jF 0 j , we have represented A as a union of -many spaces
of network weight . It is evident that any union of -many spaces of network
weight has network weight so nw.A/ . We proved that for any A X
with jAj , we have nw.A/ , i.e., X is -monolithic. The cardinal was
chosen arbitrarily so X is monolithic.
T.116. Give an example of a space which is not !-monolithic being a union of two
monolithic subspaces.
Solution. The Mrowka space M constructed in Problem 142 of [TFS] can be
represented as M D ! [ M where ! is dense in M and all points of ! are isolated
in M while the set M is uncountable and discrete in M . Thus M is a union of
two discrete (and hence monolithic) subspaces; since nw.!/ D nw.M / s.M /
jMj > !, the space M is not !-monolithic.
T.117. Show that R!1 is not !-monolithic. Hence the product of uncountably many
monolithic spaces can fail to be !-monolithic.
Solution. There exists a countable set A R!1 such that A D R!1 (see
Problem 108 of [TFS]). The space A.!1 / embeds in I!1 R!1 (see Problems 126
and 129 of [TFS]). As a consequence, nw.A/ D nw.R!1 / s.R!1 / s.A.!1 // D
!1 and hence R!1 is not !-monolithic.
T.118. Prove that any compact space is stable.
Solution. Let X be a compact space. Given an infinite cardinal assume that we
have a continuous onto map f W X ! Y such that there is a condensation g W Y !
M of Y onto some space M with w.M / . The space Y is also compact; any
condensation of a compact space is a homeomorphism (see Problem 123 of [TFS]),
so Y is homeomorphic to M and therefore nw.Y / w.Y / D w.M / D . This
shows that X is -stable for any infinite cardinal .
T.119. Prove that any pseudocompact space is !-stable.
Solution. Let X be a pseudocompact space and take any continuous onto map f W
X ! Y such that there is a condensation g W Y ! M of the space Y onto a second
countable space M . The space Y is also pseudocompact; since any condensation
of a pseudocompact space onto a second countable space is a homeomorphism (see
Problem 140 of [TFS]), the space Y is homeomorphic to M and therefore nw.Y /
w.Y / D w.M / D !. Thus X is !-stable.
T.120. Show that any compact !-monolithic space of countable tightness must be
FrchetUrysohn. Give an example of a monolithic compact space of uncountable
tightness.
Solution. Suppose that X is a compact !-monolithic space with t.X / D !. Given

any A X and any x 2 A, there is a countable B A such that x 2 B. By
!-monolithity of X , we have nw.B/ D ! and hence w.B/ D ! because B is
compact (see Fact 4 of S.307). Thus B is a FrchetUrysohn space, so there is a
sequence fxn W n 2 !g B such that xn ! x. Since we also have fxn W n 2 !g
A, we found a sequence in A which converges to x. Therefore X is a Frchet
Urysohn space.
Now let K be the set !1 C 1 with the interval topology. Then K is compact (see
Problem 314 of [TFS]); if x D !1 and A D f W < !1 g, then x 2 A. If B A is
countable, then there is < !1 such that B 0; D f W g. Since the set
0; is closed and countable, we have B 0; which shows that x B for any
countable B A. Therefore t.K/ > !.
Observe that w.K/ D !1 , so if P K and jP j D !1 , then nw.P / w.K/ D
!1 , so K is !1 -monolithic. If P K is countable, then there exists < !1 such
that P 0; [ f!1 g. Since 0; [ f!1 g is a countable closed set, the set P is also
countable and therefore K is !-monolithic. The space K has no subsets of infinite
cardinalities other than ! and !1 , so we established that K is monolithic.
T.121. Let f W X ! Y be a closed continuous onto map. Prove that if X is
-monolithic, then so is Y .
Solution. Take an arbitrary A Y with jAj and choose any B X such that
jBj and f .B/ D A. The map f is closed so f .B/ D A. We have nw.B/
because X is -monolithic. Therefore nw.A/ nw.B/ (see Problem 157 of
[TFS]) and hence Y is -monolithic.
T.122. Give an example showing that a continuous image of a monolithic space is
not necessarily !-monolithic.
Solution. Let X be the set R!1 with the discrete topology. If f W X ! R!1 is
the identity map and R!1 is considered with its natural product topology, then f
is a continuous map because X is discrete. Any discrete space is metrizable and
hence monolithic (see Problem 101). Thus the space R!1 is a continuous image of
a monolithic space X , so we only have to note that R!1 is not !-monolithic by
Problem 117.
T.123. Given an infinite cardinal , prove that any continuous image of a -stable
space is -stable. In particular, any retract of a -stable space is -stable.
Solution. Suppose that X is a -stable space and f W X ! Y is a continuous
onto map. Take any continuous map g W Y ! Z such that there is a condensation
h W Z ! M of the space Z onto a space M with w.M / . The map g f maps
X onto Z, so nw.Z/ because X is -stable. This shows that Y is also -stable.
S
T.124. Suppose that X D fX W < g and X is a -stable space for any
< . Prove that X is -stable. In particular, any space which is a countable union
of stable subspaces is stable.
Solution. Take any continuous onto map f W X ! Y such that there is a

condensation g W Y ! M of the space Y onto some space M with w.M / .
For every < the space Y D f .X / is a continuous image of X and gjY is a
onto a space of weight . Consequently, nw.Y / for each
condensation of YS
< ; since Y D fY W < g, we have nw.Y / so X is -stable.
T.125. Give an example of an !-stable space X such that some closed subspace of
X is not !-stable.
Solution. Any pseudocompact space is !-stable by Problem 119. The Mrowka
space M constructed in Problem 142 of [TFS] is pseudocompact and hence
!-stable. On the other hand there is an uncountable closed discrete D M . Since
jDj jM j c, we have w.D/ c. Since every discrete space is metrizable,
the space D can be condensed onto a second countable space (see Problem 102);
we have nw.D/ D jDj > ! which shows that D is not !-stable. Therefore D is a
closed non-!-stable subset of an !-stable space M .
T.126. Give an example of an !-stable space whose square is not !-stable.
Solution. Given a set B denote by P! .B/ the family of all countably infinite subsets
of B.
Fact 1. Let D be a discrete space of cardinality !1 . Then there exist countably
compact subspaces X and Y of the space D such that D D X \ Y and hence
X Y contains an uncountable clopen discrete subspace.
Proof. Given any set A D denote by Ad the set of all accumulation points of A
in D. Let X0 D D; assume that < !1 and we have constructed sets fX W < g
(1) X D for each < ;
(2) X0 D D and X X
whenever <
< ;
(3) jX j c for each < ;
(4) if <
< and A 2 P! .X /, then Ad \ X
;.
S
Consider the set Y D fX W < g; it is evident that jY j c and therefore
jP! .Y /j c! D c. Given any A 2 P! .Y /, the set Ad is non-empty because
D is a compact space. Choose a point xA 2 Ad for each A 2 P! .Y / and let
X D Y [ fxA W A 2 P! .Y /g. It is immediate that the properties (1)(4) hold for
the family fX W g, so our inductive construction can be continued to obtain a
family fX W < !S 1 g for which the conditions (1)(4) are satisfied for each < !1 .
The space X D fX W < !1 g has cardinality c because
jX j .supfjX j W < !1 g/
!1 c
!1 D c
[we applied the property (3)]. Furthermore, X is countably compact. Indeed, if A

P! .X /, then A X for some < !1 and hence there is x 2 Ad \ XC1 by the
property (4). It is clear that x is an accumulation point for A in X ; this proves that
every infinite subset of X has an accumulation point in X and hence X is countably
compact (see Problem 132 of [TFS]).
We claim that the space Y D D [ .DnX / is also countably compact. To prove

it take any countably infinite A Y ; there is an infinite discrete B A (see
Fact 4 of S.382). Apply Fact 2 of S.382 and Fact 2 of S.286 to conclude that B
is homeomorphic to ! and hence jBj D 2c by Problem 368 of [TFS]. Therefore
jB d j D 2c , so it is impossible that B d X because jX j c. Thus there is a point
y 2 B d \ Y ; it is clear that y is an accumulation point for A in Y . This proves that
every infinite subset of Y has an accumulation point in Y and hence Y is countably
compact.
Consider the space X Y D D. The diagonal D f.x; x/ W x 2 Dg is
closed in D and hence E D \.X Y / is closed in X Y . Now, if .x; x/ 2 X Y ,
then x 2 X \ Y D D and therefore E D f.x; x/ W x 2 Dg. Every x 2 D is
isolated in D, so .x; x/ is an isolated point of D D which implies that .x; x/
is isolated in X Y for any x 2 D. This shows that E is an uncountable clopen
discrete subspace of X Y so Fact 1 is proved. t
u
Returning to our solution observe that
(5) for any space Z if U ; is a clopen subset of Z, then U is a retract (and hence
a continuous image) of Z.
Indeed, choose any u 2 U ; let f .x/ D x for every x 2 U and f .x/ D u for each
x 2 ZnU . It is a trivial exercise that f W Z ! U is a retraction.
Consider the space K D X Y where X and Y are the countably compact
spaces from Fact 1; it is evident that K is countably compact. Observe that X Y is
a clopen subspace of K K and hence there is an uncountable clopen discrete
U K K by Fact 1. The space K is !-stable by Problem 119. If K K
is !-stable, then the clopen set U is also !-stable being a continuous image of
K K by (5) and Problem 123. The space U is discrete and hence metrizable, so
!-stability of U implies that U is separable (see Problem 106) and hence countable.
This contradiction shows that K K is not !-stable, so our solution is complete.
T.127. Prove that any Lindelf P -space is !-simple.
Solution. Let X be a Lindelf P -space. Assume that M is a second countable space
and f W X ! M is a continuous onto map. Since every point x 2 M is a G -set in
M , the set f 1 .x/ is a G -set in X ; since X is a P -space, the set f 1 .x/ is open
in X for any x 2 M . Therefore ff 1 .x/ W x 2 M g is a disjoint open cover of X .
The
S space X being Lindelf, we can choose a countable A M such that X D
ff 1 .x/ W x 2 Ag D f 1 .A/. Consequently, M D f .X / D f .f 1 .A// D A
and therefore M D A is a countable set. Thus X is !-simple.
T.128. Prove that for any scattered space X with l.X / , the Lindelf degree
of the -modification of X does not exceed . In particular, if X is a Lindelf (or
compact!) scattered space, then the !-modification of X is Lindelf.
Solution. Let D .X/ and denote by the topology of -modification of .

Given any B X let jB D S fU \ B W U 2 g, i.e., jB is the topology induced by
on B. Fix any U with U D X ; we must prove that X can be covered by a
subfamily of U of cardinality .
set O D fx 2 X W there is U 0 U and Ox 2 such that jU 0 j

Consider theS
and x 2 Ox U 0 g. If a set F is closed in .X; / and F O, then the Lindelf
number of .F; jF /Sdoes not exceed , so we can choose a set A F such that
jAj and F SfOx W x 2 Ag. For each Sx 2 A there is Ux U such that
jUx j and Ox Ux . The family U 0 D fUx W x 2 Ag U has cardinality
and covers F . Thus we have
0 0
S in .X; / and F O, then there is U U such that jU j
(1) If F is closed
and F U 0 . In particular, if O D X , then we can take F D X to conclude
that X can be covered by a subfamily of U of cardinality .
The property (1) shows that it suffices to establish that O D X . Assume the
contrary; since X is scattered, there exists a 2 F D X nO which is isolated in F .
Consequently, there is U 2 such that a 2 U and cl .U / \ F D fag. Pick any
P 2 U such that a 2 P ; since G -subsets T of X form a base in .X; / there is a
family V such that jVj and a 2 V P . Observe that FV D cl .U /nV
is a closed subset of .X; / and FV O for each V 2 V. Therefore S we can apply
(1) to find a family WVS U such that jWV j and FV WV for every
V 2 V. It is clear
S that fFV W V 2 Vg covers the set cl .U /nP and hence the
familySU 0 D fWV W V 2 Vg [ fP g U covers the set cl .U /. This implies
U U 0 which means that we have found a neighborhood of the point a in .X; /
which is covered by a subfamily of U of cardinality , i.e., a 2 O which is a
contradiction. Thus O D X and hence S (1) can be applied to conclude that there is
U1 U such that jU1 j and X D U1 . Since a -open cover of X was chosen
arbitrarily, we proved that l.X; / .
T.129. Prove that any Lindelf scattered space (and, in particular, any compact
scattered space) is a simple space.
Solution. Let X be a Lindelf scattered space and fix an infinite cardinal . Take
any continuous onto map f W X ! Y such that w.Y / . Denote by the
-modification of the topology .X/; denote by X 0 the space .X; /. We have
l.X / D ! , so Problem 128 can be applied to conclude that l.X 0 / . Since
.X/ , the identity map i W X 0 ! X is continuous and hence g D f i is a
continuous map of X 0 onto Y .
Observe that .Y / w.Y / and therefore every y 2 Y is a G -set in Y . As a
consequence, g 1 .y/ is open in X 0 being a G -subset of X 0 (see Fact 2 of S.493).
Thus the family fg 1 .y/ W y 2 Y g is an open coverS
of the space X 0 ; since l.X 0 / ,
there exists A Y such that jAj and X D fg 1 .y/ W y 2 Ag D g 1 .A/.
0
This shows that Y D g.X 0 / D g.g 1 .A// D A, so jY j D jAj which proves

that X is -simple. We proved that X is -simple for any infinite cardinal , i.e., X
is simple.
T.130. Give an example of a pseudocompact scattered non-simple space.
Solution. Recall that a family A is called almost disjoint if every A 2 A is infinite
and A \ A0 is finite for any distinct A; A0 2 A. A Mrowka space M D M.M/
is constructed as follows: M D ! [ M where M is a maximal almost disjoint

subfamily of infinite subsets of !. The points of ! are isolated in M and the local
base at a point x 2 M is the family ffxg [ .xnF / W F is a finite subset of
xg (see Problem 142 of [TFS] for the details). Every Mrowka space is a locally
countable, locally compact and first countable pseudocompact space such that ! is
open, discrete and dense in M while M is a closed discrete subspace of M (see
Problem 142 of [TFS]).
Fact 1. Every Mrowka space M is scattered.
Proof. Take any non-empty set A M ; if A \ ! ;, then any x 2 A \ ! is
isolated in A. If A \ ! D ;, then A M and hence A is discrete, so all points of
A are isolated in A. Fact 1 is proved. t
u
It was proved in Fact 2 of S.154 that there exists an almost disjoint family M on
! such that the resulting Mrowka space X D M.M/ can be mapped continuously
onto I; since w.I/ D ! and jIj > !, the space X is not !-simple. Finally observe
that the space X is pseudocompact by Problem 142 of [TFS] and scattered by Fact 1.
T.131. Prove that the following are equivalent for any space X :
(i) the space X is weight.!/-stable;
(ii) the space X is character.!/-stable;
(iii) the space X is FrchetUrysohn.!/-stable;
(iv) the space X is sequential.!/-stable;
(v) the space X is k-property.!/-stable;
(vi) the space X is w.!/-stable;
(vii) the space X is .!/-stable;
(viii) the space X is pseudocompact.
Solution. Our solution will be based on some simple facts.
Fact 1. Given any space Z, the space Cp .Z/ considered with its usual operations
of summing functions and multiplying them by a real number is a locally convex
linear topological space.
Proof. It is clear that Cp .Z/ is a linear space. Its operations are continuous by
Problem 114 of [TFS] and Problem 115 of [TFS]; the local convexity of Cp .Z/
was established in Problem 069 of [TFS] so Fact 1 is proved. t
u
Fact 2. Every infinite closed subspace of ! has cardinality 2c and hence ! has
no nontrivial convergent sequences.
Proof. Let F be an infinite closed subspace of !. Choose an infinite discrete B
F (see Fact 4 of S.382). Apply Fact 2 of S.382 and Fact 2 of S.286 to conclude that
B is homeomorphic to ! and hence jBj D 2c by Problem 368 of [TFS]. Since
B F , we have jF j D 2c so Fact 2 is proved. t
u
Fact 3. Let M be a non-compact second countable space. Then there exists a

continuous onto map f W M ! Z such that Z is not a k-space. Consequently,
M is not k-property(!)-stable.
Proof. Since M is not compact, it is not countably compact and hence there exists a
countably infinite closed discrete D M . Take any 2 !n!; then N D ! [ fg
is not a k-space. Indeed, ! is not closed in N but K \ ! is closed for any compact
K N because all compact subsets of N are finite by Fact 2.
If g W D ! N is any surjection, then g is a continuous map because D is
discrete. The space N embeds in Cp .Cp .N // as a closed subspace by Problem 167
of [TFS], so we can consider that N Cp .Cp .N //. Therefore we have a continuous
map g W D ! Cp .Cp .N //; since Cp .Cp .N // is locally convex by Fact 1 and M
is metrizable, we can apply 104 to conclude that there exists a continuous map
f W M ! Cp .Cp .N // such that f jD D g. If Z D f .M /, then f W M ! Z
is an onto map. Observe that N D g.D/ D f .D/ Z; since N is closed in
Cp .Cp .N //, it is closed in Z, so Z is not a k-space (it is an easy exercise to prove
that a closed subset of a k-space is a k-space, so if a space contains a closed subset
which is not a k-space, then it cannot be a k-space either).
Since w.M / D !, we have nw.Z/ ! and hence the space Z condenses onto a
second countable space (see Problems 156 and 157 of [TFS]). This proves that the
space M is not k-property(!)-stable so Fact 3 is proved. t
u
Fact 4. Let M be a non-compact second countable space. Then there exists a
continuous onto map f W M ! Z such that .Z/ > !. Consequently, M is
not .!)-stable.
Proof. Since M is not compact, it is not countably compact and hence there exists
a countably infinite closed discrete D M . Now, the space Cp .I/ has uncountable
-character at zero and hence at all of its points by Problem 171 of [TFS]. It is an
easy exercise that
(1) if Y is a space and T is dense in Y , then .y; Y / D .y; T / for any y 2 T .
So if we take a countable dense A Cp .I/, then .A/ > !. If g W D ! A
is any surjection, then g is a continuous map because D is discrete. Therefore we
have a continuous map g W D ! Cp .I/; since Cp .I/ is locally convex by Fact 1
and M is metrizable, we can apply Problem 104 to conclude that there exists a
continuous map f W M ! Cp .I// such that f jD D g. If Z D f .M /, then
f W M ! Z is an onto map. Observe that A D g.D/ D f .D/ Z; since A is
dense in Cp .I/, it is dense in Z, so .Z/ D .A/ > ! by (1). Since w.M / D !,
we have nw.Z/ ! and hence Z condenses onto a second countable space (see
Problems 156 and 157 of [TFS]). This proves that M is not .!/-stable. Fact 4 is
proved. t
u
Returning to our solution, observe that
.weight !/ H) .character !/ H) .FrchetUrysohn/ H) .sequential/

and (sequential) H) .k-space) which shows that (i)H)(ii)H)(iii)H)(iv)H)(v).

Now assume that X is not pseudocompact. Then there is a countably infinite closed
discrete D X such that D is C -embedded in X (see Fact 1 of S.350). Let D D
fdn W n 2 !g and f .n/ D n for each n 2 !; then f W D ! R is a continuous map
because D is discrete. Since D is C -embedded in X , there is g 2 C.X / such that
gjD D f . If M D g.X /, then M is not compact being unbounded in R. Thus
(2) if X is not pseudocompact, then there is a continuous onto map g W X ! M of
X onto a non-compact second countable space M .
Apply Fact 3 to find a continuous onto map h W M ! Z such that Z is not a
k-space. Since nw.Z/ D !, the space Z condenses onto a second countable space.
Consequently, h g maps X onto the space Z with i w.Z/ D ! while Z is not a
k-space. Therefore X is not k-property.!/-stable which shows that (v)H)(viii)
Now assume that X is pseudocompact; given any continuous map f W X ! Y ,
the space Y is also pseudocompact. If Y condenses onto a second countable space
M , then this condensation is a homeomorphism by Problem 140 of [TFS] and hence
w.Y / D w.M / D !. Therefore the space X is weight.!/-stable, i.e., we proved
that (viii)H)(i); as a consequence, (i) (ii) (iii) (iv) (v)
(viii).
It is immediate that (weight!/ H) .-weight!/ H) .-character!), so
(i)H)(vi)H)(vii). If X is not pseudocompact, then, by (2), there is a continuous
onto map g W X ! M of X onto a non-compact second countable space M . Apply
Fact 4 to find a continuous onto map h W M ! Z such that .Z/ > !. Since
nw.Z/ D !, the space Z condenses onto a second countable space. Consequently,
h g maps X onto the space Z with i w.Z/ D ! while .Z/ > !. Therefore X is
not .!/-stable which shows that (vii)H)(viii). We saw already that (viii)H)(i),
so (i) (vi) (vii) (viii) and hence our solution is complete.
T.132. Prove that the following conditions are equivalent for any space X :
(i) for any countable A Cp .X /, the space A is -compact;
(ii) for any countable A Cp .X /, the space A is -countably compact;
(iii) for any countable A Cp .X /, the space A is -pseudocompact;
(iv) for any countable A Cp .X /, the space A is locally compact;
(v) Cp .X / is a Hurewicz space;
(vi) Cp .X / is a -locally compact space;
(vii) for any countable A Cp .X /, the space A is Hurewicz;
(viii) for any countable A Cp .X /, the space A is -locally compact;
(ix) X is finite.
Solution. If Z is a space, call a family A D fAs W s 2 S g exp.Z/ indexed
discrete if, for any z 2 Z, there is W 2 .z; Z/ such that the set fs 2 S W W \ As
;g is finite. For the sake of brevity denote the space ! ! by P. A space is zero-
dimensional if it has a base which consists of clopen sets. The expression X ' Y
says that the space X is homeomorphic to the space Y .
As usual, we identify any ordinal with the set of its predecessors and, in
particular, n D f0; : : : ; n 1g for any n 2 !. If k; n 2 ! and s 2 ! k ,
then s _ n 2 ! kC1 is defined by .s _ n/.k/ D n and .s _ n/jk D s. If we have
s 2 ! n ; t 2 ! k where n k and n; k 2 ! C 1, then s t says that tjn D s. If d
is a metric on a set M , then .d / is the topology generated by the metric d .
If we work with a metric space .M; d /, then for any x 2 X and r > 0 the set
Bd .x; r/ D fy 2 M W d.x; y/ < rg is the ball of radius r centered at x. If f
is a function, then dom.f / is its domain. Suppose that we have a set of functions
ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D fj j.dom.fi / \S dom.fj // for
any i; j 2 I . Then we can define a function f with dom.f / D i 2I dom.fi /
as follows: given any x 2 dom.f /, find any i 2 I with x 2 dom.fi / and let
f .x/ D fi .x/. It is easy to check that the value of f at x does not depend on the
choice
S of i , so we have consistently defined a function f which will be denoted by
ffi W i 2 I g
The concept of Hurewicz space is a new one, so let us get some insight into it:
Fact 1.
(1) any countable union of Hurewicz spaces is a Hurewicz space;
(2) any -compact space is Hurewicz;
(3) any closed subspace of a Hurewicz space is a Hurewicz space;
(4) any continuous image of a Hurewicz space is a Hurewicz space;
(5) the space P is not Hurewicz and hence not -compact.
S
Proof. (1) Assume that Z D fZn W n 2 !g and Zn is Hurewicz for each n 2 !.
Let fUn W n 2 !g be a sequence of open covers of the space S Z. Represent ! as
a disjoint infinite union of its infinite subsets, i.e., ! D fMn W n 2 !g where
jMn j D ! for each n 2 ! and Mk \ Ml D ; whenever k l. The collection
Wk D fUi W i 2 Mk g is a sequence of open SS covers of Zk , so we can choose a finite
Vi Ui for each i 2 Mk so that Zk . fVi W i 2 Mk g/ for each k 2 !. After
this choice is carried out for each k 2 !, we obtain a finite Vi Ui for each i 2 !
and
[ [ [ n[ [ o [
fVi W i 2 !g D fVi W i 2 Mk g W k 2 ! fZk W k 2 !g D Z;
which shows that Z is a Hurewicz space and settles (1). The statement (2) follows
easily from (1) and the evident fact that every compact space is Hurewicz. The proof
of (3) and (4) is an easy exercise for the reader so let us check (5).
Given n 2 N and s 2 ! n the set s D fx 2 ! ! such that s xg is clopen in ! ! .
The family Un D fs W s 2 ! nC1 g is an open disjoint cover S of P for every n 2 !.
Suppose that Vn is a finite subfamily of Un and let Wn D Vn for all n 2 !. The
family U0 is disjoint and infinite, so we can choose s0 2 ! 1 such that s0 V0 and
hence s0 \ W0 D ;. Assume that k > 0 and we have chosen s0 ; : : : sk1 with the
(a) si 2 ! i C1 for all i < k;
(b) s0

sk1 ;
(c) si \ Wi D ; for each i < k.
The family Uk0 D fs W s 2 ! kC1 and sk1 sg Uk is infinite and disjoint,

so there is sk 2 ! kC1 such that sk 2 Uk0 nVk and therefore sk \ Wk D ;. We
have sk1 sk by our definition of Uk0 , so the conditions (a)(c) are satisfied for
the collection fsi W i kg. Thus our inductive construction can be continued to
us a sequence fsi W i 2 !g with the properties (a)(c).
give S T It follows from (b) that
s D S fsi W i 2 !g 2 P; the condition (b) shows that s 2
S fsi W i 2 !g and hence
s fWi W i 2 !g by (c) which proves that the family fVi W i 2 !g is not a cover
of P. Thus P is not Hurewicz and Fact 1 is proved. t
u
Fact 2. Let M be a zero-dimensional second countable space. Then every closed
non-empty set F M is a retract of M .
Proof. Take any metric d on the set M which generates the topology of M ; for
each x 2 M nF , let d.x; F / D inffd.x; y/ W y 2 F g and r.x/ D 14 d.x; F /.
Consider the family U D fBd .x; r.x// W x 2 M nF g; since M is zero-dimensional,
we can choose a clopen set Ox 2 .x; M / such that Ox Bd .x; r.x// for every
x 2 M nF . Observe that Ox Bd .x; r.x// M nF for each Sx 2 M nF . Since
M nF is Lindelf, there is a countable A M nF such that fOx W x 2 Ag D
M nF . Choose
S any enumeration fUn W n 2 !g of the family fOx W x 2 Ag and let
Vn D Un n. fUi W i < ng/ for each n 2 !. By our construction, for each n 2 !,
there is a point yn 2 A such that Vn Bd .yn ; rn / where rn D 14 d.yn ; F /. Take any
an 2 F such that d.yn ; an / 54 d.yn ; F / D 5rn.
Observe that
(6) given any n 2 !, we have d.x; an / 2d.x; F / for every x 2 Vn .
To prove (6) note that d.x; F / 3rn ; for otherwise, there is a point z 2 F such that
d.x; z/ < 3rn and hence
d.yn ; F / d.yn ; z/ d.yn ; x/ C d.z; x/ < rn C 3rn D 4rn D d.yn ; F /;
which is a contradiction proving that d.x; F / 3rn . Furthermore, it follows from

d.x; an / d.x; yn / C d.yn ; an / rn C 5rn D 6rn 2d.x; F / that (6) is proved.
Now, let f .x/ D x for every x 2 F . If x 2 M nF , then there is a unique n 2 !
such that x 2 Vn ; let f .x/ D an . It is evident that f W M ! F and f .x/ D x for
any x 2 F , so we only must show that f is a continuous map.
If x 2 M nF , then x 2 Vn for some n so f .x/ D an . Since Vn is a neighborhood
of x such that f .Vn / D fan g, the function f is continuous at the point x. Now take
any x 2 F and any U 2 .x; F /; there exists " > 0 such that Bd .x; "/ \ F U .
The set W D Bd .x; 3" / is open in M and contains x. We claim that f .W /
Bd .x; "/ \ F . Indeed, if y 2 W \ F , then f .y/ D y so f .y/ 2 W \ F
Bd .x; 3" / Bd .x; "/. Now, if y 2 W nF , then there is n 2 ! such that y 2 Vn
and hence f .y/ D an . We have d.x; an / d.x; y/ C d.y; an / 3" C 2d.y; F /
"
3
C 2d.x; y/ < 3" C 2" 3
D " and hence f .W / Bd .x; "/ \ F U which proves
continuity of f at the point x. Thus f W M ! F is a retraction and Fact 2 is
proved. t
u
Fact 3. Call a space M analytic if M is a continuous image of the space P. Then

(7) every closed subspace of an analytic space is an analytic space;
(8) every continuous image of an analytic space is an analytic space;
(9) any countable product of analytic spaces is an analytic space;
(10) any countable intersection of analytic spaces is an analytic space;
(11) any countable union of analytic spaces is an analytic space;
(12) every -compact space is analytic;
(13) every second countable completely metrizable space is analytic.
Proof. Take any analytic space M and any non-empty closed A M . There is
a continuous onto map f W P ! M ; let F D f 1 .A/ and g D f jF . Then
g is a continuous map of F onto A. The space P is zero-dimensional, so there is
a retraction r W P ! F by Fact 2. Then h D g r is a continuous map and
h.P/ D A so (7) is proved. Now, if q W M ! N is a continuous onto map, then
p D q f W P ! N is a continuous map such that N D p.P/, so N is analytic and
hence we settled (8).
To prove (9) note that P! D .! ! /! ' ! !! is homeomorphic to P (see
ProblemQ103 of [TFS]). Given any family fMn W n 2 !g of analytic spaces, let
M D fMn W n 2 !g and fix a continuous onto map fn W P ! Mn for all
n 2 !. Let g.x/.n/ D fn .x.n// for any x 2 P! and n 2 !. It is immediate that
g W P! ! M (called the product of the family ffn W n 2 !g) is an onto map.
Continuity of any product of maps was proved in Fact 1 of S.271 so g is continuous.
Since P! ' P, the space M is analytic and we proved (9).
T (10) take an arbitrary family fMn W n 2 !g of analytic spaces and let
To prove
M D fMnQW n 2 !g. Observe that the space M embeds as a closed subspace
in the space fMn W n 2 !g by Fact 7 of S.271, so we can apply (7) and (9) to
conclude that M is analytic.
S As to (11), assume that Mn is an analytic space for every n 2 ! and let M D
fMn W n 2 !g. It is evident that the spaceLPn D fx 2 P W x.0/ D ng is
homeomorphic to P for each n 2 ! and P ' fPn W n 2 !g. Since each Mn is
analytic, we can fix a continuous onto map fn W Pn ! Mn for every n 2 !. Given
any x 2 P there is a unique n 2 ! such that x 2 Pn ; let f .x/ D fn .x/. Then
f W P ! M is an onto map and the family fPn W n 2 !g is an open cover of P such
that f jPn D fn is continuous for each n 2 !. Therefore f is continuous by Fact 1
of S.472. This shows that M is analytic so (11) is proved.
Now observe that K D f0; 1g! is a closed subset of P and hence K is a
continuous image of P by Fact 2. Since every metrizable compact space is a
continuous image of K (see Problem 128 of [TFS]), we can apply (8) to conclude
that every compact metrizable space is analytic. Applying (11) we conclude that
every -compact second countable space is analytic so (12) is verified.
Finally, take any completely metrizable second countable space M . Then M is
homeomorphic to a closed subspace in R! by Problem 273 of [TFS]. The space R
is analytic being -compact by (12). Applying (9) and (7) we conclude that R! and
M are analytic, so (13) is settled and Fact 3 is proved. t
u
Fact 4. Let M be an analytic second countable non--compact space. Then there

is a closed subspace F M such that F is homeomorphic to P.
Proof. The space P is completely metrizable (see Problem 204 and 207 of [TFS]),
so we can fix a complete metric on P such that ./ D .P/. Take any metric
d on M with .d / D .M /. Since M is analytic, there is a continuous onto map
f W P ! M . We will need the family S D fA M W there is a -compact P M
such that A P g. Let A; D P; assume that n > 0, and for each k < n, we have a
family fAs W s 2 ! k g of closed subsets of P with the following properties:
(14) if 0 < k < n, then diam .As / < 2k for all s 2 ! k ;
(15) if 0 < k < n, then diamd .f .As // < 2k for all s 2 ! k ;
(16) if 0 k < n, then fAs W s 2 ! k g is an indexed discrete family in P;
(17) if 0 k < n, then ff .As / W s 2 ! k g is an indexed discrete family in M ;
(18) if 0 k < l < n; s 2 ! k ; t 2 ! l and s t, then At As ;
(19) if 0 k < n, then f .As / S for any s 2 ! k .
Fix s 2 ! n1 and let Bs D fx 2 As W f .U / S for any U 2 .x; As /g. For
each x 2 As nBs there is Wx 2 .x; As / such that f .Wx / 2 S. The open cover
W D fWx W x 2 As nBsS g of the Lindelf space As nBs has a countable subcover
W 0 ; it is evident that f . W 0 / D f .As nBs / 2 S. Since f .As / S, we have
f .Bs / S and hence there exists an infinite set D f .Bs / which is discrete and
closed in M . It is easy to find a set Vy 2 .y; M / such that diamd .Vy / < 2n for
each y 2 D and the family fVy W y 2 Dg is indexed discrete in M . The family
ff 1 .Vy / W y 2 Dg is indexed discrete in P; pick a point x.y/ 2 f 1 .y/ \ Bs for
each y 2 D. It is clear that x.y/ 2 f 1 .Vy /, so we can find Uy 2 .x.y/; P/ such
that Uy f 1 .Vy / and diam .Uy / < 2n for every y 2 D.
Take a faithful enumeration fyn W n 2 !g of the set D and let As _ n D
Uyn \ Bs for each n 2 !. It follows from Uyn \ Bs ; that f .As _ n / S
for each n 2 !. It is clear that diam .As _ n / diam .Uyn / < 2n . Besides,
diamd .f .As _ n // diamd .f .Uyn // diamd .Vyn / < 2n which shows that (14)
and (15) are fulfilled for the sets As _ n for all n 2 !. The properties (16) and (17)
hold because the families U D fUyn W n 2 !g and V D fVyn W n 2 !g are indexed
discrete and the families fAs _ n W n 2 !g and ff .As _ n / W n 2 !g are obtained
by shrinking U and V respectively. After we define the family fAs _ n W n 2 !g for
each s 2 ! n1 , we have the collection fAt W t 2 ! n g. This method of construction
guarantees the property (18). We also assured that f .As _ n / S for each s 2 ! n1
and n 2 ! which shows that (19) is also true for the family fAt W t 2 ! n g.
Thus our inductive construction can be completed for all natural n to give us
families fAs W s 2 ! n g with (14)(19) for all n 2 !. Now we are ready to describe a
closed subset of M which is homeomorphic to P. Given any x 2 P let sn D xjn for
each n 2 !; it follows from (14) and (18) that fAsn W n 2 !g is a decreasing family
T of P such that diam .Asn / ! 0. As a consequence there is a unique
of closed subsets
point g.x/ 2 fAsn W n 2 !g (see Problem 236 of [TFS]). Let h.x/ D f .g.x//;
this gives us a map h W P ! M . We must check that F D h.P/ is closed in M and
the map h W P ! F is a homeomorphism.
Take any point x 2 P and an arbitrary " > 0; there exists m 2 ! such that
2m < ". The set U D xjm is an open neighborhood of the point x in P, and for
any y 2 U , we have yjm D xjm and hence h.y/ D f .g.y// 2 f .Axjm /. Since
diamd .f .Axjm // 2m < ", we have d.h.y/; h.x// diamd .f .Axjm // < ",
which shows that h.U / Bd .h.x/; "/ and hence h is continuous at the point x.
If x and y are distinct points of P, then there is n 2 ! such that xjn yjn.
The property (17) implies that f .Axjn / \ f .Ayjn / D ;; since h.x/ 2 f .Axjn / and
h.y/ 2 f .Ayjn /, we have h.y/ h.x/ and hence h is a condensation.
To see that h1 W F ! P is continuous, take any yT2 F . There exists a sequence
fsi W i 2 !g such that si 2 ! i for all i 2 ! and y 2 ff .A Ssi / W i 2 !g. It follows
from (17) that si si C1 for all i 2 ! and hence x D fsi W i 2 !g is well-
defined. It is immediate that x D h1 .y/; observe that the family fxjn W n 2 Ng
is a local base at x in P, so if we take any W 2 .x; P/, then there is n 2 ! such
that xjn W . The set U D f .Asn / \ F 3 y is open in F by (17). If z 2 U , then
h1 .z/jn D sn by (16); this implies h1 .z/jn D xjn, i.e., h1 .z/ 2 xjn W . The
point z 2 U was chosen arbitrarily so h1 .U / xjn . Thus we proved that for any
W 2 .x; P/ there exists U 2 .y; F / such that h1 .U / W . Therefore h1 is
continuous at every y 2 F and h is a homeomorphism. S
To see that F is closed inTM take any point y 2 F . If Fn D ff .As / W s 2 ! n g
for each n 2 !, then F D fFn W n 2 !g. Therefore y 2 F n for each n 2 !; the
family ff .As / W s 2 ! n g being discrete, there is sn 2 ! n such that y 2 f .Asn / for all
n 2 !. The propertiesS (17) and (18) imply that sn snC1 for any T n 2 ! and hence
we can define x D fsn W n 2 !g. Now observe that h.x/ 2 ff .Asn / W n 2 !g;
the property (15) shows that d.h.x/; y/ diamd f .Asn / D diamd .f .Asn // <
2n for every n 2 !. Thus d.h.x/; y/ D 0, i.e., h.x/ D y and hence y 2 F . This
proves that F is closed in M so Fact 4 is proved. t
u
Fact 5. Given any space Z assume that U D fUt W t 2 T g .Z/ is an indexed
discrete family and xt 2 Ut for every t 2 T . Let D D fxt W t 2 T g and take any
function bt 2 C.Z/ suchP that bt .xt / D 1 and bt .ZnUt / D f0g for all t 2 T . Define
a map ' by '.h/.z/ D t 2T h.xt /bt .z/ for every h W D ! R and z 2 Z. Then
'.h/ 2 C.Z/ and '.h/jD D h for each h 2 RD ; besides, the map ' W RD !
Cp .Z/ is linear and continuous.
Proof. For any z 2 Z there is W 2 .z; Z/ such that W intersects at most
one element of U, say Ut . Consequently, '.h/jW D h.xt /bt jW is a continuous
function, so we can apply Fact 1 of S.472 to conclude that '.h/ is continuous. By
the definition of '.h/ we have '.h/.xt / D h.xt /bt .xt / D h.xt / for each t 2 T so
'.h/jD D h.
The linearity of ' is straightforward; to prove that it is continuous take any point
z 2 Z and let z .f / D f .z/ for every f 2 Cp .Z/. Observe that the mapping
z is the restriction of the natural projection of RZ onto its factor determined by
z. Let qt W RD ! R be the natural projection onto the factor determined by xt ,
i.e., qt .h/ D h.xt / for every t 2 T . Note that ' maps RD to the product RZ , so
it suffices to prove that z ' is continuous for each z 2 Z (see Problem 102 of
[TFS]).
S
If z 2 Zn. U/, then S '.h/.z/ D 0 for all h 2 R , so z ' is continuous being
D
a constant map. If z 2 U, then z 2 Ut for some t 2 T and hence bu .z/ D 0 for all
u 2 T nftg. As a consequence, '.h/.z/ D h.xt /bt .z/ for any h 2 RD which shows
that z ' is continuous because it coincides with bt .z/qt . Thus ' is continuous and
hence Fact 5 is proved. t
u
Fact 6. If a space Z is not pseudocompact, then Cp .Z/ ' Cp .Z/ R! . In
particular, R! embeds in Cp .Z/ as a closed subspace.
Proof. Since the space Z is not pseudocompact, we can fix an indexed discrete
family U D fUn W n 2 !g .Z/ and pick xn 2 Un for every n 2 !; let
D D fxn W n 2 !g. Choose a function bn 2 Cp .Z; 0; 1 / such thatPbn .xn / D 1 and
bn .ZnUn / D f0g for all n 2 !. Define a map ' by '.h/.z/ D n2! h.xn /bn .z/
for every h W D ! R and z 2 Z. Then '.h/ 2 C.Z/ and '.h/jD D h for each
h 2 RD ; besides, the map ' W RD ! Cp .Z/ is continuous (see Fact 5).
We will show that the space I D ff 2 Cp .Z/ W f .D/ D f0gg is a factor
of Cp .Z/, namely, that I RD ' Cp .Z/. To do so, let u.p/ D g C '.h/ for
any p D .g; h/ 2 I RD . It follows from Fact 5 that u W I RD ! Cp .Z/.
Given any f 2 Cp .Z/, let h D f jD and g D f '.h/. It is immediate that
p D .g; h/ 2 I RD and u.p/ D f , i.e., u is an onto map. Furthermore, let
w.f / D .f '.f jD/; f jD/ for each f 2 Cp .Z/. It is straightforward that w W
Cp .Z/ ! I RD and w is the inverse function of u. Thus u is a bijection, so it
suffices to prove continuity of u and w.
Given any p D .g; h/ 2 I RD , let u0 .p/ D g and u1 .p/ D g; it is clear
that u0 W I RD ! I and u1 W I RD ! RD are continuous maps. For any
pair .f1 ; f2 / 2 Cp .Z/ Cp .Z/, let m.f1 ; f2 / D f1 C f2 ; then m W Cp .Z/
Cp .Z/ ! Cp .Z/ is also continuous (see Problem 115 of [TFS]). Now observe that
u D m .fu0; ' u1 g/ and hence u is continuous being obtained from continuous
maps u0 ; u1 ; ' and m by applying compositions and -product.
To prove continuity of w, observe that the map D W Cp .Z/ ! RD defined by
D .f / D f jD is continuous as well as the map r W Cp .Z/ Cp .Z/ ! Cp .Z/
defined by r.f1 ; f2 / D f1 f2 for any pair .f1 ; f2 / 2 Cp .Z/ Cp .Z/. If id W
Cp .Z/ ! Cp .Z/ is the identity map, then w D fr .fid; ' D g/; D g is
continuous because it is obtained from continuous maps D ; '; r and id by applying
compositions and -products. Thus u is a homeomorphism which shows that
Cp .Z/ ' I RD ' I R! ' I R! R! ' Cp .Z/ R! ;
u
Fact 7. If Z is an infinite space, then there is f 2 C.Z/ such that f .Z/ is infinite.
Proof. We can choose a faithfully indexed discrete D D fzn W n 2 Ng Z (see

Fact 4 of S.382) and a disjoint family U D fUn W n 2 Ng .Z/ such that zn 2 Un
for each n 2 N (see Fact 1 of S.369). Using the Tychonoff property of Z, for each
n 2 N, we can find fn 2 C.Z; 0; Pn / such that fn .zn / D n and fn .ZnUn / D f0g.
1 1
We claim that the function f D n2N fn W Z ! R is continuous.

To see it, take any z 2 Z; if z 2 Un for some n 2 N, then f jUn D fS n jUn is a

continuous map, so f is continuous at the point z. Now take any z 2 Zn. U/ and
" > 0. The set Fn D fn1 ."; n1 / is closed in Z and Fn Un for each n 2 N. Take
any m 2 N such that m1 < " and observe that W D Zn.F1 [

[ Fm / is an open
neighborhood of z. Given any y 2 W consider the following cases:
Case 1. y 2 Un for some n m. Then y Fn and hence f .y/ D fn .y/ < ";
Case 2. y 2 Un for
S some n > m. Then f .y/ D fn .y/ n < m < ";
1 1
Case 3. y 2 Zn. U/. Then f .y/ D 0 < ".

Thus we established that f .W / 0; "/ .f .x/ "; f .x/ C "/ and hence f
is continuous at the point z. Since z 2 Z was chosen arbitrarily, the function f is
continuous; we have f n1 W n 2 Ng f .Z/, so the set f .Z/ is infinite and Fact 7 is
proved. t
u
Returning to our solution observe that it is trivial that (i)H)(ii)H)(iii). Besides,
(iii) implies that X is pseudocompact because otherwise Cp .X / has a closed
subspace F homeomorphic to R! by Fact 6. The space R! is separable, so we
can take a dense countable A F . Then F D A ' R! has to be -pseudocompact
which contradicts Fact 2 of S.399.
Furthermore the condition (viii) also implies pseudocompactness of X because
otherwise R! embeds in Cp .X /S as a closed subspace. Since R! is separable, it must
be -locally compact, so R D fLn W n 2 !g where Ln is locally compact for any
!
n 2 !. It is an easy exercise that every Lindelf locally compact space is -compact;

since nw.Ln / nw.R! / D !, the space Ln is Lindelf and hence -compact for
each n 2 !. This implies that R! is -compact which is a contradiction (see Fact 2
of S.399). This proves that (viii)H)(ix); since (iv)H)(viii) and (vi)H)(viii), the
conditions (iv) and (vi) also imply pseudocompactness of X .
The same reasoning shows that the condition (vii) also implies pseudocompact-
ness of X because otherwise R! embeds in Cp .X / as a closed subspace; since R!
is separable, it has to Hurewicz by (vii). The space P is completely metrizable,
so it embeds in R! as a closed subspace by Problem 273 of [TFS]. Thus P is also
Hurewicz which contradicts Fact 1. Since (v)H)(vii), the condition (v) also implies
pseudocompactness of X . Thus, we established that each one of the conditions
(i)(viii) implies pseudocompactness of X . It is evident that (ix) implies each and
every one of the conditions (i)(viii).
Now assume (iii); if X is infinite, there exists f W X ! R such that M D
f .X / is infinite (see Fact 7). The map f W Cp .M / ! Cp .X / embeds Cp .M /
in Cp .X / as a closed subspace because X is pseudocompact and hence f is R-
quotient (see Problem 163 of [TFS] and Fact 3 of S.154). The space Cp .M / (and
hence f .Cp .M //) is separable, so if A f .Cp .M // is a dense countable subset
of f .Cp .M //, then A D f .Cp .M // is -pseudocompact by (iii). Thus Cp .M /
is -pseudocompact and hence -compact because nw.Cp .M // D !. This shows
that M must be finite by Problem 186 of [TFS] which is a contradiction. Thus
(iii)H)(ix) and therefore (i) (ii) (iii) (ix).
Now assume (viii); if X is infinite, then there exists f W X ! R such that

M D f .X / is infinite (see Fact 7). The map f W Cp .M / ! Cp .X / embeds
Cp .M / in Cp .X / as a closed subspace because X is pseudocompact and hence f is
R-quotient (see Problem 163 of [TFS] and Fact 3 of S.154). The space Cp .M / (and
hence f .Cp .M //) is separable, so if A f .Cp .M // is a dense countable subset
of f .Cp .M //, then A D f .Cp .MS // is -locally compact by (viii). Thus Cp .M /
is -locally compact, so Cp .M / D fLn W n 2 !g where Ln is locally compact
for any n 2 !. Since nw.Ln / nw.Cp .M // D !, the space Ln is Lindelf and
hence -compact for each n 2 !. This implies that Cp .M / is -compact which is
a contradiction because M is infinite (see Problem 186 of [TFS]). This proves that
(viii)H)(ix). It is evident that (iv)H)(viii) and (vi)H)(viii) so (iv) (vi)
(viii) (ix).
Finally assume that (vii) holds. If X is infinite, then there exists f W X ! R such
that M D f .X / is infinite (see Fact 7). The map f W Cp .M / ! Cp .X / embeds
Cp .M / in Cp .X / as a closed subspace because X is pseudocompact and hence f
is R-quotient (see Problem 163 of [TFS] and Fact 3 of S.154). The space Cp .M /
(and hence f .Cp .M //) is separable, so if A f .Cp .M // is a dense countable
subset of f .Cp .M //, then A D f .Cp .M // is Hurewicz by (vii). Thus Cp .M /
is Hurewicz. The space M being an infinite metrizable compact space, there is a
nontrivial convergent sequence S M (we take it with its limit so S is infinite,
compact and has a unique non-isolated point). The restriction map S W Cp .M / !
Cp .S / is continuous and maps Cp .M / onto Cp .S / so Cp .S / is also a Hurewicz
space by Fact 1.
Given f; g 2 C.S / D C .S / let d.f; g/ D supfjf .x/ g.x/j W x 2 S g.
Then d is a metric on C.S / and T D .C.S /; d / is a complete metric space (see
Problem 248 of [TFS]). We claim that T is separable; to show this, denote by a the
unique non-isolated point of S and consider the set D D ff 2 C.S / W f .S / Q
and the set fx 2 S W f .x/ f .a/g is finiteg. It is easy to see that D is countable;
to see that D is dense in T , take an arbitrary f 2 C.S / and any " > 0. Since f is
continuous at the point a, there is a finite P S nfag such that jf .a/ f .x/j < 2"
for all x 2 S nP . Take any rational number r 2 .f .a/; f .a/ C 2" / and let g.x/ D r
for all x 2 S nP . Given any x 2 P choose any g.x/ 2 .f .x/; f .x/ C "/ \ Q. It
is clear that g 2 D and jg.x/ f .x/j < " for each x 2 P . If x 2 S nP , then
jg.x/ f .x/j jg.x/ f .a/j C jf .a/ f .x/j D jr f .a/j C 2" < 2" C 2" D ".
This shows that Bd .f; "/ \ D ; for each f 2 C.S / and " > 0. Therefore D is
dense in T , so T is a completely metrizable second countable space. This implies
that T is analytic by Fact 3. It is clear that the identity map from T to Cp .S / is
continuous, so Cp .S / is analytic being a continuous image of the analytic space
T (we used Fact 3 again). The space S is infinite, so Cp .S / is not -compact by
Problem 186 of [TFS]. Since Cp .S / is second countable, we can apply Fact 4 to
find a closed F Cp .S / such that F is homeomorphic to P. Any closed subspace
of a Hurewicz space is Hurewicz by Fact 1, so F is a Hurewicz space whence P is
Hurewicz, a contradiction with another item of Fact 1. This proves that X is finite
and hence (vii)H)(ix). It is evident that (v)H)(vii) so (v) (vii) (ix) and
hence our solution is complete.
T.133. Prove that any non-scattered countably compact space can be mapped
continuously onto I.
Solution. We denote by D the set f0; 1g with the discrete topology. As usual, we
will identify any ordinal with the set of its predecessors and, in particular, n D
f0; : : : ; n 1g for any n 2 !. If k 2 !; i 2 D and s 2 Dk , then s _ i 2 DkC1 is
defined by .s _ i /.k/ D i and .s _ i /jk D s. If we have s 2 Dn ; t 2 Dk where n k
and n; k 2 ! C 1, then s t says that tjn D s. For any k 2 ! and s 2 Dk , let
s D fx 2 D! W s xg. Given a point x 2 D! , it is immediate that the family
fxjn W n 2 !g is a local base at x in the space D! .
Let X be a non-scattered countably compact space. If X is not zero-dimensional,
then it can be mapped continuously onto the space 0; 1 R (see Fact 4 of T.063).
Since 0; 1 is homeomorphic to I, the space X also maps continuously onto I so
this case is clear.
Now assume that X is zero-dimensional and denote by C the family of all clopen
subsets of X . Since X is not scattered, we can find a non-empty subspace A X
which is dense-in-itself. Let F; D X and assume that, for some n > 0, we have
constructed families fFs W s 2 Dk g for all k < n with the following properties:
(1) given any k < n we have Fs 2 C for any s 2 Dk ;
S any k < nkand any distinct s; t 2 D we have Fs \ Ft D ;;
k
(2) for
(3) fFs W s 2 D g D X for any k < n;
(4) for any k < n we have Fs \ A ; for all s 2 Dk ;
(5) if k < m < n; s 2 Dk ; t 2 Dm and s t, then Ft Fs .
Fix any s 2 Dn1 ; since Fs \A ;, the set Fs \A is a non-empty open subset of
A, so it is dense-in-itself and hence infinite. Take distinct point x; y 2 A \ Fs . The
space X being zero-dimensional, there exists U 2 C such that x 2 U and y U .
The sets Fs _ 0 D U \ Fs and Fs _ 1 D Fs nU are clopen in X and disjoint. Since
x 2 Fs _ 0 and y 2 Fs _ 1 , we have Fs _ 0 \ A ; and Fs _ 1 \ A ;; besides,
Fs D Fs _ 0 [ Fs _ 1 . After we construct the pair fFs _ 0 ; Fs _ 1 g for all s 2 Dn1 we
obtain the family fFs W s 2 Dn g for which the property (5) holds. The condition
(1) is satisfied because both sets Fs _ 0 ; Fs _ 1 are clopen in X for any s 2 Dn1 . The
property (3) follows easily from the fact that Fs D Fs _ 0 [ Fs _ 1 for each s 2 Dn1 .
The property (4) takes place because Fs _ 0 \ A ; and Fs _ 1 \ A ; for all
s 2 Dn1 .
To show that the property (2) holds as well, take any distinct s; t 2 Dn ; if s 0 D
sj.n 1/ t 0 D tj.n 1/ then we have Fs 0 \ Ft 0 D ; by the induction hypothesis,
so Fs \ Ft Fs 0 \ Ft 0 D ; (we used the property (5) here) which proves that
Fs \ Ft D ;. Now, if sj.n 1/ D t.n 1/ D u, then s D u_ i and t D u_ .1 i /
for some i 2 D and therefore Fs \ Ft D Fu_ 0 \ Fu_ 1 D ;, so the property (2) holds
in all possible cases.
Consequently, we can construct the family fFs W s 2 Dk g for every k 2 ! in such
a way that the conditions (1)(5) are satisfied. Given any x 2 X and k 2 !, there
is a unique sk 2 Dk such that x 2 Fsk by (2) and (3). Observe that sk skC1 for
S
each k 2 ! by (2) and (5), so the point y D fsi W i 2 !g 2 D! is well-defined;
let f .x/ D y. Observe that it follows from our definition of f .x/ that
(6) if k 2 !; s 2 Dk and x 2 Fs , then s f .x/.
We claim that the map f W X ! D! is continuous. To prove it take any x 2 X
and W 2 .f .x/; D! /. There is n 2 ! such that f .x/jn W . Let s D f .x/jn;
then U D Fs is an open neighborhood of the point x. Given any y 2 U we have
s f .y/ by (6) and hence f .y/jn D s D f .x/jn, i.e., y 2 f .x/jn . This proves
that f .U / f .x/jn W and therefore f is continuous at the point x.
The map f is surjective; indeed, given any y 2 D! , the family
T fFyjn W n 2 !g is
decreasing by (5). Since X is countably compact, there is x 2 fFyjn W n 2 !g. It is
immediate that f .x/ D y; the point y was taken arbitrarily, so we proved that f
maps X onto D! . There exists a continuous onto map g W D! ! I by Problem 128
of [TFS], so h D g f maps X continuously onto I.
T.134. Prove that the following conditions are equivalent for any compact X :
(ii) A is a FrchetUrysohn space for any countable A Cp .X /;
(iii) X is scattered.
Solution. The implication (i)H)(ii) is obvious because the FrchetUrysohn
property is hereditary. Now, if (ii) holds and X is not scattered, then there is a
continuous onto map f W X ! I by Problem 133. Since f is closed, the set
f .Cp .I// is a closed subspace of Cp .X / homeomorphic to Cp .I/ (see Problem 163
of [TFS]). There is a dense countable A f .Cp .I// and hence f .Cp .I// D A is
FrchetUrysohn. Therefore Cp .I/ is also FrchetUrysohn; this contradiction with
Problem 147 of [TFS] proves that (ii)H)(iii).
For the implication (iii)H)(i) assume that X is scattered; we have to note first
that X is !-simple. Indeed, the !-modification Y of the space X is a Lindelf
P -space by Problem 128 and the identity map of Y onto X is continuous. Thus
every second countable continuous image of X is also a continuous image of Y , so
it must be countable by Problem 127.
Now take an arbitrary set A Cp .X /. Given any g 2 A, there is a countable
B A such that g 2 B (see Problem 149 of [TFS]). For each x 2 X let '.x/.f / D
f .x/ for any f 2 B. Then '.x/ 2 Cp .B/ for every x 2 X and ' W X ! Cp .B/
is a continuous map (see Problem 166 of [TFS]). Weight of Cp .B/ is countable,
so Z D '.X / is countable because X is !-simple. The map ' is closed so ' W
Cp .Z/ ! Cp .X / embeds Cp .Z/ into Cp .X / as a closed subspace by Problem 163
of [TFS]. It is easy to see that B ' .Cp .Z// so B ' .Cp .Z//. The space
' .Cp .Z// is second countable being homeomorphic to Cp .Z/ so B is also second
countable. Then fgg [ B is second countable as well and hence there is a sequence
S D fgn W n 2 !g B such that gn ! g. Thus we found a sequence S A which
converges to g and hence Cp .X / is FrchetUrysohn.
T.135. Prove that for any Lindelf P -space X , the space Cp .X / is a strongly !-
Solution. Take any countable A Cp .X / and let '.x/.f / D f .x/ for any x 2 X
and f 2 A. Then '.x/ 2 Cp .A/ for any x 2 X and the map ' W X ! Cp .A/ is
continuous (see Problem 166 of [TFS]). If Y D '.X /, then jY j D ! because X is
!-simple by Problem 127. Every y 2 Y is a G -set in Y so ' 1 .y/ is open in X
being a G -set in X .
Let Z be the set Y with the discrete topology; then the identity map i W Z ! Y
is continuous. Furthermore, the map D i 1 ' W X ! Z is also continuous
because 1 .z/ D ' 1 .i.z// is open in X for any z 2 Z. Every map onto a discrete
space is open so is open and hence R-quotient. The map W Cp .Z/ ! Cp .X /
embeds Cp .Z/ in Cp .X / as a closed subspace (see Problem 163 of [TFS]). It is
easy to verify that A ' .Cp .Y // .Cp .Z// and therefore A .Cp .Z//
which shows that w.A/ w. .Cp .Z/// D w.Cp .Z// D !.
This proves that Cp .X / is strongly !-monolithic. Since X n is Lindelf for each
n 2 N (see Fact 1 of T.112), we have t.Cp .X // D ! (see Problem 149 of [TFS]).
Now take an arbitrary set A Cp .X /. Given any f 2 A, there is a countable
B A such that f 2 B. The space Cp .X / is strongly !-monolithic so B is second
countable. Then ff g [ B is second countable as well and hence there is a sequence
S D ffn W n 2 !g B such that fn ! f . Thus we found a sequence S A which
converges to f and hence Cp .X / is FrchetUrysohn.
T.136. Prove that for any Lindelf scattered space X , the space Cp .X / is a strongly
Solution. Observe that the !-modification .X /! of the space X is a Lindelf P -
space by Problem 128. Since the identity map i W .X /! ! X is continuous, the dual
map i W Cp .X / ! Cp ..X /! / embeds Cp .X / in Cp ..X /! / (see Problem 163 of
[TFS]). Apply Problem 135 to conclude that Cp ..X /! / is a FrchetUrysohn space
so Cp .X / is FrchetUrysohn as well.
To prove strong monolithity of Cp .X / fix an infinite cardinal and let Y be the
-modification of X ; then Y is a P -space (i.e., every G -subset of Y is open in Y )
and l.Y / by Problem 128. The topology of Y contains the topology of X , so
the identity map i W Y ! X is continuous. The map i W Cp .X / ! Cp .Y / is an
embedding by Problem 163 of [TFS] and therefore Cp .X / is homeomorphic to a
subspace of Cp .Y /. Since strong -monolithity is a hereditary property (we leave
the easy proof as an exercise for the reader), it suffices to prove that Y is strongly
-monolithic.
Take any A Cp .Y / with jAj and let '.y/.f / D f .y/ for any y 2 Y
and f 2 A. Then '.y/ 2 Cp .A/ for any y 2 Y and the map ' W Y ! Cp .A/ is
continuous (see Problem 166 of [TFS]). If T D '.Y /, then w.T / and therefore
every y 2 T is a G -set. As a consequence, ' 1 .y/ is open in Y being a G -subset
of Y . The family f' 1 .y/ W y 2 T g is an open cover
S of the space Y ; since l.Y / ,
there is E Y such that jEj and Y D f' 1 .y/ W y 2 Eg D ' 1 .E/.
Therefore T D '.Y / D '.' 1 .E// D E which proves that jT j D jEj .
Let Z be the set T with the discrete topology; then the identity map j W Z ! T
is continuous. Furthermore, the map D j 1 ' W Y ! Z is also continuous
because the set 1 .z/ D ' 1 .j.z// is open in Y for any z 2 Z. Every map onto a
discrete space is open so is open and hence R-quotient. Thus the dual mapping W
Cp .Z/ ! Cp .Y / embeds Cp .Z/ in Cp .Y / as a closed subspace (see Problem 163
of [TFS]). It is easy to verify that A ' .Cp .T // .Cp .Z// and therefore
A .Cp .Z// which shows that w.A/ w. .Cp .Z/// D w.Cp .Z// D .
This proves that Cp .Y / is strongly -monolithic. The infinite cardinal was chosen
arbitrarily, so Cp .X / is strongly -monolithic for any infinite cardinal , i.e., Cp .X /
is strongly monolithic.
T.137. Let X be a Lindelf P -space. Prove that A is Cech-complete for any
countable A Cp .X /.
Solution. Take any countable A Cp .X / and let '.x/.f / D f .x/ for any x 2 X
and f 2 A. Then '.x/ 2 Cp .A/ for any x 2 X and the map ' W X ! Cp .A/ is
continuous (see Problem 166 of [TFS]). If Y D '.X /, then jY j D ! because X is
!-simple by Problem 127. Every y 2 Y is a G -set in Y , so ' 1 .y/ is open in X
being a G -set in X .
Let Z be the set Y with the discrete topology; then the identity map i W Z ! Y
is continuous. Furthermore, the map D i 1 ' W X ! Z is also continuous
because 1 .z/ D ' 1 .i.z// is open in X for any z 2 Z. Every map onto a discrete
space is open so is open and hence R-quotient. The map W Cp .Z/ ! Cp .X /
embeds Cp .Z/ in Cp .X / as a closed subspace (see Problem 163 of [TFS]). It is
easy to verify that A ' .Cp .Y // .Cp .Z// and therefore A .Cp .Z//.
But Z is a countable discrete space, so Cp .Z/ D RZ is homeomorphic to R! which
is Cech-complete. Consequently, .Cp .Z// is also Cech-complete and hence so is
A being a closed subspace of .Cp .Z//.
T.138. Let X be a pseudocompact space. Suppose that A is Cech-complete for any
countable A Cp .X /. Prove that X is finite.
Solution. If X is infinite, then there is a continuous function f W X ! R such that
Y D f .X / is infinite (see Fact 7 of T.132). The map f W X ! Y is R-quotient (see
Fact 3 of S.154) and hence the dual map f W Cp .Y / ! Cp .X / embeds Cp .Y /
in Cp .X / as a closed subspace (see Problem 163 of [TFS]). The space Cp .Y / is
separable and hence so is f .Cp .Y //; if A f .Cp .Y // is countable and dense in
f .Cp .Y //, then the space f .Cp .Y // D A must be Cech-complete. This implies
that Cp .Y / is also Cech-complete and hence Y is discrete by Problem 265 of [TFS].
However, Y is an infinite compact subspace of R which cannot be discrete; this
contradiction shows that X has to be finite.
T.139. Suppose that A is normal for any countable A Cp .X /. Prove that A is
collectionwise normal for any countable A Cp .X /, i.e., if the space Cp .X / is
normal.!/-monolithic, then it is collectionwise-normal.!/-monolithic.
Solution. We will need some general facts which involve non-Tychonoff spaces.
Assume thatS Z is a set and T is a collection of topologies on Z. It is evident that
the family T generates a topology on Z as a subbase; this topology is denoted by
sup T and called the least upper bound of topologies from T . If is a topology on the
set Z, then a set H Z is -closed (-open) if H is closed (or open respectively)
in the space .Z; /. Analogously, a function f W Z ! R is -continuous if f is
continuous on the space .Z; /.
Fact 1. The least upper bound of any family of completely regular (not necessarily
Tychonoff) topologies on a set Z is a completely regular topology on Z.
Proof. Take any family T of completely regular topologies on Z and let D sup T .
Assume that we have z 2 Z and F Z such that F is -closed and z F .
Therefore z 2 U D ZnF 2 ; since T is a subbase of ,Tthere are 1 ; : : : ; n 2 T
and U1 ; : : : ; Un such that Ui 2 i for each i n and x 2 fUi W i ng U .
As i is completely regular, there are i -continuous functions fi W Z ! 0; 1
such that fi .z/ D 1 and fi .ZnUi / f0g for all i n. Each function fi is also
-continuous because fi1 .H / 2 i for every open set H 0; 1 . Therefore
the function f D f1
: : :
fn W Z ! 0; 1 is -continuous while f .z/ D 1 and
f .F / f0g whence .Z; / is a completely regular space. Fact 1 is proved. t
u
Fact 2. Assume that Z and T are Tychonoff spaces and f W Z ! T is a continuous
onto map. Then there exists a Tychonoff space T 0 such that for some R-quotient
continuous onto map g W Z ! T 0 and a condensation h W T 0 ! T , we have
f D h g.
Proof. Let F D fp 2 RT W p f is continuousg; then C.T / F RT . Given any
p 2 F , let p D fp 1 .O/ W O 2 .R/g. Any p is a completely regular topology on
T (see Problems 097 and 098 of [TFS]), so DS supfp W p 2 F g is a completely
regular topology on T by Fact 1. Observe that fp W p 2 C.T /g is a base for
.T / (see Fact 1 of S.437) and hence .T / D supfp W p 2 C.T /g . This
implies that every .T /-closed set is also -closed; since Z is Tychonoff, the set fzg
is .Z/-closed and hence -closed for each z 2 Z. Thus T 0 D .T; / is a Tychonoff
space and the identity map h W T 0 ! T is a condensation. If g D h1 f , then
f D h g, so we only must prove that g is continuous and R-quotient.
Observe that f and g coincide if considered as mappings between sets, so it
suffices to establish that f is continuous
S and R-quotient considered as a map from
Z to .T; /. The family B D fp W p 2 F g is a subbase for .T; /, so it suffices
to show that f 1 .U / 2 .Z/ for any U 2 B. Now, U 2 p for some p 2 F and
hence U D p 1 .O/ for some O 2 .R/. Therefore f 1 .U / D f 1 .p 1 .O// D
.p f /1 .O/ is open in Z because p f is continuous by the definition of F . Thus
the map f is continuous.
To see that f is R-quotient, take any function p W T ! R such that p f
is continuous. Then p 2 F and hence p 1 .O/ 2 p for every O 2 .R/.
Thus p is continuous considered as a function from .T; / to R and hence the map
f W Z ! .T; / is R-quotient so Fact 2 is proved. t
u
Returning to our solution fix a countable set A Cp .X / and let '.x/.f / D

f .x/ for any x 2 X and f 2 A. Then '.x/ 2 Cp .A/ for any x 2 X and ' W
X ! Cp .A/ is a continuous map by Problem 166 of [TFS]. If Y D '.X /, then
' W X ! Y is a surjective map, so we can apply Fact 2 to find a space Z for which
there exists an R-quotient continuous onto map p W X ! Z and a condensation
i W Z ! Y such that i p D '.
The dual map p W Cp .Z/ ! Cp .X / embeds Cp .Z/ in Cp .X / as a closed
subspace and the dual map i W Cp .Y / ! Cp .Z/ embeds Cp .Y / in Cp .Z/
as a dense subspace (see Problem 163 of [TFS]). Since w.Y / w.Cp .A// D
!, the space Cp .Y / is separable and therefore so is Cp .Z/. If we choose a
countable set B p .Cp .Z// which is dense in the subspace p .Cp .Z//, then
B D p .Cp .Z//, so the space p .Cp .Z// is normal by normal.!/-monolithity of
Cp .X /. As a consequence, the space Cp .Z/ is also normal and hence collectionwise
normal by Problem 295 of [TFS]. Recalling again that p .Cp .Z// and Cp .Z/
are homeomorphic we conclude that p .Cp .Z// is collectionwise normal. It is
straightforward that A ' .Cp .Y // p .Cp .Z// and therefore A p .Cp .Z//.
Since any closed subspace of a collectionwise normal space is collectionwise
normal, the space A is collectionwise normal so our solution is complete.
T.140. Prove that, under MAC:CH, there exists a normal.!/-monolithic space
which is not collectionwise-normal.!/-monolithic.
Solution. If Z is a set and is a topology on Z, then a set is called -closed if it is
closed in .Z; /; a -interior of a set A Z is the interior of A in the space .Z; /.
Denote by D the discrete space of cardinality !1 . The set D is dense in the space
D, so w.D/ 2!1 (see Fact 2 of S.368) and therefore D is homeomorphic
!
to a subspace of I2 1 (see Problem 209 of [TFS]). Thus we can assume, without
!
loss of generality, that D I2 1 . Since MAC:CH holds, we have 2!1 D c (see
Problem 056), so we can consider that D Ic .
The set D is nowhere dense in Ic ; for otherwise, the closure of the discrete
subspace D contains some W 2 .Ic / which implies that W has isolated points
and hence Ic has isolated points which is false (we leave the proof of this easy fact
to the reader). The space Ic is separable by Problem 108 of [TFS], so let us fix a
dense countable A Ic nD; it is clear that A is also dense in Ic .
Let Z D A [ D and S consider the family U D ffag W a 2 Ag [ fU \ Z W U 2
.Ic /g. It is clear that U D Z, so there is a unique topology on Z for which U
is a subbase (see Problem 008 of [TFS]). It follows from the definition of U that the
topology Z on Z induced from Ic is contained in . Since .Z; Z / is Hausdorff and
Z , the space .Z; / is also Hausdorff and hence T1 (it is another trivial exercise
to prove that if a topology is Hausdorff, then any stronger one is also Hausdorff).
Let us show that X D .Z; / is a normal space. Observe first that
. / if P; Q D and P \ Q D ;, then there are U; V 2 such that P U; Q
V and U \ V D ;.
To prove . / note that clD .P / \ clD .Q/ D ; (see Fact 1 of S.382) so clD .P /
and clD .Q/ are disjoint closed subsets of Ic . The space Ic being normal, there are
disjoint U 0 ; V 0 2 .Ic / such that clD .P / U 0 and clD .Q/ V 0 . It is clear that
U D U 0 \ Z and V D V 0 \ Z are as promised so . / is proved.
Now take any disjoint -closed F; G Z and let P D F \ D; Q D G \ D.
Applying . / to the sets P and Q we can find disjoint U1 ; V1 2 such that P U1
and Q V1 . Then U D .U1 nG/ [ F 2 and V D .V1 nF / [ G 2 . To see it
observe that G is -closed so U1 nG 2 and hence every z 2 U1 nG belongs to the
-interior of U ; besides, any point of F n.U1 nG/ is isolated in .Z; / and hence it
also belongs to the -interior of U . An identical argument shows that every point of
V belongs to the -interior of V , i.e., V 2 . Finally, observe that F U; G V
and U \ V D ; so the space X D .Z; / is a T4 -space.
Next observe that the countable set A is dense in X . Indeed, T if x 2 D and
W 2 .x; X/, then there are U1 ; : : : ; Un 2 U such that x 2 i n Ui W . Since
c
x 2 Ui \ D, it is impossible that Ui D fag for some T a 2 A, so there is Vi 2c .I /
such that Vi \ X D Ui for each i n. If V D i n Vi , then V 2 .x; I / and
hence V \ A ; because A is dense in Ic . If a 2 V \ A, then a 2 U \ A W \ A,
so we proved that W \ A ; for any W 2 .x; X/, i.e., x 2 A. The point x 2 D
was taken arbitrarily, so D A and hence A is dense in X .
The family D D ffd g W d 2 Dg is discrete in X and consists of closed subsets of
X while there is no disjoint family W D fWd W d 2 Dg with d 2 Wd for each
d 2 D. Indeed, if such a family W exists, then W .X / which is impossible
because X is separable and hence c.X / D ! which implies that no disjoint family
of non-empty open sets is uncountable. This proves that X is not collectionwise
normal; being separable, the space X is not collectionwise-normal.!/-monolithic.
Since normality implies being normal.!/-monolithic, X is a normal.!/-monolithic
space which is not collectionwise-normal.!/-monolithic.
T.141. Suppose that A is normal for any countable A Cp .X /. Prove that A is
countably paracompact for any countable A Cp .X /.
Solution. Take any countable A Cp .X /; let '.x/.f / D f .x/ for any x 2 X
and f 2 A. Then '.x/ 2 Cp .A/ and ' W X ! Cp .A/ is a continuous map by
Problem 166 of [TFS]. If Y D '.X /, then ' W X ! Y is a surjective map, so we
can apply Fact 2 of T.139 to find a space Z for which there exists an R-quotient
continuous onto map p W X ! Z and a condensation i W Z ! Y such that
i p D '.
B D p .Cp .Z//, so the space p .Cp .Z// is normal by normal.!/-monolithity of
Cp .X /. As a consequence, the space Cp .Z/ is also normal and hence countably
paracompact by Problem 289 of [TFS]. Recalling again that p .Cp .Z// and Cp .Z/
are homeomorphic, we conclude that p .Cp .Z// is also countably paracompact.

It is straightforward that A ' .Cp .Y // p .Cp .Z// and therefore A
p .Cp .Z//. Since any closed subspace of a countably paracompact space is
countably paracompact, the space A is countably paracompact.
T.142. Suppose that A is hereditarily normal for any countable A Cp .X /. Prove
that A is perfectly normal for any countable A Cp .X /.
Solution. Take any countable A Cp .X /; let '.x/.f / D f .x/ for any x 2 X
and f 2 A. Then '.x/ 2 Cp .A/ and ' W X ! Cp .A/ is a continuous map by
Problem 166 of [TFS]. If Y D '.X /, then ' W X ! Y is a surjective map, so we
can apply Fact 2 of T.139 to find a space Z for which there exists an R-quotient
continuous onto map p W X ! Z and a condensation i W Z ! Y such that
i p D '.
B D p .Cp .Z//, so the space p .Cp .Z// is hereditarily normal by hereditary-
normality.!/-monolithity of Cp .X /. As a consequence, the space Cp .Z/ is also
hereditarily normal and hence perfectly normal by Problem 292 of [TFS]. Recalling
again that p .Cp .Z// and Cp .Z/ are homeomorphic, we conclude that p .Cp .Z//
is perfectly normal. It is straightforward that A ' .Cp .Y // p .Cp .Z// and
therefore A p .Cp .Z//. Since any subspace of a perfectly normal space is
perfectly normal (see Problem 003), the space A is perfectly normal.
T.143. Given an infinite cardinal , prove that Cp .X; I/ is -stable if and only if
Cp .X / is -stable.
Solution. If Z is a space and E C.Z/ say that E separates points from closed
sets in Z if, for any z 2 Z and any closed F Z with z F , there is f 2 E
such that f .z/ f .F /. Given any set A X let A W Cp .X / ! Cp .A/ be the
restriction map defined by A .f / D f jA for any f 2 Cp .X /. We will also need the
space Cp .AjX / D A .Cp .X // Cp .A/. If A B X , then AB W Cp .BjX / !
Cp .AjX / is also the restriction map, i.e., AB .f / D f jB for any f 2 Cp .BjX /.
The space Cp .X; I/ is a continuous image of Cp .X / (see Problem 092 of [TFS]),
so if Cp .X / is -stable, then Cp .X; I/ is also -stable by Problem 123.
Now assume that the space Cp .X; I/ is -stable and take a continuous onto map
' W Cp .X / ! Y such that there is a condensation i W Y ! M for which w.M / .
We have a continuous map D i ' W Cp .X / ! M . The space Cp .X / is a dense
subspace of RX ; given any A X , the restriction map A W Cp .X / ! Cp .AjX /
coincides with the restriction to Cp .X / of the natural projection of RX onto its face
RA , so we can apply Fact 1 of T.109 to find a set A X and a continuous map
W Cp .AjX / ! M such that jAj and A D . Let B D A; then AB is
a condensation of Cp .BjX / onto Cp .AjX / because it is the restriction to Cp .BjX /
of the restriction map of Cp .B/ to Cp .A/ which is injective (see Problem 152
of [TFS]). Let T D B .Cp .X; I//; then AB jT is also a condensation of T onto
AB .T / Cp .A/ and hence w.AB .T // jAj . The space Cp .X; I/ being
-stable, we have nw.T / .
Given any x 2 B let ~.x/.f / D f .x/ for any f 2 T . Then ~.x/ 2 Cp .T /
for each x 2 B. The Tychonoff property of X implies that Cp .X; I/ separates
points from closed sets in X ; an immediate consequence is that T D B .Cp .X; I//
separates points from closed sets in B, so we can apply Problem 166 of [TFS] to
conclude that ~ W B ! Cp .T / is a homeomorphism. Consequently, nw.B/ D
nw.~.B// nw.Cp .T // D nw.T / . Furthermore,
D AB W Cp .BjA/ !
M is a continuous onto map such that
B D . We claim that the map
" D i 1
W Cp .BjA/ ! Y is continuous.
To see it take an arbitrary set W 2 .Y /; then W1 D ' 1 .W / 2 .Cp .X // and
hence B .W1 / is open in Cp .BjX / because the map B W Cp .X / ! Cp .BjX / is
open (see Problem 152 of [TFS]). It is easy to see that " B D ' and therefore
"1 .W / D B .W1 / is open in Cp .BjX /, so the map " is indeed continuous. Thus
Y is a continuous image of Cp .BjX / and therefore nw.Y / nw.Cp .BjX //
nw.Cp .B// nw.B/ which shows that nw.Y / and hence Cp .X / is
-stable.
T.144. Given an infinite cardinal , prove that Cp .X; I/ is -monolithic if and only
if Cp .X / is -monolithic.
Solution. If Cp .X / is -monolithic, then Cp .X; I/ has to be -monolithic being
a subspace of Cp .X / (see Problem 113). Now, if Cp .X; I/ is -monolithic, then
the space Cp .X; . 1; 1// Cp .X; I/ is also -monolithic by Problem 113. Finally
observe that the space Cp .X; . 1; 1// is homeomorphic to Cp .X / (see Fact 1 of
S.295) so Cp .X / is also -monolithic.
T.145. Let be any cardinal function. Prove that the Hewitt realcompactification
X of a space X is .!/-stable if and only if X is .!/-stable. In particular, X is
!-stable if and only if X is !-stable.
Solution. Suppose that the space X is .!/-stable and take any continuous onto
map f W X ! Y such that i w.Y / ! and hence .Y / !. If there is any
point y 2 Y nf .X /, then fyg is a G -set in Y and hence P D f 1 .y/ X nX is
a non-empty G -set in X . However, every non-empty G -subset of X intersects
X (see Problem 417 of [TFS]); this contradiction shows that f .X / D Y . Since the
space X is .!/-stable, we have .Y / ! and therefore X is .!/-stable.
Now assume that X is .!/-stable and take any continuous onto mapping f W
X ! Y such that i w.Y / !. Every space of countable i -weight is realcompact
(see Problem 446 of [TFS]), so Y is realcompact and hence there is a continuous
map g W X ! Y such that gjX D f (see Problem 413 of [TFS]). Since g.X /
f .X / D Y , we have g.X / D Y and hence we can apply .!/-stability of X to
conclude that .Y / !. Thus X is .!/-stable.
T.146. Let and be cardinal functions such that .Z/ D .Cp .Z// for any
space Z. Suppose that is hereditary, i.e., for any space Z and any Y Z, we
have .Y / .Z/. Prove that for any infinite cardinal , a space X is ./-stable
if and only if Cp .X / is ./-monolithic.
Solution. Assume that the space X is ./-stable and fix any set A Cp .X / with
jAj . Given any point x 2 X let '.x/.f / D f .x/ for all f 2 A. Then '.x/ 2
Cp .A/ for any x 2 X and ' W X ! Cp .A/ is a continuous map (see Problem 166
of [TFS]); let Y D '.X /. Apply Fact 2 of T.139 to get a space Z such that there
exists an R-quotient map W X ! Z and a condensation i W Z ! Y for which
i D '. We have w.Y / w.Cp .A// D jAj so we can apply ./-stability of
X to conclude that .Z/ . By our hypothesis, .Cp .Z// D .Z/ . Observe
that the dual map ' W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / and the dual
map W Cp .Z/ ! Cp .X / embeds Cp .Z/ in Cp .X / as a closed subspace (see
Problem 163 of [TFS]). It is easy to check that A ' .Cp .Y // .Cp .Z//,
so A .Cp .Z// and hence .A/ . .Cp .Z/// D .Cp .Z// D .Z/
which shows that the space Cp .X / is ./-monolithic.
Now assume that the space Cp .X / is ./-monolithic and take a continuous
onto map ' W X ! Y of the space X onto a space Y such that i w.Y / .
The dual mapping ' W Cp .Y / ! Cp .X / embeds the space Cp .Y / in Cp .X /
(see Problem 163 of [TFS]) and d.' .Cp .Y /// D d.Cp .Y // D i w.Y / ; as
an immediate consequence, there exists a set A ' .Cp .Y // such that jAj
and ' .Cp .Y // A. Thus .' .Cp .Y /// .A/ by ./-monolithity
of Cp .X / and the fact that is hereditary. Consequently, .Y / D .Cp .Y // D
.' .Cp .Y /// which shows that X is ./-stable.
T.147. Let and be cardinal functions such that .Z/ D .Cp .Z// for any
space Z. Suppose that is closed-hereditary, i.e., for any space Z and any closed
Y Z, we have .Y / .Z/. Prove that for any infinite cardinal , a space X is
./-R-quotient-stable if and only if Cp .X / is ./-monolithic.
Solution. Assume that the space X is ./-R-quotient-stable and fix an arbitrary
set A Cp .X / with jAj . Given any point x 2 X let '.x/.f / D f .x/ for all
f 2 A. Then '.x/ 2 Cp .A/ for any x 2 X and ' W X ! Cp .A/ is a continuous
map (see Problem 166 of [TFS]); let Y D '.X /. Apply Fact 2 of T.139 to get a
space Z such that there exists an R-quotient map W X ! Z and a condensation
i W Z ! Y for which i D '. The map is R-quotient and w.Y / w.Cp .A// D
jAj , so we can apply the fact that X is ./-R-quotient-stable to conclude that
.Z/ . By our hypothesis, .Cp .Z// D .Z/ . Observe that the dual map
' W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / and the dual map W Cp .Z/ !
Cp .X / embeds Cp .Z/ in Cp .X / as a closed subspace (see Problem 163 of [TFS]).
It is easy to check that we have the inclusions A ' .Cp .Y // .Cp .Z//, so
A .Cp .Z// and we can use the fact that is closed-hereditary to conclude
that .A/ . .Cp .Z/// D .Cp .Z// D .Z/ which shows that the space
Cp .X / is ./-monolithic.
Now assume that the space Cp .X / is ./-monolithic and take any R-quotient
continuous onto mapping ' W X ! Y of the space X onto a space Y such that
i w.Y / . The dual map ' W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / as a
closed subspace (see Problem 163 of [TFS]) and d.' .Cp .Y /// D d.Cp .Y // D
i w.Y / which shows that we can find a set A ' .Cp .Y // such that jAj
and ' .Cp .Y // D A. Thus .' .Cp .Y /// D .A/ by ./-monolithity of
Cp .X /. Consequently, .Y / D .Cp .Y // D .' .Cp .Y /// and hence X is
./-R-quotient-stable.
T.148. Suppose that and are cardinal functions such that:
(i) for any space Z, if a space Y is a continuous image of Z, then .Y / .Z/;
Prove that for an arbitrary space X , the space Cp .X / is ./-stable if and only
if X is ./-monolithic.
Solution. Given any set B X let B W Cp .X / ! Cp .B/ be the restriction
map defined by B .f / D f jB for any f 2 Cp .X /. We will also need the space
Cp .BjX / D B .Cp .X // Cp .B/. If B A X , then BA W Cp .AjX / !
Cp .BjX / is also the restriction map, i.e., BA .f / D f jB for any f 2 Cp .AjX /.
Assume that Cp .X / is ./-stable and take any set B X with jBj ;
let A D B. The map A W Cp .X / ! Cp .AjX / is continuous and onto while
the restriction BA condenses Cp .AjX / onto the space Cp .BjX /; it is clear that
w.Cp .BjX // w.Cp .B// D jBj . The space Cp .X / being ./-stable, we
have .Cp .AjX // and therefore .A/ D .Cp .AjX // which shows that
X is ./-monolithic.
Now, if the space X is ./-monolithic, then consider any continuous onto map
' W Cp .X / ! Y for which there is a condensation i W Y ! M such that w.M / .
If D i ', then W Cp .X / ! M and hence there exists a set B X such that
jBj and there is a continuous map W Cp .BjX / ! M for which D B
(see Fact 1 of T.109). If A D B then .A/ by ./-monolithity of the space X .
Therefore .Cp .AjX // D .A/ .
Our next step is to prove that the map
D i 1 BA is continuous. Observe
first that
W Cp .AjX / ! Y and
A D '. Given any W 2 .Y / the set ' 1 .W /
is open in Cp .X /, so the set
1 .W / D A .' 1 .W // is open in Cp .AjX / because
A W Cp .X / ! Cp .AjX / is an open map (see Problem 152 of [TFS]). Thus
is
continuous and hence Y is a continuous image of Cp .AjX /. By the property (i) we
have .Y / .Cp .AjX // which proves that Cp .X / is ./-stable.
(i) for any space Z, if a space Y is a quotient image of Z, then .Y / .Z/;
Prove that for an arbitrary space X , the space Cp .X / is ./-quotient-stable if
and only if X is ./-monolithic.

Assume that Cp .X / is ./-quotient-stable and take any set B X with jBj
; let A D B. The map A W Cp .X / ! Cp .AjX / is open, continuous and onto
(and hence quotient, see Problem 152 of [TFS]) while the restriction BA condenses
Cp .AjX / onto the space Cp .BjX /; it is clear that w.Cp .BjX // w.Cp .B// D
jBj . The space Cp .X / being ./-quotient-stable, we have .Cp .AjX //
and therefore .A/ D .Cp .AjX // which shows that X is ./-monolithic.
Now, if the space X is ./-monolithic, then consider any quotient mapping
(see Fact 1 of T.109). If A D B, then .A/ by ./-monolithity of the space X .
Therefore .Cp .AjX // D .A/ .
D i 1 BA is quotient. Observe
first that
W Cp .AjX / ! Y and
A D ' which implies that
is surjective.
Given any W 2 .Y / the set ' 1 .W / is open in Cp .X /, so the set
1 .W / D
A .' 1 .W // is open in Cp .AjX / because A W Cp .X / ! Cp .AjX / is an open map
(see Problem 152 of [TFS]). Thus
is continuous. Now, if H Y and
1 .H / is
open in Cp .AjX /, then ' 1 .H / D A1 .
1 .H // is open in Cp .X / and hence H is
open in Y because ' is a quotient map.
Therefore
is a quotient map and hence Y is a quotient image of Cp .AjX /. By
the property (i) we have .Y / .Cp .AjX // which proves that Cp .X / is
./-quotient-stable.
(i) for any space Z, if a space Y is an R-quotient continuous image of Z, then
.Y / .Z/;
(ii) Ffor any space Z and any closed A Z we have .A/ D .Cp .AjZ//.
Prove that for an arbitrary space X , the space Cp .X / is ./-R-quotient-stable
if and only if X is ./-monolithic.
Assume that Cp .X / is ./-R-quotient-stable and take any set B X with
jBj ; let A D B. The map A W Cp .X / ! Cp .AjX / is open, continuous
and onto (and hence R-quotient, see Problem 152 of [TFS]) while the restriction
BA condenses Cp .AjX / onto the space Cp .BjX /; it is clear that w.Cp .BjX //
w.Cp .B// D jBj . The space Cp .X / being ./-R-quotient-stable, we have
.Cp .AjX // and therefore .A/ D .Cp .AjX // which shows that X is
./-monolithic.
Now, if the space X is ./-monolithic, then consider any R-quotient mapping

(see Fact 1 of T.109). If A D B, then .A/ by ./-monolithity of the space
X . Therefore .Cp .AjX // D .A/ .
D i 1 BA is R-quotient. Observe
first that
W Cp .AjX / ! Y and
A D ' which implies that
is surjective.
Given any W 2 .Y /, the set ' 1 .W / is open in Cp .X /, so the set
1 .W / D
A .' 1 .W // is open in Cp .AjX / because A W Cp .X / ! Cp .AjX / is an open map
(see Problem 152 of [TFS]). Thus the mapping
is continuous. Now, if u W Y ! R
and u
is continuous, then u ' D u .
A / D .u
/ A is a continuous
map and hence u is continuous because ' is an R-quotient map.
Therefore
is an R-quotient map and hence Y is an R-quotient image of
Cp .AjX /. By the property (i) we have .Y / .Cp .AjX // which proves
that Cp .X / is ./-R-quotient-stable.
(i) for any space Z, if Y is an open continuous image of Z, then .Y / .Z/;
Prove that for an arbitrary space X , the space Cp .X / is ./-open-stable if and
only if X is ./-monolithic.
Assume that Cp .X / is ./-open-stable and take any set B X with jBj ;
let A D B. The map A W Cp .X / ! Cp .AjX / is open, continuous and onto (see
Problem 152 of [TFS]) while the restriction BA condenses Cp .AjX / onto Cp .BjX /;
it is clear that w.Cp .BjX // w.Cp .B// D jBj . The space Cp .X / being ./-
open-stable, we have .Cp .AjX // and therefore .A/ D .Cp .AjX //
which shows that X is ./-monolithic.
Now, assume that the space X is ./-monolithic and consider any open map
(see Fact 1 of T.109). If A D B, then .A/ by ./-monolithity of the space
X . Therefore .Cp .AjX // D .A/ .
D i 1 BA is open (and hence
continuous and onto). Observe first that
W Cp .AjX / ! Y and
A D ' which
implies that
is surjective. Given any W 2 .Y / the set ' 1 .W / is open in Cp .X /,
so
1 .W / D A .' 1 .W // is open in Cp .AjX / because A W Cp .X / ! Cp .AjX /
is an open map (see Problem 152 of [TFS]). Thus

is continuous. Now, if U is open
in Cp .AjX /, then
.U / D '.A1 .U // is open in Y because ' is an open map.
Therefore
is an open map and hence Y is an open continuous image of
Cp .AjX /. By the property (i) we have .Y / .Cp .AjX // which proves
that Cp .X / is ./-open-stable.
T.152. Let X be an arbitrary space and an infinite cardinal. Prove that X is
-monolithic if and only if Cp .X / is -stable. In particular, X is monolithic if and
only if Cp .X / is stable.
Solution. If Z is a space and E C.Z/, say that E separates points from closed
restriction map defined by A .f / D f jA for any f 2 Cp .X /. We will also need
the space Cp .AjX / D A .Cp .X // Cp .A/.
To apply the result of Problem 148 let D D nw. Then ./-stability coin-
cides with -stability and ./-monolithity coincides with -monolithity. Network
weight is not increased by continuous images (see Problem 157 of [TFS]) which
shows that satisfies the condition (i) from Problem 148.
For any A X we have nw.A/ D nw.Cp .A// nw.Cp .AjX //. Given an
arbitrary point x 2 A let '.x/.f / D f .x/ for each f 2 E D Cp .AjX /. Then
'.x/ 2 Cp .E/ and ' W A ! Cp .E/ is a continuous map (see Problem 166
of [TFS]). It is easy to see that E separates points from closed sets in A, so the
map ' is actually an embedding (this was also proved in Problem 166 of [TFS]).
As a consequence, nw.A/ nw.Cp .E// D nw.E/ D nw.Cp .AjX // and hence
nw.A/ D nw.Cp .AjX // for any A X . This shows that the condition (ii) from
Problem 148 is also satisfied (in a stronger form) so we can apply Problem 148 to
the cardinal functions D D nw. Thus Cp .X / is nw./-stable if and only if X is
nw./-monolithic, i.e., Cp .X / is -stable if and only if X is -monolithic.
T.153. Prove that if Cp .X / is a stable space, then .Cp .X // is also a stable space
for any cardinal .
Solution. The expression P ' Q says that the spaces P and Q are homeomorphic.
The stability of the space Cp .X / is equivalent to monolithity of X by Problem 152.
Let X be a homeomorphic
L copy of the space X for each < and consider the
space Y D fX W < g in which we identify each X with the respective
clopen subspace of Y (see Problem 113 of [TFS]). Then .Cp .X // ' Cp .Y / (see
Problem 114 of [TFS]) and hence stability of .Cp .X // is equivalent to monolithity
of Y by Problem 152. It is evident that fX W < g is a locally finite (in fact,
discrete) closed cover of Y ; since each X is monolithic, we can apply Problem 115
to conclude that Y is also monolithic and hence Cp .Y / ' .Cp .X // is stable.
T.154. Suppose that X is an arbitrary space and is an infinite cardinal. Prove
that Cp .X / is -monolithic if and only if X is -stable. In particular, Cp .X / is
monolithic if and only if X is stable.
Solution. To apply the result of Problem 146 let D D nw. Then ./-stability
coincides with -stability and ./-monolithity coincides with -monolithity. Net-
work weight is hereditary (see Problem 159 of [TFS]) and nw.Cp .Z// D nw.Z/ for
any space Z which shows that the pair . ; / satisfies the hypothesis of Problem 146.
Therefore Problem 146 is applicable and hence Cp .X / is nw./-monolithic if and
only if X is nw./-stable, i.e., Cp .X / is -monolithic if and only if X is -stable.
T.155. Prove that X is a monolithic space if and only if so is Cp Cp .X /.
Solution. The space X is monolithic if and only if Cp .X / is stable by Problem 152.
Now Cp .X / is stable if and only if Cp .Cp .X // is monolithic by Problem 154
applied to the spaces Y D Cp .X / and Cp .Y / D Cp .Cp .X //. As a consequence, X
is monolithic if and only if so is Cp .Cp .X //.
T.156. Prove that X is a stable space if and only if so is Cp Cp .X /.
Solution. The space X is stable if and only if Cp .X / is monolithic by Problem 154.
Now Cp .X / is monolithic if and only if Cp .Cp .X // is stable by Problem 152
applied to the spaces Y D Cp .X / and Cp .Y / D Cp .Cp .X //. As a consequence, X
is stable if and only if so is Cp .Cp .X //.
T.157. Prove that X is -simple if and only if Cp .X / is strongly -monolithic.
Solution. It is an easy exercise to see that the space X is -simple if and only if
X is cardinality./-stable; besides, strong -monolithity of X coincides with its
w./-monolithity. Now observe that weight is hereditary and jZj D w.Cp .Z//
for any infinite space Z (see Problem 169 of [TFS]). Therefore, if D w and
D card i nali ty, then the pair . ; / satisfies the hypothesis of Problem 146.
Therefore Problem 146 is applicable and hence Cp .X / is w./-monolithic if and
only if X is cardinality./-stable, i.e., X is -simple if and only if Cp .X / is strongly
-monolithic.
(i) Cp .X / is strongly -monolithic;
(ii) Cp .X / is w./-monolithic;
(iii) Cp .X / is ./-monolithic;
(iv) Cp .X / is ./-monolithic.
Solution. As usual, given an arbitrary space Z, points z1 ; : : : ; zn 2 Z and sets
O1 ; : : : ; On 2 .R/, let z1 ; : : : ; zn I O1 ; : : : ; On D ff 2 Cp .Z/ W f .zi / 2 Oi for
all i ng. The sets z1 ; : : : ; zn I O1 ; : : : ; On are called standard open subsets of
Cp .Z/. If U D z1 ; : : : ; zn I O1 ; : : : ; On is a standard open subset of Cp .Z/, then
let supp.U / D fz1 ; : : : ; zn g.
Fact 1. For any infinite space Z we have w.Cp .Z// D w.Cp .Z// D .Cp .Z//.
Proof. We have .Cp .Z// w.Cp .Z// w.Cp .Z// D jZj (see Problem 169
of [TFS]) so it suffices to prove that jZj .Cp .Z//. Let u 2 Cp .Z/ be the
function which is identically zero on Z. Suppose that .Cp .Z// D and fix a
-base B with jBj at the point u in the space Cp .Z/. The family C of all
standard open subsets of Cp .Z/ is a base of the space Cp .Z/, so each U 2 B
contains a non-empty OU 2 C; it is evident that B 0 D fOU W U 2 Bg is also a
-base at theSpoint u.
If Y D fsupp.W / W W 2 B 0 g, then jY j jB 0 j jBj . Furthermore,
if x 2 ZnY , then G D x; . 1; 1/ is an open neighborhood of the point u. For
any W D z1 ; : : : ; zn I O1 ; : : : ; On 2 B 0 we have x supp.W / and hence there
exists a function f 2 C.Z/ such that f .zi / 2 Oi and f .x/ D 1. It is immediate
that f 2 W nG, which proves that W nG ; for any W 2 B 0 , i.e., G witnesses
that B 0 is not a -base at u. This contradiction proves that Y D Z and therefore
jZj . We have established that jZj .Cp .Z// and hence .Cp .Z// D
w.Cp .Z// D w.Cp .Z// D jZj so Fact 1 is proved. t
u
Returning to our solution observe that for an arbitrary space Z, we have w.Z/
H) w.Z/ H) .Z/ , so (i)H)(ii)H)(iii). It is also evident
that w.Z/ H) .Z/ H) .Z/ for any space Z, so
(i)H)(iv)H)(iii). Consequently, it suffices to show that (iii)H)(i).
Assume that Cp .X / is ./-monolithic and take any set A Cp .X / with
jAj ; let '.x/.f / D f .x/ for any x 2 X and f 2 A. Then '.x/ 2 Cp .A/ and
' W X ! Cp .A/ is a continuous map by Problem 166 of [TFS]. If Y D '.X /, then
' W X ! Y is a surjective map, so we can apply Fact 2 of T.139 to find a space
Z for which there exists an R-quotient continuous onto map p W X ! Z and a
condensation i W Z ! Y such that i p D '.
subspace and the dual map i W Cp .Y / ! Cp .Z/ embeds Cp .Y / in Cp .Z/ as
a dense subspace (see Problem 163 of [TFS]). Since w.Y / w.Cp .A// ,
the space Cp .Y / is has density and therefore so does Cp .Z/. If we choose
a set B p .Cp .Z// with jBj and dense in the subspace p .Cp .Z//, then
B D p .Cp .Z// so .p .Cp .Z/// by ./-monolithity of Cp .X /. As a
consequence, we have .Cp .Z// and hence w.Cp .Z// by Fact 1.
Recalling again that the spaces p .Cp .Z// and Cp .Z/ are homeomorphic we
conclude that w.p .Cp .Z/// ; it is straightforward that A ' .Cp .Y //
p .Cp .Z// and therefore A p .Cp .Z// whence w.A/ . This proves that
Cp .X / is strongly -monolithic, i.e., (iii)H)(i), so our solution is complete.
T.159. Given a space X and an infinite cardinal , prove that X is s ./-monolithic
if and only if Cp .X / is s ./-stable.
Solution. If Z is a space and E C.Z/, say that E separates points from closed
To apply the result of Problem 148 let D D s . The invariant s is not
increased by continuous images because if Y is a continuous image of a space Z,
then Y n is also a continuous image of Z n for every n 2 N (see Problem 157 of
[TFS]) which shows that satisfies the condition (i) from Problem 148.
For any A X we have s .A/ D s .Cp .A// s .Cp .AjX // (see

Problem 025). Given an arbitrary point x 2 A let '.x/.f / D f .x/ for each
f 2 E D Cp .AjX /. Then '.x/ 2 Cp .E/ and ' W A ! Cp .E/ is a continuous map
(see Problem 166 of [TFS]). It is easy to see that E separates points from closed sets
in A, so the map ' is actually an embedding (this was also proved in Problem 166
of [TFS]). As a consequence, s .A/ s .Cp .E// D s .E/ D s .Cp .AjX //
and hence s .A/ D s .Cp .AjX // for any A X . This shows that the condition
(ii) from Problem 148 of [TFS] is also satisfied (in a stronger form), so we can apply
it to the cardinal functions D D s . Thus Cp .X / is s ./-stable if and only if
X is s ./-monolithic.
T.160. Given an arbitrary space X and an infinite cardinal , prove that if Cp .X /
is spread./-stable, then X is spread./-monolithic.
Solution. Given a space Z, points z1 ; : : : ; zn 2 Z and sets O1 ; : : : ; On 2 .R/,
let z1 ; : : : ; zn I O1 ; : : : ; On Z D ff 2 Cp .Z/ W f .zi / 2 Oi for all i ng. Recall
that the sets z1 ; : : : ; zn I O1 ; : : : ; On Z are called standard open subsets of Cp .Z/.
If we have any set A Z, let A W Cp .Z/ ! Cp .A/ be the restriction map defined
by A .f / D f jA for any f 2 Cp .Z/. We will also need the space Cp .AjZ/ D
A .Cp .Z// Cp .A/. If A B Z, then AB W Cp .BjX / ! Cp .AjX / is also the
restriction map, i.e., AB .f / D f jA for any f 2 Cp .BjX /.
Fact 1. For an arbitrary space Z and any A Z we have s.A/ s.Cp .AjZ//.
Proof. Assume that s.Cp .AjZ// D and take any discrete D A. For any d 2 D
there is Ud 2 .d; Z/ such that Ud \ D D fd g. Fix any fd 2 Cp .Z/ such that
fd .d / D 1 and fd .ZnUd / f0g for every d 2 D. The set Wd D d; .0; 2/ A is an
open neighborhood of the function gd D A .fd / in Cp .A/ for each d 2 D. Observe
also that e 2 Dnfd g implies ge .d / D fe .d / D 0 and hence ge Wd . In particular,
the map A jD is an injection and the set E D A .D/ D fgd W d 2 Dg Cp .AjZ/
is discrete because Wd \ E D fgd g for every point d 2 D. As a consequence,
jDj D jEj s.Cp .AjZ// D which shows that s.A/ so Fact 1 is proved. u t
Returning to our solution take any A X with jAj and let B D A. The
space Cp .BjX / is a continuous image of Cp .X / and the map AB condenses the
space Cp .BjX / onto the space Cp .AjX /. Since w.Cp .AjX // jAj , we can
use spread./-stability of Cp .X / to conclude that s.Cp .BjX // . Finally, apply
Fact 1 to see that s.A/ D s.B/ s.Cp .BjX // and therefore the space X is
spread./-monolithic.
T.161. Give an example of a spread-monolithic space X such that Cp .X / is not
spread(!)-stable.
Solution. Let X be the Sorgenfrey line (see Problem 165 of [TFS]). Then s.X / D
! and hence X is spread-monolithic because for any infinite cardinal and any
A X with jAj , we have s.A/ s.X / D ! . The space Cp .X / has
an uncountable closed discrete subspace and hence s.Cp .X // > !; since X is

separable, Cp .X / condenses onto a second countable space. This, together with
s.Cp .X // > !, implies that Cp .X / is not spread.!/-stable.
T.162. Given an arbitrary space X and an infinite cardinal , prove that X is
s ./-stable if and only if Cp .X / is s ./-monolithic.
Solution. To apply the result of Problem 146 let D D s . The invariant
s is hereditary (see Problem 159 of [TFS]) and s .Cp .Z// D s .Z/ for any
space Z (see Problem 025) which shows that the pair . ; / satisfies the hypoth-
esis of Problem 146. Therefore Problem 146 is applicable and hence Cp .X / is
s ./-monolithic if and only if X is s ./-stable.
T.163. Prove that if Cp .X / is spread()-monolithic, then X is spread()-stable.
Solution. Take a continuous onto map f W X ! Y for which i w.Y / . The
dual map f W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / (see Problem 163 of
[TFS]). Furthermore, d.f .Cp .Y /// D d.Cp .Y // D i w.Y / (we applied
Problem 174 of [TFS] and the fact that f .Cp .Y // is homeomorphic to Cp .Y /).
Now apply spread./-monolithity of the space Cp .X / to conclude that we have
s.f .Cp .Y /// . As a consequence, s.Y / s.Cp .Y // D s.f .Cp .Y ///
(here we applied Problem 016 and the fact that Cp .Y / is homeomorphic to
f .Cp .Y //); this proves that Y is spread./-stable.
T.164. Give an example of a spread-stable space X such that the space Cp .X / is
not spread(!)-monolithic.
Solution. Let X be the Sorgenfrey line (see Problem 165 of [TFS]). Then s.X / D
! and hence X is spread-stable because for every continuous image Y of the space
X , we have s.Y / s.X / D ! (see Problem 157 of [TFS]) for any infinite
cardinal (the condensations of Y do not matter). On the other hand, the space
Cp .X / has an uncountable closed discrete subspace and hence s.Cp .X // > !. The
identity map condenses X onto R so d.Cp .X // D i w.X / D !. This, together with
s.Cp .X // > !, implies that Cp .X / is not spread.!/-monolithic.
T.165. Given an arbitrary space X and an infinite cardinal prove that X is
hd ./-monolithic if and only if Cp .X / is hl ./-stable.
Solution. If Z is a space and E C.Z/ say that E separates points from closed
To apply the result of Problem 148 let D hl and D hd . The invariant

hl is not increased by continuous images because if Y is a continuous image of a
space Z, then Y n is a continuous image of Z n for every n 2 N (see Problem 157 of
[TFS]). Thus satisfies the condition (i) from Problem 148.
For any A X we have hd .A/ D hl .Cp .A// hl .Cp .AjX // (see

Problem 027). Given an arbitrary point x 2 A let '.x/.f / D f .x/ for each
f 2 E D Cp .AjX /. Then '.x/ 2 Cp .E/ and ' W A ! Cp .E/ is a continuous map
(see Problem 166 of [TFS]). It is easy to see that E separates points from closed sets
in A, so the map ' is actually an embedding (this was also proved in Problem 166 of
[TFS]). As a consequence, hd .A/ hd .Cp .E// D hl .E/ D hl .Cp .AjX //
and hence hd .A/ D hl .Cp .AjX // for any A X . This shows that the condition
(ii) from Problem 148 of [TFS] is also satisfied (in a stronger form), so we can apply
it to the cardinal functions D hl and D hd . Thus Cp .X / is hl ./-stable if
and only if X is hd ./-monolithic.
T.166. Given an arbitrary space X and an infinite cardinal prove that if Cp .X / is
hl./-stable, then X is hd./-monolithic.
Solution. Given a space Z, points z1 ; : : : ; zn 2 Z and sets O1 ; : : : ; On 2 .R/,
let z1 ; : : : ; zn I O1 ; : : : ; On Z D ff 2 Cp .Z/ W f .zi / 2 Oi for all i ng. Recall
that the sets z1 ; : : : ; zn I O1 ; : : : ; On Z are called standard open subsets of Cp .Z/.
If we have any set A Z, let A W Cp .Z/ ! Cp .A/ be the restriction map defined
by A .f / D f jA for any f 2 Cp .Z/. We will also need the space Cp .AjZ/ D
A .Cp .Z// Cp .A/. If A B Z, then AB W Cp .BjX / ! Cp .AjX / is also the
restriction map, i.e., AB .f / D f jA for any f 2 Cp .BjX /.
Fact 1. For an arbitrary space Z and any A Z we have hd.A/ hl.Cp .AjZ//.
Proof. Assume that hl.Cp .AjZ// D and we have a set D D fd W < C g A
which is left-separated by its indexation. For any < C there is U 2 .d ; Z/
such that U \ D fd W g. Fix any f 2 Cp .Z/ such that f .d / D 1
and f .ZnU / f0g for every < C . The set W D d ; .0; 2/ A is an open
neighborhood of g D A .f / in Cp .A/ for each < C . Observe also that <
implies g .d / D f .d / D 0 and hence g W . This implies that, the map A jD
is an injection. Besides, the set E D A .D/ D fg W < C g Cp .AjZ/ is right-
separated because W \ E fg W g for each < C . As a consequence,
jEj D jDj D C > D hl.Cp .AjZ// which is a contradiction with Problem 005.
Thus A has no left-separated subsets of cardinality C (see Fact 2 of T.004) which
proves that hd.A/ by Problem 004. Fact 1 is proved. t
u
Returning to our solution take any A X with jAj and let B D A. The
space Cp .BjX / is a continuous image of Cp .X / and the map AB condenses the
space Cp .BjX / onto the space Cp .AjX /. Since w.Cp .AjX // jAj , we can
use hl./-stability of Cp .X / to conclude that hl.Cp .BjX // . Finally, apply
Fact 1 to see that hd.A/ D hd.B/ hl.Cp .BjX // and therefore the space X
is hd./-monolithic.
T.167. Give an example of an hd -monolithic space X such that Cp .X / is not hl.!/-
stable.
Solution. Let X be the Sorgenfrey line (see Problem 165 of [TFS]). Then
hd.X / D ! and hence X is hd -monolithic because for any infinite cardinal and
any A X with jAj , we have hd.A/ hd.X / D ! . The space Cp .X /

has an uncountable closed discrete subspace and hence hl.Cp .X // > !; since X
is separable, Cp .X / condenses onto a second countable space. This, together with
hl.Cp .X // > !, implies that Cp .X / is not hl.!/-stable.
hd ./-stable if and only if Cp .X / is hl ./-monolithic.
Solution. To apply the result of Problem 146 let D hl and D hd . The
invariant hl is hereditary (see Problem 159 of [TFS]) and hl .Cp .Z// D hd .Z/
for any space Z (see Problem 027) which shows that the pair . ; / satisfies the
hypothesis of Problem 146. Therefore Problem 146 is applicable and hence Cp .X /
is hl ./-monolithic if and only if X is hd ./-stable.
T.169. Given an arbitrary space X and an infinite cardinal , prove that if Cp .X /
is hl./-monolithic, then X is hd./-stable.
Solution. Take a continuous onto map f W X ! Y for which i w.Y / . The dual
map f W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / (see Problem 163 of [TFS]).
Furthermore, d.f .Cp .Y /// D d.Cp .Y // D i w.Y / (we applied Problem 174
of [TFS] and the fact that f .Cp .Y // is homeomorphic to Cp .Y /). Now apply
hl./-monolithity of the space Cp .X / to conclude that we have hl.f .Cp .Y ///
. As a consequence, hd.Y / hl.Cp .Y // D hl.f .Cp .Y /// (here we
applied Problem 017 and the fact that Cp .Y / is homeomorphic to f .Cp .Y //);
this proves that Y is hd./-stable.
T.170. Give an example of an hd -stable space X such that the space Cp .X / is not
hl.!/-monolithic.
Solution. Let X be the Sorgenfrey line (see Problem 165 of [TFS]). Then hd.X / D
! and hence X is hd -stable because for every continuous image Y of the space X ,
we have hd.Y / hd.X / D ! (see Problem 157 of [TFS]) for any infinite
Cp .X / has an uncountable closed discrete subspace and hence hl.Cp .X // > !.
The identity map condenses X onto R, so d.Cp .X // D i w.X / D !. This, together
with hl.Cp .X // > !, implies that Cp .X / is not hl.!/-monolithic.
hl ./-stable if and only if Cp .X / is hd./-monolithic.
Solution. To apply the result of Problem 146 let D hd and D hl .
The invariant hd is hereditary and hd.Cp .Z// D hl .Z/ for any space Z
(see Problem 030) which shows that the pair . ; / satisfies the hypothesis
of Problem 146. Therefore Problem 146 is applicable and hence Cp .X / is
hd./-monolithic if and only if X is hl ./-stable.
T.172. Give an example of an hl-stable space X such that the space Cp .X / is not
hd.!)-monolithic.
Solution. Let X be the Sorgenfrey line (see Problem 165 of [TFS]). Then hl.X / D
! and hence X is hl-stable because for every continuous image Y of the space X ,
we have hl.Y / hl.X / D ! (see Problem 157 of [TFS]) for any infinite
Cp .X / has an uncountable closed discrete subspace and hence hd.Cp .X // > !.
The identity map condenses X onto R so d.Cp .X // D i w.X / D !. This, together
with hd.Cp .X // > !, implies that Cp .X / is not hd.!/-monolithic.
hl ./-monolithic if and only if Cp .X / is hd./-stable.
let nij .Z/ S
Call a set W 2 .Z n / marked if W D W1

Wn where fWi W i 2 Mn g .Z/
and Wi \ Wj D ; for any distinct i; j 2 Mn . If Y is a space and Z Y , let
Z W Cp .Y / ! Cp .Z/ be the restriction map defined by Z .f / D f jZ for any
f 2 Cp .Y /. We will also need the space Cp .ZjY / D Z .Cp .Y // Cp .Z/.
Fact 1. Suppose that Y is a space and Z Y ; given n 2 N and any marked set
W Z n , we have hl.W / hd.Cp .ZjY //.
Proof. Assume that hd.Cp .ZjY // D ; we have W D W1

Wn where the
family fWi W i 2 Mn g consists of open subsets of Z and Wi \ Wj D ; for any
distinct i; j 2 Mn . If hl.W / > , then there exists a set P D fz W < C g W
which is right-separated by its indexation (see Problem 005 and Fact 2 of T.005).
For any < C , we have z D .z1 ; : : : ; zn / where zi 2 Wi for any i 2 Mn .
For each < C , there exist disjoint O1 ; : : : ; On 2 .Y / with the following
properties:

(3) Oi \ Z Wi for all i 2 Mn .
For any < C and i 2 Mn , take a function fi 2 C.Y; 0; 1 / such that
D 1 and fi .Y nOi / f0g; let f D f1 C

C fn and g D Z .f /.
fi .zi /
We will prove that the set E D fg W < C g Cp .ZjY / is left-separated by its
indexation.
Let U D ff 2 Cp .ZjY / W f .zi / > 0 for all i 2 Mn g. It is clear that U is
an open subset of Cp .ZjY / and g 2 U for all < C . Assume that < and
j
g 2 U . Then, for any i 2 Mn , we have g .zi / > 0 and hence zi 2 O for some
j
j 2 Mn . However, zi 2 Wi and O \ Wi D ; for any j i . Therefore zi 2 Oi
for all i 2 Mn whence z 2 O which contradicts (2). This contradiction shows that
we have
. / g U for any < < C and, in particular, the map z ! g is a bijection
between P and E.
Take any < C ; given any < , we have g U by . /. Therefore

U \ E D ; where E D fg
W
< g. This shows that the set E is closed in
E for each < C , and hence E is a left-separated (by its indexation) subspace of
Cp .ZjY /. Since jEj D C by . /, we have a contradiction with hd.Cp .ZjY //
(see Problem 004). Fact 1 is proved. t
u
Fact 2. Let Y be a space; then, for any Z Y , we have l .Z/ D t.Cp .ZjY //.
Proof. We have t.Cp .ZjY // t.Cp .Z// D l .Z/ by Problem 149 of [TFS]. To
prove the converse inequality assume that t.Cp .ZjY // D and take any !-cover U
of the space Z (i.e., U .Z/, and for any finite K Z, there is U 2 U such that
K U ). Consider the set A D ff 2 Cp .ZjY / W supp.f / D f 1 .Rnf0g/ U for
some U 2 Ug.
If K Z is finite, then there exists U 2 U with K U ; take any V 2 .Y /
such that V \ Z D U and a function g 2 Cp .Y / for which g.K/ f1g and
g.Y nV / f0g. For the function f D Z .g/ we have supp.f / U and f jK 1.
The finite set K was chosen arbitrarily, so we proved that w 2 A for the function
w 2 Cp .ZjY / defined by w.z/ D 1 for every z 2 Z. There is B A with jBj
and w 2 B; for every f 2 B, there is U.f / 2 U such that supp.f / U.f /. The
family U 0 D fU.f / W f 2 Bg U has cardinality . If K Z is finite, then the
set H D ff 2 Cp .ZjY / W f .z/ > 0 for all z 2 Kg is an open neighborhood of w in
Cp .ZjY / and hence H \ B ;. If f 2 B \ H , then f .z/ > 0 for every z 2 K and
hence K supp.f / U.f / 2 U 0 . Thus U 0 is an !-cover of Z, so we proved that
every !-cover of Z has an !-subcover of cardinality . Therefore l .Z/ by
Problem 148 of [TFS] and Fact 2 is proved. t
u
Fact 3. Suppose that Y is a space and Z Y ; given any natural number n 2, we
have hl.Z n nU / hd.Cp .ZjY // for any U 2 .n .Z/; Z n /.
Proof. Let hd.Cp .ZjY // D ; observe that t.Cp .ZjY // hd.Cp .ZjY // D
and therefore l .Z/ by Fact 2. The set Z n nU is closed in Z n , so l.Z n nU / .
For every z 2 F D Z n nU , there exists a marked set Vz 2 .z; Z n /. The family
V D fVz W z 2 FS g is an open cover of the space F , so there is V 0 V such
that jVSj and V 0 F . We have hl.V / for each V 2 V 0 by Fact 1,
0
so hl. V 0 / (it is an easy exercise that a union of -many spaces with

hereditary LindelfS number has hereditary Lindelf number ). Therefore
hl.Z n nU / hl. V 0 / and Fact 3 is proved. t
u
Fact 4. We have .n .Z/; Z n / .Z/ for any space Z.
Proof. Fix n 2 N with T n 2; take an arbitrary family U .Z Z/ such that

jUj D .Z/ and U D D 2 .Z/. Given distinct i; j 2 Mn , let qij W
Z n ! Z Z be the natural projection onto the face defined by i and j , i.e., for
any z D .z1 ; : : : ; zn / 2 Z n , we have qij .z/TD .zi ; zj / 2 Z Z. It is clear that
nij .Z/ D qij1 ./ and therefore nij .Z/ D Uij where Uij D fqij1 .U / W S U 2 Ug.
.i; j / 2 Bn g W Uij 2 Uij for all .i; j / 2 Bn g consists of open subsets of Z n and
T
V D n .Z/. It is evident that jVj , so Fact 4 is proved. t
u
Fact 5. If Y is a space and Z Y , then hl .Z/ D hd.Cp .ZjY //.

Proof. We have hd.Cp .ZjY // hd.Cp .Z// D hl .Z/ by Problem 030. To
prove the converse inequality assume that hd.Cp .ZjY // D ; since Cp .ZjY / is
dense in Cp .Z/ (see Problem 152 of [TFS]), we have i w.Z/ D d.Cp .Z//
(see Problem 174 of [TFS]). It is an easy exercise that .Z/ i w.Z/, so we have
.Z/ and hence .n .Z/; Z n / T for any n 2 by Fact 4.
S nFix a family V .Z n
/ such that V D n .Z/; it is clear that Z n nn .Z/ D
fZ nV W V 2 Vg. For any set F 2 S F D fZ n nV W V 2 Vg, we have hl.F / by
Fact 3. It follows from jF j and F D Z n nn .Z/ that hl.Z n nn .Z// .
Now apply Fact 0 of T.021 to conclude that hl.Z n / D hl.Z n nn .Z// . Thus
hl.Z n / for every n 2 and hence hl .Z/ . This implies hl .Z/ D so
Fact 5 is proved. t
u
Returning to our solution, let D hd and D hl . It is evident that hereditary
density is not raised by continuous maps, so the condition (i) of Problem 148
is satisfied. Furthermore, Fact 5 shows that the condition (ii) of Problem 148 is
satisfied as well (even in a stronger form). Therefore it is applicable to the pair
. ; /, so Cp .X / is hd./-stable if and only if X is hl ./-monolithic.
T.174. Give an example of an hl-monolithic space X such that Cp .X / is not
hd.!)-stable.
!, and hence X is hl-monolithic because for any infinite cardinal and any A X
with jAj , we have hl.A/ hl.X / D ! . The space Cp .X / has an
uncountable closed discrete subspace and hence hd.Cp .X // > !; since X is
separable, Cp .X / condenses onto a second countable space. This, together with
hd.Cp .X // > !, implies that Cp .X / is not hd.!/-stable.
p./-stable if and only if Cp .X / is a./-monolithic.
Solution. Let D a and D p. It is evident that the Alexandroff number is
hereditary; besides, a.Cp .Z// D p.Z/ for any space Z by Problem 178 of [TFS].
Consequently, the pair . ; / satisfies the hypothesis of 146 which shows that X is
p./-stable if and only if Cp .X / is a./-monolithic.
l ./-stable if and only if Cp .X / is t./-monolithic.
Solution. Let D t and D l . It is evident that tightness is hered-
itary; besides, t.Cp .Z// D l .Z/ for any space Z by Problem 149
of [TFS]. Consequently, the pair . ; / satisfies the hypothesis of Prob-
lem 146 which shows that X is l ./-stable if and only if Cp .X / is
t./-monolithic.
d./-stable if and only if Cp .X / is i w./-monolithic.
Solution. Let D i w and D d . It is evident that the i -weight is hereditary;

besides, i w.Cp .Z// D d.Z/ for any space Z by Problem 173 of [TFS]. Conse-
quently, the pair . ; / satisfies the hypothesis of Problem 146 which shows that X
is d./-stable if and only if Cp .X / is i w./-monolithic.
T.178. Given an arbitrary space X and an infinite cardinal , prove that the
following conditions are equivalent:
(i) Cp .X / is i w./-monolithic;
(ii) Cp .X / is ./-monolithic;
(iii) Cp .X / is ./-monolithic.
Solution. It is evident that .i w.Z/ / H) ..Z/ / H) . .Z/ / for
any space Z which shows that (i)H)(ii)H)(iii).
Now assume that Cp .X / is ./-monolithic and take any A Cp .X / with
jAj . Let '.x/.f / D f .x/ for any x 2 X and f 2 A. Then '.x/ 2 Cp .A/
and ' W X ! Cp .A/ is a continuous map by Problem 166 of [TFS]. If Y D '.X /,
then ' W X ! Y is a surjective map, so we can apply Fact 2 of T.139 to find a space
Z for which there exists an R-quotient continuous onto map p W X ! Z and a
condensation i W Z ! Y such that i p D '.
subspace and the dual map i W Cp .Y / ! Cp .Z/ embeds Cp .Y / in Cp .Z/ as a
dense subspace (see Problem 163 of [TFS]). Since w.Y / w.Cp .A// , the
space Cp .Y / has density and therefore d.Cp .Z// .
If we choose a set B p .Cp .Z// with jBj which is dense in the
subspace p .Cp .Z//, then B D p .Cp .Z//, so .p .Cp .Z/// by ./-
monolithity of Cp .X /. As a consequence, .Cp .Z// and hence i w.Cp .Z//
by Problem 173 of [TFS]. Recalling again that p .Cp .Z// and Cp .Z/ are
homeomorphic, we conclude that i w.p .Cp .Z/// . It is straightforward that
A ' .Cp .Y // p .Cp .Z// and therefore A p .Cp .Z// which implies
i w.A/ . Thus Cp .X / is i w./-monolithic which proves (iii)H)(i) and hence
(i) (ii) (iii).
T.179. Give an example of a space X which is pseudocharacter-monolithic but not
diagonal-number-monolithic.
Solution. Let X be the double arrow space (see Problem 384 of [TFS]). Then
X is a separable first countable non-metrizable space. The space X is even -
monolithic because for any infinite cardinal and any A X with jAj , we
have .A/ .X / D ! . On the other hand, .X / > ! because X is not
metrizable (see Problem 091). This, together with separability of X , implies that X
is not .!/-monolithic.
T.180. Give an example of a space X which is diagonal-number-monolithic but not
i -weight-monolithic.
S
Solution. Call a space Z strongly -discrete if Z D fZn W n 2 !g where Zn
is closed and discrete in Z for each n 2 !. It is easy to see that every subset of
a strongly -discrete space Z is an F -subset of Z. Another easy fact (which we

leave to the reader as an exercise) is that the square of a strongly -discrete is also
strongly -discrete. Thus, if a space Z is strongly -discrete, then every subset of
Z 2 is an F -set. In particular, if D f.z; z/ W z 2 Zg Z 2 is the diagonal of Z,
then Z 2 n is an F -subset of Z 2 and hence is a G -set in Z 2 , i.e., .Z/ D !.
Now let X be the Mrowka space (see Problem 142 of [TFS]). Then X is
a pseudocompact non-compact space such that X D ! [ M where all points
of ! are isolated in X and M is closed and discrete in X . Consequently, the
space X is strongly -discrete and therefore .Z/ D !. This shows that Z
is -monolithic because for any infinite cardinal and any A X , we have
.A/ .X / D ! .
Observe that i w.X / > ! because any condensation of a pseudocompact space
onto a second countable one is a homeomorphism (see Problem 140 of [TFS]), so if
i w.X / D !, then X is metrizable and hence compact which is a contradiction. The
set ! is dense in X , so X is separable which, together with .X / > !, implies that
X is not .!/-monolithic.
tm ./-R-quotient-stable if and only if Cp .X / is q./-monolithic.
Solution. Let D q and D tm . The HewittNachbin number q is closed-
hereditary by Problem 422 of [TFS]; besides, q.Cp .Z// D tm .Z/ for any space
Z by Problem 429 of [TFS]. Consequently, the pair . ; / satisfies the hypothesis of
Problem 147 which shows that X is tm ./-R-quotient-stable if and only if Cp .X / is
q./-monolithic.
T.182. Give an example of a space X for which Cp .X / is q-monolithic and X is not
tm .!/-stable.
Solution. Let X be a discrete space of cardinality c; then q.Cp .X // D ! (see
Problem 429 of [TFS]). Consequently, for any infinite cardinal and any set A
Cp .X /, we have q.A/ q.Cp .X // D ! (see Problem 422 of [TFS]) so Cp .X /
is q-monolithic.
Now if M is the Mrowka space (see Problem 142 of [TFS]), then M is
a pseudocompact non-compact separable space. This implies that q.M / > !
by Problem 407 of [TFS] and hence we have tm .Cp .M // D q.M / > ! (see
Problem 434 of [TFS]). Since M is separable, the space Cp .M / condenses onto
a second countable space by Problem 173 of [TFS] and therefore jCp .M /j c.
As a consequence, there is a surjective map ' W X ! Cp .M / which is continuous
because X is discrete. Thus Cp .M / is a continuous image of the space X such that
i w.Cp .M // D ! < tm .Cp .M // which proves that X is not tm .!/-stable.
T.183. Given an infinite cardinal , prove that X is l ./-monolithic if and only if
Cp .X / is t./-quotient-stable.
Solution. Let D t and D l . Tightness t is an invariant of quotient maps by
Problem 162 of [TFS]; this shows that the condition (i) of Problem 149 is satisfied
for . Besides, t.Cp .AjX // D l .A/ for any A X by Fact 2 of T.173; hence
the condition (ii) is also satisfied in a stronger form. Therefore Problem 149 is
applicable to the pair . ; / and hence X is l ./-monolithic if and only if Cp .X /
is t./-quotient-stable.
T.184. Given an infinite cardinal , suppose that Cp .X / is l./-monolithic. Prove
that the space X is tightness ./-R-quotient stable.
Solution. Take any R-quotient continuous surjective map ' W X ! Y such that
i w.Y / . The map ' W Cp .Y / ! Cp .X / embeds Cp .Y / in Cp .X / as a closed
subspace (see Problem 163 of [TFS]). We have d.' .Cp .Y /// D d.Cp .Y // D
i w.Y / , so there is a set A ' .Cp .Y // such that jAj and A D
' .Cp .Y //. The space Cp .X / being l./-monolithic, we have l.' .Cp .Y ///
and hence l.Cp .Y // . Now apply Problem 189 of [TFS] to conclude that
t .Y / which shows that X is t ./-R-quotient-stable.
T.185. Given an arbitrary space X , suppose that Cp .X / is t./-monolithic for some
infinite cardinal . Prove that it is t ./-monolithic.
Solution. Take any A Cp .X / with jAj ; for any x 2 X let '.x/.f / D f .x/
for every f 2 A. Then '.x/ 2 Cp .A/ for each x 2 X and the map ' W X ! Cp .A/
is continuous (see Problem 166 of [TFS]). If Y D '.X /, then we can apply Fact 2
of T.139 to find a space Z and a continuous R-quotient onto map p W X ! Z such
that i p D ' for some condensation i W Z ! Y .
If we choose a set B p .Cp .Z// with jBj which is dense in the
subspace p .Cp .Z//, then B D p .Cp .Z//, so t.p .Cp .Z/// by t./-
monolithity of Cp .X /. As a consequence, t.Cp .Z// and hence t .Cp .Z//
by Problem 150 of [TFS]. Recalling again that p .Cp .Z// and Cp .Z/ are
homeomorphic, we conclude that t .p .Cp .Z/// . It is straightforward that
A ' .Cp .Y // p .Cp .Z// and therefore A p .Cp .Z// which implies
t .A/ . Thus Cp .X / is t ./-monolithic
T.186. Suppose that Cp .X / is FrchetUrysohn./-monolithic for some infinite
cardinal . Prove that it is FrchetUrysohn./-monolithic.
Solution. Take any A Cp .X / with jAj ; for any x 2 X , let '.x/.f / D f .x/
for every f 2 A. Then '.x/ 2 Cp .A/ for each x 2 X and the map ' W X ! Cp .A/
If we choose a set B p .Cp .Z// with jBj which is dense in
the subspace p .Cp .Z//, then B D p .Cp .Z//, so p .Cp .Z// is a Frchet
Urysohn space by FrchetUrysohn./-monolithity of Cp .X /. As a consequence,
Cp .Z/ is also a FrchetUrysohn space and hence .Cp .Z//! is also Frchet
Urysohn by Problem 145 of [TFS]. Recalling again that p .Cp .Z// and Cp .Z/
are homeomorphic, we conclude that .p .Cp .Z///! is FrchetUrysohn; since
A ' .Cp .Y // p .Cp .Z//, we have A p .Cp .Z// which implies that .A/!
is a FrchetUrysohn space. Thus Cp .X / is FrchetUrysohn./-monolithic.
T.187. Given an infinite cardinal , prove that X is -scattered if and only if Cp .X /
is w./-open-stable.
Solution. The definition of -scattered space shows that being -scattered is the
same as being cardinality./-monolithic. Let D w and Dcardinality. Since open
continuous maps do not increase weight (see Problem 161 of [TFS]), the condition
(i) of Problem 151 is satisfied for .
Fact 1. Let Z be an arbitrary space. If Y is dense in Z, then w.Y / D w.Z/.
Proof. If B is a -base in Z, then fU \ Y W U 2 Bg is a -base in Y , so w.Y /
w.Z/. Now take any -base C in Y . For each V 2 C, fix OV 2 .Z/ such that
OV \ Y D V . We claim that the family B D fOV W V 2 Cg is a -base in Z. To
see this take any W 2 .Z/; there is W1 2 .Z/ such that W 1 W . Choose any
V 2 C with V W1 \ Y . Then U D OV 2 B and U U D OV W 1 W ,
and hence B is a -base in Z. Since jBj jCj, we showed that w.Z/ w.Y /
u
Fact 2. If Z is a space and Y Z, then w.Cp .Y jZ// D w.Cp .Y jZ// D jY j.
Proof. It is evident that w.Cp .Y jZ// w.Cp .Y jZ// w.Cp .Y // D jY j. On
the other hand, Cp .Y jZ/ is dense in Cp .Y / (see Problem 152 of [TFS]), so we
can apply Fact 1 to conclude that w.Cp .Y // D w.Cp .Y jZ//. Now we have
jY j D w.Cp .Y // D w.Cp .Y // by Fact 1 of T.158. This implies w.Cp .Y jZ// D
w.Cp .Y jZ// D jY j, so Fact 2 is proved. t
u
Returning to our solution observe that Fact 2 shows that the condition (ii) of
Problem 151 is also satisfied (in a stronger form) for the cardinal functions and .
Thus Problem 151 is applicable to the pair . ; / and therefore X is -scattered if
and only if Cp .X / is w./-open-stable.
T.188. Let X be a d.!/-stable space such that X n is Hurewicz for all n 2 N. Prove
that for any A Cp .X / and any f 2 AnA, there is a discrete D A such that f
is the only accumulation point of D.
Solution. Given a space Z say that a family U is an open !-cover of Z if U .Z/
and for any finite K Z, there is U 2 U with K U .
Fact 1. The following properties are equivalent for any space Z:

(1) Z n is a Hurewicz space for any n 2 N.
(2) If Uk is an open !-cover of Z for allSk 2 !. then we can choose a finite Uk0 Uk
for each k 2 ! in such a way that fUk0 W k 2 !g is an !-cover of Z.
Proof. (1)H)(2).
S Find a disjoint family A D fAk W k 2 Ng of infinite subsets of !
such that A D !. Given any n 2 N, the collection Vkn D fU n W U 2 Uk g is an
open cover of Z n for any k 2 !, so fVkn W k 2 An g is a countably infinite family of
open covers of Z n . By theS Hurewicz property of Z n , for all k 2 An , there exists a
finite Wk Vk such that fWkn W k 2 An g is a cover of Z n . Fix a finite Ukn Uk
n n
such that Wkn D fU n W U 2 Ukn g for all n 2 N and k 2 An .

Once we have the collections fUkn W k 2 An g for each n 2 N, take any k 2 !
and find the unique n D n.k/ 2 N such that k 2 An . Letting Uk0 D Ukn , we obtain a
finite Uk0 Uk for every k 2 !. Given any finite set K D fz1 ; : : : ; zn g Z we have
z D .z1 ; : : : ; zn / 2 Z n and hence there is k 2 An and W 2 Wkn such that z 2 W .
We have WSD U n for some U 2 Ukn D Uk0 , so z 2 U n and hence K U . This
shows that fUk0 W k 2 !g is an !-cover of Z.
(2)H)(1). Fix an arbitrary n 2 N and take any collection fVk W k 2 !g of
open covers of Z n . Call a finite family U of open subsets of Z k-small if for any
U1 ; : : : ; Un 2 U (which are not necessarily distinct), there is V 2 Vk such that
U1

Un V . It is easy to see that for each k-small family .Z/, there
exists a finite W./ Vk such that for any U1 ; : : : ; Un 2 there is V 2 W./ for
which U1

Un V .
If F Z is a finite set, then for any z1 ; : : : ; zn 2 F we have z D .z1 ; : : : ; zn / 2 V
for some V 2 Vk , so there are U1 ; : : : ; Un 2 .Z/ such that zi 2 Ui for all i n and
U1

Un V . Taking all possible n-tuples of the elements of F and intersecting
the respective neighborhoods S of every z 2 F , we obtain a k-small family
S D fWz W
z 2 F g such that F . As a consequence, the family Uk D f W .Z/
is k-smallg is an !-cover of Z for every k 2 !. The property (2) guarantees
S that we
can choose a finite Uk0 Uk for every k 2 ! in such a way that fUk0 W k 2S !g is
an !-cover ofS Z. For each U 2 Uk0 , take a k-small family U such that U D U
and let Vk0 D fW.U / W U 2 Uk0 g. It is evident that Vk0 Vk and jVk0 j < ! for all
k 2 !. S
To see that fVk0 W k 2 !g is a cover of Z n , take any z D .z1 ; : : : ;S zn / 2
Z n . There is k 2 ! and U 2 Uk0 such that F D fz1 ; : : : ; zn g U D U .
Therefore we can choose U1 ; : : : ; Un 2 U such that zi 2 Ui for all i n. There
is V 2 W.U / with W D U1

Un V ; since W.US / Vk0 , we have
z 2 W V 2 Vk . The point z 2 Z was chosen arbitrarily, so fVk0 W k 2 !g is
0 n
a cover of Z n which proves that Z n is a Hurewicz space for each n 2 N. Fact 1 is

proved. t
u
Fact 2. The following properties are equivalent for any space Z:
(3) Z n is a Hurewicz space for any n 2 N; T
(4) if Ak Cp .Z/ for every k 2 ! and u 2 fAk W k 2 !g, then for each k 2 !,
S
we can choose a finite Bk Ak in such a way that u 2 fBk W k 2 !g.
Proof. (3)H)(4). Since Cp .Z/ is homogeneous (i.e., for any f; g 2 Cp .Z/, there
is a homeomorphism ' W Cp .Z/ ! Cp .Z/ such that '.f / D g (see Problem 079
of [TFS])), it suffices to prove (4) only for the function u which is identically zero
on Z. Let Ukn D ff 1 .. n1 ; n1 // W f 2 Ak g for all k 2 ! and n 2 N. Since u 2 Ak
for each k 2 !, the family Ukn is an !-cover of Z for each n 2 N. Fact 1 shows that
we can use the property (2) of Fact 1 instead of (3). For any n 2 N apply (2) to the
countable sequence fUkn W k nS 1g of !-covers of Z to find, for all k n 1, a
finite family Vkn Ukn such that fVkn W k n 1g is an !-cover of Z.
For any U 2 Vkn , there is a function fU 2 Ak such that U D fU1 . . n1 ; n1 //. The
set Bk D ffU W U 2 Vkn W n k C 1g Ak is finite for each k 2 !. We claim
S
that u 2 fBk W k 2 !g. To see this, take S any finite P Z and " > 0; it suffices
to show that there is a point f 2 B D fBk W k 2 !g such S that jf .z/j < " for all
z 2 P . Fix any n 2 N with n1 < "; since the family W D fVkn W k n 1g is an
!-cover of Z, there is k n 1 and U 2 Vkn such that P U . Then fU 2 Bk and
jfU .z/j < n1 < " for each z 2 P . This proves that u 2 B.
(4)H)(3). Let w 2 Cp .Z/ be defined by w.z/ D 1 for all z 2 Z. By Fact 1, it
suffices to show that (4)H)(2), so take any sequence fUk W k 2 !g of open !-covers
of Z. If Ak D ff 2 Cp .Z/ W supp.f / D f 1 .Rnf0g/ U for some U 2 Uk g,
then it is an easy exercise that w 2 Ak for each k 2 !. ApplyS (4) to choose, for each
k 2 !, a finite Bk Ak such that w 2 B where B D fBk W k 2 !g. For each
f 2 Bk there is U.f / 2 Uk such that supp.f / U.f /. The familyS Uk0 D fU.f / W
f 2 Bk g Uk is finite for each k 2 !. To see that the family U D fUk0 W k 2 !g
0
is an !-cover of Z, take any finite P Z. The set O D ff 2 Cp .Z/ W f .z/ > 0

for each z 2 P g is an open neighborhood of w, so there are k 2 ! and f 2 Bk such
that f 2 O. Therefore P supp.f / U.f / 2 Uk0 U 0 which shows that U 0 is
an !-cover of Z and Fact 2 is proved. t
u
Returning to our solution, say that vet.Cp .Z// D ! to abbreviate the fact that
(4) is fulfilled for the space Z. We have vet.Cp .X // D ! by Fact 2. Observe that
any Hurewicz space is Lindelf so t.Cp .X // D ! by Problem 149 of [TFS]; this
implies that there is a countable B A such that f 2 BnB. Since X is d.!/-
stable, the space Cp .X / is i w.!/-monolithic by Problem 177 and hence the space
B has countable pseudocharacter. Therefore we can find aTfamily fOn W n 2 !g 2
.f; Cp .X // such that O nC1 On for each n 2 ! and . fOn W n 2 !g/ \ B D
ff g. If Cn D On \ B, then f 2 C n for all n 2 !. Since vet.Cp .X // D !, we
S
can find a finite Dn Cn such that f 2 D where D D fDn W n 2 !g. To see
that D is discrete, observe that D B and hence all accumulation points of D
belong to B. Furthermore, the set Fn D .Cp .X S/nO n / \ D D0 [

[ Dn1 is
finite and Fn 2 .D/ for each n 2 !. Since fFn W n 2 !g D D, each element
of D has a finite neighborhood in D which shows that D is discrete. Finally, if
g 2 Bnff g, then g 2 Cp .X /nO n for some n 2 ! and hence g has a neighborhood
whose intersection with D is finite. Therefore f is the unique accumulation point
of D and our solution is complete.
T.189. Let X be an l.!/-monolithic space of countable spread. Prove that X is
Lindelf.
Solution. Let U be an open cover of the space X . Apply Fact 1 of T.007

S to find
a countable family U 0 U and a discrete D X such that X D . U 0 / [ D.
The set D is countable because s.X / D !; the space X being l.!/-monolithic,
S D
is Lindelf and therefore there is a countable U 00 U such that D U 00 . The
family V D U 0 [ U 00 is, evidently, a countable subcover of U which proves that X
is Lindelf.
T.190. Let X be an hl.!/-monolithic space of countable spread. Prove that X
is hereditarily Lindelf. In particular, if X is an !-monolithic space of countable
spread, then X is hereditarily Lindelf.
Solution. It is evident that hl.!/-monolithity is hereditary and, in particular, any
Y X is l.!/-monolithic. We also have s.Y / s.X / D ! which shows that
Problem 189 can be applied to conclude that Y is Lindelf. Therefore every Y X
is Lindelf, i.e., X is hereditarily Lindelf.
T.191. Let X be an !-monolithic space such that Cp .X / has countable spread.
Prove that hl .X / D ! and hence hd .Cp .X // D !.
Solution. We have s.X X / s.Cp .X // D ! by Problem 016. The space X X is
!-monolithic by Problem 114, so we can apply Problem 190 to conclude that hl.X
X / D ! and hence .X / D !. Now Problems 025 and 028 show that s .X / D
s .Cp .X // D !. For any n 2 N, the space X n is !-monolithic by Problem 114;
since s.X n / D !, we can apply Problem 190 again to see that hl.X n / D !. Thus
hl .X / D ! and hence hd .Cp .X // D ! by Problem 026.
T.192. Suppose that a space X is !-monolithic and !-stable. Prove that the equality
s.Cp .X // D ! implies that nw.X / D !.
Solution. Since X is !-stable, the space Cp .X / is !-monolithic (see Problem 154).
This, together with s.Cp .X // D !, implies that hl.Cp .X // D ! (see Problem 190)
and hence d.X / hd.X / D ! by Problem 017. Thus X is !-monolithic and
separable; an immediate consequence is that nw.X / D !.
T.193. Assume SA. Prove that for any space X , if Cp .X / is s.!/-monolithic, then
it is hl .!/-monolithic. In particular, if Cp .X / is hl.!/-monolithic, then it is
hl .!/-monolithic.
Solution. Take any countable A Cp .X /; for any x 2 X , let '.x/.f / D f .x/ for
every f 2 A. Then '.x/ 2 Cp .A/ for each x 2 X and the map ' W X ! Cp .A/
dense subspace (see Problem 163 of [TFS]). Since w.Y / w.Cp .A// D !, the
space Cp .Y / is separable and therefore d.Cp .Z// D !.
If we choose a countable set B p .Cp .Z// which is dense in the subspace

p .Cp .Z//, then B D p .Cp .Z//, so s.p .Cp .Z/// by s.!/-monolithity of

Cp .X /. As a consequence, s.Cp .Z// D ! and hence hl .Cp .Z// D ! by SA and

Problem 036. Recalling again that p .Cp .Z// and Cp .Z/ are homeomorphic, we
conclude that hl .p .Cp .Z/// D !. It is straightforward that A ' .Cp .Y //
p .Cp .Z// and therefore A p .Cp .Z// which implies hl .A/ !. Thus the
space Cp .X / is hl .!/-monolithic.
T.194. Give an example of an hl.!/-monolithic non-hl .!/-monolithic space.
!, and hence X is hl-monolithic because for any infinite cardinal and any A X
with jAj , we have hl.A/ hl.X / D ! . The space X X has an
uncountable closed discrete subspace and hence hl .X / hl.X X / > !. This,
together with separability of X , implies that X is not hl .!/-monolithic.
T.195. Assume SA. Let X be a !-stable space such that s.Cp .X // D !. Prove that
X has a countable network.
Solution. Since X is !-stable, the space Cp .X / is !-monolithic by Problem 154.
The axiom SA and s.Cp .X // D ! imply that hd .Cp .X // D ! (see Problem 036).
In particular, the space Cp .X / is separable, so nw.Cp .X // D ! by !-monolithity
of Cp .X /. Finally, nw.X / D nw.Cp .X // D ! by Problem 172 of [TFS].
T.196. Assume SA. Prove that if s.Cp .X // D !, then for any Y X we have
s.Cp .Y // D !.
Solution. Let Y W Cp .X / ! Cp .Y / be the restriction map defined by the equality
Y .f / D f jY for each f 2 Cp .X /. As usual, Cp .Y jX / D Y .Cp .X // Cp .Y /.
Observe that SA and s.Cp .X // D ! imply hd.Cp .X // ! by Problem 036.
Consequently, hd.Cp .Y jX // ! and hence hl .Y / D hd.Cp .Y jX // ! (see
Fact 5 of T.173). Thus we can apply Problem 026 to conclude that s.Cp .Y //
hd .Cp .Y // D hl .Y / !.
T.197. Assume MAC:CH. Let X be an !-monolithic space such that the spread of
Cp .X / is countable. Prove that X has a countable network.
Solution. We have s.X X / s.Cp .X // D ! (see Problem 016). The space X X
being !-monolithic (see Problem 114), we have hl.X X / D ! (see Problem 190)
and therefore .X / D !. Thus s .Cp .X // D ! (see Problem 028) and hence
hl .Cp .X // D ! by Problem 059. As a consequence, hd .X / D hl .Cp .X // D
! (see Problem 027) and, in particular, X is separable. Since every separable
!-monolithic space has a countable network, we have nw.X / D !.
T.198. Assume MAC:CH and suppose that Cp .X / contains no uncountable free
sequences. Prove that for any Y X , the space Y is hereditarily Lindelf if and
only if it is hereditarily separable.
Solution. For an arbitrary set P we denote by Fin.P / the family of all non-
empty finite subsets of P . For the sake of brevity, a free sequence of length !1
is called a free !1 -sequence. Given a space Z call a set fz W < !1 g Z
a weak free !1 -sequence if fz W g \ fz W > g D ; for any <
!1 . It is easy to see that every free !1 -sequence is a weak free !1 -sequence
while the space !1 with its interval topology is an example of a weak free
!1 -sequence which is not a free !1 -sequence. However, the following fact shows
that for our purposes we can restrict ourselves to weak free !1 -sequences.
Fact 1. A space Z does not contain an uncountable free sequence if and only if it
does not contain a weak free !1 -sequence.
Proof. Every free !1 -sequence is a weak free !1 -sequence so sufficiency is trivial.
To prove necessity take any weak free !1 -sequence fz W < !1 g Z. It is easy
to find a function ' W !1 ! !1 such that './ is a successor ordinal for each
< !1 and < implies './ < './. Now if y D z'./ for all < !1 , then
S D fy W < !1 g is a free !1 -sequence in Z. To see it, fix any < !1 ; then
'./ D
C 1 for some
< !1 . Observe that fy W < g A D fz W
g
and fy W g B D fz W >
g and hence fy W < g \ fy W g
A \ B D ; which proves that S is a free !1 -sequence. Fact 1 is proved. t
u
Returning to our solution, suppose that some Y X is a hereditarily Lindelf
non-hereditarily separable space. Then there is a set fy W < !1 g Y which is
left-separated by its indexation (see Problem 004 and Fact 2 of T.004), so there is
no loss of generality to assume that Y D fy W < !1 g; let Y D fy W < g
for all < !1 . For every y 2 Y there is a unique < !1 such that y D y ; let
.y/ D and fix sets Uy ; Vy 2 .y; X/ and a function fy 2 C.X; 0; 1 / with the
(1) U y Vy and fy .X nVy / D f0g;
(2) fy jU y 1 and z Vy whenever .z/ < .y/.
Let P D fp 2 Fin.Y / W for any y; z 2 p, if .y/ < .z/, then z Uy g. The order
on P is the reverse inclusion, i.e., given any p; q 2 P we declare that p q if
and only if q p. Observe first that
(3) the partially ordered set .P; / is not ccc.
Indeed, if .P; / is ccc, then MA+:CH implies that for any uncountable R P
there is an uncountable centered subset of R (see Problem 049). In particular, there
is an uncountable Q Y such that the family ffyg W y 2 Qg is centered. As a
consequence,
(4) if y; z 2 Q and .y/ < .z/, then z Uy
because there is p 2 P with p fyg and p fzg which implies fy; zg p, so
(4) holds by the definition of P. Next observe that if y 2 Q, then Uy \ Q D fyg;
indeed, if z 2 Q and .z/ < .y/, then z Vy by the definition of Vy Uy . Now if

.z/ > .y/, then z Uy by (4). Therefore Q is an uncountable discrete subspace
of a hereditarily Lindelf space Y ; this contradiction proves (3).
By (3), there is an uncountable antichain A in .P; /. We can apply the Delta-
lemma (see Problem 038) to find a finite set w Y (which may be empty) and
an uncountable A0 A such that p \ q D w for any distinct p; q 2 A0 . Since
the family B D fpnw W p 2 A0 g is disjoint, only countably many elements of
B intersect Y for each < !1 . This makes it possible to choose, by an evident
transfinite induction, a set B 0 D fp W < !1 g B such that < < !1 implies
.t/ < .y/ < .z/ for any S y 2 p ; z 2 p and t 2 w. For eachS ordinal < !1
consider the sets U.p / D fUy W y 2 p g and V .p / D fVy W y 2 p g;
observe that if < and U.p / \ p D ;, then q D w [ p [ p 2 P. Moreover,
we have q w[p and q w[p which contradicts the incompatibility of w[p
and w [ p . Therefore
(5) if < < !1 , then V .p / \ p D ; and U.p / \ p ;.
P
Now let g D ffy W y 2 p g for every < !1 . It is clear that g .x/ 1
for every x 2 p and g .X nV .p // D f0g. Let G D ff 2 Cp .X / W f .x/ > 12
for some x 2 p g and H D ff 2 Cp .X / W f .x/ > 0 for some x 2 p g. It is
straightforward that G and H are open in Cp .X / and G H for each < !1 .
Now take any < !1 ; if , then it follows from (5) that there exist y 2 p
and z 2 p such that z 2 Uy and therefore g .z/ fy .z/ D 1 > 12 which shows
that g 2 G . On the other hand, if > , then y Vz for any z 2 p , and hence
fz .y/ D 0 for all y 2 p and z 2 p which proves that g H . As a consequence,
fg W g \ fg W > g G \ Cp .X /nH D G \ .Cp .X /nH / D ;
for every < !1 which shows that S D ff W < !1 g is a weak free !1 -

sequence in Cp .X /. Thus there exists a free !1 -sequence in Cp .X / by Fact 1
which is a contradiction. Consequently, every hereditarily Lindelf subspace of X
is hereditarily separable.
Now suppose that some Y X is a hereditarily separable non-hereditarily
Lindelf space. There is a set fy W < !1 g Y which is right-separated by its
indexation (see Problem 005 and Fact 2 of T.005), so there is no loss of generality
to assume that Y D fy W < !1 g; let Y D fy W < g for all < !1 .
For every y 2 Y there is a unique < !1 such that y D y ; let .y/ D .
Fix sets Uy ; Vy 2 .y; X/ and a function fy 2 C.X; 0; 1 / with the following
properties:
(6) U y Vy and fy .X nVy / D f0g;
(7) fy jU y 1 and z Vy whenever .z/ > .y/.
Let P D fp 2 Fin.Y / W for any y; z 2 p, if .y/ > .z/, then z Uy g. The order
on P is the reverse inclusion, i.e., given any p; q 2 P we declare that p q if
and only if q p. Observe first that
(8) the partially ordered set .P; / is not ccc.

Indeed, if .P; / is ccc, then MAC:CH implies that for any uncountable R P
there is an uncountable centered subset of R (see Problem 049). In particular, there
is an uncountable Q Y such that the family ffyg W y 2 Qg is centered. As a
consequence,
(9) if y; z 2 Q and .y/ > .z/, then z Uy
because there is p 2 P with p fyg and p fzg which implies fy; zg p so
(9) holds by the definition of P. Next observe that if y 2 Q, then Uy \ Q D fyg;
indeed, if z 2 Q and .z/ > .y/, then z Vy by the definition of Vy Uy . Now if
.z/ < .y/, then z Uy by (9). Therefore Q is an uncountable discrete subspace
of a hereditarily separable space Y ; this contradiction proves (8).
By (8), there exists an uncountable antichain A in .P; /. We can apply the
Delta-lemma (see Problem 038) to find a finite set w Y (which may be empty)
and an uncountable A0 A such that p \ q D w for any distinct p; q 2 A0 .
Since the family B D fpnw W p 2 A0 g is disjoint, only countably many elements
of B intersect Y for each < !1 . This makes it possible to choose, by an evident
transfinite induction, a set B 0 D fp W < !1 g B such that < < !1 implies
.t/ < .y/ < .z/ Sfor any y 2 p ; z 2 p and t 2Sw. For each < !1 consider
the sets U.p / D fUy W y 2 p g and V .p / D fVy W y 2 p g; observe that
if > and U.p / \ p D ;, then q D w [ p [ p 2 P. Moreover, we have
q w [ p and q w [ p which contradicts the incompatibility of w [ p and
w [ p . Therefore,
(10) if < < !1 , then V .p / \ p D ; and U.p / \ p ;.
P
Now let g D ffy W y 2 p g for every < !1 . It is clear that g .x/ 1 for
any x 2 p and g .X nV .p // D f0g. Let G D ff 2 Cp .X / W f .x/ > 12 for some
x 2 p g and H D ff 2 Cp .X / W f .x/ > 0 for some x 2 p g. It is straightforward
that G and H are open in Cp .X / and G H for each < !1 .
Now take any < !1 ; if , then it follows from (10) that there exist y 2 p
and z 2 p such that z 2 Uy and therefore g .z/ fy .z/ D 1 > 12 which shows
that g 2 G . On the other hand, if < , then y Vz for any z 2 p and hence
fz .y/ D 0 for all y 2 p and z 2 p which proves that g H . As a consequence,
fg W < g \ fg W g G \ Cp .X /nH D G \ .Cp .X /nH / D ;
for every < !1 which shows that S D ff W < !1 g is a free !1 -sequence

in Cp .X / which is a contradiction. Consequently, every hereditarily separable
subspace of X is hereditarily Lindelf so our solution is complete.
T.199. Assume MAC:CH. Let X be an !-monolithic space with l .X / D !.
Suppose that Cp .X / is Lindelf and Y X has countable spread. Prove that Y
has a countable network.
Solution. We have hl.Y / D ! because the space Y is !-monolithic and has

countable spread (see Problem 190). It follows from l .X / D ! that t.Cp .X // D !
(see Problem 149 of [TFS]). Now assume that S D ff W < !1 g is a free sequence
in Cp .X /; let S D ff W < g and T D S nS for every < !1 . The family
T D fT W < !1 g is countably centered (i.e., the intersection of any countable
subfamily of T is non-empty) and consists of non-empty
T closed subsets of Cp .X /.
The space Cp .X / being Lindelf, we have T D T ; (see Fact 2 of S.336).
If f 2 T , then f 2 S , and hence there is a countable S 0 S with f 2 S 0 . Choose
< !1 such that S 0 S and observe that f 2 S \ T , a contradiction with the
fact that S is a free sequence.
Therefore Cp .X / has no uncountable free sequences and hence we can apply
Problem 198 to the hereditarily Lindelf space Y to conclude that Y is separable.
Any separable !-monolithic space has a countable network so nw.Y / D !.
T.200. Assume MAC:CH. Prove that if Cp .X / is hereditarily stable, then X has a
countable network.
Solution. A discrete space is !-stable only if it is countable by Problem 106,
so Cp .X / contains no uncountable discrete subspaces, i.e., s.Cp .X // D !. The
space X is monolithic because Cp .X / is stable (see Problem 152). Now apply
Problem 197 to conclude that X has a countable network.
T.201. Prove that any continuous image of a hereditarily normal compact space is
hereditarily normal.
Solution. Assume that Z is a space and f W Z ! Y is a closed continuous onto
map. For any open U Z, let f # .U / D Y nf .ZnU /. It is straightforward that
f # .U / is an open subset of Y and f # .U / D fy 2 Y W f 1 .y/ U g. An immediate
consequence is that
(1) if U; V .Z/ and U \ V D ; then f # .U / \ f # .V / D ;.
Now it is easy to prove the following statement.
Fact 1. A closed continuous image of a normal space is a normal space.
Proof. Let Z be a normal space; suppose that f W Z ! Y is a closed continuous
onto map. If we have disjoint closed sets F; G Y , then the sets F 0 D f 1 .F / and
G 0 D f 1 .G/ are closed in Z and disjoint. By normality of Z there exist disjoint
U 0 ; V 0 2 .Z/ such that F 0 U 0 and G 0 V 0 . Given any point y 2 F , we
have f 1 .y/ f 1 .F / D F 0 U 0 and therefore y 2 f # .U 0 /; this shows that
F f # .U 0 /. Analogously, G f # .V 0 /, so we can apply (1) to conclude that the
sets U D f # .U 0 / and V D f # .V 0 / are open disjoint neighborhoods of F and G
respectively. Thus Y is normal and Fact 1 is proved. t
u
Returning to our solution take any compact hereditarily normal space X and a
continuous onto map f W X ! Y . The map f is closed (Problem 122 of [TFS])
and hence, for any A Y , the map fA D f jf 1 .A/ W f 1 .A/ ! A is also closed
by Fact 1 of S.261. By hereditary normality of X the space f 1 .A/ is normal and
hence A is normal by Fact 1. Therefore every A Y is normal, i.e., Y is hereditarily
normal.
T.202. Let X be a compact space for which X 2 is hereditarily normal. Prove that X
is perfectly normal and hence first countable.
Solution. We can assume that X is infinite and hence the space X has a countable
non-closed subset by Fact 2 of T.090. Therefore Fact 1 of T.090 is applicable to the
spaces Y D Z D X to conclude that there is no closed F X which is not a
G -set in X . In other words, every closed subset of X is a G -set in X and hence X
is perfectly normal.
T.203. Let X be a compact space and denote by the diagonal of X , i.e., D
f.x; x/ W x 2 X g X 2 . Prove that if X 2 n is paracompact, then X is metrizable.
TD f.z; z/ W z 2 Zg Z is its diagonal; a

2
Solution. Given a space Z the set .Z/
family U exp.Z/ is T1 -separating if fU 2 U W z 2 U g D fzg for every z 2 Z.
The family U is called point-countable if every z 2 Z belongs to at most countably
many elements of U. Given a set A Z let UA D fU 2 U W U \ A ;g; for any
z 2 Z we will write Uz instead of Ufzg .
Fact 1. A countably compact space Z is metrizable if and only if there exists a
point-countable T1 -separating family U .Z/.
Proof. If the space Z is metrizable, then it has a countable base U which is,
evidently, point-countable and T1 -separating so necessity is clear.
To prove sufficiency fix a point-countable T1 -separating family U .Z/ and
take any z0 2 Z; let A0 D fz0 g. Assume that n > 0 and we have countable sets
A0 ; : : : ; An1 with the following properties:
(1) A0

An1 Z; S S
(2) if k < n 1 and V UAk is finite, then V Z implies AkC1 n. V/ ;.
The set An1 being countable, the family UAn1 has to be countable as well
because U is point-countable. Therefore the family
[
S D fV W V is a finite subfamily of UAn1 such that V Zg
S
is countable. For any family V 2 S we can choose a point z.V/ 2 Zn. V/; then
An D An1 [ fz.V/ W V 2 Sg is countable and the sets A0 ; : : : ; An still satisfy
the conditions (1) and (2). Therefore our inductive construction can be continued to
provide a sequence fAn W n S2 !g with the properties (1) and (2) for each n 2 !.
We claim that the set A D fAn W n 2 !g is dense in Z. Observe first that an open
set intersects A if and only if it intersects A. Therefore U A D UA is countable.
If A Z, then take any u 2 ZnA; the family U being T1 -separating, for every z 2
A, there is Wz 2 Uz such that Wz 63 u. Since Wz 2 UA for every z 2 A, the family
W D fWz W z 2 Ag is a countable openScover of a countably Scompact space A.
Take any finite V W such that A V. Since UA D fUAnS W n 2 !g
and (1) Sholds, there is k 2 ! such that S V UA k . We have u 2 Zn. V/ which
implies V S Z and hence A kC1 n. V/ ; by (2) which is a contradiction with
AkC1 A V.
Thus A is dense in Z and hence U D UA is countable; therefore the family

F D fZnU W UT2 Ug is also countable and T1 -separating because U is
T1 -separating. Thus F z D fzg for every z 2 Z. Observe also that F consists of
closed subspaces of Z. Given any z 2 Z take an enumeration fFn W n 2 !g of F z ;
if U 2 .z; Z/, then let Gn D F0 \T

\ Fn \ .ZnU / for all n 2 !. It is clear
that GnC1 Gn for all n 2 ! and n2! Gn D ;. The space Z being countably
compact, we have Gn D ; for some n 2 ! and therefore F0 \

\ Fn U .
This
T proves that for any z 2 Z and U 2 .z; Z/ there is a finite F 0 F z such
0
that F U . Consequently, the family of all finite intersections of the elements
of F is a countable network in Z. Therefore Z condenses onto a second countable
space (see Problem 156 of [TFS]); the respective condensation is a homeomorphism
(see Problem 140 of [TFS]), so Z is second countable and hence metrizable. Fact 1
is proved. t
u
Fact 2. If is an uncountable cardinal and Z D A./, then Z 2 n.Z/ is not
normal.
Proof. Denote by a the unique non-isolated point of Z; let w D .a; a/ 2 Z 2 . The
sets F D .Z fag/nfwg and G D .fag Z/nfwg are closed in Z 2 n.Z/ and
disjoint. Assume that there are disjoint U; V 2 .Z 2 / such that F U and G V .
Choose any faithfully indexed sequence fan W n 2 !g Znfag; since .an ; a/ 2 U
S a finite Bn Z such that fan g .ZnBn / U for all n 2 !. The set
there is
B D fBn W n 2 !g is countable, so there is z 2 Zn.B [ fag/. It is evident that
un D .an ; z/ 2 U for each n 2 !; since the sequence S D fun W n 2 !g converges
to .a; z/ 2 G, we have S \ G ; which contradicts the fact that S U Z 2 nV
and hence S Z 2 nV Z 2 nG. Therefore the disjoint closed sets F and G cannot
be separated by disjoint open sets in Z 2 n.Z/, i.e., Z 2 n.Z/ is not normal and
hence Fact 2 is proved. t
u
Fact 3. If Z is a locally compact paracompact space, then there isS
a disjoint family
U of clopen subsets of Z such that each U 2 U is -compact and U D Z.
Proof. Each z 2 Z has an open neighborhood Oz such that O z is compact. If W is
a locally finite refinement of the open cover fOz W z 2 Zg of the space Z, then W is
compact for any W 2 W. Call a sequence C D fW1 : : : ; Wn g of elements of W a
chain if Wk \ WkC1 ; for every k D 1; : : : ; n 1. Given y; z 2 Z, let y z if
there is a chain C D fW1 ; : : : ; Wn g of elements of W such that y 2 W1 and z 2 Wn
(we will say that the chain C connects the points y and z). It is evident that is an
equivalence relationshipSon Z; let U be the set of the respective equivalence classes.
Then U is disjoint and U D Z.
For any z 2 Z denote by Uz the equivalence class of z. Given any t 2 Uz there
is a chain C D fW1 ; : : : ; Wn g of elements of W such that z 2 W1 and t 2 Wn . It is
clear that the same chain C connects z and any point t 0 2 Wn . Therefore Wn Uz
and hence every t 2 Uz lies in Uz together with its neighborhood Wn which proves
that Uz is open in Z for any z 2 Z. The family U being disjoint, each U 2 U is also
closed in Z.
The last thing we must prove is that every U 2 U is -compact. Observe that if a
set F Z is compact, then WF is finite because the family fW \ F W W 2 Wg
is a locally finite family of open sets in a compact space F (see Problem 136 of
[TFS]). Take any z0 2 U , choose W0 2 W with z0 2 W0 and let A0 D fW0 g.
Assume that n > 0 and we have finite families A0

An1 W such that
S
(3) W Ak AkC1 for any k < n 1.
S S
The set Vn D fV W V 2 An1 g is compact, so An D WVn D W An1
is finite and the sequence fA0 ; : : : ; An g still satisfies (3). Thus we can construct a
S fAn W n 2 !g with the property (3) for all n
sequence S2 !. It is clear that the family
A D fAn W n 2 !g is countable and hence A D fW W W 2 Ag is -compact.
Given any z 2 U D Uz0 , there is a chain C D fW1 ; : : : ; Wn g W which connects
S z. It is immediate from (3) that Wi 2SAi for each i D 1; : : : ; n and hence
z0 and
z 2 An . As a consequence, the space U A A is -compact being a closed
subset of a -compact set A. Fact 3 is proved. t
u
Returning to our solution assume that X is not metrizable and observe that X 2 n
is a locally compact paracompact space, so S we can apply Fact 3 to find a disjoint
family W of open -compact sets such that W D X 2 n. For any W 2 W and
any z D .x; y/ 2 W , we have x y and hence there exist disjoint Uz ; Vz 2 .X/
such that x 2 Uz ; y 2 Vz and Uz Vz W . The space W being -compact,
there is a countable subcover of the cover fUz Vz W z 2 U g. Thus we can choose,
for each W 2 W, countable families UW D fU.W; n/ W n 2 !g .X/ and
VW D fV .W; S n/ W n 2 !g .X/ such that U.W; n/ \ V .W; n/ D ; for each
n 2 ! and fU.W; n/ V .W; n/S W n 2 !g D W .
We claim that the family U D fUW W W 2 Wg is T1 -separating in X . Indeed,
if x 2 X and y x, then .x; y/ 2 X 2 n and hence .x; y/ 2 W for some
W 2 W which implies that there is n 2 ! such that .x; y/ 2 U.W; n/ V .W; n/
and therefore x 2 U.W; n/ and y 2 V .W; n/. The sets U.W; n/ and V .W; n/ being
disjoint, we have x 2 U.W; n/ 63 y and hence U is T1 -separating. Since X is not
metrizable, the family U cannot be point-countable by Fact 1. Take any x 2 X such
that Ux is uncountable; there exists a faithfully indexed family fW W < !1 g
W and n 2 ! such that x 2 U.W ; n/ for each < !1 . The family fW W < !1 g
is discrete in X 2 n and U.W ; n/ V .W ; n/ W for each < !1 . Therefore the
family fU.W ; n/V .W ; n/ W < !1 g is also discrete in X 2 n. Let Xx D fxgX ;
it is clear that Xx \.U.W ; n/V .W ; n// D fxgV .W ; n/ for every < !1 which
implies that ffxg V .W ; n/ W < !1 g is discrete in Xx nf.x; x/g and therefore the
family fV .W ; n/ W < !1 g is discrete in X nfxg. Pick any x 2 V .W ; n/ for every
< !1 . Then the set D D fx W < !1 g is closed and discrete in X nfxg and hence
K D D [ fxg is homeomorphic to A.!1 / being a compact space of cardinality !1
with a unique non-isolated point. It is easy to see that K 2 n.K/ is a closed subset of
X 2 n.X / and therefore K 2 n.K/ is paracompact which contradicts Fact 2. This
contradiction shows that X is metrizable and hence our solution is complete.
T.204. Observe that any FrchetUrysohn space is Whyburn. Prove that any
countably compact Whyburn space is FrchetUrysohn.
Solution. If X is a FrchetUrysohn space, then for any A X and any x 2 AnA,

there exists a sequence fan W n 2 !g A such that an ! x. It is immediate that
F D fan W n 2 !g is an almost closed set with x 2 F so X is Whyburn.
Now assume that X is a countably compact Whyburn space. Given a set A X
and x 2 AnA choose an almost closed F A with x 2 F . Take a maximal family
U .F / such that U \ V S D ; for any distinct U; V 2 U and x U for every
U 2 U. It is evident that G D U is dense in F and hence x 2 G.
Use the Whyburn property of X again to find an almost closed H G with
x 2 H . Let HU D H \ U for every U 2 U. Every U 2 U is closed in G, so
it follows from x H U that HU is closed (maybe empty) in X for each U 2 U.
0
S family U 0D fU 2 U W HU ;g has to be infinite; for otherwise, x
The
fHU W U 2 U g D H which is a contradiction.
Now, take any y 2 X nfxg; if y H , then W D X nH 2 .y; X/ and W \HU D
; for any U 2 U 0 . If y 2 H , then there is a unique U 2 U 0 such that y 2 HU . Then
U 2 .y; X/ and U \ HV D ; for any V 2 U 0 nfU g. This shows that the family
fHU W U 2 U 0 g is discrete in X nfxg.
Thus for any set W 2 .x; X/ the family HW D fHU nW W .HU nW / ; and
U 2 U 0 g is discrete in X . Since X is countably compact, HW has to be finite and
hence HU W for all but finitely many U 2 U 0 . Pick a point xU 2 HU for
any U 2 U 0 . Then the set B D fxU W U 2 U 0 g is infinite and BnW is finite for
any W 2 .X/ with x 2 W . This shows that for any faithfully indexed sequence
S D fan W n 2 !g B we have S A and S ! x, i.e., X is a FrchetUrysohn
space.
T.205. Give an example of a pseudocompact Whyburn space which is not Frchet
Urysohn.
Solution. Take an arbitrary countably infinite set D which will be identified with
the discrete space whose underlying set is D. The letter K stands for the Cantor set
(see Problem 128 of [TFS]); the mappings D W D K ! D and K W D K ! K
are the respective natural projections. If A D K, then A.d / denotes the set
K .A \ .fd g K// for any d 2 D. A set A D K is called admissible if it has
the following properties:
(1) the set i.A/ D D .A/ is infinite;
(2) the set A.d / is clopen in K for every d 2 i.A/ and the family fA.d / W d 2
i.A/g is disjoint.
A family A of admissible subsets of D K is called essentially disjoint if for
any distinct A; B 2 A, the set fd 2 D W A.d / \ B.d / ;g is finite. Denote by
A.D/ any maximal essentially disjoint family of admissible subsets of D K (we
leave to the reader a simple verification that such a maximal family exists). Assign
some point pA D K to every set A 2 A.D/ in such a way that pA pB for
any distinct A; B 2 A.D/.
We can now define our auxiliary space P .D/ D .D K/ [ fpA W A 2 A.D/g.
The local bases at the points of D K are given by their usual local bases in D K
(remember that D is endowed with the discrete topology). Given A 2 A.D/, a local
base at the point pA is the family ffpA g [ .An.F K// W F D is finiteg.
Fact 1. The space P .D/ is Tychonoff, first countable, pseudocompact and zero-
dimensional; D K is an open subspace of P .D/ and the set fpA W A 2 A.D/g is
closed and discrete.
Proof. We omit a simple proof that the local bases are well-defined. The elements
of the local base of any w 2 D K are all contained in D K, which shows that
every w 2 D K belongs to D K together with its neighborhood; thus D K is
open in P .D/.
Let us check that the space P .D/ is Hausdorff. If x and y are distinct points
from D K, then they have open disjoint neighborhoods in the space D K; since
the same neighborhoods are open in P .D/, any pair of distinct points from D K
can be separated by disjoint open neighborhoods in P .D/.
Now, if x 2 DK and y 2 fpA W A 2 A.D/g, then y D pA for some A 2 A.D/
and hence the sets fpA g [ .An.fD .x/g K// and fD .x/g K are disjoint open
neighborhoods of the points y and x respectively.
Finally, if x D pA and y D pB for distinct A; B 2 A.D/, then F D i.A/ \ i.B/
is finite because A.D/ is essentially disjoint. Therefore fpA g [ .An.F K// and
fpB g [ .Bn.F K// are open disjoint neighborhoods of x and y respectively. This
proves that P .D/ is a Hausdorff space.
Next observe that K is zero-dimensional. Indeed, if x 2 K and U 2 .x; K/,
then there is an interval .a; b/ R such that x 2 .a; b/ \ K U (we consider
that K R (see Problem 128 of [TFS])). The interval .x; b/ cannot be contained
in K because the interior of K is empty. Thus there is c 2 .x; b/nK; analogously,
there exists a point d 2 .a; x/nK. Consequently, W D .c; d / \ K D c; d \ K
is a clopen subset of K and x 2 W .a; b/ \ K U which shows that clopen
subsets form a base in K, i.e., K is zero-dimensional. It is evident that fd g U is
a clopen subset of D K for any d 2 D and clopen U K. Since the family
O D ffd g U W d 2 D and U is clopen in Kg is a base in D K, the space D K
is also zero-dimensional. It follows from the definition of local bases in P .D/ that
every O 2 O is clopen in P .D/ which proves that the family of all clopen subsets
of P .D/ contains local bases at all points of D K.
Observe that given any A 2 A.D/, the set U \ .D K/ is clopen in D K for
any basic neighborhood U of any point pA and hence no point from D K belongs
to U nU . Furthermore, if B 2 A.D/nfAg, then F D i.A/ \ i.B/ is finite because
A.D/ is essentially disjoint; therefore fpB g[.Bn.F K// is an open neighborhood
of pB which does not intersect U . As a consequence, pB does not belong to U nU ,
i.e., U is a clopen subspace of P .D/. Therefore P .D/ is zero-dimensional and
hence Tychonoff by Fact 1 of S.232.
It is evident that P D fpA W A 2 A.D/g is closed; to see that it is discrete,
observe that U \ P D pA for any basic neighborhood U of the point pA .
Let us finally check that P .D/ is pseudocompact. The set D K is open and
dense in P .D/, so if fWn W n 2 !g .P .D// is a discrete family, then we can
choose a non-empty clopen Un Wn \ .D K/ for each n 2 !. It is evident that
the family fUn W n 2 !g is also Discrete, so it suffices to show that for any infinite
family U of non-empty clopen subsets of D K, there exists an accumulation point
for U, i.e., a point x 2 P .D/ such that every neighborhood of x meets infinitely
many elements of U.
If the family Ud D fU 2 U W U.d / ;g is infinite for some d 2 D, then we
have an infinite family U 0 D fU \ .fd g K/ W U 2 Ud g of non-empty open sets in
the compact space fd g K. Thus there is an accumulation point x for U 0 in fd g K
which, evidently, is an accumulation point for U in P .D/.
If every Ud is finite, then the set fd 2 D W U.d / ; for some U 2 Ug is
infinite and hence we can find an admissible set B D K and an injection
d ! Ud of the set i.B/ into U such that, B.d / Ud .d / for each d 2 i.B/. The
family A.D/ being maximal essentially disjoint, we can find A 2 A.D/ such that
A.d / \ B.d / ; for infinitely many d s. To finish our proof observe that each
neighborhood of the point pA intersects infinitely many elements of U. Fact 1 is
proved. t
u
Returning to our solution denote by L the subset of cC which consists of limit
ordinals of countable cofinality. Considering L with its order topology it is natural
to say that a countably infinite set S L is a sequence converging to 2 L if
S ; sup S D and S \ is finite for any < .
If 2 L and D C ! for some < cC (i.e., if is an isolated point of L), let
D D f; /g, where ; / D f
W
< g. If is not isolated in L, define D
to be any maximal almost disjoint family of sequences
S converging to . We leave it
to the reader as an exercise to prove that D D fD W 2 Lg is a maximal almost
disjoint family of countable subsets of cC .
Apply Fact 1 to fix the family A.D/, the points fpA W A 2 A.D/g and the space
P .D/ for every set D 2 D. Choose a new point p and define
X D fpg [ fpA W A 2 A.D/; D 2 Dg [ .cC K/
to be the underlying set of the space we want to construct. Given an arbitrary point
x D .; z/ 2 cC K, take a clopen local base fWn W n 2 !g of z in the space K and
let Un D fg Wn for each n 2 !. We declare the family Bx D fUn W n 2 !g to be
a local base of the point x in X . Observe that the local bases thus defined generate
the product topology on cC K if cC is considered with the discrete topology.
If A 2 A.D/ for some D 2 D, then the local base Bx at the point x D pA
is the same as in the space P .D/ (recall that P .D/ D D K cC K is a
subset of X ). Finally, declare the local base of X at the point p to be the family
Bp D fV W 2 Lg where
[
V D fpg [ .f W < < cC g K/ [ fpA W A 2 A.D/; D 2 fD W > gg
for each 2 L.
We claim that X is a pseudocompact Whyburn space which is not Frchet
Urysohn. It is immediate that X is a T1 -space, so to prove that X is Tychonoff, it
suffices to show that it is zero-dimensional (see Fact 1 of S.232). We will establish

a stronger fact, namely, that all elements of the local bases defined above are clopen
in X .
If x D .; z/ 2 cC K and U 2 Bx , then W D .cC K/nU is open in cC K
and therefore W is a neighborhood of any y 2 .cC K/nU . Given A 2 A.D/ for
some D 2 D, it is immediate from the definition of the local base at y D pA that
fpA g [ ..cC K/nU / is an open neighborhood of y which does not meet U . Now,
if y D p, then V \ U D ; for any 2 L; > and this finishes the proof that
U is a clopen set.
We will show next that if x D pA for some A 2 A.D/; D 2 D, then every
U 2 Bx is a clopen set. Observe first that W D .cC K/nU is a clopen set in
cC K, so every point from W is not in the closure of U . If y D pB is distinct from
x, then either B 2 A.D/ or i.A/ \ i.B/ is finite because the family D is almost
disjoint. Now observe that pB belongs to the closure of a set T cC K if and
only if T intersects infinitely many sets of the family ffd g B.d / W d 2 i.B/g. If
B 2 A.D/, this is impossible for T D A because the family A.D/ is essentially
disjoint, and if B 2 A.D 0 / for some D 0 D, then i.B/ \ i.A/ is finite being
contained in the finite set D \ D 0 . Finally, to see that p is not in the closure of U
observe that V \ U D ; for any 2 L with > sup.D/.
To establish that V is clopen for any 2 L, observe that V \.cC K/ is clopen
in cC K, so no point of .cC K/nV can be in the closure of V . If pA V ,
then A 2 A.D/ for some D 2 D with . By definition of D , we have
sup.D/ and therefore fpA g [ .D K/ is a neighborhood of pA which misses
V . Thus we have finally proved that X is a Tychonoff zero-dimensional space.
It is easy toS ffg K W < cC g is a family of clopen subsets of X
see that U D S
such that p 2 U but p U 0 for any U 0 U with jU 0 j c. This shows that X
has uncountable tightness and hence it is not FrchetUrysohn.
To establish that X is pseudocompact take any discrete family fUn W n 2 !g
of non-empty open subsets of X . Since cC K is open and dense in X , we can
assume (choosing a non-empty clopen subset in each Un \ .cC K/ if necessary)
that Un is a clopen subset of cC K for each n 2 !. Since each K D fg K
is compact, only finitely many Un s can intersect each K . Thus, choosing smaller
clopen sets and passing to an appropriate infinite subfamily, we can construct a
discrete family fVn W n 2 !g of non-empty clopen subsets of the space X such that,
for all m; n 2 ! we have ji.Vn /j D 1; i.Vn / \ i.Vm / D ; if n m and the family
fK .Vn / W n 2 !g is disjoint.S By maximality of D there exists a set D 2 D such
0
that the setS D D D \ . fi.Vn/ W n 2 !g/ is infinite. It is easy to see that the
set B D fVn W i.Vn / D 0 g is admissible. By maximality of the family A.D/
there exists A 2 A.D/ such that A./ \ B./ ; for infinitely many s. As a
consequence, pA is an accumulation point of the family fVn W n 2 !g which is a
contradiction. Thus X is pseudocompact.
Finally we show that X is a Whyburn space. It is evident that all points of
X except for p have countable local bases, so we must only check the Whyburn
property at p. Let T be the set fpA W A 2 A.D/; D 2 Dg. Suppose that p 2 C nC
for some C X . If p 2 C \ T , then B D C \ T is an almost closed subset of C
with p 2 B. If not, then p 2 C \ .cC K/ and therefore the cardinality of the set
C 0 D C \ .cC K/ is equal to cC . Since there are only c points in K, there exists
t 2 K such that the set B D f < cC W .; t/ 2 C 0 g has cardinality cC . It is easy
to check that F D B ftg is an almost closed subset of C with p 2 F . We finally
established that X is a Whyburn pseudocompact space of uncountable tightness so
T.206. Observe that a continuous image of a Whyburn space need not be Whyburn.
Prove that any image of a Whyburn space under a closed map is a Whyburn space.
Prove that the same is true for weakly Whyburn spaces.
Solution. The space X D .!1 C 1/ with the interval topology is not Whyburn
because t.X / > ! while any compact Whyburn space is FrchetUrysohn (see
Problem 204). However, X is a continuous image of a discrete space D of
cardinality !1 which is Whyburn vacuously because there exists no x 2 D with
x 2 AnA for some A D. This shows that the Whyburn property is not preserved
by continuous maps.
Now assume that X is a Whyburn space and take any closed continuous onto
map f W X ! Y . If A Y and y 2 AnA, then let B D f 1 .A/. Observe that
P D f .B/ is a closed set with A P and hence A P . Pick any x 2 B with
f .x/ D y; it is clear that x 2 BnB, so there is an almost closed G B such that
x 2 G. The set F D f .G/ is contained in A; besides, F [ fyg D f .G [ fxg/ is a
closed set which together with y 2 F implies that F is almost closed and hence Y
is Whyburn.
To prove that the weak Whyburn property is also preserved by closed maps, take
any weakly Whyburn space X and a closed continuous onto map f W X ! Y . If
A Y is not closed in Y , then B D f 1 .A/ is not closed in X and hence there
is x 2 BnB and an almost closed G B such that x 2 B. It is immediate that
y D f .x/ 2 AnA; the set F D f .G/ is contained in A and y 2 F . Furthermore
F [ fyg D f .G [ fxg/ is a closed set which implies that F is almost closed and
hence Y is weakly Whyburn.
T.207. Prove that every space with a unique non-isolated point is Whyburn. In
particular, there exist Whyburn spaces of uncountable tightness.
Solution. Let X be a space with a unique non-isolated point a. Given a set A X
and x 2 AnA, we must have x D a and hence F D A is an almost closed subset of
X with x 2 F , i.e., X is Whyburn.
Now, if X D L.!1 / is the Lindelfication of a discrete space of cardinality !1 ,
then X is a Whyburn space of uncountable tightness. A much more complicated
example of a Whyburn space of uncountable tightness is constructed in T.205.
T.208. Prove that any submaximal space is Whyburn.
Solution. Let X be a submaximal space; assume that x 2 AnA for some A X .
Observe first that if a set P X has empty interior, then X nP is dense in X and
hence open. Therefore every P X with Int.P / D ; is closed and discrete in X .
In particular, the set AnA is closed and discrete in X , so there is U 2 .x; X/ such
that U \ .AnA/ D fxg. By regularity of X there is V 2 .x; X/ with V U ; let

F D V \ A. It is immediate that x 2 F nF ; assume that y x and y 2 F . Then
y 2 V \ A U \ A; however, the unique point of U which can belong to AnA is x
which is distinct from y. Therefore y 2 A and hence y 2 V \ A D F , i.e., x is the
unique point of F nF . Thus F is an almost closed subset of A with x 2 F which
proves that X is a Whyburn space.
T.209. Prove that any radial space is weakly Whyburn.
Solution. Let X be a radial space and take any non-closed A X . Denote by
the minimal cardinality of a set B A such that B is not contained in A. Fix a set
B A such that jBj D and BnA ;; pick any x 2 BnA. The space X being
radial, there is a regular cardinal and a -sequence S D fx W < g B with
S ! x. Since x 2 S and S B, the choice of B shows that S jS j and
hence D ; let S D fx W < g for each < . The set F D fS W < g
contains S and hence x 2 F . Observe also that jS j < and hence S A for
each < . Consequently, F A and the last thing we must prove is that F is
almost closed.
Given any y 2 X nfxg take disjoint U; V 2 .X/ such that x 2 U and y 2 V .
Since S ! x, there is < such that S nS U and therefore y S nS . Thus
y 2 F implies y 2 S F which proves that x is the unique point of F nF , i.e.,
F is almost closed.
T.210. Prove that any k-space, which is a Whyburn, is FrchetUrysohn. In
particular, any sequential Whyburn space as well as any Cech-complete Whyburn
space is FrchetUrysohn.
Solution. Assume that X is a Whyburn k-space; take any A X and x 2 AnA.
There is an almost closed F A such that x 2 F . Since F is not closed, there is a
compact K X such that P D K \ F is not closed in K. The set F [ fxg is closed
and hence so is K \.F [fxg/ D .K \F /[fxg D P [fxg. Therefore P D P [fxg,
i.e., the unique point of P nP is x. Furthermore, P is a compact Whyburn space (it
is compact because it is a closed subspace of K and it is an easy exercise that a
closed subspace of a Whyburn space is Whyburn). Thus P is a FrchetUrysohn
space by Problem 204 and hence there is a sequence fan W n 2 !g P A with
an ! x. This proves that X is a FrchetUrysohn space.
Fact 1. Every Cech-complete space is a k-space.
Proof. Take any Cech-complete space Z and assume that A Z and A \ K is

closed in K for any compact K Z. To obtain a contradiction, suppose that A is not
closed in Z and fix a point z 2 AnA. ThereT exists a family fOn W n 2 !g .Z/
such that OnC1 On for all n 2 ! and fOn W n 2 !g D Z. Choose a sequence
fUn W n 2 !g .z; Z/ such that U0 D O0 and U nC1 Un \ On for every
n 2 !T(the bar denotes theT closure in the space Z). It is easy to see that the set
P D fUn W n 2 !g D fU n W n 2 !g is a compact subspace of Z. The set
K D P \ A is closed in P and hence compact, so there is V 2 .z; Z/ such that
V \ K D ;.
Our next step is to construct a family fVn W n 2 !g T.z; Z/ such that T V0 D V

and V nC1 Vn \ Un for every n 2 !. The set Q D fVn W n 2 !g D fV n W
n 2 !g is again a compact subspace of Z such that Q \ A D ; and it is an easy
consequence of Fact 1 of S.326 that the family V D fVn W n 2 !g is an outer base
of Q in Z.
Pick a point yn 2 Vn \ A for each n 2 !; the subspace L D fyn W n 2 !g [ Q is
compact. Indeed, if U S .Z/ is an open cover of L, then there is a finite U 0 U
such that Q W D U 0 . Since V is an outer base of Q, we have Vn W
for some n 2 ! and therefore LnW fy0 ; : : : ; yn1 g is a finite set which can be
covered by a finite U 00 U. It is immediate that U 0 [ U 00 is a finite subcover of L
and hence L is compact. The set Y D L \ A D fyn W n 2 !g is not closed in L;
for otherwise, we have Y \ Q D ; and we can apply again the fact that V is an
outer base of Q to find n 2 ! such that Vn \ Y D ; which is a contradiction with
yn 2 Y \ Vn . Thus L is a compact subspace of Z such that L \ A is not closed in
L; this contradiction shows that A is closed in Z and hence Z is a k-space. Fact 1
is proved. t
u
Returning to our solution, observe that if X is sequential and A X is not
closed, then there is a sequence S D fan W n 2 !g A such that S ! x for some
x A. It is clear that K D S [ fxg is a compact subset of X such that K \ A is
not closed in K. Consequently, every sequential space is a k-space. Finally, every
Cech-complete space is a k-space by Fact 1 so our solution is complete.
T.211. Prove that any compact weakly Whyburn space is pseudoradial but not
necessarily sequential.
Solution. To prove that every compact weakly Whyburn space is pseudoradial,
we will have to establish existence of sufficiently many convergent transfinite
sequences.
Fact 1. Let Z be a compact space. If z is a non-isolated point of Z, then there exists
a regular cardinal and a -sequence S D fz W < g Znfzg such that S ! z.
Proof. Let .z; Z/ D and fix a base B D fB W < g .z; Z/ at the point z in
the space Z. There is nothing to prove if D !, so we assume that is uncountable.
For each < , there is a closed G -set F such that z 2 F B (see Fact 2 of
S.328).
The ordinal D cf./ is regular and we can choose T an increasing -sequence
f W < g which is cofinal in . Let H D fF W < g for all
< . Since each H is the intersection of < -many open sets, it is impossible
that fzg D H because .z; Z/ D .z; Z/ D (see Problem
T 327 of [TFS]). Choose
a point z 2 H nfzg for each < . We have fzg D fH W < g, so Fact 1 of
S.326 can be applied to conclude that H D fH W < g is a net at the point z, i.e.,
for any U 2 .z; Z/, there is < such that H U ; an immediate consequence
is that z 2 H H U for any which shows that the -sequence
fz W < g Znfzg converges to the point z. Fact 1 is proved. t
u
Fact 2. The space Z D .!1 C 1/ with its interval topology is radial.
Proof. Take any A Z and x 2 AnA. If x !1 , then .x; Z/ D !, so there is a

sequence S D fxn W n 2 !g A with S ! x. Now, if x D !1 , then A is cofinal in
!1 , and hence we can choose an increasing !1 -sequence T D fx W < !1 g A
which is cofinal in !1 . It is an easy exercise that T ! !1 so Fact 2 is proved. t
u
Given a weakly Whyburn space X let A be a non-closed subset of X . There is
an almost closed F A such that x 2 F for some x 2 X nA. We can apply Fact 1
to the compact space Z D F [ fxg to find a regular cardinal and a -sequence
S D fx W < g F A such that S ! x. This shows that X is pseudoradial.
Finally, observe that the compact space .!1 C 1/ is radial by Fact 2 and hence
weakly Whyburn by Problem 209. However .!1 C 1/ does not even have countable
tightness. Therefore not every weakly Whyburn compact space is sequential.
T.212. Give an example of a Whyburn space which is not pseudoradial.
Solution. Take any 2 !n! and let X D ! [ fg. Assume that S is a nontrivial
convergent sequence in X . Then S must converge to and we can choose infinite
disjoint sets A; B S \ !. The point is still the limit of both sequences A and
B and hence 2 A \ B while A \ B D ; by Fact 1 of S.382. This contradiction
shows that X has no nontrivial convergent sequences.
The space X is Whyburn because it has a unique non-isolated point (see
Problem 207). If X is pseudoradial, then 2 clX .!/ implies that there is a regular
cardinal and a -sequence S D fx W < g ! with S ! . The space X
being countable, if > !, then there is x 2 X such that jf < W x D xgj D
and hence the set X nfxg 2 .; X/ witnesses the fact that S does not converge
to . Therefore D ! and hence S is a nontrivial convergent sequence in X ,
i.e., we obtained a contradiction again. Thus X is a Whyburn space which is not
pseudoradial.
T.213. Prove that any scattered space is weakly Whyburn.
Solution. If Z is a space and A Z, call the set A Whyburn closed if F A for
any almost closed F A. It is evident that a space Z is weakly Whyburn if and
only if every Whyburn closed subspace of Z is closed.
Let X be a scattered space; by the remark above, it suffices to show that every
Whyburn closed subset of X is closed. For any Y X denote by i.Y / the set of
isolated points of the subspace Y . To obtain a contradiction assume that some non-
closed set A X is Whyburn closed in X . Let A0 D ; and B0 D A; suppose that
> 0 and we have families fA W < g and fB W < g with the following
properties:
(1) A \ B D ; and A [ B D A for any < ;
(2)
< < implies A
A ;
(3) if C 1 < , then AC1 D A [ i.BS /;
(4) if < is a limit ordinal, then A D fA
W
< g.
If is a successor ordinal, then take the ordinal
such that D
S
C 1 and let
A D A
[ i.B
/; B D AnA . If is a limit ordinal, let A D fA W < g
T
and B D fB W < g. It is evident that in both cases the properties (1)(3)
hold for all . Therefore this inductive construction can be accomplished for
all < D jAjC .
The space X being scattered, the property (3) implies that for each < , the
inclusion A AC1 is strict, i.e., A AC1 if B ;. If B ; for all < ,
then fAC1 nA W < g is a disjoint family of cardinality of non-empty subsets
of A which contradicts jAj < . Therefore B D ; for some < and hence
A D A by (1). Consequently, the ordinal D minf < W A \ .AnA/ ;g is
well-defined. It follows from A0 D ; and (4) that D
C 1 for some
< .
Pick a point x 2 A nA. Since A
A, we have x 2 i.B
/nA by (3). The
point x being isolated in B
, there is an open U X such that U \ B
D fxg.
Take any V 2 .X/ with x 2 V V U and consider the set F D V \ A.
Since x 2 A, we have x 2 F . We claim that F is an almost closed set. Indeed,
F A D A
[ B
and F \ B
V \ B
D fxg. Therefore F nfxg A
A
and therefore F nfxg V \ A D F which proves that F is almost closed; since
x 2 F nA, the set A is not Whyburn closed which is a contradiction. Thus every
Whyburn closed A X is closed in X and hence X is weakly Whyburn.
T.214. Observe that any sequential space is a k-space. Prove that any hereditarily
k-space (and hence any hereditarily sequential space) is FrchetUrysohn.
Solution. If X is sequential and A X is not closed, then there is a sequence
S D fan W n 2 !g A such that S ! x for some x A. It is clear that
K D S [ fxg is a compact subset of X such that K \ A is not closed in K.
Consequently, every sequential space is a k-space.
Now assume that X is a hereditarily k-space. Given any A X and x 2 AnA,
the space B D A [ fxg is a k-space and hence there is a compact K B such that
F D K \A is not closed in B. It is evident that F K D .K \A/[fxg D F [fxg;
besides, x 2 F because F is not closed. Consequently, F is an almost closed subset
of A such that x 2 F which proves that X is a Whyburn space. Any Whyburn
k-space is FrchetUrysohn by Problem 210, so X is a FrchetUrysohn space.
T.215. Prove that there exist hereditarily weakly Whyburn spaces which are not
Whyburn.
Solution. The compact space K D .!1 C 1/ is scattered because it cannot be
continuously mapped on I (see Fact 1 of S.319 and Problem 133). Therefore K
is hereditarily weakly Whyburn (see Problem 213) because every subspace of a
scattered space is scattered. On the other hand, the space K is not Whyburn because
it is not FrchetUrysohn (see Problem 204).
T.216. Prove that if X is d.!/-stable and X n is a Hurewicz space for each natural
n, then Cp .X / is a Whyburn space. In particular, if X is a -compact space, then
Cp .X / is a Whyburn space.
Solution. Given a space Z say that vet.Cp .Z// D ! if it has the following
property:
T
. / if Ak Cp .Z/ for every k 2 ! and f 2 fAk W k 2 !g, then for each k 2 !,
S
we can choose a finite Bk Ak in such a way that f 2 fBk W k 2 !g.
It was proved in Fact 2 of T.188 that vet.Cp .Z// D ! is equivalent to the fact
that Z n is a Hurewicz space for all n 2 N.
Thus it follows from our hypothesis vet.Cp .X // D !; any Hurewicz space is
Lindelf so t.Cp .X // D ! by Problem 149 of [TFS]. Take any A Cp .X / and
f 2 AnA. There is a countable B A such that f 2 BnB. Since X is d.!/-
stable, the space Cp .X / is i w.!/-monolithic by Problem 177 and hence the space
B has countable pseudocharacter. Therefore we can find aTfamily fOn W n 2 !g 2
.f; Cp .X // such that O nC1 On for each n 2 ! and . fOn W n 2 !g/ \ B D
ff g. If Cn D On \ B, then f 2 C n for all n 2 !. Since vet.Cp .X // D !, we can
S
find a finite Dn Cn such that f 2 D where D D fDn W n 2 !g A. Observe
that D B and hence all accumulation points of D belong to B. Furthermore, the
set .Cp .X /nO n / \ D D0 [

[ Dn1 is finite for any n 2 !; since for any
g 2 Bnff g we have g 2 W D Cp .X /nO n for some n 2 !, the function g has
a neighborhood W whose intersection with D is finite. This proves that f is the
unique accumulation point of D and hence D is almost closed.
We have proved that for any A Cp .X / and any f 2 AnA, there is an almost
closed D A such that f 2 D. Therefore Cp .X / is Whyburn. To finish our
solution, observe that if X is -compact, then X n is also -compact for all n 2 N.
Since every -compact space is Hurewicz by Fact 1 of T.132, the space Cp .X / is
Whyburn if X is -compact.
T.217. Prove that for a paracompact space X , if Cp .X / is Whyburn, then X is a
Hurewicz space. In particular, if X is metrizable and Cp .X / is Whyburn, then X is
separable.
Solution. If we have a space Z and A Z, then A W Cp .Z/ ! Cp .A/ is the
restriction map defined by A .f / D f jA for every f 2 Cp .Z/. A retraction in Z
is a map r W Z ! Z such that r r D r. A set R Z is a retract of Z if there is a
retraction r W Z ! Z such that r.Z/ D R.
Fact 1. Given a space Z and a non-empty discrete family U .Z/, choose a
point zU 2 U and a function fU 2 Cp .Z/ such that fU .zU / D 1 and fU .ZnU / D
f0g for every U 2 U. For the set A D fzU W U 2 Ug Z define a mapping ' by
'.f /.z/ D ff .zU /
fU .z/ W U 2 Ug for every f 2 RA and z 2 Z. Then
(1) ' maps RA into Cp .Z/ and A .'.f // D f for any f 2 RA ;
(2) ' is a linear homeomorphism between RA and L D '.RA /;
(3) L is a closed linear subspace of Cp .Z/ and the map r D ' A is a linear
retraction of Cp .Z/ onto L;
(4) for the set IA D ff 2 Cp .Z/ W f jA 0g define W L IA ! Cp .Z/ by
.f; g/ D f C g for any .f; g/ 2 L IA ; then is a linear homeomorphism.
In particular, Cp .Z/ is linearly homeomorphic to RA IA and RA is linearly
homeomorphic to a closed linear subspace of Cp .Z/.
Proof. For any z 2 Z there is a set V 2 .z; Z/ such that V intersects at most one
element of U, so we can find U 2 U such that V \ W D ; for any W 2 UnfU g.
Then '.f /.t/ D fU .t/
f .zU / for every t 2 V and hence '.f /jV D .f .zU /
fU /jV
is a continuous function. Thus '.f / is continuous by Fact 1 of S.472 and hence
' W RA ! Cp .Z/. Furthermore, '.f /.zU / D f .zU /
fU .zU / D f .zU / for every
f 2 RA and U 2 U which shows that '.f /jA D f for any f 2 RA , i.e., (1) is
proved.
It is easy to see that the map ' is linear; since A .'.f // D f for any f 2 RA
by (1), for any f; g 2 RA if '.f / D '.g/, then f D A .'.f // D A .'.g// D g
and therefore ' W RA ! L is a bijection while A jL is the inverse map for '.
Since ' maps RA into the product RZ , to prove continuity of ', it suffices to show
that z ' is continuous for any z 2 Z. Here z W RZ ! R is the natural projectionS
to the factor determined by z; recall that z .h/ D h.z/ for every h 2 RZ . If z U,
then fU .z/ D 0 for every U 2 U and hence z .'.f // D '.f /.z/ D 0 for every
f 2 RA , so the map z ' is constant and hence continuous. If z 2 U 2 U, then
z .'.f // D '.f /.z/ D fU .z/
f .zU / and hence z ' W RA ! R is continuous
because it coincides with the function fU .z/
qU where qU W RA ! R is the natural
projection to the factor determined by zU . This shows that the map ' is continuous;
since the inverse of ' is a continuous map A jL (see Problem 152 of [TFS]), the
map ' is a linear homeomorphism and we settled (2).
It follows from the linearity of ' that L is a linear subspace of Cp .Z/ and r
is also a linear map. For any f 2 Cp .Z/, we have A .'.A .f /// D A .f / by
(1) and hence r.r.f // D '.A .r.f /// D '.A .'.A .f //// D '.A .f // D
r.f / for every f 2 Cp .Z/, i.e., r is indeed a retraction. It is straightforward that
r.Cp .Z// D L, so L is closed being a retract of Cp .Z/ (see Fact 1 of S.351); this
completes the proof of (3).
It is evident that is a linear continuous map. For any function h 2 Cp .Z/, let
.h/ D .'.A .h//; h '.A .h///; we omit a trivial verification of the fact that
W Cp .Z/ ! L IA is a linear continuous map and is the identity map on
L IA . Thus the maps and are linear homeomorphisms which settles (4) and
completes the proof of Fact 1. t
u
Fact 2. Suppose that Cp .Z/ is a Whyburn space. Then any discrete U .Z/ is
countable.
Proof. If there is an uncountable discrete family U .Z/, then we can assume,
without loss of generality, that jUj D !1 and hence R!1 embeds in Cp .Z/ by Fact 1.
The Whyburn property being hereditary, we conclude that R!1 is Whyburn and
hence every subspace of R!1 is also Whyburn. The space Y D .!1 C 1/ has weight
!1 and hence it embeds in R!1 . Since Y is compact and has uncountable tightness,
it cannot be Whyburn by Problem 204. The obtained contradiction shows that R!1
is not Whyburn and therefore no discrete family of non-empty open subsets of Z is
uncountable. Fact 2 is proved. t
u
Fact 3. If Z is paracompact and Cp .Z/ is a Whyburn space, then Z is Lindelf.
Proof. Every open cover U of Z has a -discrete open refinement V by Problem 230
of [TFS]. Since discrete families of non-empty open subsets of Z are countable by
Fact 2, the family V is countable and hence Z is Lindelf. Fact 3 is proved. t
u
Fact 4. Suppose that Z is normal and Cp .Z/ is a Whyburn space. Assume that we
have a sequence fUn W n 2 Ng of open covers of Z with the following properties:
(5) Un D fUmn W m 2 !g and Umn UmC1 n
for each m 2 ! and n 2 N;
(6) for each n 2 N there exists a closed cover Fn D fFmn W m 2 !g of the space Z
such that Fmn Umn and Fmn FmC1
n
for all m 2 !.
Then we can choose Wn 2 Un for each n 2 N in such a way that fWn W n 2 Ng is an
!-cover of Z, i.e., for any finite K Z, there is n 2 N such that K Wn .
Proof. For each pair .m; n/ 2 ! N, choose fmn 2 Cp .X / such that fmn jFmn n1
and fmn j.X nUmn / 1. It is clear that the sequence Sn D ffmn W m 2 !g converges
S hn n . As a consequence, the function h 0 is in the closure of
1
to the function
the set S D fSn W n 2 Ng. Apply the Whyburn property of the space Cp .X / to
find an almost closed F S such that h 2 F . Observe that for any n 2 N, the set
Fn D F \ Sn cannot be infinite because otherwise hn 2 F nF . Therefore, for each
n 2 N, we have a natural m.n/ such that Fn ffmn W m m.n/g. For each n 2 N
let Wn D Um.n/
n
. We claim that the family fWn W n 2 Ng is an !-cover of X .
Indeed, let K be a finite subset of X . Since h 2 F , there exists fmn 2 F such
that fmn .x/ < 1 for every x 2 K and therefore K \ .X nUmn / D ;. Consequently,
K Umn Um.n/ n
D Wn and Fact 4 is proved. t
u
Returning to our solution apply Fact 3 to conclude that X is a Lindelf space.
Let fLn W n 2 Ng be a sequence of open covers of the space X . Since X is Lindelf,
without loss of generality, we may assume that each Ln is countable
S and choose an
enumeration fWmn W m 2 !g of Ln for each n 2 N. Let Umn D fWin W i mg for
all m 2 ! and n 2 N. It is clear that the family Un D fUmn W m 2 !g is a cover of
the space X for every n 2 N and Umn UmC1 n
for all m 2 !.
For each n 2 N there exists a precise closed shrinking fGm
n
W m 2 !g of the cover
Ln , i.e., fGm W m 2 !g is a closed cover of X and Gm Wmn for
n n
S all m 2 ! and
n 2 N (see Fact 1 of S.219 and Fact 2 of S.226). Now if Fmn D fGin W i mg,
then the covers Un D fUmn W m 2 !g and Fn D fFmn W m 2 !g, constructed for all
n 2 N, satisfy the hypothesis of Fact 4. Therefore we can choose Wn 2 Un so that
fWn W n 2 Ng is a(n) (!-)cover of X . Since each Wn is covered S by finitely many
elements of Ln , there exist finite families Nn Ln such that fNn W n 2 Ng D X .
Thus X is Hurewicz and our solution is complete.
T.218. Given a space X such that Cp Cp .X / is Whyburn prove that X has to be
finite if it is either countably compact or has a countable network.
Solution. If we have a space Z and A Z then A W Cp .Z/ ! Cp .A/ is the
restriction map defined by A .f / D f jA for every f 2 Cp .Z/.
Fact 1. If Z is any space and K Z is compact, then K is C -embedded in Z, i.e.,

K .Cp .Z// D Cp .K/.
Proof. Let f 2 Cp .K/; since Z Z, the set K is a closed subspace of Z. The
space Z being normal there is h 2 Cp .Z/ such that hjK D f . It is clear that
g D hjZ is continuous on Z and gjK D f so Fact 1 is proved. t
u
Returning to our solution, assume first that nw.X / D !. Then nw.Cp .X // D
! and hence Cp .X / is Lindelf; apply Problem 217 to conclude that Cp .X / is
Hurewicz and hence X is finite by Problem 132. Thus,
. / if nw.X / D ! and Cp .Cp .X // is Whyburn, then X is finite.
To prove the same for a countably compact space X note that X embeds in
Cp .Cp .X // by Problem 167 of [TFS] which implies that X is Whyburn and
hence FrchetUrysohn by Problem 204. Assume towards a contradiction that X
is infinite. Then it has a nontrivial convergent sequence S D fxn W n 2 !g. If x is
the limit of S , then K D S [fxg is a countably infinite compact subspace of X . The
map K W Cp .X / ! Cp .K/ is surjective by Fact 1, and hence the dual map .K /
embeds Cp .Cp .K// in Cp .Cp .X // (see Problem 163 of [TFS]) which proves that
Cp .Cp .K// is also Whyburn. The set K is countable, so nw.K/ D ! and hence K
is finite by . /. This contradiction shows that X also has to be finite.
T.219. Prove that there exists a separable metrizable space X such that Cp .X / is
not weakly Whyburn.
Solution. In this solution no axioms of separation are assumed by default for the
topological spaces under consideration. A space Z is called crowded if every set
U 2 .Z/ is infinite. Observe that a T1 -space is crowded if and only if it has no
isolated points.
Call a space Z maximal if it is crowded while any strictly stronger topology on
Z fails to be crowded. The space Z is maximal Tychonoff if Z is Tychonoff and
dense-in-itself while any stronger Tychonoff topology on Z has isolated points. The
space Z is nodec if every nowhere dense subset of Z is closed. Call Z irresolvable
if it is crowded and no two dense subsets of Z are disjoint. A crowded space Z
is resolvable if it is not irresolvable, i.e., there are disjoint A; B Z such that
A D B D Z. A crowded space Z is hereditarily irresolvable if every crowded
subspace of Z is irresolvable. A space Z is extremally disconnected if the closure
of any open subset of Z is open. If Z is crowded and the closures of any two disjoint
crowded subspaces of Z are disjoint, then Z is called ultradisconnected. The space
Z is perfectly disconnected if no point of z is an accumulation point of two disjoint
subsets of Z, i.e., if A; B Z and A \ B D ;, then S .AnA/ \ .BnB/ D ;.
Given a set Z and a family A exp.Z/ with A D Z, we denote by hAi the
topology generated by A as a subbase.
Fact 0. Any maximal space is T1 .
Proof. If Z is a maximal space, then the family B D fU nF W U 2 .Z/ and
F Z is finiteg generates a topology on Z as a base; it is clear that .Z/ .
Since all elements of B are infinite, the topology is crowded and hence .Z/
by maximality of Z. Thus D .Z/ and hence all finite sets are closed in Z, i.e.,
Z is a T1 -space. Fact 0 is proved. t
u
S 1. Let Z be any space; if A is a family of resolvable subspaces of Z, then

Fact
A is also resolvable.
S
Proof. In the space Y D A take a maximal disjoint
S family D of non-empty
resolvable subspaces
S of Y . We claim that the set D is dense in Y . Indeed, if
W D Y nclY . D/ ;, then there is A 2 A with W \ A ;. It is evident that
any non-empty open subspace of a resolvable space is resolvable, so A0 D W \ A is
resolvable and hence D D0 D D [fA0 g D while D0 is still disjoint and consists
S subspaces of Y ; this gives a contradiction with the maximality of D.
of resolvable
Thus T D D is dense in Y . For each D 2 D choose disjoint ES D ; FD D such
that clD .ED
S / D clD .F D / D D. It is straightforward that E D fED W DS2 Dg
and F D fFD W D 2 Dg are disjoint dense subsets of Y . Thus Y D A is
resolvable so Fact 1 is proved. t
u
Fact 2. If Z is an irresolvable space, then there is a non-empty open hereditarily
irresolvable U Z.
Proof.SLet A be the family of all non-empty resolvable subspaces of Z. The space
Y D A is resolvable by Fact 1. If a space has a dense resolvable subspace, then it
is resolvable, so Y cannot be dense in Z. It is evident that U D ZnY is as promised
u
Fact 3. A space Z is ultradisconnected if and only if for any crowded A Z, if
ZnA is also crowded, then A is clopen. In particular, every ultradisconnected T1 -
space is irresolvable.
Proof. Necessity follows from the fact that A \ ZnA D ; implies that both A
and ZnA have to be clopen in Z. To prove sufficiency assume that our hypothesis
holds and take any disjoint crowded A; B Z. If A1 D AnB and B1 D BnA1 ,
then A A1 A and B B1 B which implies that both A1 and B1 are
crowded. The spaces C D A [ B and ZnC are also crowded which shows that C is
clopen in Z. Furthermore, A1 \ B1 D ; and A1 [ B1 D A [ B which implies that
ZnA1 D B1 [ .ZnC / and therefore ZnA1 is also crowded being a union of two
crowded subspaces. Thus A1 is clopen and hence ZnA1 is clopen as well; therefore
A D A1 and B ZnA1 so A \ B D ;.
To see that any ultradisconnected T1 -space Z is irresolvable, observe that given
disjoint dense A; B Z, the subspaces A and B have to be crowded because,
in any T1 -space, a dense subspace of a crowded space is crowded. Since Z is
ultradisconnected, we have A \ B D ; which is a contradiction with A D B D Z.
Thus Z is irresolvable and Fact 3 is proved. t
u
Fact 4. Given a crowded space Z suppose that A and ZnA are crowded subspaces
of Z for some A Z. Then the topology D h.Z/ [ fA; ZnAgi is also crowded
and .Z; / is homeomorphic to the space A .ZnA/.
Proof. Let B D ZnA; if CA D fU \A W U 2 .Z/g and CB D fU \B W U 2 .Z/g,

then C D CA [ CB is a base of ; an immediate consequence is that the topology of
subspace of .Z; / on each of the sets A and ZnA coincides with the topology on
the respective set induced from Z. Since A and ZnA are open in .Z; /, the space
.Z; / is homeomorphic to A .ZnA/.
Finally, observe that every non-empty element of CA (CB ) is infinite being an
open subset of a crowded space A (or B respectively). Thus no element of C is finite
which implies that is crowded and Fact 4 is proved. t
u
Fact 5. The following conditions are equivalent for any crowded T1 -space Z:
(1) Z is maximal;
(2) Z is perfectly disconnected;
(3) every crowded subspace of Z is open;
(4) Z is ultradisconnected and nodec;
(5) Z is ultradisconnected and every discrete subspace of Z is closed;
(6) Z is submaximal and extremally disconnected;
(7) Z is extremally disconnected, hereditarily irresolvable and nodec.
Proof. Assume that (7) holds and consider any dense A Z. If ZnA has non-
empty interior, then take any V 2 .Z/ with V ZnA and observe that the sets
.ZnA/\V and A\V are disjoint dense subsets of V which contradicts the fact that
Z is hereditarily irresolvable. Thus Int.ZnA/ D ;; consequently, ZnA is nowhere
dense and hence closed in Z because Z is nodec. This proves that A is open in Z,
so Z is submaximal and we have established that (7)H)(6).
Assume that Z is submaximal and extremally disconnected. If Y is a subspace
of Z with Int.Y / D ;, then X nY is open being dense in Z. Consequently, every
subspace of Z with empty interior is closed and discrete. In particular, if D is a
discrete subspace of Z, then it has empty interior and hence D is closed. Now
assume that A and ZnA are crowded subspaces of Z. Let U D Int.A/; suppose that
there exists a point z 2 AnU . The set E D AnU has empty interior and hence E
is closed and discrete. There is W 2 .z; Z/ such that W \ E D fzg; it is clear
that V D W nU 2 .z; Z/ and V \ A D fzg which contradicts the fact that A
is crowded. Therefore A U ; analogously, if G D Int.ZnA/, then ZnA G.
The space Z being extremally disconnected, the set U is open and U \ G D ;.
Therefore G \ U D ; which implies that A \ ZnA D ; and hence A and ZnA are
clopen in Z which shows that Z is ultradisconnected by Fact 3. Thus (6)H)(5).
Now assume that (5) holds for Z and take a nowhere dense A Z. The set B D
ZnA is dense in Z and hence crowded because Z is a T1 -space. As a consequence,
every B 0 B is also crowded, so if the set I of isolated points of A is not dense in
A, then C D AnI is a non-empty crowded subspace of A, and by the above remark,
the set B 0 D ZnC is also crowded which implies that C is clopen because Z is
ultradisconnected. However, C is nowhere dense because C A; this contradiction
shows that the discrete subspace I is dense in A and hence A D I because (5) holds
for Z and hence I has to be closed. Thus every nowhere dense subspace of Z is
closed, i.e., Z is nodec so we settled (5)H)(4).
If Z is ultradisconnected and nodec, then take any crowded A Z and let E to

be the set of all isolated points of ZnA. The set E is nowhere dense and hence closed
in Z because Z is nodec. Furthermore, the set F D .ZnA/nE is crowded and hence
F \A D ; because Z is ultradisconnected. Consequently, ZnA D F [E D F [E
is disjoint from A which shows that ZnA D ZnA, i.e., A is open in Z and hence
(4)H)(3).
Now suppose that every crowded subspace of Z is open and take any disjoint
sets A; B Z such that z 2 .AnA/ \ .BnB/. If U D Int.A/, then P D AnU has
empty interior and hence ZnP is crowded being dense in Z (we applied once more
the fact that Z is T1 ). By our hypothesis, the set ZnP is open and therefore P is
closed in Z. Since z 2 A D P [ U D P [ U , we have z 2 U because z A while
P A. Thus the crowded space U is dense in U [ fzg whence V D U [ fzg is
crowded and hence open. It turns out that V is a neighborhood of z which is disjoint
from B. Therefore z B; this contradiction proves that (3)H)(2).
Assume that Z is perfectly disconnected and take any crowded A Z. If z 2 A,
then z 2 Anfzg because A is crowded. Now perfect disconnectedness of Z implies
that z ZnA; the point z 2 A was taken arbitrarily, so ZnA \ A D ;
and therefore ZnA D ZnA, i.e., A is open in Z. This proves that (2)H)(3) and
hence (2) (3).
Now, if (3) holds for Z and is a crowded topology on Z with .Z/ ,
assume that there is A 2 n.Z/. Then A is crowded in .Z; / and hence in
Z because for every U 2 .Z/ we have U \ A 2 and hence U \ A is
infinite whenever it is non-empty. Apply (3) to conclude that A 2 .Z/ which is
a contradiction. Thus D .Z/ and hence Z is maximal, i.e., (3)H)(1).
Finally, assume that Z is maximal. If A Z and both sets A and ZnA are
crowded then the topology D h.Z/ [ fA; ZnAgi is crowded by Fact 4 and
hence D .Z/ which shows that A 2 .Z/ and ZnA 2 .Z/, i.e., A is clopen
in Z. Therefore Z is ultradisconnected by Fact 3. Given any open U Z both
sets U and ZnU are crowded, so U is also open which shows that Z is extremally
disconnected.
Now if A is a crowded subspace of Z and B; C A are disjoint and dense in
A, then they are both crowded because Z is a T1 -space. However, we have A
B \ C D ; because Z is ultradisconnected; this contradiction shows that Z is
hereditarily irresolvable. Finally, if A is nowhere dense in Z, then D D ZnA is
dense in Z. Consider the topology D h.Z/ [ fDgi. It is clear that .Z/ and
the base in .Z; / is given by the family E D .Z/ [ fU \ D W U 2 .Z/g. The
T1 -property of Z implies that U \ D is infinite for every U 2 .Z/ and hence
all elements of E are infinite which shows that is a crowded topology on Z. By
maximality of Z we have D .Z/ and therefore D 2 .Z/, i.e., A is closed in
Z. Thus every nowhere dense subset of Z is closed in Z, i.e., Z is nodec and we
established that (1)H)(7) so Fact 5 is proved. t
u
Fact 6. For any Tychonoff crowded space Z, there exists a maximal Tychonoff
topology on the set Z such that .Z/ .
Proof. Let T be the family of all TychonoffS crowded topologies on Z such that
.Z/ . If C is a chain in T , then D h Ci contains all topologies from C and
is Tychonoff by Problem 099 of [TFS]. The base of is given by all (non-empty)
intersections U D U1 \

\ Un where Ui 2 i and i 2 C for all i n. The
collection C being a chain, there is 2 C such that i and hence Ui 2 for
all i n. Therefore U 2 and hence U is infinite. Consequently, the topology
is crowded because it has a base all elements of which are infinite. Thus Zorns
lemma is applicable to T to conclude that there is a maximal element 2 T . It is
immediate that is as promised so Fact 6 is proved. t
u
Fact 7. A space Z is maximal Tychonoff if and only if Z is Tychonoff and
ultradisconnected.
Proof. Suppose that Z is a maximal Tychonoff space and take any A Z such that
both A and ZnA are crowded. The topology D h.Z/ [ fA; ZnAgi is stronger
than .Z/ and the space .Z; / is homeomorphic to A .ZnA/ (see Fact 4) which
shows that .Z; / is Tychonoff and crowded. By maximality of .Z/ in the family
of all Tychonoff crowded topologies on Z, we have A 2 .Z/ and ZnA 2 .Z/,
i.e., A is clopen in Z, so Z is ultradisconnected by Fact 3 which proves necessity.
To prove sufficiency, assume that Z is an ultradisconnected Tychonoff space and
take any crowded Tychonoff topology on the set Z such that .Z/ . By
regularity of .Z; / the topology has a base B such that U D Int.U / for every
U 2 B (the bar denotes the closure in the space .Z; / and the interior is also taken
in .Z; /).
Given any U 2 B, the set X nU is dense in X nU ; for otherwise, there is a non-
empty W 2 with W U such that W is not contained in U which contradicts the
fact that U D Int.U / is the largest set from contained in U . Thus ZnU contains a
dense crowded subspace ZnU and hence ZnU is crowded in .Z; / for any U 2 B.
It turns out that both sets U and ZnU are crowded in .Z; / and hence in Z for
any U 2 B. Since Z is ultradisconnected, we have U 2 .Z/ for any U 2 B and
therefore D .Z/. This shows that Z is maximal Tychonoff and finishes the proof
of Fact 7. t
u
Fact 8. Suppose that Z is a countable Tychonoff hereditarily irresolvable space.
Then the set T D fz 2 Z W z 2 DnD for some discrete subspace D Zg is
nowhere dense in Z.
Proof. Observe that
. / if Int.T / D ;, then T is nowhere dense.
Indeed if Int.T / D ;, then ZnT is dense in Z, so if T \ U is dense in some
subspace U 2 .Z/, then T \ U and .ZnT / \ U are disjoint dense subsets of U
which contradicts the fact that Z is hereditarily irresolvable.
Therefore it suffices to prove that T has empty interior; to obtain a contradiction,
assume that U T for some U 2 .Z/. Then we can choose for any z 2 U a
discrete E.z/ U such that z 2 E.z/nE.z/. Take any z0 2 U and let D0 D fz0 g.
Suppose that n 0 and we have sets D0 ; : : : ; Dn with the following properties:
(i) Di U and Di is discrete for any i n;

(ii) D i \ Dj D ; if i < j n;
(iii) Di D j for any i < j n.
To construct DnC1 observe that we can choose Ud 2 .d; Z/ for each d 2 Dn
such that the family fUd W d 2 Dn g is disjoint (see Fact 1 of S.369) and Ud \D i D ;
for every i < n. Observe that Ud \ .D n nDn / D ; for every d 2 Dn and let Ad D
E.d / \ US d for every d 2 Dn . It is evident that Ad is discrete and d 2 Ad nAd . Let
DnC1 D fAd W d 2 Dn g. The conditions (ii) and (iii) are obviously satisfied. As
to (i), we must only prove that DnC1 is discrete. Pick any z 2 DnC1 ; then z 2 Ad for
some d 2 Dn . Since Ad is discrete, there is W 2 .z; Z/ such that W \ Ad D fzg;
it is immediate that .W \ Ud / \ DnC1 D fzg, so DnC1 is also discrete and hence our
inductive construction can be continued to provide a sequence fDn W n 2 !g such
for all n 2 !.
that (i)(iii) hold S
The set D D fDn W n 2 !g is dense-in-itself because given any d 2 D, we
have dS2 Dn for some n 2 ! and hence S d 2 D nC1 nDnC1 Dnfd g by (iii). Let
G D fD2n W n 2 !g and H D fD2nC1 W n 2 !g; again, if d 2 D, then
d 2 Dn for some n 2 ! and hence d 2 D 2n \ D 2nC1 by (iii). Therefore G and
H are disjoint dense subsets of D which contradicts the fact that Z is hereditarily
irresolvable. This contradiction proves that Int.T / D ; and hence T is nowhere
dense by . /. Fact 8 is proved. t
u
Fact 9. There exists a countable Tychonoff maximal space.
Proof. By Fact 6 there exists a maximal Tychonoff topology on the set Q such
that .Q/ . The space Q D .Q; / is ultradisconnected by Fact 7 and hence
irresolvable by Fact 3. Apply Fact 2 to find a hereditarily irresolvable open R Q.
The space T D fz 2 R W z 2 DnD for some discrete D Rg is nowhere dense in R
by Fact 8, so there is S 2 .R/ such that S \ T D ;. It is obvious that no point of
S is an accumulation point of a discrete subspace of S , so every discrete subspace
of S is closed. The space Q being ultradisconnected, S is also ultradisconnected
and hence maximal by Fact 5. Fact 9 is proved. t
u
Fact 10. If Z is a Tychonoff submaximal countable space, then Z A.!/ is not
weakly Whyburn.
Proof. Let s D 0 and sn D nC1 1
for all n 2 !; then the space S D fsn W n 2 !g[fsg,
with the topology induced from R, is homeomorphic to A.!/, so it suffices to show
that Z S is not weakly Whyburn. Let Sn D fs0 ; : : : ; sn g for any n 2 ! and take a
bijection ' W Z ! S nfsg; then its graph Y D f.z; '.z// W z 2 Zg Z S is not
closed in Z S . Indeed, take any point t D .z; s/ 2 Z fsg. If U 2 .t; Z S /, then
there is n 2 ! and U1 2 .z; Z/ such that U1 .S nSn / U . The set P D ' 1 .Sn /
is finite and U1 is infinite, so there is z 2 U1 nP . It is obvious that .z; '.z// 2 U \ Y
which shows that U \ Y ; for any U 2 .t; Z S / and hence t 2 Y nY .
If Z S is weakly Whyburn, then there is an almost closed F Y such that
u 2 F for some u 2 .Z S /nY . If u D .z; sn / for some n 2 ! and z 2 Z, then
W D Z fsn g is an open neighborhood of u with jW \ Y j 1, so u cannot be
an accumulation point for F . Therefore u D .z; s/ for some z 2 Z. It is clear that

u 2 F 1 where F1 D F nf.z; '.z//g; let W Z S ! Z be the natural projection.
Then, for the set H D .F1 / we have z 2 H nH . If Int.H / D ;, then ZnH is
dense in Z; since Z is submaximal, the set ZnH is open and hence H is closed
which contradicts z 2 H nH .
Therefore O D Int.H / ; and hence we can choose y 2 Onfzg. Given any
open neighborhood V of the point u0 D .y; s/, there is m 2 ! such that V1
.S nSm / V for some V1 2 .y; Z/ with V1 O. The set B D F1 \ .Z Sm /
is finite, so there is w D .a; sk / 2 F1 nB such that a 2 O. It is immediate that
w 2 V \F1 and therefore V \F1 ; for each V 2 .u0 ; Z S /. As a consequence,
u0 2 F 1 F , which is a contradiction because u0 2 .Z S /nY and u0 u. Thus
Z S is not weakly Whyburn and Fact 10 is proved. t
u
Returning to our solution, observe that it follows from Facts 9 and 10 that there
exists a countable space T which is not weakly Whyburn. The space X D Cp .T /
is separable and metrizable while T embeds in Cp .X / as a closed subspace (see
Problem 167 of [TFS]). It is an easy exercise that a closed subspace of a weakly
Whyburn space is weakly Whyburn; therefore Cp .X / is not weakly Whyburn and
T.220. Prove that there exists a compact weakly Whyburn space which is not
hereditarily weakly Whyburn.
Solution. If and are ordinals such that , then ; D f
W
g
and .; / D f
W <
< g. Analogously, ; / D f
W
< g and
.; D f
W <
g. Given a space Z and a set A Z, say that A is
Whyburn closed if F A for any almost closed F A. It is evident that a space
Z is weakly Whyburn if and only if any Whyburn closed subset of Z is closed in
Z. Since any nontrivial convergent sequence is almost closed, any Whyburn closed
A Z is sequentially closed in Z, i.e., if S D fan W n 2 !g A is a sequence
with S ! x, then x 2 A. It is also immediate that a space Z is sequential if and
only if any sequentially closed subspace of Z is closed in Z.
Fact 1. If M is an uncountable second countable space, then there is an uncountable
M 0 M such that every U 2 .M 0 / is uncountable.
Proof. Let A D fx 2 M W there is a countable Ux 2 .x; M /g. Since A is Lindelf,
the open cover U D fUx W x 2 Ag of the set A has a countable
S subcover U 0 U.
0
Since each element of U is countable, we have jAj j Uj D ! and therefore the
set M 0 D M nA is uncountable.
Given any non-empty open subset U of the space M 0 , find a set V 2 .M / with
V \ M 0 D U . Then V is uncountable because there is a point x 2 V \ M 0 , and, by
definition of M 0 , every open neighborhood of x is uncountable. Thus U D V nA is
also uncountable so Fact 1 is proved. t
u
Fact 2. For any second countable uncountable space M the space Z D L.!1 / M
is not weakly Whyburn.
Proof. Recall that L.!1 / D !1 [ fag where a is the unique non-isolated point of
L.!1 / and the neighborhoods of a are the complements of countable subsets of !1 .
We will need the set I D fag M ; for any t 2 M let xt D .a; t/ 2 I . Choose any
injection ' W !1 ! M ; then its graph A D f.; '.// W 2 !1 g is not closed in Z.
Indeed, the set P D '.!1 / M is uncountable, so there is a point t 2 P such that
U \ P is uncountable for any U 2 .t; M / (see Fact 1). Given any W 2 .xt ; Z/
there is a countable C !1 and U 2 .t; M / such that .L.!1 /nC / U W .
Since U \ P is uncountable, there is s 2 .U \ P /n'.C /. If D ' 1 .s/, then
C and therefore .; s/ 2 ..L.!1 /nC / U / \ A W \ A which shows that
W \ A ; for any W 2 .xt ; Z/ and hence xt 2 AnA.
Now assume that F A is almost closed non-closed subset of Z; if x D .; t/
for some < !1 and t 2 M , then G D fg M is an open neighborhood of
the point x such that jG \ F j 1, so x F nF . Therefore F is uncountable and
F nF I . The set H D '.F / M being uncountable, we can apply Fact 1
again to find two distinct points t; s 2 M such that U \ H is uncountable for any
U 2 .M / with U \ ft; sg ;. Given any W 2 .xt ; Z/, there is a countable
C !1 and U 2 .t; M / such that .L.!1 /nC / U W . Since U \ '.F / is
uncountable, there is t 0 2 .U \ '.F //n'.C /. If D ' 1 .t 0 /, then C and hence
.; t 0 / 2 .L.!1 /nC / U / \ A W \ A which shows that W \ A ; for any
W 2 .xt ; Z/ and hence xt 2 F nF . Analogously, xs 2 F nF which contradicts
the fact that F is an almost closed subset of Z. Thus Z is not weakly Whyburn and
Fact 2 is proved. t
u
Returning to our solution note that the space X D .!1 C 1/ I is compact and
Y D X n.f!1 g I/ D !1 I is first countable and hence sequential. To see that X
is weakly Whyburn take any Whyburn closed non-closed A X . By the remark
above, A is sequentially closed and hence A \ Y is sequentially closed in Y ; by
sequentiality of Y , the set A0 D A \ Y is closed in Y . Analogously, if I D f!1 g I,
then I is second countable and hence sequential so G D A \ I is closed in I .
As a consequence, AnA I ; take any x D .!1 ; t/ 2 AnA and fix " > 0 such
that .f!1 g .t "; t C "// \ G D ;. It is easy to see that the set P D .!1 ftg/ \ A
is almost closed in X . If P is uncountable, then x 2 P nP which contradicts the
fact that A is Whyburn closed in X .
Now assume that P is countable and hence there exists < !1 such that .; t/
A for any . Let K D f.
; t/ W
g D ; ftg for each
2 !1 n. It is clear that K is compact; since A0 is closed in Y and K \ A0 D ;
for any , we can apply Fact 3 of S.271 to conclude that there is n./ 2 N
such that ; .t n./
1
; t C n./
1
/ does not intersect A0 . There is an uncountable
R ; !1 / and n 2 N such that n./ D n for every 2 R. It is immediate that the
set H D ; !1 / .t n1 ; t C n1 / does not intersect A0 and hence H \ A D ;. Now,
if D minf n1 ; "g, then W D .; !1 .t ; t C / is an open neighborhood of x;
since W H [ .f!1 g .t "; t C "//, we have W \ A D ; which is a contradiction
with x 2 A. Therefore the set A has to be closed and hence X is weakly Whyburn.
To see that the space X is not hereditarily weakly Whyburn, consider its subspace
Z D .L [ f!1 g/ I where L D f < !1 W D C 1 for some < !1 g.
Observe that M D L [ f!1 g is homeomorphic to L.!1 / because !1 is the unique

non-isolated point of M and a set U M is a neighborhood of !1 if and only if
!1 2 U and M nU is countable. Consequently, the space Z is homeomorphic to
L.!1 / I and therefore Z is not weakly Whyburn by Fact 2. Thus X is a compact
weakly Whyburn space while Z X is not weakly Whyburn which shows that our
T.221. Prove that any metrizable space is a p-space and a -space at the same
time.
Solution. Let .X; d / be a metric space. Then C D ffxg W x 2 X g is a closed cover
of X with compact subsets; the space X has a -discrete base B (see Problem 221
of [TFS]) which is, evidently, a network with respect to C. This proves that X is a
-space.
For each x 2 X and n 2 N, take Uxn 2 .X / such that Uxn \ X D Bd .x; n1 /.
Then Un D fUxn W xT2 X g .X / is a cover of X for each n 2 !. Assume that
z 2 X nX and z 2 fSt.x; Un / W n 2 Ng for some x 2 X . Take any W 2 .x; X /
such that z W (the bar denotes the closure in X ) and find n 2 N for which
Bd .x; n2 / W \ X . Since z 2 St.x; Un /, there is y 2 X such that x 2 Uyn and
z 2 Uyn . Consequently, d.x; y/ < n1 , and given any t 2 Bd .y; n1 /, we have d.x; t/
d.x; y/ C d.y; t/ < n2 which shows that t 2 Bd .x; n2 /. Therefore Bd .y; n1 /
Bd .x; n2 / W which implies
T z 2 Uyn D Bd .y; n1 / W , a contradiction with the
choice of W . Therefore fSt.x; Un / W n 2 Ng X for each x 2 X and hence X is
a p-space.
T.222. Prove that Cp .X / is a p-space if and only if X is countable.
Solution. If X is countable, then Cp .X / is second countable and hence metrizable
which shows that Cp .X / is a p-space by Problem 221.
Fact 1. If Z is a p-space, then it is of pointwise countable type, i.e., for every
z 2 Z, there is a compact K Z such that z 2 K and .K; Z/ !.
Proof. Let fUn W n 2 !g be the sequence of open (in Z) covers of Z which
witnesses that Z is a p-space. Choose Un 2 Un such that z 2 Un for all n 2 !. It
is easy to construct a sequence fVn W n 2 !g .z; Z/ such that Vn T Un and
V nC1 VnT(the bar denotes the closure in Z) for all n 2 !. The set K D fVn W
n 2 !g D fV n W n 2 !g is compact and
\ \
z2K fUn W n 2 !g fSt.z; Un / W n 2 !g Z:
Since K is a G -set in Z, we have .K; Z/ D ! by Problem 327 of [TFS] and

hence .K; Z/ !. Fact 1 is proved. t
u
Returning to our solution assume that Cp .X / is a p-space. Then it is of pointwise
countable type by Fact 1 an hence there is a compact K Cp .X / with
.K; Cp .X // D !. Now apply Problem 170 of [TFS] to conclude that X is
countable.
T.223. Prove that every Lindelf p-space is a Lindelf -space. Give an example
of a p-space which is not a -space.
Solution. Suppose
S that we have a space Z and a family A exp.Z/; then
St.z; A/ D fA W A 2 A and z 2 Ag. Given a family Un .Z/ for every S
n 2 !, call the sequence
T fUn W n 2 !g a feathering of the space Z if Z Un
for each n 2 ! and fSt.z; Un / W n 2 !g Z for every z 2 Z. It is clear that a
sequence fUn W n 2 !g is a feathering of Z if and only if it witnesses the fact that
Z is a p-space. Given families U and V of subsets of Z, say that U is (strongly)
inscribed in V if for any U 2 U there is V 2 V such that U V (or U V
respectively). If A exp.Z/ and z 2 Z, then A.z/ D fA 2 A W z 2 Ag and
ord.z; A/ D jA.z/j; given any Y Z we let AjY D fA \ Y W A 2 Ag.
Call Z a strong -space if there exists a compact cover C of the space Z and a
-discrete family F exp.Z/ which is a network with respect to C (it is evident
that strong -spaces are -spaces). A space Z is subparacompact if every open
cover of Z has a (not necessarily open) -discrete refinement.
Fact 1. Suppose that SL is a Lindelf subspace of a space Z. Then, for any family
S
U .Z/ with L U, there exists a countable V .Z/ such that L V,
the family V is strongly inscribed in U and ord.y; V/ < ! for any y 2 L.
Proof. For every y 2 L take Uy 2 U such that y 2 Uy and use regularity of Z to
find Vy 2 .y; Z/ such that clZ .Vy / Uy . If V 0 D fVy W y 2 Lg, then V 0 jL is an
open cover of L, so by paracompactness of L (see Fact 1 of S.219), there is locally
finite open (in the space L) refinement V 00 of the cover V 0 jL. By Lindelfness of L
we can assume that V 00 is countable. For any W 2 V 00 , there is VW 2 V 0 such that
W VW \L; take any OW 2 .Z/ such that OW \L D W and let GW D VW \OW .
It is straightforward that V D fGW W W 2 V 00 g is as promised so Fact 1 is proved.
t
u
Fact 2. Every strong -space is subparacompact.
Proof. Assume that Z is a strong -space and take an open cover U of the space Z.
Let C be a compact cover of Z such thatSthere exists a -discrete family F which is a
network with respect to C. Thus F D fFn W n 2 !g where Fn is discrete for every
n 2 !.S A set F 2 F will be called U-small if there is a finite UF U such that
F UF . In particular, there is kF 2 ! such that UF D fU.F; 0/; : : : ; U.F; kF /g
for each U-small F 2 F ; let H.F; i / D F \ U.F; i / for all i kF . The family
H.i; n/ D fH.F; i / W F 2 Fn is U-small and i kF g is discrete for any i; n 2 !
because each H.i; n/ is obtained from Fn by shrinking some of its elements. It
is immediate fromSthe definition that H.i; n/ is inscribed in U for all i; n 2 !.
The family H D fH.i; n/ W i; n 2 !g being -discrete, it suffices to show that
S
H D Z. Take any z 2 Z; there is K 2 C such that z 2 K. The set K being
compact there is a finite U 0 U such that K US0
. Since the family F is a network
with respectSto C, there is F 2 F with K F U 0 . Consequently, F S is U-small;
since F D fH.F; i / W i kF g, there is i kF with z 2 H.F; i / H. Thus
H is a -discrete refinement of U and Fact 2 is proved. t
u
Fact 3. Every locally compact space is p-space.

Proof. If Z is locally compact, then is open in Z, and we can take Un D fZg for
any n 2 !. It is evident that the sequence fUn W n 2 !g is a feathering of Z so Fact 3
is proved. t
u
Fact 4. The space .A.!2 / A.!2 //nf.a; a/g is not subparacompact.
Proof. Recall that A.!2 / D !2 [ fag where all points of !2 are isolated in A.!2 /
and the neighborhoods of the point a !2 are the complements of finite subsets of
!2 . Let H D A.!2 /fg and V D fgA.!2 / for every < !2 . If h D f.a; /g
and v D f.; a/g for each < !2 , then H is a compact open neighborhood of
h and V is a compact open neighborhood of v . Let H D fH W < !2 g and
V D fV W < !2 g; to obtain a contradiction assume that the open S cover H [ V
of the space .A.!2 / A.!2 //nf.a; a/g has a refinement F D fFn W n 2 !g
such that Fn is discrete for all n 2 !. Every F 2 F is contained in an element of
H [ V, so F D FH [ FV where FH D fF 2 F W F H for some < !2 g and
FV D fF 2 F W F V for some < !2 g.
Observe that jF \ H j 1 for everyS F 2 FV ; besides, h 2 DnD for every
infinite D H nfh g and h Bn D .Fn \ FV / for all < !2 and n 2 !.
The family Fn \ FV being discreteS for all n 2 !, the set Bn \ H is finite for all
< !2 . Analogously,
S if An D n \ FH /, then An \ VS
.FS is finite for any < !2 .
Let A D fAn W n 2 !g and B D fBn W n 2 !g. Then F D A [ B while the
sets A \ V and B \ H are countable for all < !2 .
For every < !1 , there is ./ < !2 such that .; / A \ V for every
./. There is
< !2 such that ./ <
for all < !1 . Consequently, the
set P D f.;
/ W < !1 g does not intersect A and therefore P H
\ B which
is a contradiction because H
\ B has to be countable. Thus the open cover H [ V
of the space .A.!2 / A.!2 //nf.a; a/g has no -discrete refinement and Fact 4 is
proved. t
u
Returning to our solution, suppose that X be a Lindelf p-space and fix its
0
S 0 fUn W n 2 !g. Observe that if Un0 .X / is inscribed in Un and
feathering
X Un for all n 2 !, then the sequence fUn W n 2 !g is still a feathering of X .
The space X being a Lindelf subspace of X , we can construct inductively, using
Fact 1, a sequence fVn W n 2 !g with the following properties:
S
(1) Vn .X /; X Vn and Vn is countable for each n 2 !;
(2) VnC1 is strongly inscribed in the family Un ^ Vn D fU \ V W U 2 Un and
V 2 Vn g for any n 2 !;
(3) ord.x; Vn / < ! for any x 2 X and n 2 !.
Observe that it follows from (1) and (2) that the sequence fVn W n 2 !g is a
feathering of X . Furthermore, (2) and (3) imply that
[ [
St.x; VnC1 / D fV W V 2 VnC1 .x/g fV W V 2 Vn .x/g D St.x; Vn /
for anyTpoint x 2 X and n 2 T ! (the bar denotes the closure in X ). Therefore

Kx D fSt.x; Vn / W n 2 !g D fSt.x; Vn / W n 2 !g is a compact subset of X and
it follows from Fact 1 of S.326 that Cx D fSt.x; Vn / W n 2 !g is an outer base of Kx
in X . Consequently, Cx jX is an outer base of Kx in X . Since eachS Vn .x/ is finite,
the family C 0 of all finite unions of the elements of the family V D fVn W n 2 !g
is a countable network in X with respect to the compact cover F D fKx W x 2 X g
of the space X . Therefore C D C 0 jX is a countable (and hence -discrete) network
with respect to the family F in X which shows that X is a -space and proves that
any Lindelf p-space is a Lindelf -space.
As to an example of a p-space which is not a -space, observe that the space
Z D .A.!2 / A.!2 //nf.a; a/g is locally compact being open in the compact space
A.!2 / A.!2 /. Therefore Z is a p-space by Fact 3. Let us show that
. / every closed countably compact subspace of Z is compact.
Let p1 W Z ! A.!2 / and p2 W Z ! A.!2 / be the restrictions of the relevant
natural projections of the space A.!2 / A.!2 / onto its first and second factors
respectively. If K Z is countably compact, then there exists a finite Q !2
such that K p11 .Q/ [ p21 .Q/. Indeed, if such Q does not exist, then we can
choose inductively an infinite W !2 and an injective map ' W W ! !2 such
that G D f.; '.// W 2 W g K. It is immediate that G is a closed discrete
subspace in Z and hence in K which contradicts countable compactness of K. Thus
K R D p11 .Q/ [ p21 .Q/ for some finite Q A.!2 /. Since R is compact and
K is closed in R, the space K is compact, i.e., . / is established.
An immediate consequence of . / is that if Z is a -space, then it is a strong
-space and hence subparacompact by Fact 2. Since Z is not subparacompact by
Fact 4, the space Z is an example of a p-space which is not a -space, so our
T.224. Prove that
(i) any closed subspace of a -space is a -space. In particular, any closed
subspace of a Lindelf -space is a Lindelf -space;
(ii) any closed subspace of a p-space is a p-space. In particular, any closed
subspace of a Lindelf p-space is a Lindelf p-space.
Solution. Suppose that K is a compact space and Z is dense in K. If Un .K/
for every n 2 !, call the sequence
S fUn W n 2 !g a feathering
T of the space Z in its
compactification K if Z Un for each n 2 ! and fSt.z; Un / W n 2 !g Z for
every z 2 Z.
Fact 1. A space Z is a p-space if and only if there exists a compactification K of
the space Z such that Z has a feathering in K.
Proof. If Z is a p-space, then it has a feathering in Z so necessity is easy. Now,
assume that Z has a feathering fUn W n 2 !g in some compactification K. There
is a continuous onto map f W Z ! K such that f .z/ D z for each z 2 Z
(see Problem 258 of [TFS]). The family Vn D ff 1 .U / W U 2 Un g consists of
S
open subsets of Z and Z Vn for all n 2 !. Given any z 2 Z observe that
St.z; Vn / D f 1 .St.z; Un // for each n 2 ! and hence
\ \
fSt.z; Vn / W n 2 !g D ff 1 .St.z; Un // W n 2 !g f 1 .Z/ D Z
(see Fact 1 of S.259). Thus the sequence fVn W n 2 !g is a feathering of Z in Z

and hence Z is a p-space. Fact 1 is proved. t
u
(i) Assume that X is a -space and take a closed cover C of the space X for which
there exists a -discrete network F with respect to C. If P X is closed, then
the family C 0 D fC \P W C 2 Cg is a closed cover of P with countably compact
subsets of P and it is straightforward that the family F 0 D fF \ P W F 2 F g
is a -discrete network with respect to C 0 . Thus P is a -space.
(ii) Suppose that X is a p-space and F is a closed subset of X . Fix a feathering
fUn W n 2 !g of the space X in X and let Vn D fU \ clX .F / W U 2 Un g for
each n 2 !. It is immediate that the sequence fVn W n 2 !g is a feathering of F
in its compactification K D clX .F /, so F is a p-space by Fact 1.
T.225. Prove that X is a Lindelf -space if and only if X has a countable network
with respect to a compact cover C.
Solution. Suppose that X is a Lindelf -space and fix a closed cover C of X with
countably compact subspaces of X such that there exists a -discrete network F
with respect to C. Every closed subspace of a Lindelf space is Lindelf and every
Lindelf countably compact space is compact, so C consists of compact subsets
of X . Furthermore, ext.X / l.X / D ! and hence every discrete family in X
is countable. Consequently, F is a countable network with respect to the compact
cover C of the space X . This proves necessity.
Now assume that X has a compact cover C for which there is a countable network
F with respect to C. It is clear that X is a -space, so we only must establish that
X is Lindelf. Let U be an open cover of the space X . A set F S 2 F will be called
U-smallS if there is a finite family U F U for which F UF . The family
U 0 D fUF W F 2 F is U-smallg is countable and U 0 U. Given any x 2 X ,
there is C 2 C such
S that x 2 C . The set C being compact there is a finite V U
such that C V. Since the family
S F is a network with respect to C, there is
F 2 F for which S C S F V. It is evident that F is U-small and hence
x 2 C F UF U 0 which shows that we found a countable subcover U 0
of the cover U so X is Lindelf and sufficiency is proved.
S
T.226. Suppose that X D fXn W n 2 !g where Xn is countably compact and
closed in X for any n 2 !. Prove that X is a -space. As a consequence, every
-compact space is a Lindelf -space.
Solution. Let F D C D fFn W n 2 !g; then C is a closed cover of X with countably
compact subspaces ofSX and F is a network with respect to C. This proves that X
is a -space. If Y D fKn W n 2 !g where each Kn is compact, then Kn is closed
in Y , so Y is a -space by the previous observation. Every -compact space is

Lindelf so Y is a Lindelf -space.
T.227. Prove that the Sorgenfrey line and the space L.!1 / are examples of Lindelf
spaces which are not Lindelf .
Solution. A cover C of a space Z is called compact if all elements of C are compact;
C is a -cover of Z if there exists a -discrete F exp.Z/ which is a network
with respect to C.
Fact 1. If X and Y are Lindelf -spaces, then X Y is also a Lindelf -space.
Proof. Apply Problem 225 to find compact -covers CX and CY of the spaces X
and Y respectively and let FX and FY be the respective -discrete networks. The
spaces X and Y being Lindelf, the families FX and FY are countable. It is evident
that C D fK L W K 2 CX and L 2 CY g is a compact cover of the space X Y .
To see that C is a -cover, let F D fF G W F 2 FX and G 2 FY g. If C 2 C
and O 2 .C; X Y /, then C D K L for some K 2 CX and L 2 CY , so we can
apply Fact 3 of S.271 to find U 2 .K; X/ and V 2 .L; Y / such that U V O.
Since FX and FY are networks for CX and CY , respectively, there exist F 2 FX and
G 2 FY such that K F U and L G V . Then H D F G 2 F and
C H O which proves that F is a network with respect to C. The family F is
countable, so we can apply Problem 225 again to conclude that X Y is a Lindelf
-space. Fact 1 is proved. t
u
Fact 2. Assume that X is a Lindelf -space such that every compact subspace of
X is finite. Then X is countable.
Proof. Apply Problem 225 to take a compact cover C of the space X such that there
exists a countable network F with respect to C. We do not lose generality assuming
that F is closed under finite intersections. Let C 0 D fC 2 C W C 2 F g; it is clear that
C 0 is countable. Assume for a moment that there is an element C 2 CnC 0 ; since F
is a network for C, there is a sequence S D fFn W n 2 !g F such that FnC1 Fn
for all n 2 ! and S is a network at C in the sense that for any U 2 .C; X/ there
is n 2 ! for which C Fn U . Since Fn C for all n 2 !, we can pick
xn 2 Fn nC for each n 2 !.
The space K D C [ fxn W n 2 !g is compact. Indeed, ifSU is an open (in X )
cover of K, then there is a finite U 0 U such that C U D U 0 . There is m 2 !
such that Fm U and hence xn 2 Fn Fm U for all n m which S 00shows
that KnU is finite and hence there is a finite U 00 U with KnU U . It is
immediate that U 0 [ U 00 is a finite subcover of U so K is, indeed, compact.
The set P D fxn W n 2 !g cannot be finite; for otherwise, X nP 2 .C; X/ and
hence xn 2 Fn X nP for some n 2 ! which is a contradiction. Thus K is an
0
infinite compact subset of X ; this contradiction shows thatS C D C and hence C is
countable. Since every element of C is finite, the set X D C is countable so Fact 2
is proved. t
u
Returning to our solution, recall that it was proved in Problem 165 of [TFS] that
Sorgenfrey line S is Lindelf while S S is not Lindelf. Thus S cannot be a
Lindelf -space by Fact 1. The space L.!1 / is Lindelf by Problem 354 of [TFS]
and it is easy to see that every compact subspace of L.!1 / is finite, so L.!1 / is not
Lindelf by Fact 2.
T.228. Prove that any space with a -discrete network is a -space. In particular,
if nw.X / !, then X is a Lindelf -space.
Solution. If F is a -discrete network of a space Y , then it is a network with
respect to the cover C D ffyg W y 2 Y g of the space Y which consists of closed
compact subsets of Y . Thus Y is a -space. Every countable network is -discrete,
so nw.X / ! implies that X is a -space. Since every space with a countable
network has to be Lindelf, the space X is Lindelf .
T.229. Let X be a metrizable space. Prove that Cp .X / is a -space if and only if
X is second countable.
Solution. If we have a set P , then Fin.P / is the family of all finite subsets of P .
Given a space Z say that a cover C of Z is compact (closed) if all elements of C are
compact (or closed in Z, respectively). A cover C of a space Z is called a -cover
of Z if there exists a -discrete F exp.Z/ which is a network with respect to
C. Thus Z is a -space if and only if it has a closed -cover whose elements are
countably compact. Call a space Z a strong -space if Z has a compact -cover.
If A exp.Z/, then A.z/ D fA W A 2 A and z 2 Ag. Given a set A !1 and
s 2 ! A let s; A D ft 2 ! !1 W tjA D sg. It is evident that the family B D fs; A W
A 2 Fin.!1 / and s 2 ! A g is a base of the space ! !1 .
S
theS choice of i , so we have consistently defined a function f which will be denoted
by ffi W i 2 I g
Fact 1. Suppose that C is a compact -cover of a space Z. Then there is a -
discrete network F with respect to C such that each F 2 F is closed in Z.
S
Proof. There is some network G with respect to C such that G D fGn W n 2 !g
and Gn is discrete for all n 2 !. Let Fn D fG W G 2SGn g; then the family Fn is
discrete for all n 2 !, so it suffices to show that F D fFn W n 2 !g is a network
with respect to C.
Take any C 2 C; if O 2 .C; Z/, then there is V 2 .C; Z/ such that V O
(this is an easy exercise on using compactness of C ). The family G being a network
with respect to C there is G 2 G such that C G V . Then F D G 2 F and
C F V O, so F is a -discrete network with respect to C and Fact 1 is
proved. t
u
Fact 2. Let Z be a strong -space. Then there exists a -discrete

T family F
exp.Z/ such that all elements of F are closed in Z and Cz D F .z/ is compact
for any z 2 Z.
Proof. Take a compact -cover C of the space Z and a -discrete network F with
respect to C such that all elements of F are closed in Z. This is possible by Fact 1.
Given any z 2 Z there is C 2 C with z 2 C . The family FC D fF 2 FTW C F g
is contained in FT.z/; since F is a network
T with respect to C, we have FC D C
and hence Cz D fF 2 F W z 2 F g FC D C . Therefore Cz is compact being
a closed subspace of a compact space C . Fact 2 is proved. t
u
Returning to our solution observe that if w.X / !, then nw.Cp .X // D ! (see
Problem 172 of [TFS]) and therefore Cp .X / is a Lindelf -space by Problem 228.
Now, if Cp .X / is a -space and w.X / > !, then ext.X / > ! and hence R!1
embeds in Cp .X / as a closed subspace by Fact 1 of S.215. This implies that R!1 is
a -space by Problem 224 and hence ! !1 is also a -space because it embeds in
R!1 as a closed subspace. To obtain a contradiction observe first that
. / every closed countably compact subspace of ! !1 is compact
because it is realcompact and pseudocompact (see Problems 401, 403 and 407
of [TFS]); consequently, if C is a closed -cover of ! !1 by countably compact
subspaces of ! !1 , then each C 2 C has to be compact by . /. Thus ! !1 has a
compact -cover, i.e., ! !1 is a strong -space. Apply T Fact 2 to find a -discrete
family F of closed subspaces
S of ! !1
such that C s D F .s/ is compact for every
s 2 ! !1 . We have F D fFn W n 2 !g where the family Fn is discrete for each
n 2 !. Let n D F0 [

[ Fn for all n 2 !. We have the following important
property:

. / If U 2 S .! / and G is a -discrete family of closed subspaces of ! with
!1 !1
U G, then there is G 2 G such that Int.G \ U / ;.

S
To prove . / recall that we have G D S fGn W n 2 !g where the family Gn
for every n 2 !. The set Gn D
is discrete S Gn is closed in ! !1 for each n 2 !
and U D fGn \ U W n 2 !g. Since the space ! !1 has the Baire property (see
Problems 464, 465 and 470 of [TFS]), some Gn \ U is not nowhere dense in U ;
being a closed subset of U , the set Gn \ U has a non-empty interior W in U ; it is
evident that W 2 .! !1 /. Each G 2 Gn is open in Gn because Gn is discrete, so
take any G 2 Gn with W 0 D W \ G ;; then W 0 is open in ! !1 , non-empty and
W 0 G \ U.S S \ U / ; so . / is proved.
This shows that Int.G
S Let F n D F n and P n D fFi W i < ng for all n 2 !; it follows from
F D ! !1 and the Baire property of ! !1 that the number m0 D minfn 2 ! W
Int.Fn / ;g is well-defined. Apply . / to choose a set H0 2 Fm0 such that
W D Int.H0 / ;. The set Pm0 is nowhere dense in ! !1 , so there is V 2 .! !1 /
such that V H0 nPm0 . Therefore there exists a finite set A0 !1 and a function
s0 2 ! A0 such that s0 ; A0 H0 nPm0 .
Assume that n 2 ! and we have constructed a sequence m0 ; : : : ; mn of natural

numbers, finite sets A0 ; : : : ; An , functions f0 ; : : : ; fn and sets H0 ; : : : ; Hn with the
(1) A0

An !1 and fi 2 ! Ai for all i n;
(2) m0 <

< mn and Hi 2 Fmi for all i n;
(3) fj jAi D fi whenever 0 i < j n;
(4) fi ; Ai Hi for every i n;
(5) fi ; Ai \ Fj D ; for any i n and j 2 f0; : : : ; mn gnfm0 ; : : : ; mn g.
Observe that it follows from (1) and (3) that fj ; Aj fi ; Ai T if 0 i <
j n. Besides, fn ; An Hn which shows, together with (5), that mn .s/
fn ; An for any s 2 fn ; An . If there
T is some s 2 fn ; An such that s Fk for all
k > mn , then the compact set Cs D mn .s/ contains Sa non-compact closed subset
fn ; An which is a contradiction. Thus fn ; An fFk W k > mn g, so we can
apply . / to conclude that the number mnC1 D minfk > mn W Int.Fk \fn ; An /
;g is well-defined. Take any W 2 .! !1 / with W 2 FmnC1 \ fn ; An ; there is
G 2 FmnC1 such that G \ W ;. The set G is open in FmnC1 because FmnC1 is
discrete. Therefore W 0 D WS\ G is open in ! !1 and contained in G \ fn ; An .
Observe that the set Q D fFk \ fn ; An W mn < k < mnC1 g is closed and
nowhere dense, so we can find V 2 B such that V W 0 nQ. By definition of B,
there is and finite AnC1 !1 and a function fnC1 2 ! AnC1 such that An AnC1
and fnC1 jAn D fn . It is immediate that the conditions (1)(5) are satisfied for the
natural numbers m0 ; : : : ; mnC1 , finite sets A0 ; : : : ; AnC1 , functions f0 ; : : : ; fnC1 and
sets H0 ; : : : ; HnC1 . Therefore our inductive construction can be continued to obtain
a sequence fmi W i 2 !g !, a collection fAi W i 2 !g of finite sets, a sequence
ffi W i 2 !g of functions and a family fHi W i 2 !g such that the conditions (1)(5)
are satisfied for S all n 2 !.
Let A D S fAi W i 2 !g. It follows from the conditions (1) and (3) that the
function f D ffi W i 2 !g is well-defined on A and f jAT i D fi for all i 2 !.
Therefore f; A fi ; Ai for all i 2 ! and hence f; A fHi W i 2 !g by the
property (4). T If s 2 f; A , then s Fk for all k 2 !nfmn W n 2 !g which shows
that Cs D fHi W i 2 !g and hence a non-compact closed set f; A is contained
in a compact set Cs which is a contradiction. Therefore ! !1 is not a -space; this
final contradiction shows that R!1 is not a -space and hence Cp .X / cannot be a
-space either. Thus X has to be separable and our solution is complete.
T.230. Prove that any p-space is a k-space. Give an example of a Lindelf -space
which is not a k-space.
Solution. Suppose that K is a compact space and Z is dense in K. If Un .K/
for every n 2 !, call the sequence
S fUn W n 2 !g a feathering
T of the space Z in its
every z 2 Z. It is clear that every p-space Z has a feathering in Z.
Take any p-space X and assume that A X and A \ K is closed in K for
any compact K X . To obtain a contradiction, suppose that A is not closed in X
and fix a point z 2 AnA. Let F D fUn W n 2 !g be a feathering of X in X ; take
T
On 2 Un such that z 2 On for every n 2 !. Since F is a feathering, we have fOn W
n 2 !g X . Choose a sequence fUn W n 2 !g .z; X / such that U0 D O0 and
U nC1 Un \ O0 \

\ On for every n 2T! (the bar denotes Tthe closure in the
space X ). It is easy to see that the set P D fUn W n 2 !g D fU n W n 2 !g is a
compact subspace of X . The set K D P \ A is closed in P and hence compact, so
there is V 2 .z; X / such that V \ K D ;.
Our next step is to construct a family fVn W n 2 !g T.z; X / such that TV0 D V
and V nC1 Vn \ Un for every n 2 !. The set Q D fVn W n 2 !g D fV n W
n 2 !g is again a compact subspace of X such that Q \ A D ; and it is an easy
consequence of Fact 1 of S.326 that the family V D fVn W n 2 !g is an outer base
of Q in X .
Pick a point yn 2 Vn \ A for each n 2 !; the subspace L D fyn W n 2 !g [ Q is
compact. Indeed, if U S .X / is an open cover of L, then there is a finite U 0 U
such that Q W D U 0 . Since V is an outer base of Q, we have Vn W
for some n 2 ! and therefore LnW fy0 ; : : : ; yn1 g is a finite set which can be
covered by a finite U 00 U. It is immediate that U 0 [ U 00 is a finite subcover of L
and hence L is compact. The set Y D L \ A D fyn W n 2 !g is not closed in L;
for otherwise, we have Y \ Q D ; and we can apply again the fact that V is an
outer base of Q to find n 2 ! such that Vn \ Y D ; which is a contradiction with
yn 2 Y \ Vn . Thus L is a compact subspace of X such that L \ A is not closed
in L; this contradiction shows that A is closed in X and hence X is a k-space. This
proves that every p-space is a k-space.
Any countable space is Lindelf because it has a countable network (see
Problem 228), so to finish our solution it suffices to construct a countable space
which is not a k-space. Take any 2 !n! and consider the space N D ! [ fg.
All compact subspaces of N are closed in !; since every infinite closed subspace
of ! is uncountable (see Fact 2 of T.131), no compact subset of N is infinite. It
is clear that ! is dense in N while K \ ! is closed in K for any compact K N
because K has to be finite. This shows that N is a Lindelf -space which is not a
k-space.
T.231. Give an example of a countable space which is not a p-space. Note that this
example shows that not every Lindelf -space is a p-space.
Solution. Any countable space is Lindelf because it has a countable network
(see Problem 228). Take any 2 !n! and consider the space N D ! [ fg. All
compact subspaces of N are closed in !; since every infinite closed subspace of
! is uncountable (see Fact 2 of T.131), no compact subset of N is infinite. It is
clear that ! is dense in N while K \ ! is closed in K for any compact K N
because K has to be finite. This shows that N is a countable space (and hence a
Lindelf -space) which is not a k-space. Thus N is not a p-space either because
every p-space is a k-space by Problem 230.
T.232. Prove that any Cech-complete space is a p-space. Give an example of a
p-space which is not Cech-complete.
Solution.
T If X is Cech-complete, fix a family fUn W n 2 !g .X S / such that
fUn W n 2T!g D X and let Un D fUnT g for all n 2 !. Then X Un for each
n 2 ! and fSt.x; Un / W n 2 !g D fUn W n 2 !g D X so X is a p-space.
This proves that every Cech-complete space is a p-space. Finally, the space Q is a
p-space being second countable (see Problem 221) while it is not Cech-complete
because it is of first category in itself (see Problem 274 of [TFS]).
(i) for an arbitrary compactification bX of the space X , there exists a countable
family of compact subspaces of bX which separates X from bX nX ;
(ii) there exists a compactification bX of the space X and a countable family of
compact subspaces of bX which separates X from bX nX ;
(iii) there exists a compactification bX of the space X and a countable family of
Lindelf -subspaces of bX which separates X from bX nX ;
(iv) there exists a space Z such that X is a subspace of Z and there is a countable
family of compact subspaces of Z which separates X from ZnX ;
(v) X is a Lindelf -space.
Solution. A family C of subsets of a space Z is compact if all elements of C are
compact. A cover C of Z is called a -cover if there is a -discrete F exp.Z/
which is a network with respect to C. Observe that in a Lindelf space Z, any -
discrete family is countable, so if Z is Lindelf and C is a -cover of Z, then there
is actually a countable network with respect to C.
The implication (i)H)(ii) is trivial; (ii)H)(iii) is an easy consequence of the
fact that each compact space is Lindelf . To prove (iii)H)(iv) assume (iii) and
let F be a countable family of Lindelf -subspaces of bX which separates X
from bX nX . For any element F 2 F there is a compact -cover CF of the space
F (see Problem 225); let NF be the relevant countable network in F with respect
to CF . The family G D fP W P 2 NF for some F 2 F g consists ofScompact
subsets of bX (the bar denotes S in bX ). Observe that F D
Sthe closure NF for
any F 2 F and hence X F G which shows that G separates X from
bX nX if bX nX D ;. Now, if x 2 X and y 2 bX nX , then there is F 2 F such
that x 2 F and y F . There is C 2 CF such that x 2 C ; it is clear that y C
and hence there is U 2 .C; bX / such that y U . The family NF being a network
with respect to CF , there is P 2 NF such that C P U \ F . Consequently,
x 2 C P U bX nfyg which shows that x 2 P while y P . Thus G is
a countable family of compact subsets of Z D bX which separates X from ZnX ,
i.e., (iii)H)(iv) is proved.
To prove the implication (iv)H)(v) take a countable family F 0 of compact
subsets of Z which separates X from ZnX . We can assume, without loss of
generality that the family F 0 is closed under finite intersections. Let F D fF \ X W
F 2 F 0 g.T Then F is a countable family of subsets of X . Given any point x 2 X
let Cx D fF W F 2 F 0 and x 2 F g. Since F 0 separates X from ZnX , we have
Cx X and hence C D fCx W x 2 X g is a compact cover of X . Take any x 2 X
and U 2 .Cx ; X /; there is V 2 .Z/ such T that V \ X D U . For the family

Fx0 D fF W F 2 F 0 and x 2 F g, we have Cx D Fx0 V and hence we can apply
Fact 1 of S.326 (and the fact that F 0 is closed under finite intersections) to conclude
that there is F 2 Fx0 F 0 such that Cx F V and hence Cx F \ X U
which, together with F \ X 2 F , shows that F is a network with respect to C.
Therefore X is a Lindelf -space by Problem 225 and hence we established that
(iv)H)(v).
Finally, assume that X is a Lindelf -space and fix a compact cover C of the
space X such that there is a countable network F with respect to C. The family
G D fF W F 2 F g is countable and consists of compact subsets of bX (the bar
denotes the closure in bX ). Since F is a cover of X , so is G which shows that we
have the required separation of X from bX nX when bX nX D ;. If bX nX ;,
take any x 2 X and y 2 bX nX ; there is C 2 C with x 2 C . Since y C , there
is U 2 .C; bX / such that y U . Since F is a network with respect to C, there is
F 2 F such that C F U \ X . Then G D F 2 G and we have x 2 C G
while G U and hence y G. Thus G is a countable family of compact subsets of
bX which separates X from bX nX , so (v)H)(i) is proved and hence our solution
is complete.
T.234. Let X be a space of countable tightness such that Cp .X / is a -space. Prove
that if Cp .X / is normal, then it is Lindelf.
Solution. Since t.X / D !, we have tm .X / t.X / D ! (see Problem 419
of [TFS]). Therefore q.Cp .X // D tm .X / D !, i.e., Cp .X / is realcompact (see
Problem 429 of [TFS]). The space Cp .X / being a -space, there is a cover C
of Cp .X / such that all elements of C are closed and countably compact while
there exists a -discrete network F with respect to C. Every closed subspace of a
realcompact space is realcompact and every realcompact countably compact space
is compact (see Problems 403 and 407 of [TFS]), so all elements of C are compact.
The space Cp .X / being normal, we have ext.Cp .X // D ! (see Problem 295 of
[TFS]) and hence the family F has to be countable. Thus we can apply Problem 225
to conclude that Cp .X / is a Lindelf -space.
T.235. Let X be a -space with a G -diagonal. Prove that X has a -discrete
network.
S
Solution. If Z is a space and A exp.Z/, then St.z; A/ D fA 2 A W z 2 Ag; let
^A be the family of all finite intersections of the elements of A. If A1 ; : : : ; An are
families of subsets of Z, then A1 ^

^ An D fA1 \

\ An W Ai 2 Ai for all
i ng. The space Z is subparacompact if every open cover of Z has a -discrete
(not necessarily open) refinement.
T if and only if there is a sequence fDn W n 2 !g

Fact 1. A space Z has a G -diagonal
of open covers of Z such that fSt.z; Dn / W n 2 !g D fzg for any z 2 Z. The
sequence fDn W n 2 !g is called a G -diagonal sequence for Z.
Proof. Let D f.z; z/ W z 2 Zg be the diagonal of Z. Given S a G -diagonal

sequence fDn W n 2 !g for the space Z, consider the set On D fU U W U 2 Dn g;
it is evident that On is an open subset of Z Z for every n 2 !. For any z 2 Z and
n 2 !, there is U 2 Dn such that z 2 U and hence .z; z/ 2 U U On . Therefore
On for every n 2 !. If z D .x; y/ 2 .Z Z/n, then x y and hence
there exists n 2 ! such that y St.x; Dn /. Now, if U 2 Dn and z 2 U U , then
fx; yg U which shows that y 2 St.x; Dn / which is a contradiction. T Therefore
.x; y/ U U for any U 2 Dn , i.e., z On which proves that D fOn W n 2 !g
and hence .Z/ D !.
Now assume that .Z/ D ! and fix T a family fOn W n 2 !g .Z Z/ such
that OnC1 On for all n 2 ! and fOn W n 2 !g D . Given any n 2 !,
consider the family Dn D fU 2 .Z/ W U U On g. If z 2 Z, then .z; z/ 2 On
and hence there is U 2 .z; Z/ such that U U On which shows that Dn is
a cover of Z for all n 2 !. Given distinct x; y 2 Z, there is n 2 ! such that
.x; y/ On ; if U 2 Dn , then .x; y/ U U and therefore fx; yg is not contained
in U forTany U 2 Dn . It is easy to see that this implies that y St.x; Dn / and hence
fxg D fSt.x; Dn / W n 2 !g for any x 2 Z, i.e., fDn W n 2 !g is a G -diagonal
sequence in Z, so Fact 1 is proved. t
u
Fact 2. Every countably compact space Z with .Z/ D ! is compact and hence
metrizable.
Proof. Apply Fact 1 to fix a G -diagonal sequence fDn W n 2 !g of open covers
of the space Z. It is obvious that if Dn0 is an open refinement of Dn for any n 2 !,
then the family fDn0 W n 2 !g is also a G -diagonal sequence in Z. Thus, letting
Dn0 D D0 ^

^Dn for each n 2 !, we obtain a G -diagonal sequence fDn0 W n 2 !g
such that Di0 is a refinement of Dj0 whenever j < i . To simplify the notation, we
will assume that Dn D Dn0 , i.e., Di is a refinement of Dj whenever j < i .
To prove compactness of Z, it suffices to show that Z is Lindelf. To obtain a
contradiction, assume that there is an open cover U of the space Z which has no
countable Ssubcover. Call a set Y Z small if there is a countable U 0 U such
that Y U 0 . Otherwise the set Y will be called large. Observe
S that a countable
union of small sets is a small set, and hence if a set Y D fYn W n 2 !g is large,
then some Yn has to be large. Observe also that S if0 a set Y is small and closed in Z,
then there is a finite U 0 U such that Y S U because Y is countably compact.
Choose any z0 2 Z; since the set Znfz0 g D fZnSt.z0 ; Dn / W n 2 !g is large,
there exists n0 2 ! such that ZnSt.z0 ; Dn0 / is large. Suppose that 0 < < !1 and
we have chosen sets fz W < g Z and fn W < g ! with the following
properties:
S
(1) z
fSt.z ; DnS / W <
g for any
< ;
(2) the set L
D Zn. fSt.z ; Dn / W
g/ is large for any
< .
If D
C 1 forS some
< !1 , then choose any point z 2 L
and observe that
the set L
nfz g D fL
nSt.z ; Dn / W n 2 !g is large and hence there is n 2 !
such that L
nSt.z ; Dn / is large. It is immediate that the properties (1) and (2) are
fulfilled for all
.
T
Now if is a limit ordinal, then the set M D fL
W
< g is large. Indeed, S 0
suppose that there is a countable family U 0 U such that M U D U;
then fL
nU W
< g T is decreasing and consists of closed subsets of a countably
compact space Z and fL
nU W
< g D ;. Take any increasing sequence
f
n W n 2 !g T such that lim
n D . If Fn D L
n nU , then FnC1 Fn for
every n 2 ! and fFn W n 2 !g D ;. By countable compactness of Z, there is
n 2 ! such that Fn D ; and hence L
n U which shows that L
n is small, a
contradiction. Therefore M is large and reasoning as in the case of a successor ,
we can choose z 2 M and n 2 ! such that the set L D M nSt.z ; Dn / is
large. It is clear that in this case we also have the properties (1) and (2) for all

and therefore our construction can be continued to give us sets fz W < !1 g Z
and fn W < !1 g ! with the properties (1) and (2) for each < !1 .
There is n 2 ! and an uncountable A !1 such that n D n for all 2 A. The
set E D fz W 2 Ag is closed and discrete in Z because, given z 2 Z, there is
U 2 Dn with z 2 U ; it easily follows from (1) that U cannot contain more than one
element of E. The space Z being countably compact, this is a contradiction which
proves that Z is Lindelf and hence compact. Finally, apply Problem 091 to see that
Z is metrizable. Fact 2 is proved. t
u
Fact 3. If Z is a space and F exp.X / is -discrete, then the family G D ^F is
also -discrete.
S
Proof. We have F D fFn W n 2 !g where Fn is discrete for all n 2 !. If G 2 G,
then there is n 2 ! and distinct j0 ; : : : ; jn 2 ! such that G D F0 \

\ Fn for
some F0 2 Fj0 ; : : : ; F
Sn 2 Fjn .
Therefore ^F D fFj0 ^

^ Fjn W j0 ; : : : ; jn 2 ! are distinctg, so it suffices
to show that the family Fj0 ^

^ Fjn is discrete for any distinct j0 ; : : : ; jn 2 !.
Take any z 2 Z and i n; since Fji is discrete, there is Ui 2T.z; Z/ and Fi 2 Fji
such that Ui \ F D ; for any FT2 Fji nfFi g. Then U D i n Ui 2 .z; Z/. If
Gi 2 Fji for all i n and G D i n Gi , then U \ G ; implies U \ Gi ;
and therefore Gi D Fi for every i n, i.e., U intersects at most one element of
Fj0 ^

^ Fjn . Fact 3 is proved. t
u
Fact 4. A space Z is subparacompact if and only if any open cover of Z has a
closed -discrete refinement.
Proof. Sufficiency
S is obvious, so take any subparacompact space Z and U .Z/
such that U D Z. For every z 2 Z take Uz 2 U with z 2 U and Vz 2 .z; Z/
such that V z Uz . The cover fVz W z 2 Zg has a -discrete refinement F ; if
G D fF W F 2 F g, then G is a closed -discrete refinement of U so Fact 4 is
proved. t
u
Returning to our solution fix a G -diagonal sequence fDn W n 2 !g in the space
X (which exists by Fact 1). Let C be a closed cover of X such that every C 2 C
is countably compact and there is a -discrete network F with respect to C. The
property .X / D ! implies that .C / D ! for any C 2 C and therefore C
is compact and metrizable by Fact 2. Therefore, for every U 2 .C; X/, there is
V 2 .C; X/ such that V U ; as a consequence, F1 D fF W F 2 F g is also a

-discrete network with respect to the family C.
It follows from compactness of all elements of C that X is subparacompact (see
Fact 2 of T.223). Apply Fact 4 to find a -discrete closed
S refinement En of the cover
Dn for every n 2 !. It is clear that the family E D fEn W n 2 !g is also -discrete
and therefore G D ^E is -discrete as well by Fact 3; besides, G is closed under
finite intersections.
Let N D G ^ F1 ; the family N is -discrete being contained in ^.G [ F1 / (see
Fact 3), so it suffices to show that N is a network in the space X . Pick any x 2 X
and U 2 .x; X/; there is C 2 C such that x 2 C . T Furthermore, for each n 2 !,
thereTis En 2 En such that x 2 En ; it follows from fSt.x; Dn / W n 2 !g D fxg
that fEn W n 2 !g D fxg. Therefore
T the family S D fEn \ C W n 2 !g consists
of compact subsets of X and S D T fxg. Now apply Fact 1 of S.326 to conclude
that there is a finite S 0 0
TS 0such that S U whichTshows that there is a finite
E E for which x 2 . E / \ C U . The set P D E 0 belongs to G; since P
0
is closed, the set W D .X nP / [ U is an open neighborhood of C . The family F1

being a network with respect to C, there is F 2 F1 such that C F W . Now
N D P \ F 2 N and x 2 N . Besides, N F W implies that N U [ .X nP /
which together with N P implies N U . Therefore, for any x 2 X and any
U 2 .x; X/, there is N 2 N such that x 2 N U . Consequently, N is a
-discrete network of X , so our solution is complete.
T.236. Let X be a Lindelf -space of countable pseudocharacter. Prove that
jX j c.
Solution. Take a compact cover C of the space X such that there is a countable
network F with respect to the family C (see Problem 225). It is easy to see that
every C 2 C is the intersection of a family FC F . Therefore jCj j exp F j
j exp !j D c. We have .C / D .C / .X / D ! (see Problem 327 of [TFS])
and hence jC j 2.C / D 2! D c for any C 2 C (see Problem 329 of [TFS]). This
implies jX j jCj
c D c
c D c.
T.237. Prove that, under CH, there exists a hereditarily separable compact space X
such that Cp .X / does not have a dense -subspace.
Solution. It was proved in Problem 098 that, under CH, there exists a non-
metrizable compact space X such that hd .X / D !. We have hl .Cp .X // D !
by Problem 027, so if S Cp .X / is a dense -subspace of Cp .X /, then S is
Lindelf .
Given any x 2 X let '.x/.f / D f .x/ for any f 2 S . The function '.x/ is
continuous on S for any x 2 X and the map ' W X ! Cp .S / is continuous by
Problem 166 of [TFS]. It follows from the density of S is Cp .X / that S separates
the points of X and hence ' W X ! Y D '.X / is a condensation by Fact 2 of S.351.
Since every condensation of a compact space is a homeomorphism, the space Y is
homeomorphic to X , i.e., X embeds in Cp .S /.
Fact 1. Every continuous image of a Lindelf -space is a Lindelf -space.
Proof. Suppose that Z is a Lindelf -space and f W Z ! T is a continuous

onto map. By Problem 225 we can choose a compact cover C of the space Z such
that there exists a countable network F with respect to C. It is evident that D D
ff .C / W C 2 Cg is a compact cover of T ; we claim that G D ff .F / W F 2 F g is
a (countable) network with respect to D. Indeed, if D 2 D and U 2 .D; T /, then
there is C 2 C with f .C / D D and therefore V D f 1 .U / 2 .C; Z/. The family
F being a network with respect to C, there is F 2 F such that C F V ; then
G D f .F / 2 G and D G U which shows that G is a countable network with
respect to the compact cover D of the space T . Thus T is a Lindelf -space and
Fact 1 is proved. t
u
Fact 2. Every Lindelf -space is !-stable.
Proof. Let Z be a Lindelf -space and take any continuous onto map f W Z ! T
such that there is a condensation of the space T onto a second countable space. It is
immediate that .T / i w.T / D !; the space T is Lindelf by Fact 1, so we
can apply Problem 235 to conclude that T has a -discrete network F . In a Lindelf
space, every -discrete family is Countable, so nw.T / jF j D ! which shows that
Z is !-stable and hence Fact 2 is proved. t
u
Returning to our solution, observe that the space S being !-stable (see Fact 2) the
space Cp .S / is !-monolithic (see Problem 154) and hence any separable subspace
of Cp .S / has a countable network. Since Y is separable, we have w.X / D nw.X / D
nw.Y / D ! and hence X is metrizable which is a contradiction (see Fact 4 of
S.307). Thus X is a hereditarily separable compact space such that there is no dense
Lindelf -subspace in the space Cp .X /.
T.238. Prove that, under CH, the space Cp .!/ is not a -space.
Solution. It was proved in Problem 237 that, under CH, there exists a separable
compact space X such that there is no dense -subspace in Cp .X /. In particular,
Cp .X / is not a -space. Take any countable dense D X ; if p W ! ! D is any
surjection, then p is continuous and hence there is a continuous map ' W ! ! X
such that 'j! D p. Since '.!/ contains a dense subset D of the space X , it is
dense in X and hence '.!/ D X .
The dual map ' W Cp .X / ! Cp .!/ embeds Cp .X / in Cp .!/ as a closed
subspace (see Problem 163 of [TFS]) because the map ' is closed and hence
R-quotient. If Cp .!/ is a -space, then ' .Cp .X // is also a -space (see
Problem 224) which is a contradiction because ' .Cp .X // is homeomorphic to
Cp .X / while Cp .X / is not a -space.
T.239. Prove that for any metrizable X , the space Cp .X / has a dense Lindelf -
subspace.
Solution. Since X is metrizable, the space Cp .X / has a dense -compact subspace
by Problem 313 of [TFS]. Since every -compact subspace is Lindelf by
Problem 226, the space Cp .X / has a dense Lindelf -subspace.
T.240. Let p W X ! Y be compact-valued upper semicontinuous onto map. Prove

that l.Y / l.X /.
Solution. Let l.X / D and take any open cover U of the space Y . Given any
point x 2 X , the set p.x/Sis compact and hence there is a finite family Ux U
such that p.x/ Ux D Ux . The mapping p being upper semicontinuous, the
set Vx D fx 0 2 X W p.x 0 / Ux g is open in X and x 2 Vx . The open cover
fVx W x 2 X g of the space XShas a subcover of cardinality
S , so there is A X
such that jAj and X D fVx W x 2 Ag. If U 0 D fUx W x 2 Ag, then U 0 U
and jU 0 j . Pick any y 2 Y and take x
S2 X with y 2 p.x/. There isS
a 2 A such
that x 2 Va and hence p.x/ Ua U 0 . Therefore y 2 p.x/ U 0 which
0
shows that U is a subcover of U of cardinality . Thus l.Y / D l.X /.
that if X is compact, then so is Y .
Solution. Take an open cover U of the space Y . Given any x 2 X , the Sset p.x/
is compact and hence there is a finite Ux U such that p.x/ Ux D Ux . The
mapping p being upper semicontinuous, the set Vx D fx 0 2 X W p.x 0 / Ux g
is open in X and x 2 Vx . The open cover fVx W x 2 X g ofSthe space X has a
finite subcover,
S so there is a finite A X such that X D fVx W x 2 Ag. If
U 0 D fUx W x 2 Ag, then U 0 U and jU 0 j < !. Pick any y 2 Y and take xS2 X
with y 2 p.x/. There is S a 2 A such that x 2 Va and hence p.x/ Ua U 0 .
Therefore y 2 p.x/ U 0 which shows that U 0 is a finite subcover of U. We
proved that every open cover of Y contains a finite subcover of Y , i.e., Y is compact.
that if X is a Lindelf -space, then so is Y .
Solution. Fix a compact cover C of the space X such that there S is a countable
network F with respect to C. For any A X , we let p.A/ D fp.x/ W x 2 Ag;
if B Y , then p 1 .B/ D fx 2 X W p.x/ Bg. Given C 2 C, let C 0 D p.C /;
we claim that the map q D pjC W C ! C 0 is upper semicontinuous. Indeed, if
U 2 .C 0 /, then take any V 2 .Y / with V \ C 0 D U and observe that q 1 .U / D
p 1 .U / \ C is an open subset of C . It is clear that q is compact-valued and onto, so
C 0 is compact by Problem 241. Therefore D D fp.C / W C 2 Cg is a compact cover
of Y .
Let G D fp.F / W F 2 F g; if D 2 D and U 2 .D; Y /, then there is C 2 C
such that D D p.C / and therefore C V D p 1 .U /. The family F being a
network with respect to C there is F 2 F such that C F V . It is immediate
that G D p.F / 2 G and D G U which shows that G is a countable network
with respect to the compact cover D of the space Y . Thus Y is a Lindelf -space.
T.243. Prove that
(i) any continuous image of a Lindelf -space is a Lindelf -space;
(ii) any perfect preimage of a Lindelf -space is a Lindelf -space.
Solution. The statement of (i) was proved in Fact 1 of T.237. To prove (ii) assume
that f W X ! Y is a perfect map and Y is a Lindelf -space. Let p.y/ D f 1 .y/
for every y 2 Y . Then S
p W Y ! X is a compact-valued map because f is perfect.
Furthermore, p.Y / D fp.y/ W y 2 Y g D f 1 .Y / D X and hence p is an onto
map. Given an open U X observe that
p 1 .U / D fy 2 Y W p.y/ U g D fy 2 Y W f 1 .y/ U g D Y nf .X nU /
is an open subset of Y because f is a closed map. Thus p is upper semicontinuous

and therefore X is a Lindelf -space by Problem 242 so (ii) is proved.
T.244. Prove that w.X / D nw.X / D i w.X / for any Lindelf p-space X . In
particular, any Lindelf p-space with a countable network has a countable base.
Solution. Given a set Z, a family A exp.Z/ and P Z, recall that St.P; A/ D
S
fA 2 A W A \ P ;g. If z 2 Z, then A.z/ D fA 2 A W z 2 Ag; we write
St.z; A/ instead of St.fzg; A/. Given any Z 0 Z, let AjZ 0 D fA \ Z 0 W A 2 Ag;
the family A is locally finite in Z 0 at a point z 2 Z 0 if there is U 2 .z; Z 0 / such
that U intersects only finitely many elements of A.
If Z is a space and we are given families A; B exp.Z/, say that A is (strongly)
inscribed in B if for any A 2 A there is B 2 B such that A B (or A B
respectively). Say that A is (strongly) barycentrically inscribed in B if the family
fSt.z; A/ W z 2 Zg is (strongly) inscribed in B.
Suppose that K is a compact space and Z is dense in K. If Un .K/ for
every n 2 !, call the sequenceS fUn W n 2 !g a feathering T of the space Z in its
compactification K if Z Un for each n 2 ! and fSt.z; Un / W n 2 !g Z
for every z 2 Z. A space Z is a p-space if and only if it has a feathering in some
compactification of Z (see Fact 1 of T.224). It is easy to see that
(1) if fUn W n 2 !g a feathering of a space Z in Z and S we are given a sequence
S 0 D fUn0 W n 2 !g such that Un0 .Z/; Z Un0 and Un0 is inscribed in
Un for each n 2 !, then S 0 is also a feathering of Z in Z.
S
Fact 1. Suppose that Z is a paracompact space. If SU .Z/ and Z U, then
S V .Z/ such that Z V, the family V is locally finite
there exists a family
in Z at each z 2 V and V is strongly barycentrically inscribed in U.
Proof. For any W 2 .Z/, choose a set O.W / 2 .Z/ such that O.W / \ Z D
W . Let U 0 D fU 0 2 .Z/ W U 0 S U for some U 2 Ug (the bar denotes the
closure in Z). It is clear that Z U 0 and hence there is a family W .Z/
such that W is a barycentric refinement of U 0 jZ (see Problem 230 of [TFS]). Use
paracompactness of Z again to find a locally finite (in Z) refinement W 0 of the
cover W. Observe that the family V 0 D fO.W / W W 2 W 0 g is locally finite in Z
at all points of Z. Indeed, given z 2 Z, there is U 2 .z; Z/ such that the family
A D fW 2 W 0 W U \ W ;g is finite. It is an immediate consequence of density
of Z in Z that fW 2 W 0 W O.U / \ O.W / ;g D A, i.e., the set O.U / witnesses
that V 0 is locally finite at z in Z. It is clear that the set G D fz 2 Z W V 0 is locally
finite at z in Zg is open in ZSand Z G. Let V D fV \ G W V 2 V 0 g; it is

evident that V .Z/; S Z V and V is locally finite in Z at all points of G
and hence at all points of V.
To prove that V is strongly barycentrically inscribed in U, take any z 2S Z; the
case when St.z; V/ D ; is trivial, so assume that St.z; V/ ;. Then z 2 V and
therefore V is locallyTfinite at z in Z. In particular, the family V.z/ is finite and
hence the set H D V.z/ is open in Z and non-empty; take any t 2 H \ Z.
It turns out that every element of V which contains z also contains t, so St.z; V/
St.t; V/. For every V 2 V.t/ we have WV D V \ Z 2 W 0 . By definition of W 0 ,
there is GV 2 W such that WV GV . The family W being a barycentric refinement
of U 0 , we have St.t; W/ U 0 for some U 0 2 U 0 ; by definition of U 0 there is U 2 U
such that U 0 U . Consequently,
[ [
St.z; V/ St.t; V/ D fV W V 2 V.t/g fW V W V 2 V.t/g
[
fG V W V 2 V.t/g St.t; W/ U 0 U
and hence V is strongly barycentrically inscribed in U, i.e., Fact 1 is proved. t

u
Returning to our solution, observe that we have i w.Z/ nw.Z/ w.Z/ for
any space Z, so it suffices to show that w.X / D i w.X /. The space X is
paracompact being Lindelf, so we can apply (1) and Fact 1 to fix a feathering
fUn W n 2 !g of SX in X such that each Un is countable and locally finite in X
at all points of Un and UnC1 is strongly barycentrically inscribed in Un for every
n 2 !.
There exists a condensation f W X ! Y such that w.Y / ; let B be a base
in Y such that any finite intersection of elements of B belongs to B and jBj . It
follows from the fact that B is a base in Y that the family C D ff 1 .U / W U 2 Bg
has the following property:
T
(2) fU W U 2 C.x/g D fxg for any x 2 X (the bar denotes the closure in X ).
S
Since the family U of all finite unions of elements of fUn W n 2 !g is countable,
the family E of all finite intersections of elements of C [ .UjX / has cardinality ;
we claim that E is a base in X . To prove it fix any x 2 X and O 2 .x; X/. For
every n 2 !, let Un D St.x; Un /; the family Un .x/ is finite, so it follows from the
fact that UT
nC1 is strongly inscribed
T in Un that clX .UnC1 / Un for all n 2 !. The
set K D fUn W n 2 !g D fclX .Un / W n 2 !g is contained in X and compact
because fUn W n 2 !g is a feathering of X in X . It is an easy consequence of Fact 1
of S.326 that the family fUn W n 2 !g is an outer base of K in X and therefore
(3) the family fUn \ X W n 2 !g is an outer base of K in X .
T
Apply Fact 1 of S.326 to the family F D fU \ K W U 2 C.x/g; since F D
fxg, there is U 2 C.x/ such that U \ K O. Thus the set U nO is closed in
X and disjoint from K. The property (3) implies that there is n 2 ! such that
.Un \ X / \ .U nO/ D ;. Therefore V D .Un \ X / \ U D Un \ U 2 E and
x 2 V O which proves that E is a base in X . Thus w.X / i w.X / whence

w.X / D nw.X / D i w.X / and therefore our solution is complete.
T.245. Prove that any perfect image and any perfect preimage of a Lindelf p-space
is a Lindelf p-space. Give an example of a closed continuous onto map f W X ! Y
such that X is a Lindelf p-space and Y is not a p-space.
Solution. Given a set Z, a family A exp.Z/ and P Z, recall that St.P; A/ D
S
fA 2 A W A \ P ;g. If z 2 Z, then A.z/ D fA 2 A W z 2 Ag; we write
St.z; A/ instead of St.fzg; A/. Given any Z 0 Z, let AjZ 0 D fA \ Z 0 W A 2 Ag;
the family A is locally finite in Z 0 at a point z 2 Z 0 if there is U 2 .z; Z 0 / such
that U intersects only finitely many elements of A.
If Z is a space and we are given families A; B exp.Z/, say that A is (strongly)
inscribed in B if for any A 2 A there is B 2 B such that A B (or A B
respectively). Say that A is (strongly) barycentrically inscribed in B if the family
fSt.z; A/ W z 2 Zg is (strongly) inscribed in B.
every n 2 !, call the sequenceS fUn W n 2 !g a feathering T of the space Z in its
compactification K if Z Un for each n 2 ! and fSt.z; Un / W n 2 !g Z
for every z 2 Z. A space Z is a p-space if and only if it has a feathering in some
compactification of Z (see Fact 1 of T.224). If f W Z ! T is a closed (and hence
continuous and onto) map between the spaces Z and T , then f # .U / D T nf .ZnU /
for any U Z. It is easy to see that f # .U / D ft 2 T W f 1 .t/ U g for any
U Z and the set f # .U / is open in T (maybe empty) for any open U Z.
Fact 1. Let Z be a paracompact p-space; if f W Z ! T is a perfect map (recall
that our definition of a perfect map implies that f is continuous and onto), then T
is also a paracompact p-space. In other words, a perfect image of a paracompact
p-space is a paracompact p-space.
Proof. The space T is paracompact by Fact 4 of S.226. There exists a continuous
map g W Z ! T such that gjZ D f (see Problem 258 of [TFS]). Furthermore,
g is onto and g.ZnZ/ D T nT (see Fact 3 of S.261).
Using paracompactness of Z, we will improve its feathering in Z which exists
because Z is a p-space. It is easy to see that
(1) if fUn W n 2 !g a feathering of the space Z in Z andS we are given a sequence
S 0 D fUn0 W n 2 !g such that Un0 .Z/; Z Un0 and Un0 is inscribed in
Un for each n 2 !, then S 0 is also a feathering of Z in Z.
We also have the following property:
S
(2) if U
S.Z/ and Z U, then there exists a family S V .Z/ such that
Z V, the family V is locally finite in Z at each z 2 V and V is strongly
barycentrically inscribed in U
which is an immediate consequence of Fact 1 of T.244.
Now apply properties (1) and (2) to choose a feathering U D fU Sn W n 2 !g of

Z in Z such that each Un is locally finite in Z at all points of Un and UnC1
T barycentrically inscribed in Un for all n 2 !. For any A Z, let
is strongly
UA D fSt.A; Un / W n 2 !g; if A D fzg for some z 2 Z, then we write Uz
instead of Ufzg . Observe that
(3) Uz ZnZ for any z 2 ZnZ,
because if y 2 Uz \ Z, then there is Un 2 Un such that fy; zg Un for all n 2 !
which implies that fy; zg Uy Z (the last inclusion holds because U is a
feathering of Z in Z). This contradiction proves (3).
Furthermore,
(4) if K Z is compact, then UK Z and the family fSt.K; Un / W n 2 !g is an
outer base of UK in Z.
For any n 2 !, the family Un .K/ D fU 2 Un W U \ K ;g is finite being Un jK
a locally finite family in a compact space K. If n 2 ! and U 2 UnC1 , then there is
VU 2 Un such that U VU (the bar denotes the closure in Z). Since only finitely
many elements of UnC1 intersect K, we have
[ [
St.K; UnC1 / D fU W U 2 UnC1 .K/g fVU W U 2 UnC1 .K/g St.K; Un /
T
for each n 2 ! and therefore UK D fSt.K; Un / W n 2 !g. Now apply Fact 1
of S.326 to conclude that for any set W 2 .UK ; Z/, there is n 2 ! such that
St.K; Un / St.K; Un / W , and hence fSt.K; Un / W n 2 !g is an outer base
of UK in Z. To show that UK Z, assume that there is z 2 UK nZ. This
implies that for any n 2 !, there is Un 2 Un such that z 2 Un and Un \ K ;.
As a consequence, St.z; Un / \ K ; for every n 2 !. Since the family UnC1 is
strongly barycentrically inscribed in Un , we have St.z; UnC1 / St.z; Un / for each
n 2 ! and therefore fSt.z; Un / \ K W n 2 !g is a decreasing sequence of T non-empty
compact subsets of K which shows Tthat there exists a point y
T 2 K \ . fSt.z; Un / W
n 2 !g/ Z. However, we have fSt.z; Un / W n 2 !g D fSt.z; Un / W n 2 !g D
Uz ZnZ by (3); this contradiction finishes the proof of the property (4).
Let VnS D fSt.f 1 .t/; Un / W t 2 T g for all n 2 !. It is clear that Vn .Z/
and Z Vn for all n 2 !. We claim that
T
(5) VK D fSt.K; Vn / W n 2 !g Z for any compact K Z.
Indeed, assume that z 2 ZnZ for some point z 2 VK ; let t D g.z/.
There exists a sequence ftn W n 2 !g Y such that z 2 St.f 1 .tn /; Un / and
St.f 1 .tn /; Un / \ K ; for all n 2 !. Therefore we can find Vn ; Wn 2 Un such
that z 2 Wn ; Wn \ f 1 .tn / ;; Vn \ f 1 .tn / ; and Vn \ K ; for all
n 2 !. Choose any an 2 Wn \ f 1 .tn / and bn 2 Vn \ f 1 .tn / for each n 2 !. It
follows from (3) and (4) that F D UK and G D Uz are compact subsets of Z
and ZnZ respectively and hence g.F / \ g.G/ D ; because g.F / D f .F / T
and g.G/ T nT . Take disjoint U 2 .g.F /; T / and V 2 .g.G/; T /; then

U 0 D g 1 .U / and V 0 D g 1 .V / are disjoint open (in Z) neighborhoods of F and
G respectively. Observe that bn 2 Vn St.K; Un / for each n 2 ! which, together
with (4), implies that there is m 2 ! such that bn 2 St.K; Un / U 0 for all n m.
The family UnC1 is strongly barycentrically inscribed in Un for all n 2 !, so it
follows from Fact 1 of S.326 that the family fSt.z; Un / W n 2 !g is an outer base of
Uz in Z. Thus there is k 2 ! such that an 2 Wn St.z; Un / V 0 for all n k.
For n D m C k we have an 2 V 0 and bn 2 U 0 and hence tn D g.an / D g.bn / 2
U \ V which is a contradiction. The property (5) is proved.
Clearly, Wn D fg # .V / W V 2 Vn g is a family of open subsets of T for every
n 2 !. We claim that the sequence W D fWn W n 2 !g is a feathering of T in T .
To prove it observe first that g 1 .t/ D f 1 .t/ St.f 1 .t/; Un / and therefore
tS2 g # .St.f 1 .t/; Un // 2 Wn for any t 2 T and n 2 ! which shows that T
Wn for all n 2 !.
Now for an arbitrary t 2 T , let Vtn DSSt.f 1 .t/; Un / and Wtn D g # .Vtn /. Then
for any t 2 T we have St.t; Wn / D fWsn W t 2 Wsn g. For every s 2 T , if
t 2 Wsn , then f 1 .t/ g 1 .WT 1
s / Vs . Thus g .St.t; Wn // St.f
n n 1
.t/; Vn /
1 1
for any n 2 !.T Therefore g . fSt.t; Wn / W n 2 !g/ Vf .t/ Z by (5).
Consequently, fSt.t; Wn / W n 2 !g T , i.e., the sequence fWn W n 2 !g is
indeed a feathering of T in T , so T is a p-space and Fact 1 is proved. t
u
Fact 2. Let Z be a normal space and assume that F is a non-empty closed subset of
Z; for any A Z, let A D .AnF / [ fF g. Given z 2 Z, let pF .z/ D z if z 2 ZnF
and pF .z/ D F if z 2 F . It is clear that pF W Z ! ZF D fF g [ .ZnF /. Then
(i) the family F D fU 2 .Z/ W U ZnF g [ fU W U 2 .F; Z/g is a topology
on the set ZF ;
(ii) the space Z=F D .ZF ; F / is T1 and normal (and hence Tychonoff) and the
map pF W Z ! Z=F is continuous, closed and onto.
The operation of obtaining the space Z=F from a space Z is called collapsing
the set F to a point.
Proof. (i) Since ; 2 .Z/ and ; ZnF , we have ; 2 F . Since ZF D Z , we
have ZF 2 F , so the first axiom of topology is satisfied. Let U D fU 2 .Z/ W
U ZnF g and V D fU W U 2 .F; Z/g. It is immediate that U \ U 0 2 U
for any U; U 0 2 U and V \ V 0 2 V for any V; V 0 2 V. Now, if U 2 U and
V 2 V, then U \ V 2 U which shows that the intersection of any two elements
of F is again in F . S S
As to the third axiom of topology, observe first that U 0 2 U and V 0 2 V
for any U 0 U and V 0 V. Now, S if0 W F , then let
0
TU 0 D W \ U and
0
V D W \ V. We have U D
S U 2 U and V D V 2 V. Besides,
W D U [ V 2 V which shows that the union of any subfamily of F
belongs to F , i.e., F is a topology on ZF and hence (i) is proved.
(ii) The surjectivity of pF is obvious; if U 2 U, then pF1 .U / D U is an open

subset of Z; if V 2 V, then V D U for some U 2 .F; Z/ and therefore
pF1 .V / D U is again an open subset of Z. Since F D U [ V, this shows that
the map pF is continuous. To prove that pF is closed, take any closed G Z. If
G \ F D ;, then pF .G/ D G is a closed subset of Z=F because .Z=F /nG D
.ZnG/ is an open subset of Z=F . If G \ F ;, then pF .G/ D G [ fF g is
closed in Z=F because .Z=F /n.G [ fF g/ D Zn.F [ G/ 2 U is also an open
subset of Z=F . Thus the map pF is closed.
It is straightforward to check that Z=F is a T1 -space; let us prove that it is normal.
Take any disjoint closed sets H; G Z=F . Then H 0 D pF1 .H / and G 0 D pF1 .G/
are disjoint closed subsets of Z. By normality of the space Z, there are U 0 ; V 0 2
.Z/ such that G 0 U 0 ; H 0 V 0 and U 0 \ V 0 D ;. It is immediate that U D
pF# .U 0 / 2 F ; V D pF# .V 0 / 2 F and U \ V D ;, so Z=F is a T4 -space and Fact 2
is proved. t
u
Returning to our solution, assume that Z is a Lindelf p-space and T is a
perfect image of Z. Then T is a p-space by Fact 1 because any Lindelf space
is paracompact. Since any continuous image of a Lindelf space is Lindelf, the
space T is a Lindelf p-space and hence any perfect image of a Lindelf p-space is
a Lindelf p-space.
Now assume that f W Z ! T is a perfect map and T is a Lindelf p-space. Then
Z is a Lindelf -space by Problem 243. Let fVn W n 2 !g be a sequence of open
(in T ) covers of T which witnesses that T is a p-space. There is a continuous
map g W Z ! T such that gjZ D f (see Problem 258 of [TFS]); besides,
g.ZnZ/ T nT (see Fact 3 of S.261) and therefore
(6) for any A T , we have A T if and only if g1 .A/ Z.
Let Un D fg 1 .U / W U 2 Vn g for all n 2 !. It is clear that fUn W n 2 !g is a
sequence of open (in Z) covers of Z. Furthermore, St.z; Un / D g 1 .St.g.z/; Vn //
for any n 2 ! and z 2 Z. This implies, together with the property (6), that
T
fSt.z; Un / W n 2 !g Z for any z 2 Z, i.e., Z is a p-space. Thus we have
established that any perfect preimage of a Lindelf p-space is a Lindelf p-space.
To construct the promised example, let X D R and F D ! R. The space X
is Lindelf p by Problem 221; let Y D X=F and f D pF . Then f is a closed
map by Fact 2. It is clear that Y has a countable network, so if Y is a p-space, then
w.Y / D ! and, in particular, .F; Y / D ! (according to the context, the set F is
considered either as a point of Y or a subset of X ). Let W be a countable local base
at the point F in Y . Given any U 2 .F; X/ the set V D f # .U / is open in Y and
F 2 V . Therefore there is W 2 W such that F 2 W V . As a consequence, F
f 1 .W / f 1 .V / U which shows that the family V D ff 1 .W / W W 2 Wg
is a countable outer base of the set F in X . Let fVn W n 2 !g be an enumeration of
V. For each n 2 !, the set Vn is an open neighborhood of n in R and hence there is
"n 2 .0; 13 / such that .n "n ; n C "n / Vn ; choose any rn 2 .n "n ; n C "n /. The
set W D Rnfrn W n 2 !g is open in R because frn W n 2 !g is closed and discrete.
Since F W , there is n 2 ! such that F Vn W and hence rn Vn which
is a contradiction. Thus .F; R/ > ! and hence Y is not first countable. Therefore
nw.Y / D ! < w.Y / which shows that Y is not a p-space (see Problem 244) and
makes our solution complete.
T.246. Suppose that Cp .X / is a closed continuous image of a Lindelf p-space.
Solution. Given spaces Y and Z call a continuous onto map h W Y ! Z irreducible
if, for any closed F Y with F Y , we have h.F / Z. For any U Y , let
h# .U / D Znh.Y nU /. It is easy to see that h# .U / D fz 2 Z W h1 .z/ U g; if the
map h is closed, then h# .U / is open (maybe empty) for any U 2 .Y /. Another easy
observation is that a closed map h W Y ! Z is irreducible if and only if h# .U / ;
for any U 2 .Y /. A space Y is of pointwise countable type if for every y 2 Y
there is a compact P Y such that y 2 P and .P; Y / D !.
Let Y be a space; given any points y1 ; : : : ; yn 2 Y and sets O1 ; : : : ; On 2
.R/, the set y1 ; : : : ; yn I O1 ; : : : ; On D ff 2 Cp .Y / W f .yi / 2 Oi for
all i ng is called a standard open subset of Cp .Y /. Standard open sets
y1 ; : : : ; yn I O1 ; : : : ; On where n 2 N; y1 ; : : : ; yn 2 Y and O1 ; : : : ; On 2
.R/ form a base in the space Cp .Y / (see Problem 056 of [TFS]). If U D
y1 ; : : : ; yn I O1 ; : : : ; On is a standard open subset of Cp .Y /, then supp.U / D
fy1 ; : : : ; yn g.
Fact 1. Let Y be a paracompact space. Suppose that Z is a space in which any point
is a limit of a nontrivial convergent sequence. Then any closed map h W Y ! Z is
irreducible on some closed subset of Y , i.e., there is a closed F Y such that
h.F / D Z and hF D hjF is irreducible.
Proof. For every y 2 Z fix a sequence Sy D fyn W n 2 !g Znfyg converging to
S
y. We will prove first that the set Py D h1 .y/ \ fh1 .yn / W n 2 !g is compact
for every y 2 Z. Indeed, if for some y 2 Z the set Py is not compact, then it is not
countably compact being closed in Y and hence paracompact (it is an easy exercise
that any countably compact paracompact space is compact). Therefore there is a
countably infinite closed discrete set D D fxn W n 2 !g Py . Since Y is
paracompact, it is collectionwise normal and hence we can find a discrete family
D fUn W n 2 !g .Y / with xn 2 Un \ Py for all n 2 !.

If ASis an arbitrary finite subset of !, then for each natural number n, we have
1
Un \ . fhS .zk / W k 2 !nAg/ ;. This makes it possible to choose a point
zn 2 Un \ . fh1 .zk / W k 2 !g/ in such a way that h.zm / h.zn / if n m.
The family
being discrete the set D D fzn W n 2 !g is closed and discrete in
Y . The set h.D/ is also closed because h is a closed map. Note that h.D/ has also
to be discrete because h.C / is closed for any C D. However h.D/ is a nontrivial
sequence converging to y, a contradiction with the fact that h.D/ is closed and
discrete. This proves Py is compact for all y 2 Z.
Claim. Suppose that H is a closed subset of Y such that h.H / D Z. Then H \
Py ; for all y 2 Z.
Proof of the claim. Fix any y 2 Z; it follows from h.H / D Z that it possible to
choose a point tn 2 H \ h1 .yn / for all n 2 !. The map h is closed and therefore
ftn W n 2 !g \ h1 .y/ ;. But H ftn W n 2 !g and ftn W n 2 !g \ h1 .y/ Py .
Thus H \ Py ; and the claim is proved. t
u
Suppose that we have a family F of closed subsets of Y such T that F is totally
ordered by inclusion and h.H / D Z for every H 2 F . Then h. F / D Z. Indeed,
HT\ Py ; for anyTy 2 Z and H 2 F . We proved that the set P Ty is compact, so
. F /\h1 .y/ . F /\P T y ; for all y 2 Z; consequently,
T . F /\h 1
.y/
; which implies y 2 h. F / for every y 2 Z, i.e., h. F / D Z. Finally, use
Zorns lemma to find a closed F Y which is maximal (with respect to the inverse
inclusion) in the family of all closed sets H Y such that h.H / D Z. It is evident
that hF is irreducible so Fact 1 is proved. t
u
Returning to our solution, suppose that Y is a Lindelf p-space for which there is
a closed continuous onto map ' W Y ! Cp .X /. Given f 2 Cp .X / observe that the
sequence ff C n1 g is nontrivial and converges to f . Therefore Fact 1 is applicable to
the map ' W Y ! Cp .X / to obtain a closed F Y such that '.F / D Cp .X / and
'jF is irreducible. The space F is also Lindelf p by Problem 224 which shows that
Cp .X / is a closed irreducible image of a Lindelf p-space. Thus we can assume,
without loss of generality, that the map ' is irreducible. The space Y is of pointwise
countable type by Fact 1 of T.222, so there is a non-empty compact P Y such
that .P; Y / D !. Fix a decreasing outer base fUn W n 2 !g of the set P in Y . The
space K D '.P / Cp .X / is compact and the family f' # .Un / W n 2 !g consists of
non-empty open subsets of Cp .X / with the following property:
(1) for any W 2 .K; Cp .X // there is m 2 ! such that Vn D ' # .Un / W for
every n m.
To see that (1) holds observe that Um ' 1 .W / for some m 2 ! because the
family fUn W n 2 !g is an outer base of P in Y and P ' 1 .W /. Therefore
Un Um ' 1 .W / and hence Vn W for all n m. Observe that if Vn0 Vn
for all n 2 !, then (1) still holds for the sequence fVn0 W n 2 !g. Therefore there
exists a family O D fOn W n 2 !g of standard non-empty open subsets of Cp .X /
Sfor O.
such that (1) holds
The set A D fsupp.On / W n 2 !g is countable; to prove that X D A assume
that z 2 X nA. The map pz W Cp .X / ! R defined by pz .f / D f .z/ for all f 2
Cp .X / is continuous (see Problem 166 of [TFS]) and therefore Q D pz .K/ is a
compact and hence bounded subspace of R. Take any bounded H 2 .R/ such that
Q H . The set W D z; H is open in Cp .X / and contains K, so there is n 2 !
for which On W . Take any f 2 On ; we have g 2 On for any g 2 Cp .X / with
gjsupp.On / D f jsupp.On /. Since supp.On / A, we have z B D supp.On / and
hence there is g 2 Cp .X / such that gjB D f jB and g.z/ H (see Problem 034 of
[TFS]). As a consequence, g 2 On nW which is a contradiction. Therefore X D A
is countable and our solution is complete.
T.247. Show that an open continuous image of a p-space is not necessarily a p-

space. Supposing that Cp .X / is an open continuous image of a p-space, prove that
X is countable (and hence Cp .X / is a p-space).
Solution. Let S be the Sorgenfrey line (see Problem 165 of [TFS]). Then S is a
Lindelf space with i w.S / D ! < nw.S /. Thus S is not a p-space by Problem 244.
The space S is first countable, so there is a metrizable space M and an open
continuous onto map ' W M ! S (see Problem 223 of [TFS]). The space M is
a p-space (see Problem 221) which proves that S is an open image of a p-space
which fails to be a p-space.
Now assume that Cp .X / is an open continuous image of a p-space Y . The space
Y is of pointwise countable type, i.e., for every y 2 Y , there is a compact P Y
such that y 2 P and .P; Y / D ! (see Fact 1 of T.222). It is an easy exercise
that any open image of a space of pointwise countable type is a space of pointwise
countable type. Thus Cp .X / is of pointwise countable type and therefore there is
a non-empty compact P Cp .X / such that .P; Cp .X // D !. Finally, apply
Problem 170 of [TFS] to conclude that X is countable.
T.248. Prove that X is a Lindelf -space if and only if there exists a second
countable space M and a compact K such that X is a continuous image of a closed
subspace of K M .
Solution. Suppose that there exists a compact space K and a second countable
space M such that some closed F K M maps continuously onto X . Observe
first that K and M are Lindelf -spaces by Problems 226 and 221. Therefore
K M is also a Lindelf -space by Fact 1 of T.227 and hence so is the space F
by Problem 224. Since every continuous image of a Lindelf -space is Lindelf
by Problem 243, the space X is Lindelf as well. This proves sufficiency.
Now assume that X is a Lindelf -space and fix a family F D fFn W n 2
!g of compact subsets of X which separates TX from X nX (this is possible by
Problem 233). Let M D fs 2 ! ! W Ps D fFs.n/ W n 2 !g X g; it is clear
that M is a second countable space. For K D X consider the set F D f.x; s/ 2
K M W x 2 Ps g; it is clear that F K M . If z D .x; t/ 2 .K M /nF , then
x Pt and hence there is n 2 ! such that x Ft .n/ . Then V D KnFt .n/ 2 .x; K/
and W D fs 2 M W s.n/ D t.n/g is an open subset of M with t 2 W . Thus
U D V W is an open neighborhood of z; if .y; s/ 2 U , then y 2 V D KnFt .n/
while Fs.n/ D Ft .n/ which shows that y Fs.n/ and therefore y Ps which implies
.y; s/ F . This proves that every z 2 .K M /nF has a neighborhood U with
U \ F D ;, i.e., F is a closed subspace of K M .
Let p W K M ! K be the natural projection, i.e., p.z/ D x for any point
z D .x; s/ 2 K M . Then f D pjF W F ! X is a continuous map; for any z D
.x; s/ 2 F , we have x 2 Ps X by the definition of M and hence x D p.z/ 2 X
whence f .F / TD p.F / X . Now, given any x 2 X , let A D fn 2 ! W x 2 Fn g;
then x 2 P D fFn W n 2 Ag X because the family F separates X from X nX .
Choose any s 2 ! ! such that s.!/ D A; then Ps D P and therefore z D .x; s/ 2 F
while f .z/ D p.z/ D x. This shows that f .F / D X and hence X is a continuous

image of F . We settled necessity so our solution is complete.
(i) there exists a second countable space M and a space Y such that Y maps
perfectly onto M and continuously onto X ;
(ii) there exists an upper semicontinuous compact-valued onto map ' W M ! X
for some second countable space M ;
(iii) X is a Lindelf -space.
Solution. Given spaces Z; T and a map h W Z ! T , let h# .U / D T nh.ZnU / for
any U Z. It is easy to see that h# .U / D ft 2 T W h1 .t/ U g; if the map h is
closed, then h# .U / is open (maybe empty) for any U 2 .Z/.
(i)H)(ii). Assume that w.M / D ! while, for some space Y , there exists a
perfect map f W Y ! M and a continuous onto map g W Y ! X . Given any
s 2 M , let '.s/ D g.f 1 .s//. It is obvious that ' W M ! X is a compact-valued
map. If x 2 X , then g.y/ D x for some y 2 Y and hence x 2 g.f 1 .f .y///, i.e.,
x 2 '.s/ for s D f .y/ which shows that the map ' is onto. To see that ' is upper
semicontinuous, take any U 2 .X/. Then
' 1 .U / D fs 2 M W '.s/ U g D fs 2 M W f 1 .s/ g 1 .U /g D f # .g 1 .U //
is an open subset of M and therefore ' W M ! X is an upper semicontinuous

compact-valued onto map.
(ii)H)(iii). Assume that there is a second countable space M and an upper
semicontinuous compact-valued onto map ' W M ! X . The space M is Lindelf
by Problem 221, so we can apply Problem 242 to conclude that X is a Lindelf
-space.
(iii)H)(i). If X is a Lindelf -space, then there exists a compact space K
and a second countable space N such that X is a continuous image of some closed
subspace Y of the space K N (see Problem 248). Let p W K N ! N be the
natural projection; then p is a perfect map by Fact 3 of S.288. If M D p.Y /, then
f D pjY W Y ! M is a perfect map because Y is closed in K N . Thus Y
maps perfectly onto a second countable space M and continuously onto X . This
completes the proof of (iii)H)(i) so our solution is complete.
T.250. Give an example of a space X which embeds into Cp .Y / for some Lindelf
p-space Y and is not embeddable into Cp .Z/ for any K -space Z.
Solution. For any n 2 N, let Mn D f1; : : : ; ng. A space Z is K-analytic if it is a
continuous image of a K -space. The Cantor set K is the space f0; 1g! . Call a space
Z uniformly uncountable if every non-empty open subset of Z is uncountable. A set
F Cp .Z/ is called D-separating if for any closed P Z and finite K Z with
K \ P D ;, if " > 0, then there exists a function f 2 F such that f .K/ . "; "/
and f .P / 34 ; 1 . The function uZ 2 RZ is defined by uZ .z/ D 0 for all z 2 Z.
Given a space Z and a point p 2 Z, let D.Z; p/ D ff 2 IZ W f .p/ D 0 and

f .U / 12 ; 12 for some U 2 .p; Z/g. We consider D.Z; p/ to be a space with
the topology induced from IZ .
If C is a class of spaces, then Y 2 C says that a space Y belongs to C; furthermore,
Y 2 ImC if Y is a continuous image of a space from C. The statement Y 2 UnC
says that Y is a countable union of spaces from C and the expression Y 2 ClC is
a short way of saying that Y is a closed subspace of a space from C. Besides, we
write Y 2 ProdC if Y is a finite product of spaces from C.
We will stick to the usual agreements of set theory in what concerns operations
on classes of spaces. In particular, given classes C and D, we write C D if every
spaceS from C belongs to D; also, if Ct is a class of spaces for all t 2 T , then the
class fCt W t 2 T g consists of spaces Y such that Y 2 Ct for some t 2 T .
Let Z be a space; then a space Y belongs to the class K0 .Z/ if and only if Y D Z
or Y is compact. Assume that < !1 and we have S defined a class K .Z/ for all
< . If is a limit ordinal, then let K .Z/ D fK .Z/ W < g; if D C 1
for some < !1 , then let K .Z/ D ImK .Z/ [ ClK .Z/ [ UnK S .Z/ [
ProdK .Z/ . Once we have defined K .Z/ for all < !1 , let K.Z/ D fK .Z/ W
< !1 g.
Fact 1. Let CS be the class of compact spaces. For any space Z, the class K.Z/ is
the minimal class of spaces which contains fZg [ CS and is invariant under finite
products, countable unions, closed subspaces and continuous images. To put it more
rigorously, if C is a class of spaces such that
fZg [ CS C and UnC D ClC D ProdC D ImC D C;
then K.Z/ C.
Proof. If Y 2 K.Z/ and T is a continuous image of Y , then Y 2 K .Z/ for some
< !1 and therefore T 2 KC1 .Z/. If Y 2 K.Z/ and T is a closed subspace of
Y , then Y 2 K .Z/ for some < !1 and therefore T 2 KC1 .Z/. If Y1 ; : : : ; Yn 2
K.Z/ and T D Y1

Yn , then for any i 2 Mn , we have Yi 2 Ki .Z/ for some
i < !1 . If D maxfi W i 2 Mn g, then Yi 2 K .Z/ for all i 2 Mn and hence
T 2 KC1 .Z/. t
u
S
Proof. Now, if fYi W i 2 !g K.Z/ and T D fYi W i 2 !g, then for any i 2 !,
we have Yi 2 Ki .Z/ for some i < !1 . If > supfi W i 2 !g, then Yi 2 K .Z/
for all i 2 ! and hence T 2 KC1 .Z/. This proves that F K.Z/ D K.Z/ for
all F 2 fIm; Cl; Un; Prodg. Finally, if C is a class such that fZg [ CS C and
UnC D ClC D ProdC D ImC D C, then by an evident induction, K .Z/ C
for all < !1 and therefore K.Z/ C. Fact 1 is proved. t
u
L
Fact 2. If Zi is a K -space for all i 2 !, then Z D fZi W i 2 !g is also a
K -space.
T
Proof. For each i 2 !, there is a space Yi such that Zi Yi and Zi D fYni W
n 2 !g where Yni is a -compact subspace of Yi for all n 2 !. It is evident that
L S
Z is a subspace of Y D fYi W i 2 !g. Furthermore,
T Yn D fYni W i 2 !g is
a -compact subspace of Y for every n 2 ! and fYn W n 2 !g D Z, so Z is a
K -space and Fact 2 is proved. t
u
Fact 3.
(i) every closed subspace of a K-analytic space is K-analytic;
(ii) any countable product of K-analytic spaces is a K-analytic space;
(iii) any continuous image of a K-analytic spaces is a K-analytic space;
(iv) any countable union of K-analytic spaces is a K-analytic space.
Proof. (i) If Z is K-analytic and P is a closed subspace of Z, take any K -space
Y such that there is a continuous onto map f W Y ! Z. The set Y 0 D f 1 .P /
is a K -space being closed in Y (see Problem 338 of [TFS]) and f jY 0 maps
Y 0 continuously onto P . Thus P is K-analytic and (i)Q is proved.
(ii) If Zi is a K-analytic space for all i 2 ! and Z D fZi W i 2 !g, take a
K -space
Q Yi and a continuous onto map fQ i W Yi ! Zi for all i 2 !. Then
f D ffi W i 2 !g maps the space Y D fYi W i 2 !g continuously onto
Z (see Fact 1 of S.271); since Y is a K -space by Problem 338 of [TFS], the
space Z is K-analytic and (ii) is proved.
(iii) Assume that Z is K-analytic and fix a K -space Y and a continuous onto map
f W Y ! Z. If g W Z ! T is a continuous onto map and then g f maps Y
continuously onto T , so T is also S
K-analytic and (iii) is proved.
(iv) Let Z be a space such that Z D fZi W i 2 !g where Zi is K-analytic for
all i 2 !. For everyL i 2 !, take a K -space Yi which maps continuously
onto Zi . Then Y D fYi W i 2 !g is a K -space (see Fact 2) which maps
continuously onto Z, so Z is K-analytic and Fact 3 is proved.
t
u
Fact 4. Let M be a second countable uncountable K-analytic space. Then the
Cantor set K embeds in M .
Proof. Fix a metric d on M with .d / D .M / and a continuous map ' W ZT! M
of some K -space Z onto M . Let Y Z be any space such that Z D fYn W
n 2 Ng and each Yn is a -compact subspace of Y . A set B Z will be called n-
precompact if clY .B/ is a compact subset of Yn . Given a set A M , the symbol A
denotes the closure of A in M . For each k 2 N, denote by Ck the set of all functions
from k D f0; : : : ; k 1g to f0; 1g. For every k 2 N, we will construct by induction
families fPf W f 2 Ck ; k 2 Ng exp.M / and fQf W f 2 Ck ; k 2 Ng exp.Z/
(1) Pf is uniformly uncountable and diam.Pf / k1 for any k 2 N and f 2 Ck ;
(2) Qf is k-precompact, closed in Z and '.Qf / D Pf for any f 2 Ck and k 2 N;
(3) the family fPf W f 2 Ck g is disjoint for any k 2 N;
(4) if m; k 2 N; m < k and f 2 Ck , then Qf jm Qf .
Since the set Y1 Z is a countable union of compact spaces, there is a compact

K1 Y1 such that the set '.K1 \ Z/ is uncountable; it is clear that Z1 D K1 \ Z
is closed in Z; let '1 D 'jZ1 W Z1 ! M1 D '.Z1 /. Take a uniformly uncountable
L1 M1 which is closed in M1 (this is possible by Fact 1 of S.343). Pick distinct
x0 ; x1 2 L1 and choose U 2 .x0 ; M /; V 2 .x1 ; M / such that U \ V D ;.
There is " 2 .0; 1=2/ such that B.x0 ; "/ U and B.x1 ; "/ V . We have C1 D
ff0 ; f1 g where fi .0/ D i for i 1; let Pf0 D clM1 .B.x0 ; "/ \ L1 / and Pf1 D
clM1 .B.x1 ; "/ \ L1 /. Observe that Pfi L1 because L1 is closed in M1 ; since Pfi
is closed in M1 , the set Qfi D '11 .Pfi / is closed in Z1 and hence in Z for all
i 2 f0; 1g.
Since the closure of any open set in a uniformly uncountable space is uniformly
uncountable, the space Pfi is uniformly uncountable for every i 2 f0; 1g. Since
Qfi K1 Y1 , every Qfi is 1-precompact. Furthermore, diam.Pfi / 1, because
Pfi B.xi ; "/ and diam.B.z; r// 2r for any z 2 M and r > 0. Therefore the
properties (1)(4) are satisfied for the families fPf W f 2 C1 g and fQf W f 2 C1 g.
Suppose that for each k n, we defined Pf for all f 2 Ck so that the properties
(1)(4) hold. Any function f 2 CnC1 is an extension of the function f jn and there
g g
are exactly two such extensions. This shows that CnC1 D ff0 ; f1 W g 2 Cn g where
g g
fi jn D g and fi .n/ D i for i D 0; 1.
Now, take an arbitrary function g 2 Cn ; observe that Qg is contained in YnC1
which is -compact. Therefore there is a compact KnC1 YnC1 such that the set
MnC1 D '.KnC1 \ Qg / Pg is uncountable. It is clear that ZnC1 D KnC1 \ Qg
is closed in Qg and hence in Z. Let 'nC1 D 'jZnC1 W ZnC1 ! MnC1 . Apply
Fact 1 of S.343 to find a uniformly uncountable LnC1 MnC1 which is closed in
MnC1 . The space LnC1 has no isolated points and hence we can take distinct points
x0 ; x1 2 LnC1 . Fix any U 2 .x0 ; M /; V 2 .x1 ; M / such that U \ V D ;. We
can find a number " 2 .0; 1=.2n C 2// such that B.x0 ; "/ U and B.x1 ; "/ V .
Let Pf g D clMnC1 .B.x0 ; "/ \ LnC1 / and Pf g D clMnC1 .B.x1 ; "/ \ LnC1 /. Observe
0 1
g g
that Pfi LnC1 because LnC1 is closed in MnC1 ; since Pfi is closed in MnC1 , the
g 1 g
set Qfi D 'nC1 .Pfi / is closed in ZnC1 and hence in Z for all i 2 f0; 1g. Since the
function g 2 Cn was taken arbitrarily, we indicated how to construct sets Pf g ; Pf g
0 1
and Qf g ; Qf g for all g 2 Cn . This gives the desired families fPf W f 2 CnC1 g and
0 1
fQf W f 2 CnC1 g.
Since the closure of any open set in a uniformly uncountable space is uniformly
g
uncountable, the space Pfi is uniformly uncountable for each i 2 f0; 1g. Applying
g
Fact 1 of S.236 to the sets Pfi B.xi ; "/ \ MnC1 , we conclude that
g 1
diam.Pfi /diam.B.xi ; "/ \ LnC1 /diam.B.xi ; "// D diam.B.xi ; "//2"<
nC1
so (1) is satisfied.
Since Qfig KnC1 YnC1 , every set Qfig has to be .n C 1/-precompact; the
rest of the statements of property (2) are, evidently, true by our construction. Now,
(3) has only to be checked for k D n C 1. Observe that if f; g 2 CnC1 ; f g and
f jn D gjn, then Pf \ Pg D ; by our construction. If we have f jn gjn, then

Pf \ Pg Pf jn \ Pgjn D ; by the induction hypothesis. Therefore (3) holds for
k D nC1 and all m n by the induction hypothesis. The property (4) is guaranteed
by our construction for m D n and k D n C 1. Thus our inductive construction can
be continued to give us families fPf W f 2 Cn ; n 2 Ng and fQf W f 2 Cn ; n 2 Ng
with the properties (1)(4). T
For each f 2 K, let Tf D fclY .Qf jn / W n 2 Ng and observe that the family
fclY .Qf jn / W n 2 Ng consists of non-empty decreasing compact sets by (2) and
(4) so Tf ;. Besides, (2) implies that there isTa compact set Kn Yn such that
clY .Qf jn / Kn Yn and consequently Tf fYn W n 2 Ng D Z. Since Tf
clY .Qf jn / forTevery n 2 N, we have Tf clY .Qf jn / \ Z D clZ .Qf jn / D Qf jn .
Thus Tf fQf jn W n 2 Ng; take any zf 2 Tf and let .f / D '.zf /. This
gives us a map W K ! MTsuch that .f / 2 '.Qf jn / D Pf jn for all n 2 N.
As a consequence,
T .f / 2 fPf jn W n 2 Ng; since diam.Pf jn / ! 0, we have
f.f /g D fPf jn W n 2 Ng.
The map is injective because if f g, then f jn gjn for some n 2 N;
consequently, .f / 2 Pf jn and .g/ 2 Pgjn . Since Pf jn \ Pgjn D ; by (3), we have
.f / .g/.
The map is continuous; to see this, take any f 2 K and " > 0. There exists
n 2 N such that 1=n < ". The set W D fg 2 K W gjn D f jng is open in K and
f 2 W . For any g 2 W we have .g/ 2 Pgjn D Pf jn ; since diam.Pf jn / 1=n,
we have d..g/; .f // 1=n < " and hence .W / B..f /; "/ which proves
continuity of at the point f . Thus W K ! C D .K/ is a condensation and
hence homeomorphism. This shows that K embeds in M so Fact 4 is proved. t
u
Fact 5. There is a subspace Y K which is not K-analytic.
Proof. It follows from Fact 5 of S.151 that there exist disjoint sets A; B K such
that A \ P ; B \ P for any uncountable compact P K. Observe that A has
to be uncountable. Indeed, if jAj !, then A0 D KnA is a G -subset of K and hence
A0 is K because every open subset of K is -compact. Therefore A0 is K-analytic
and uncountable, so there is K A0 homeomorphic to K by Fact 4. However,
according to the definition of A, it must intersect every uncountable compact subset
of K; this contradiction proves that A is uncountable. Now, if the space Y D A is K-
analytic, then there is P A homeomorphic to K by Fact 4. However B \ P ;
because P is uncountable, so A \ B P \ B ; which is a contradiction. Thus
Y is not K-analytic and Fact 5 is proved. t
u
Fact 6. For any space Z with nw.Z/ , there exists a space M such that w.M /
and M condenses onto Z.
Proof. Take any network A in Z with jAj . It is an easy consequence of the
the space Z that the family N D fN W N 2 Ag is also a network
regularity ofS
in Z. Since N D Z, the family N generates a topology on Z as a subbase.
Observe thatS.Z/ because for every U 2 .Z/, there is N 0 N
with U D N 0 so U 2 . Therefore the identity map i W .Z; / ! Z is a
condensation. Since .Z/ , every closed subspace of Z is also closed in .Z; /

which shows that all elements of N are clopen in M D .Z; /. The family N being
a network for the space Z, for any distinct x; y 2 Z, there are P; Q 2 N such that
x 2 P; y 2 Q and P \ Q D ;. Thus the space M is Hausdorff. The family B
of all finite intersections of the elements of N is a base in M and all elements of
B are clopen, i.e., M is zero-dimensional. Clearly, jBj and hence w.M / .
Every zero-dimensional Hausdorff space is Tychonoff (see Fact 1 of S.232), so M
is a Tychonoff space with w.Z/ which condenses onto Z. Fact 6 is proved. u t
Fact 7. If N is a space with a countable network, then there is a second countable
M such that N embeds in Cp .M /.
Proof. The space Cp .N / has a countable network by Problem 172 of [TFS], so there
is a second countable space M which can be mapped continuously onto Cp .N / (see
Fact 6). If ' W M ! Cp .N / is a continuous onto map, then the dual map ' embeds
Cp .Cp .N // in Cp .M / (see Problem 163 of [TFS]). Since N embeds in Cp .Cp .N //
by Problem 167 of [TFS], it also embeds in Cp .M / so Fact 7 is proved. t
u
Fact 8. Given a space Z suppose that T Z and p 2 T . Then the natural
projection W IZ ! IT of IZ onto its face IT maps D.Z; p/ onto D.T; p/ and
hence D.T; p/ is a continuous image of D.Z; p/.
Proof. Recall that .f / D f jT for any f 2 IZ . If f 2 D.Z; p/, then let g D
.f /; there is U 2 .p; Z/ such that f .U / 12 ; 12 . It is obvious that V D
U \ T 2 .p; T / and g.V / 12 ; 12 . Therefore .D.Z; p// D.T; p/. Now,
if g 2 D.T; p/, then there exists V 2 .p; T / such that g.V / 12 ; 12 . Choose
any U 2 .p; Z/ with U \ T D V and let f .z/ D g.z/ for all z 2 T ; if z 2 ZnT ,
then let f .z/ D 0. It is immediate that .f / D g and f .U / 12 ; 12 . Therefore
f 2 D.Z; p/ and Fact 8 is proved. t
u
Fact 9. Given a space Z suppose that F Cp .Z; I/ is a D-separating set such that
u D uZ 2 F . Then Z is homeomorphic to a closed subspace of D.F; u/.
Proof. Given any z 2 Z, let e.z/.f / D f .z/ for any f 2 F . Then e.z/ 2 Cp .F; I/
and the map e W Z ! Cp .F; I/ is continuous (see Problem 166 of [TFS]). Since
e.z/.u/ D u.z/ D 0 for any z 2 Z and e.z/ is continuous at u, we have e.z/ 2
D.F; u/ for any z 2 Z, i.e., e.Z/ D.F; u/. It is an evident consequence of the
fact that F is D-separating that for any z 2 Z and any closed P Z with z P ,
there is f 2 F such that f .z/ f .P /. Therefore e is an embedding by Problem
166 of [TFS] and we only must prove that e.Z/ is closed in D.F; u/.
Take any element ' 2 D.F; u/ne.Z/. There exists O 2 .u; F / such that
'.O/ 12 ; 12 . By definition of the pointwise convergence topology, there is a
finite K Z and " > 0 such that u 2 W D ff 2 F W f .K/ . "; "/g O and
hence '.W / 12 ; 12 . Since ' e.K/, there is U 2 .K; Z/ such that ' e.U /.
The family F being D-separating, there is g 2 F such that g.K/ . "; "/ and
g.ZnU / 34 ; 1 and, in particular, g 2 W . This implies e.z/.g/ D g.z/ 2 34 ; 1
for all z 2 ZnU while '.g/ 2 '.W / 12 ; 12 . Consequently, G D f 2 D.F; u/ W
.g/ < 34 g is an open neighborhood of ' in D.F; u/ such that G \ e.ZnU / D ; and
therefore ' e.ZnU /. This implies ' e.U / [ e.ZnU / D e.Z/. The function
' 2 D.F; u/ne.Z/ was chosen arbitrarily, so e.Z/ is closed in D.F; u/ and Fact 9
is proved. t
u
Fact 10. Given a space Z and any F Cp .Z; I/ with u D uZ 2 F , the space
D.F; u/ belongs to the class K.Z/.
Proof. Given any number n 2 N and a point z D .z1 ; : : : ; zn / 2 Z n , consider the
set B.z; n/ D f' 2 IF W '.f / 2 12 ; 12 for any f 2 F such that f .zi / 2 . n1 ; n1 /
for all i 2 Mn g. It is immediate S that B.z; n/ D.F; u/ for any n 2 N and z D
.z1 ; : : : ; zn / 2 Z S
n
. Let Bn D fB.z; n/ W z 2 Z n g for every n 2 N; it is easy to see
that D.F; u/ D fBn W n 2 Ng and hence it suffices to establish that Bn 2 K.Z/
for all n 2 N.
Fix any n 2 N and denote by i W Z n ! Z the natural projection of Z n onto
its i th factor, i.e., i .z/ D zi for any z D .z1 ; : : : ; zn / 2 Z n . Consider the set
Pn D f.z; '/ 2 Z n IF W z 2 Z n and ' 2 B.z; n/g. It turns out that
. / the set Pn is closed in Z n IF .
To prove . / take any point a D .a1 ; : : : ; an / 2 Z n and ' 2 IF such that
.a; '/ 2 .Z n IF /nPn . Then ' B.a; n/ which implies that there is a function
f 2 F such that jf .ai /j < n1 for all i 2 Mn while j'.f /j > 12 . Recalling that f
is continuous on Z and IF carries the product topology, we conclude that the set
W D f.b; / 2 Z n IF W jf .i .b//j < n1 for all i 2 Mn and j.f /j > 12 g is
open in Z n IF . It is evident that .a; '/ 2 W and W \ Pn D ;, so any point
.a; '/ 2 .Z n IF /nPn has a neighborhood W which does not meet Pn . Thus Pn is
closed in Z n IF and . / is proved.
Observe that Z n IF 2 K.Z/ because IF is compact and hence Pn 2 K.Z/ by
. /. Now, if W Z n IF ! IF is the natural projection, then .Pn / D Bn and
hence Bn isSa continuous image of Pn which implies that Bn 2 K.Z/. Therefore
D.F; u/ D fBn W n 2 Ng also belongs to K.Z/ and Fact 10 is proved. t
u
Fact 11. If Z is a K-analytic space, then every Y 2 K.Z/ is also K-analytic.
Proof. Let CS be the class of compact spaces; it is evident that every compact space
is K-analytic, so the class A of K-analytic spaces contains fZg [ CS. Besides, A
is closed under finite (and even countable) products, closed subspaces, continuous
images and countable unions by Fact 3, so K.Z/ A by Fact 1. Therefore Fact 11
is proved. t
u
Fact 12. If Z is a K-analytic space and Cp .Y / embeds in Cp .Z/, then Y is also
K-analytic.
Proof. It is easy to see that there exists an embedding e W Cp .Z/ ! Cp .Z; I/ such
that e.uZ / D uZ . The space Cp .Z/ being homogeneous (i.e., for any f; g 2 Cp .Z/,
there is a homeomorphism ' W Cp .Z/ ! Cp .Z/ such that '.f / D g (see Problem
079 of [TFS])), there is an embedding w W Cp .Y / ! Cp .Z/ such that w.uY / D uZ .
Therefore i D e w embeds Cp .Y / in Cp .Z; I/ in such a way that i.uY / D uZ . If

F D Cp .Z; I/; H D i.Cp .Y // and p D uZ , then D.H; p/ is a continuous image
of D.F; p/ by Fact 8. The space D.F; p/ belongs to the class K.Z/ by Fact 10
and hence D.F; p/ is K-analytic by Fact 11. Being a continuous image of D.F; p/
(see Fact 8), the space D.H; p/ is also K-analytic by Fact 3. Now apply Fact 9 to
conclude that Y embeds in D.H; p/ as a closed subspace and therefore Y is also
K-analytic. Fact 12 is proved. t
u
Now it is easy to finish our solution. There exists a space M K which is not K-
analytic (see Fact 5). The space X D Cp .M / has countable network and hence there
is a second countable space Y such that X embeds in Cp .Y / (see Fact 7). It is clear
that Y is a Lindelf p-space (see Problem 221). If the space X D Cp .M / embeds
in Cp .Z/ for some K-analytic space Z, then M is K-analytic by Fact 12 which
is a contradiction. Therefore X is not embeddable in Cp .Z/ if Z is K-analytic;
hence X is a space which embeds in Cp .Y / for a Lindelf p-space Y while X is
not embeddable in Cp .Z/ if Z is a K -space. Our solution is complete.
T.251. Give an example of a p-space Y and a pseudo-open (and hence quotient)
map ' W Y ! Cp .X / of Y onto Cp .X / for an uncountable space X .
Solution. The space X D A.!1 / is uncountable and it was proved in Problem 146
of [TFS] that Cp .X / is a FrchetUrysohn space. Therefore we can find a metrizable
space Y such that there is a pseudo-open map f W Y ! Cp .X / (see Problem 225
of [TFS]). Since Y is a p-space (see Problem 221), our spaces X and Y have all
required properties.
T.252. Prove that X is a Lindelf p-space if and only if it there is a perfect map of
X onto a second countable space.
Solution. If Z is a space and A Z, then A is a cozero (zero) set in Z if there is
f 2 Cp .Z/ such that A D f 1 .Rnf0g/ (or A D f 1 .0/ respectively).
every n 2 !, call the sequence
S fUn W n 2 !g a featheringT of the space Z in its
every z 2 Z. A space Z is a p-space if and only if it has a feathering in Z.
Fact 1. For any space Z:
(1) a set A Z is cozero if and only if ZnA is a zero-set;
(2) if ' W Z ! T is a continuous map and U is a cozero-set in T , then ' 1 .U / is
a cozero-set in Z;
(3) any finite intersection of cozero-sets in Z is a cozero-set in Z;
(4) any countable union of cozero-sets in Z is a cozero-set in Z;
(5) a set A Z is cozero if and only if there is f 2 Cp .Z/ and U 2 .R/ such
that A D f 1 .U /;
(6) the family Coz.Z/ of all cozero subsets of Z is a base in Z.
Proof. It is evident that a function f 2 Cp .Z/ witnesses that A is a cozero-set

if and only if f witnesses that ZnA is zero-set. This proves (1). If f 2 Cp .T /
and U D f 1 .Rnf0g/, then g D f ' 2 Cp .Z/ and ' 1 .U / D g 1 .Rnf0g/.
This proves (2). The property (3) follows from (1) and the fact that any finite union
of zero-sets is a zero-set by Fact 1 of S.499. Analogously, the property (4) holds
because any countable intersection of zero-sets is a zero-set by Fact 1 of S.499.
To prove (5) observe that in a perfectly normal space, any open set is cozero (see
Fact 2 of T.080). Therefore every U 2 .R/ is a cozero-set in R and hence we can
apply (2) to conclude that f 1 .U / is a cozero-set in Z for any f 2 Cp .Z/. This
proves sufficiency in (5); necessity in (5) is evident from the definition.
As to (6), take any z 2 Z and any W 2 .z; Z/. By the Tychonoff property
of Z, there is f 2 Cp .Z; 0; 1 / such that f .z/ D 1 and f .ZnW / f0g. It is
immediate that U D f 1 .Rnf0g/ 2 Coz.Z/ and z 2 U W . This establishes (6)
and completes the proof of Fact 1. t
u
Returning to our solution, observe that any second countable space is Lindelf p
by Problem 221, so if there is perfect map of X onto a second countable space, then
X is Lindelf p by Problem 245. This proves sufficiency.
Now assume that X is a Lindelf p-space and fix a feathering fUn W n 2 !g
of the space X in X . Given n 2 !, for every x 2 X , there is a cozero-set Ux in
the space X such that x 2 Ux U for some U 2 Un (see Fact 1); let Vn be a
countable subcover of the cover fUx W x 2 X g of the space X . It is obvious that
S the
sequence fVn W n 2 !g is also a feathering of X in X . For each V 2 V D fVn W
n 2 !g, take a function fV 2 Cp .X / with V D fV1 .Rnf0g/. The family V is
countable, so the function f D ffV W V 2 Vg maps X into a second countable
space RV and hence Y D f .X / is also second countable. Thus g D f jX maps X
onto a second countable space Y , so it suffices to show that g is perfect.
Take any x 2 X and y 2 X nX . Since fVn W n 2 !g is a feathering of X in
X , there is n 2 ! such that y St.x; Vn / and hence there is V 2 V with x 2 V
and y V . As a consequence, fV .y/ D 0 fV .x/ and therefore f .x/ f .y/.
This shows that f .X nX / \ Y D ; and hence X D f 1 .Y /. Now apply Fact 2
of S.261 to conclude that g is a perfect map of X onto a second countable space Y .
This settles necessity and makes our solution complete.
T.253. Prove that X is a Lindelf -space if and only if it is a continuous image of
a Lindelf p-space.
Solution. If X is a continuous image of a Lindelf p-space, then X is a Lindelf
-space by Problems 223 and 243. Now if X is a Lindelf -space, then there is a
compact space K, a second countable space M and a closed subspace F of K M
such that X is a continuous image of F (see Problem 248). The natural projection
W K M ! M is a perfect map by Fact 3 of S.288. The space M being a Lindelf
p-space by Problem 221, the product K M is also a Lindelf p-space because it is
a perfect preimage of M (see Problem 245). Furthermore, F is a Lindelf p-space
by Problem 224, so X is a continuous image of a Lindelf p-space F .
T.254. Prove that the class of Lindelf -spaces is the smallest one which contains
all compact spaces, all second countable spaces and is invariant with respect to
finite products, continuous images and closed subspaces.
Solution. Let L./ be the class of Lindelf -spaces. Apply Problem 226 to
see that L./ contains the class CS of compact spaces; the class M of second
countable spaces is contained in L./ by Problem 221. It was proved in Fact 1
of T.227 that L./ is invariant under finite products. The class L./ is invariant
under continuous images by Problem 243 and under closed subspaces by Problem
224, so L./ contains CS [ M and has the above mentioned invariance properties.
Now assume that a class D of spaces contains the class CS [ M and is invariant
under finite products, closed subspaces and continuous images. For any X 2 L./
there is a compact space K, a second countable space M and a closed subspace
F of K M such that X is a continuous image of F (see Problem 248). Since
K 2 CS D and M 2 M D, we have K M 2 D because D is invariant under
finite products. Besides, F 2 D because D is invariant under closed subspaces and,
finally, X 2 D because D is invariant under continuous images. This shows that
L./ D and hence L./ is the minimal class containing CS [ M with the
invariance properties in question.
T.255.
Q Suppose that Xn is a Lindelf p-space for each n 2 !. Prove that the product
fXn W n 2 !g is a Lindelf p-space.
Solution. For any n 2 !, there is a second countable space Mn and a perfect
Q map
fn W Xn ! Mn because
Q X n is Lindelf p (see Problem 252). Then f D Q ffn W
n 2 !g maps X D fXn W n 2 !g onto the second countable space M D fMn W
n 2 !g and the map f is perfect by Fact 4 of S.271. Applying Problem 252 once
more, we conclude that X is a Lindelf p-space.
Q
T.256. Suppose that Xn is a Lindelf -space for each n 2 !. Prove that fXn W
n 2 !g is a Lindelf -space.
Solution. For any n 2 !, there is a Lindelf p-space Yn and a continuous onto
map fn W Yn ! XnQbecause the space Xn is Lindelf
Q (see Problem 253). Then
the function
Q f D ffn W n 2 !g maps Y D fYn W n 2 !g onto the space
X D fXn W n 2 !g; besides, the map f is continuous and onto by Fact 1 of
S.271. The space Y is Lindelf p by Problem 255, so we can apply Problem 253
once more to conclude that X is a Lindelf -space.
S
T.257. Let X be a space such that X D fXn W n 2 !g, where each Xn is a
Lindelf -space. Prove that X is a Lindelf -space.
Solution. Lthe space X is a continuous image of the space
L It is straightforward that
Y D fXn W n 2 !g Z D fXn W n 2 !g. We consider each Xn to be
the respective clopen subspace of Z (see Problem 113 of [TFS]). For each n 2 !,
there is a countable family Fn of compact subsets of Xn whichS separates Xn from
Xn nXn (see Problem 233). It is evident that the family F D fFn W n 2 !g is
countable, consists of compact subsets of Z and separates Y from ZnY . Therefore

Y is a Lindelf -space by Problem 233 and hence X is a Lindelf -space by
Problem 243.
T that Z is a space and Xn Z is Lindelf for each n 2 !. Prove

T.258. Suppose
that X D fXn W n 2 !g is a Lindelf -space.
Solution.
Q Observe that the space X is homeomorphic to a closed subspace of Y D
fXn W n 2 !g by Fact 7 of S.271. Since Y is a Lindelf -space by Problem 256,
the space X is also Lindelf by Problem 224.
T.259. Let X be a Lindelf -space such that each compact subset of X is finite.
Solution. This is Fact 2 of T.227.
T.260. Give an example of a Lindelf p-space X such that nw.X / > ! and all
compact subsets of X are countable.
Solution. Given an arbitrary space Z recall that its Alexandroff double AD.Z/ is
the space with the underlying set Zf0; 1g in which all points of Zf1g are isolated
and the local base at any point y D .z; 0/ is given by the family By D f.U
f0; 1g/nf.z; 1/g W U 2 .z; Z/g. Let W AD.Z/ ! Z be the natural projection,
i.e., .y/ D z for any y D .z; i / 2 AD.X / with i 2 f0; 1g. It is immediate that
W AD.Z/ ! Z is a continuous map for any space Z.
For the Cantor set K R apply Fact 5 of S.151 to find disjoint A; B K
such that A \ F ; B \ F for any uncountable closed F K. The projection
W AD.K/ ! K is perfect because AD.K/ is compact (see Problem 364 of [TFS]).
If X D 1 .A/, then f D jX W X ! A is also a perfect map by Fact 2 of S.261.
Since A is second countable, the space X is Lindelf p by Problem 245. Observe
that the set ff.a; 1/g W a 2 Ag is a disjoint uncountable family of non-empty open
subsets of X and hence nw.X / c.X / > !. Now, if G is an uncountable compact
subspace of X , then F D f .G/ is an uncountable compact subset of A and hence
B \ A B \ F ; which is a contradiction. Thus every compact subspace of X
is countable, i.e., X is a Lindelf p-space of uncountable network weight in which
all compact subsets are countable.
T.261. Prove that any K -space is Lindelf . Show, additionally, that there exists
a K -space which is not Lindelf p.
Solution. If X is a K -space, then there exists a sequence fYn WQ
n 2 !g such that
Yn is -compact for all n 2 ! and X embeds in the product Y D fYn W n 2 !g as
a closed subspace (see Problem 338 of [TFS]). Each Yn is Lindelf by Problem
226, so Y is a Lindelf -space by Problem 256. Therefore X is Lindelf by
Problem 224.
It was proved in Problem 231 that there exists a countable space N which is not
a p-space. Every countable space is K being -compact. Thus N is an example
of a K -space which is not a p-space.
T.262. Let X be a K -space such that all compact subsets of X are countable and
.X / !. Prove that X is countable.
Solution. The space D is the set f0; 1g with the discrete topology. We consider
that D0 D f;g; for every n 2 !; f 2 Dn and i 2 D let .f _ i /jn D f and
.f _ i /.n/ D i . Then f _ 0; f _ 1 2 DnC1 and D nC1 D ff _ 0; f _ 1 W f 2 Dn g.
As usual, we identify any ordinal with the set of its predecessors and, in particular,
n D f0; : : : ; n 1g for every n 2 !.
Fact 1. If Z is an uncountable K -space with .Z/ D !, then there exist
uncountable closed sets F; G Z such that F \ G D ;.
Proof. Assume first that every z 2 Z has a countable open neighborhood Uz . The
S Lindelf, there is a countable subcover of the cover fUz W z 2 Zg,
space Z being
i.e., Z D fUz W z 2 Ag for some countable A Z and hence Z is countable
which is a contradiction.
Thus there is z 2 Z such that jU j > ! for any U 2 .z; Z/. It follows from
.Z/ D ! that there
T is a sequence fUn W n 2 !g .z; Z/ suchS that U nC1 Un
for all n 2 ! and fUn W n 2 !g D fzg. The set Znfzg D fZnUn W n 2 !g
being uncountable, there is n 2 ! such that F D ZnUn is uncountable. The set
G D U nC1 is also uncountable by our choice of z; it follows from U nC1 Un that
F \ G D ;, so the sets F and G are like promised and Fact 1 is proved. t
u
T assume that X is uncountable and fix a space Y such

Returning to our solution,
that X Y and X DS fYn W n 2 !g where each Yn is a -compact subspace
of Y . We have Yn D fKnm W m 2 !g where Knm is compact and Knm KnmC1
for all m; n 2 !. Since X Y0 , there exists m0 2 ! such that H; D K0m0 \ X
is uncountable. Given n 2 ! assume that we constructed fmi W i ng ! and
families Hk D fHf W f 2 Dk g for all k n with the following properties:
(1) m0

mn ;
(2) Hk is disjoint for every k n;
(3) Hf is closed in X and uncountable for any f 2 Dk and k n;
(4) if m < k T
S n, then Hf Hf jm for any f 2 Dk ;
(5) Hk fKimi \ X W i kg for any k n.
To construct the family HnC1 fix any f 2 Dn ; the set Hf being uncountable,
qf
there is qf 2 ! such that the set Z D KnC1 \ Hf is uncountable. It is clear
that Z is a K -space, so there are uncountable closed disjoint sets F; G Z by
Fact 1; let Hf _ 0 D F and Hf _ 1 D G. After we have Hf _ 0 and Hf _ 1 for all
f 2 Dn we have P the family fH f _ 0 ; Hf _ 1 W f 2 D g D fHf W f 2 D
n nC1
g; let
mnC1 D mn C fqf W f 2 D g. We omit a straightforward verification of the fact
n
that the set fmi W i nC1g and the families fHk W k nC1g satisfy the conditions
(1)(5). Continuing our inductive construction, we obtain a set fmi W i 2 !g !
and a sequence fHk W k 2 !g of families with the properties (1)(5) fulfilled for any
n 2 !.
T mi
Observe that Qn D fKS i W i ng is a non-empty compact subspace of Y
because the uncountable set Hn is contained in Qn by T the property (5). Since
the sequence fQ T n W n 2 !g is decreasing, we have Q D fQn W n 2 !g ;.
Besides, Q fYn W n 2 !g D X , so Q is a compact subspace of X . Given any
f 2 D! , observe that the sequence fHf jn W n 2 !g is decreasing by (4); the set
H f jn is compact for each n 2 ! because Hf jn Knmn for each n 2 ! by (5) (the
T
bar denotes the closure in Y ). As a consequence, Rf D fH f jn W n 2 !g ; and
T
Rf fYn W n 2 !g D X which showsTthat Rf H f jn \ X D clX .Hf jn / D
Hf jn [see (3)] for every n 2 !, so Sf D fHf jn W n 2 !g ; for any f 2 D! ;
take any xf 2 Sf . If f; g 2 D! and f g, then f jn gjn for some n 2 ! and
hence Hf jn \ Hgjn D ; by (2) which shows that xf xg . Consequently, f ! xf
is an injection of D! into Q and hence Q is a compact subspace of X of cardinality
at least continuum which contradicts the assumption that all compact subspaces of
X are countable. Thus X is countable and our solution is complete.
T.263. Suppose that X is a Lindelf -space and Cp .X / has the Baire property.
Prove that X is countable. In particular, if X is a space with a countable network
and Cp .X / is Baire, then X is countable.
Solution. All compact subsets of X are finite by Problem 284 of [TFS]. Now apply
Problem 259 to conclude that X is countable.
T.264. Prove that there exists an uncountable Lindelf space X for which Cp .X /
has the Baire property.
Solution. The space X D L.!1 / is the desired example. It is clear that X is
Lindelf and uncountable. Besides, all countable subsets of X are closed in X
which, together with normality of X , implies that all countable subsets of X are
closed and C -embedded in X . Thus Cp .X / is pseudocomplete by Problem 485 of
[TFS] and hence Baire by Problem 464 of [TFS].
T.265. Suppose that Cp .X / is a Lindelf -space and has the Baire property. Prove
Solution. Given any points x1 ; : : : ; xn 2 X and sets O1 ; : : : ; On 2 .R/, the set
x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng is called a
standard open subset of Cp .X /. Standard open sets x1 ; : : : ; xn I O1 ; : : : ; On where
n 2 N; x1 ; : : : ; xn 2 X and O1 ; : : : ; On 2 .R/ form a base in the space Cp .X /
(see Problem 056 of [TFS]). If U D x1 ; : : : ; xn I O1 ; : : : ; On is a standard open
subset of Cp .X /, then supp.U / D fx1 ; : : : ; xn g.
Choose any countable network F with respect to a compact cover C of the space
Cp .X / (see Problem 225). We can assume that F is closed with respect to finite
intersections and all elements of F are closed S in Cp .X / (see Fact 1 of T.229). Let
F 0 D fF 2 F W Int.F / D S ;g. Then F 0 Cp .X / by the Baire property of
Cp .X /; fix any u 2 Cp .X /n. F 0 /. If H D fF 2 F W u 2 F g, then Int.F / ;
for any F 2 H, so we Scan choose a non-empty standard open WF F for every
F 2 H. The set A D fsupp.WF / W F 2 Hg is countable; we claim that A D X .
T
Observe first that H is a compact subspace of Cp .X /. Indeed, there is C 2 C
with u 2 C ; it follows from the fact that F is a network with respect toTC that
T
TfF 2 F W C F g D C . Since fF T 2 F W C F g H, we have H
fF 2 F W C F g D C ,Ti.e., the set H is a closed subspace of the compact
space C and therefore K D H is compact. Thus we can apply Fact 1 of S.326 to
conclude that H is a network at K, i.e., for any U 2 .K; Cp .X //, there is H 2 H
with H U .
Now, if A X , then take any x 2 X nA. For any g 2 Cp .X / let '.g/ D g.x/;
then the map ' W Cp .X / ! R is continuous by Problem 166 of [TFS]. Therefore
'.K/ is a compact (and hence bounded) subset of R; take any m 2 N such that
'.K/ . m; m/. The set U D fh 2 Cp .X / W h.x/ 2 . m; m/g is open in Cp .X /
and K U . Since H is network at K, there is H 2 H with H U and hence
WH U . We have WH D x1 ; : : : ; xn I O1 ; : : : ; On where xi 2 A for all i n
and hence x fx1 ; : : : ; xn g. Therefore there exists a function h 2 Cp .X / such that
h.xi / 2 Oi for all i n and h.x/ D m C 1 (see Problem 034 of [TFS]). It is
immediate that h 2 WH nU which contradicts WH H U . This shows that
X D A is a countable space.
T.266. Prove that every Lindelf -space is stable, and hence, for every Lindelf
-space X , the space Cp .X / is monolithic.
Solution. If we establish that every Lindelf -space is stable, then Cp .X / is
monolithic for every Lindelf -space X by Problem 154. To show that every
Lindelf -space is stable, we will need the following fact.
Fact 1. Let Z be a space and suppose that fa W Z ! Ya is aQ continuous map for
all a 2 A. For the map f D ffa W a 2 Ag W Z ! Y D fYa W a 2 Ag, let
Z 0 D f .Z/. If the map fb is perfect for some b 2 A, then f W Z ! Z 0 is also
perfect.
if A D fbg. If Anfbg ;, consider the map
Proof. Clearly, there is nothing to proveQ
g D ffa W a 2 Anfbgg W Z ! T D fYa W a 2 Anfbgg. Since g is continuous,
the graph G.g/ D f.z; g.z// W z 2 Zg Z T is closed in Z T and the restriction
jG.g/ of the natural projection W Z T ! Z is a homeomorphism of G.g/
onto Z (see Fact 4 of S.390). If ' D .jG.g//1 , then ' W Z ! G.g/ is also
a homeomorphism. Define the identity map id W T ! T by id.z/ D z for every
z 2 T . Then id is a homeomorphism and hence perfect which shows that the map
u D fb id W Z T ! Yb T D Y is also perfect by Fact 4 of S.271. It is
immediate that Z 0 D u.G.g// and hence the map u0 D ujG.g/ W G.g/ ! Z 0 is
also perfect being a restriction of a perfect map u to a closed subset G.g/ of the
space Z T . Now observe that f D u ' and hence f is perfect as well. Fact 1 is
proved. t
u
Returning to our solution, let us show first that
(1) nw.Z/ i w.Z/ for any Lindelf -space Z.
To prove (1) take an infinite cardinal such that i w.Z/ . There is a Lindelf
p-space Y and a continuous onto map f W Y ! Z (see Problem 253). There exists
a second countable space M and a perfect map g W Y ! M by Problem 252. Let
h W Z ! T be a condensation such that w.T / .
The map ' D .h f /g W Y ! T M is perfect because so is g (see Fact 1);
if W T M ! T is the natural projection, then it is immediate from the definition
of the diagonal product of mappings that ' D h f . Besides, for the space
H D '.Y /, we have w.H / w.T M / w.T / C w.M / C ! D . The
map p D jH W H ! T is continuous and onto so r D h1 p maps H onto Z.
Given a closed subset F of the space Z, we have r 1 .F / D '.f 1 .F // because
r ' D f . The map ' being perfect, the set r 1 .F / is closed in H for every closed
F Z, i.e., r is continuous. It turns out that Z is a continuous image of the space
H ; therefore nw.Z/ w.H / and (1) is proved.
Now it is easy to see that any Lindelf -space is stable. Indeed, if Z is Lindelf
and f W Z ! Y is a continuous onto map, assume that i w.Y / for some
infinite cardinal . The space Y is also Lindelf by Problem 243, so we can apply
(1) to conclude that nw.Y / i w.Y / . This shows that Z is stable and makes
our solution complete.
T.267. Prove that if X is a Lindelf -space, then X is !-stable.
Solution. Take a continuous onto map f W X ! Y such that i w.Y / !. Apply
Problem 446 of [TFS] to conclude that Y is realcompact and hence there is a
continuous map g W X ! Y such that gjX D f (see Problem 412 of [TFS]).
Therefore Y is a Lindelf -space being a continuous image of a Lindelf -space
X (see Problem 243). Consequently, the space Y is stable by Problem 266 and
hence nw.Y / i w.Y / ! which proves that X is !-stable.
T.268. Prove that any product and any -product of Lindelf -spaces is stable.
Show that any -product of Lindelf -spaces is !-stable.
Q
Solution. Given a family fXt W t 2 T g of spaces and a point a 2 X D t 2T Xt ,
let .X; a/ D fx 2 X W jft 2 T W x.t/ a.t/gj < !g; we will also need the space
.X; a/ D fx 2 XQ W jft 2 T W x.t/ a.t/gj !g. For any S T , the map
pS W X ! XS D t 2S Xt is the natural projection defined by pS .x/ D xjS for
any x 2 X .
Fact 1. Given spaces Y and Z, if q W Y ! Z is an R-quotient map, then for any
space M , a map p W Z ! M is continuous if and only if p q is continuous.
Proof. Necessity is evident so assume that p q is continuous for some mapping
p W Z ! M . The family B D ff 1 .O/ W O 2 .R/; f 2 C.M /g is a base in M
by Fact 1 of S.437, so it suffices to show that p 1 .U / is open in Z for any U 2 B.
There is O 2 .R/ and f 2 C.Z/ such that U D f 1 .O/. Then p 1 .U / D
.f p/1 .O/.
Observe that f .p q/ D .f p/ q is continuous because p q is continuous.
Therefore f p is also continuous because q is R-quotient. Consequently, the set
p 1 .U / D .f p/1 .O/ is open in Z and hence the map p is continuous. Fact 1 is
proved. t
u
Fact 2.QAssume that we have a family fXt W t 2 T g of spaces such that theQproduct
XA D t 2A Xt is Lindelf for every finite A T . Then for any a 2 X D t 2T Xt ,
the space D .X; a/ is Lindelf.
Proof.SLet supp.x/ D ft 2 T W x.t/ a.t/g for every x 2 X and observe that
D fn W n 2 !g where n D fx 2 W jsupp.x/j ng for every n 2 !.
Therefore it suffices to show that n is Lindelf for every n 2 !. We will do this by
induction on n. If n D 0, then n D fag is a Lindelf space; assume that n > 0 and
we proved that n1 is Lindelf. Q
Call a set U X standard if U D t 2T Ut where Ut 2 .Xt / for each t 2 T
and the set supp.U / D ft 2 T W Ut Xt g is finite. The family B of all standard sets
is a base in X (see Problem 101 of [TFS]), soS to prove that n is Lindelf, it suffices
to show that for
S 0 every U B such that n U, there is a countable U 0 U such
that n U . S
By the induction
S hypothesis, there is a countable V U such that n1 V.
The set A D fsupp.V / W V 2 Vg is countable; we claim that
S
. / supp.x/ A for any x 2 n n. V/.
S
Indeed, if . / is not true, then t0 2 supp.x/nA for some x 2 n n. V/. Let
y.t/ D x.t/ for all t 2 T nft0 g and y.t0 / D a.t0 /. Then y 2 n1 and hence y 2 V
for some V 2 V. We have supp.V / A and therefore pA1 ..pA .V // D V ; since
pA .x/ D pA .y/ and y 2 V , we have x 2 V which is a contradiction showing that
. / is true. S
It follows from . / that P D n n. V/ Q D n \ .XA fpT nA .a/g/. It is
straightforward that the mapping pA jQ is a homeomorphism of Q onto the space
.n; A/ D fy 2 XA W jft S 2 T W y.t/ a.t/gj ng. It is easy to see that we have
the equality .n; A/ D fRB W B A and jBj ng where RB D XB fpAnB .a/g
is Lindelf being homeomorphic to XB for any finite B A. Any countable union
of Lindelf spaces is a Lindelf space, so .n; A/ is Lindelf and hence Q is also
Lindelf being homeomorphic to .n; A/. Since P is a closed subset of n , it is
also closed in Q which S implies that P is Lindelf. Thus we can choose a countable
W U such that S 0P W. It is clear that U 0 D V [ W is a countable subfamily
of U and n U . The inductive
S step being carried out, we proved that every n
is Lindelf and hence D n is also Lindelf so Fact 2 is proved. t
u
Q
Fact 3. Given a space Xt for every t 2 T , let X D fXt W t 2 T g. Suppose that
Y is a Lindelf subspace of the space X and f W Y ! M is a continuous map of
Y onto a space M such that w.M / D !. Then there is a set S T and a
continuous map g W pS .Y / ! M such that jS j and f D g .pS jY /.
Proof. We will also need projections between the faces of X , namely, if A B
T , then we have a natural projection pAB W XB ! XA defined by pAB .x/ D xjA
for any x 2 XB . The map pAB is open, continuous and pAB PB D pA for all
A B T.
Since w.M / , there is an embedding of M into R , so we can assume,

without loss of generality, that M R . For every < , the map W R ! R is
the natural projection of R onto its th factor; let q D jM . Given any < , we
can apply Problem 298 of [TFS] to the map f D q f W Y ! R to find a countable
set S T and a continuous
S map g W pS .Y / ! R such that f D g .pS jY /.
We claim that S D fS W < g is as promised.
Indeed, given z 2 pS .Y /, let g.z/./ D g .zjS / for each < ; this defines a
map g W pS .Y / ! R . It is immediate that g D fg .pSS jpS .Y // W < g and
hence the map g is continuous. Given any y 2 Y , we have
for every < which shows that f .y/ D g.pS .y// for each y 2 Y . As a
consequence, g W pS .Y / ! M and f D g .pS jY / so Fact 3 is proved. u
t
Fact 4. Assume that we have a family fXt W t 2 T g of Q
Q spaces such that the product
X
t 2A t is Lindelf for every finite A T . Let X D fXt W t 2 T g and suppose
that Y 2 f.X; a/; .X; a/g. Then, for any infinite cardinal and any continuous
map f W Y ! M of Y to a space M with w.M / , there is a set S T and a
Proof. If Y D .X; a/, then Y is Lindelf by Fact 1, so our statement is true for
Y by Fact 3. Now, if Y D .X; a/, then fix a set S T with jS j such that
there is a continuous map g1 W pS ..X; a// ! M for which g1 .pS j.X; a// D
f j.X; a/. Take any x; y 2 .X; a/ such that pS .x/ D pS .y/; then supp.x/\S D
supp.y/ \ S , so we can choose sets P D fti W i 2 !g T and Q D fsi W i 2 !g
T such that, for any i 2 !, we have ti 2 S if and only if si 2 S while supp.x/ P
and supp.y/ Q. Let xi .t/ D a.t/ for all t 2 T nftk W k i g and xi .t/ D x.t/
otherwise. Analogously let yi .t/ D a.t/ for all t 2 T nfsk W k i g and yi .t/ D
y.t/ otherwise. It is immediate that xi ; yi 2 .X; a/; pS .xi / D pS .yi / for all
i 2 ! and lim xi D x; lim yi D y. By our choice of S , the equality pS .z/ D pS .z0 /
implies f .z/ D f .z0 / for any z; z0 2 .X; a/ so f .xi / D f .yi / for all i 2 ! and
therefore f .x/ D f .y/ by continuity of f .
We proved that pS .x/ D pS .y/ implies f .x/ D f .y/ for any x; y 2
.X; a/. As a consequence, there exists a map g W pS ..X; a// ! M with
g .pS j.X; a// D f . Observe that .X; y/ .X; a/ for any y 2 .X; a/
and hence the map pS j.X; a/ is open by Fact 1 of T.110. Since the map f D
g .PS j.X; a// is continuous, we can apply Fact 1 to conclude that g is a
continuous map (see Problems 153 and 154 of [TFS]). Fact 4 is proved. t
u
Fact 5. Q Assume that we have a family fXt W t 2 T g of spaces suchQthat the product
XA D t 2A Xt is Lindelf for every finite A T . Let X D fXt W t 2 T g
and suppose that f W X ! M is a continuous map of X to a space M such that
w.M / . Then there is a set S T and a continuous map g W XS ! M such
that jS j and f D g pS .
Proof. Assume first that D !. Choose any a 2 X and let D .X; a/. By
Fact 4 there exists a countable set S T and a continuous mapping g W pS ./ D
XS ! M such that g .pS j/ D f j. For any x 2 X let h.x/ D g.pS .x//; then
h W X ! M is a continuous map such that hj D f j. Since is dense in X ,
we have f D h by Fact 0 of S.351. Therefore f D g pS and Fact 5 is proved for
D !.
Now assume that is an arbitrary infinite cardinal. Since w.M / , there is
an embedding of M into R , so we can assume, without loss of generality, that
M R . For every < , the map W R ! R is the natural projection of
R onto its th factor; let q D jM . Given any < , we can apply the result
proved for D ! to the map f D q f W X ! R to find a countable set S T
and aScontinuous map g W XS ! R such that f D g pS . We claim that
S D fS W < g is as promised.
It is evident that jS j . Given z 2 XS , let g.z/./ D g .zjS / for each < ;
this defines a mapping g W XS ! R . Obviously, g D fg pSS W < g and
hence the map g is continuous. Given any y 2 X , we have
for every < which shows that f .y/ D g.pS .y// for each y 2 X . As a
consequence, g W XS ! M and f D g pS , so Fact 5 is proved. u
t
Returning to our solution assume
Q that Xt is a Lindelf -space for any t 2 T
and take any point a 2 X D t 2T Xt . Let D .X; a/ and D .X; a/. Take
any infinite cardinal and a continuous onto map f W ! Z for which there is a
condensation h W Z ! M such that w.M / .
We can apply Fact 4 to the space and the map h f to find a set S T and a
continuous map g W pS ./ ! M such that jS j and g .pS j/ D h f . Denote
the map pS j W ! pS ./ by . Then is an open map because .X; y/ for
any y 2 (see Fact 1 of T.110). For the map ' D h1 g the map ' D f is
continuous, so ' is continuous by Fact 1. Therefore Z is a continuous image of the
space pS ./.
For any A S let QA D XA fpS nA .a/g. ItS is clear that QA is a subspace of
XS homeomorphic to XA . Observe that pS ./ D fQA W A is a finite subset of S g.
Each QA is a Lindelf -space by Problem 256 and hence pS ./ is a union of -
many Lindelf -spaces. Since each element of this union is stable by Problem 266,
the space pS ./ is -stable by Problem 124. Thus Z is -stable by Problem 123 and
hence nw.Z/ which proves that is -stable. This proves stability of .
To prove stability of the space X observe first that for any t 2 T , there exists a
Lindelf p-spaceQYt and a continuousQonto map qt W Yt ! Xt (see Problem 253).
Therefore q D t 2T qt maps Y D t 2T Yt continuously onto X (see Fact 1 of
S.271). Now apply Problem 123 to conclude that it suffices to prove that Y is stable,
and hence we can assume, without loss of generality, that Xt is a Lindelf p-space
for every t 2 T .
Now take any continuous onto map f W X ! Z such that there is a condensation
h W Z ! M where w.M / . Apply Fact 5 to find a set S T and a continuous
map g W XS ! M such that jS j and g pS D h f . If ' D h1 g, then the
map ' pS D f is continuous and hence ' is continuous because pS is an open
map (see Fact 1). Thus Z is a continuous image of XS .
We saw that each Xt can be assumed to be Lindelf p, so there is a perfect map
ut W Xt ! Lt of Xt onto Q a second countable space Lt for every t 2 T Q
(see Problem
252) and hence u D t 2S ut maps XS perfectly onto the space L D t 2S Lt (see
Fact 4 of S.271). It is immediate that w.L/ . Observe that D ug is perfect by
Fact 1 of T.266 and maps XS onto a subspace F of LM ; it is clear that w.F / .
Let W F ! M be the restriction of the natural projection. Then .F / D M and
r D h1 W F ! Z. Observe that r D ' is a continuous map, so r is
continuous by Fact 1. Thus the space Z is a continuous image of F and therefore
nw.Z/ nw.F / w.F / which shows that nw.Z/ and hence X is
-stable. This shows that X is a stable space.
Finally take any continuous onto map f W ! Z for which there is a
condensation h W Z ! M such that w.M / !. We can apply Fact 4 to the
space and the map h f to find a countable set S T and a continuous
map g W pS ./ D XS ! M such that g .pS j/ D h f . Denote the map
pS j W ! XS by . Then is an open map because .X; y/ for any
y 2 (see Fact 1 of T.110). For the map ' D h1 g the map ' D f is
continuous, so ' is continuous by Fact 1. Therefore Z is a continuous image of XS
which is a Lindelf -space by Problem 256. Thus Z is also Lindelf which
shows that Z is stable by Problem 266. As a consequence, nw.Z/ ! which
proves !-stability of and makes our solution complete.
T.269 (Baturovs theorem). Let X be a Lindelf -space. Prove that for any set
Y Cp .X /, we have ext.Y / D l.Y /.
Solution. Suppose that we have spaces Z; T and a map u W Z ! T . Given
any number n 2 N, let un .z1 ; : : : ; zn / D .u.z1 /; : : : ; u.zn // for any point z D
.z1 ; : : : ; zn / 2 Z n . Thus un W Z n ! T n . If P is a set, then Fin.P / is the family of all
non-empty finite subsets of P . Consider the family O D f.a; b/ W a < b; a; b 2 Qg;
given a number n 2 N let On D fO1

On W S Oi 2 O for any i D 1; : : : ; ng.
Let fOk W k 2 !g be an enumeration of the family fOn W n 2 Ng. Thus, for each
k 2 !, there is mk 2 N and O1k ; : : : ; Omk k 2 O such that Ok D O1k

Omk k .
For any x D .x1 ; : : : ; xmk / 2 X mk , let x; Ok D ff S 2 Cp .X / W f .x/ 2 O g. If
mk k
Bk D fx; Ok W x 2 X g, then the family B D fBk W k 2 !g is a base of the

mk
space Cp .X /.
There is a Lindelf p-space X 0 such that X is a continuous image of X 0 (see
Problem 253). The relevant dual map embeds Cp .X / into Cp .X 0 / which shows
that Y can be considered a subspace of Cp .X 0 /. Therefore we can assume, without
loss of generality, that X D X 0 , i.e., X is a Lindelf p-space. Fix a perfect map
p W X ! M of X onto a second countable space M (see Problem 252). Then
pk D p mk maps X mk perfectly onto the second countable space M mk for every
k 2 ! (see Fact 4 of S.271).
To prove that ext.Y / D l.Y / it suffices to show that l.Y / ext.Y / so assume
S there exists an infinite cardinal ext.Y / and U .Cp .X //
the contrary; then
such that Y U and no subfamily of U of cardinality covers Y . It is easy to
see that, without loss of generality, we can assume that U B. We will need the set
Ak D fx 2 X mk W x; Ok 2 Ug for every k 2 !; since U covers Y , we have
(1) for any f 2 Y , there is k 2 ! and x 2 Ak such that f mk .x/ 2 Ok .
On the other hand, no subfamily of U of cardinality covers Y , so we have
(2) if Bk Ak and jBk j for any k 2 !, then there is a function f 2 Y such
that f mk .Bk / \ Ok D ; for every k 2 !.
Choose a function f0 2 Y arbitrarily and let B.k; 0/ D ; for all k 2 !.
Proceeding inductively assume that 0 < < C and we have chosen a set
ff W < g Y and a family fB.k; / W < ; k 2 !g with the following
properties:
(3) B.k; / Ak and jB.k; /j for all k 2 ! and < ;
(4) if
< < , then B.k;
/ B.k; / for every k 2 !;
(5) for any < ; k 2 ! and any H 2 Fin.ff
W
< g/, the set uH .B.k; //
is dense in uH .Ak / where uH D pk .ff mk W f 2 H g/ W X mk ! M mk
Rmk jH j ;
(6) fmk .B.k; // \ Ok D ; for all < and k 2 !.
To get f , let F D ff W < g and fix any k 2 !; for every H 2 Fin.F /,
let uH D pk .ff mk W f 2 H g/ W X mk ! M mk Rmk jH j . The space uH .Ak /
being second countable, there is a countableSB.H; k/ Ak such that S uH .B.H; k//
is dense in uH .Ak /. The set B.k; / D . fB.k; / W < g/ [ . fB.H; k/ W
H 2 Fin.F /g/ has cardinality . Once we have a set B.k; / for every k 2 !,
apply (2) to find a function f 2 Y such that fmk .B.k; // \ Ok D ; for all k 2 !.
It is immediate that the properties (3)(6) still hold for the set ff W g and
the family fB.k; / W ; k 2 !g, and therefore, our inductive construction can
be continued to give us a set D D ff W < C g and a family fB.k; / W <
C ; k 2 !g such that the properties (3)(6) hold for all < C .
Assume that < < C ; it follows from (5) that .pk fmk /.B.k; // is dense
in .pk fmk /.Ak / and hence fmk .B.k; // is dense in fmk .Ak / for all k 2 !. The
property (1) shows that fmk .Ak / \ Ok ; and therefore fmk .B.k; // \ Ok ;
for some k 2 !. On the other hand, fmk .B.k; // \ Ok D ; for all k 2 ! by the
property (6). Consequently, f f and therefore jDj D C .
Our purpose is to prove that D is closed and discrete in Y , so assume, towards
a contradiction, that g is an accumulation point in Y for the set D. Recall that
l.X ! / D ! by Problem 256 and hence t.Y / t.Cp .X // D ! (see Problem 149
of [TFS]). Therefore g is also an accumulation point for some countable subset of
D, and hence the ordinal D minf < C W g is an accumulation point for F g
is well-defined. It is evident that is a limit ordinal. There is k 2 ! and y 2 Ak

such that g 2 y; Ok ; it is evident Tthat g is also an accumulation point for the set
G D F \ y; Ok . The set K D f.f mk /1 .f mk .y// W f 2 Gg is non-empty
because y 2 K.
Let W D .gmk /1 .Ok / and assume that KnW ;. Take any x 2 KnW and
observe that gmk .x/ Ok while g mk .y/ 2 Ok and therefore g mk .x/ g mk .y/.
On the other hand, f mk .x/ D f mk .y/ for all f 2 G which contradicts g 2 G. We
proved that the case KnW ; is impossible, i.e., K W .
Let Kf D .f mk /1 .f mk .y// for all f 2 G; the set N D pk1 .pk .y// is compact
because the map pk T is perfect. Therefore N \ Kf is a non-empty compact set for
all f 2 G and y 2 fN \ Kf W f 2 Gg K W . Now we canTapply Fact 1
of S.326 to conclude that there is a finite H G such that Q D fN \ Kf W
f 2 H g W . Observe that for the map uH D pk .ff mk W f 2 H g/, we have
Q D u1 H .uH .y//. Now, if Y D uH .X /, then the map uH W X ! Y is perfect
because pk is perfect (see Fact 1 of T.266).
Therefore Fact 1 of S.226 is applicable to conclude that there is U 2 .Y / such
that uH .y/ 2 U and u1 H .U / W . Let
D maxf W f 2 H g. Then
< D
C 1 < because is a limit ordinal. We have H 2 Fin.F / and therefore

uH .B.k; // is dense in uH .Ak / by (5). Furthermore, uH .y/ 2 U \ uH .Ak / which
shows that U \uH .Ak / is a non-empty open subset of uH .Ak /. The set uH .B.k; //
being dense in uH .Ak / by (5), we have uH .B.k; // \ U ; and therefore there
is z 2 B.k; / for which u1 H .uH .z// W and, in particular, z 2 W . This implies
that g mk .z/ 2 Ok .
On the other hand, the condition (6) implies that fmk .B.k; // \ Ok D ;; the
conditions (4) and (6) show that for any ordinal with < < , we have
fmk .B.k; // \ Ok fmk .B.k; // \ Ok D ;. Consequently, fmk .z/ Ok
whenever < which shows that g GnF and hence g is an accumulation
point for the set F which is a contradiction with < and the choice of . This
contradiction proves that D is a closed discrete subspace of Y . We already saw that
jDj D C > ext.Y /; this contradiction shows that l.Y / ext.Y / and completes
our solution.
T.270. Prove that every subspace of X is a Lindelf -space if and only if X has a
countable network.
Solution. Given a space Z say that a family F of subsets of Z is T0 -separating if
for any distinct x; y 2 Z there is F 2 F such that F \ fx; yg consists of exactly
T F \ fx; yg D fxg or F \ fx; yg D fyg. The family F is
one point, i.e., either
T1 -separating if fF 2 F W z 2 F g D fzg for every z 2 Z.
Fact 1. Let Z be a Lindelf -space. If there exists a countable T1 -separating
family of closed subsets of Z, then nw.Z/ D !.
Proof. Take a compact cover C be of the space Z such that some countable family
N of closed subspaces of Z is a network with respect to C (see Problem 225 and
Fact 1 of T.229). Let F be a countable T1 -separating family of closed subsets of Z.
The family M of all finite intersections of the elements of N [ F is countable, so

it suffices to show that M is a network in Z.
Take any z 2 Z and O 2 .z; Z/; let Fz D fF 2 F W z 2 F g. There is C 2 C
with z 2 C . It is evident that
T the family K D fC \ F W F 2 Fz g consists of compact
subsets of Z and fzg D K. Therefore T we can apply Fact 1 of S.326
T to conclude
finite P Fz such that fC \F W F 2 Pg D . P/\C O. It
that there exists a T
is clear that G D P 2 M. Since G \C O, the closed set GnO is disjoint from
C , so there are U; V 2 .Z/ such that C U; GnO V and U \ V D ;. The
family N being a network with respect to C, there is N 2 N such that C N U .
We have P D N \ G 2 M; it is clear that z 2 P . Besides, given any y 2 P ,
if y O, then y 2 GnO V which, together with y 2 N U gives us a
contradiction. Therefore y 2 O for any y 2 P , i.e., z 2 P O which proves that
M is a countable network in Z, so Fact 1 is proved. t
u
Fact 2. Suppose that a hereditarily Lindelf space Z has a countable T0 -separating
family of closed subsets. Then Z has a countable T1 -separating family of closed
subsets.
Proof. Take a countable T0 -separating family F of closed subsets of Z. Since Z is
hereditarily Lindelf, every open subset of Z is an F -set in Z, soS we can fix, for
any F 2 F , a countable family PF ofSclosed subsets of Z such that PF D ZnF .
It is clear that the family G D F [ . fPF W F 2 F g/ is countable and consists of
closed subsets of Z.
To see that G is T1 -separating, take any point x 2 Z. If y x, then there exists
an element F 2 F such that F \ fx; yg is a singleton. If F \ fx; yg D fxg, then
x 2 F and y F ; then P D F 2 G. Now, if F \ fx; yg D fyg, then x 2 ZnF
and hence there is Q 2 PF such that x 2 Q ZnF which shows that y Q; then
P D Q 2 G. As a consequence, T for any x 2 Z and any y x, there is P 2 G such
that x 2 P and y P . Thus fG 2 G W x 2 Gg D fxg for any x 2 Z and hence
G is a T1 -separating countable family of closed subsets of Z. Fact 2 is proved. u t
Fact 3. Let Z be a hereditarily Lindelf space. If K is a the family of all compact
subspaces of Z, then jKj c.
Proof. We have jZj c by Problem 015. Since .Z/ hl.Z/ D !, we can fix,
for every z 2 Z,Ta family fUnz W n 2 !g .z; Z/ such that clZ .UnC1
z
/ Unz for
all n 2 ! and fUn W n 2 !g D fzg. The family U D fUn W n 2 !; z 2 Zg
z z
has cardinality c, so the cardinality of the family V of all countable unions of

elements of U does not exceed c. T
Given any K 2 K, take any z 2 ZnK; we have fclZ .Unz / W n 2 !g D fzg and
hence the family of compact sets fclZ .Unz / \ K W n 2 !g has an empty intersection.
This implies that there exists m.z/ 2 ! for which clZ .Um.z/
z
/ \ K D ; and hence
S z
Um.z/ ZnK. Consequently, ZnK D fUm.z/ W z 2 ZnKg. The space ZnK
z
being Lindelf, there is a countable subfamily of the family fUm.z/

z
W z 2 ZnKg
which covers ZnK. Thus ZnK 2 V which shows that jKj D jfZnK W K 2 Kgj
jVj c. Fact 3 is proved. t
u
Fact 4. If Z is a hereditarily Lindelf space, then there exist A; B Z with the

(1) A [ B D Z;
(2) all compact subspaces of A are countable;
(3) all compact subspaces of B are also countable.
Proof. Let K be the family of all uncountable compact subsets of Z. By Fact 3, we
can enumerate K (possibly with repetitions) as K D fK W < cg. Observe also
that every K 2 K has cardinality c. Indeed, jKj jZj c (see Problem 015).
On the other hand, if jKj < c, then K cannot be mapped continuously (and even
discontinuously) onto I because jIj D c. Therefore K is scattered by Problem 133.
Now apply Problem 128 to conclude that the !-modification L of the space K is
Lindelf. However, L is discrete because .K/ .Z/ D !. Since a discrete
space is Lindelf only if it is countable, we have jKj D jLj ! and hence K K
Now, choose arbitrarily distinct points x0 ; y0 2 K0 ; assume that < c and we
have chosen sets A D fx W < g and B D fy W < g such that
(a) A \ B D ;;
(b) fx ; y g K for all < .
Since jK j D c, we can choose distinct points x ; y 2 K n.A [ B /. It is
immediate that the sets fx W g and fy W g still satisfy (a) and (b), so
our inductive construction can be continued to furnish us sets P D fx W < cg
and Q D fy W < cg with the properties (a) and (b) fulfilled for all < c. We
claim that A D P and B D ZnP are as promised.
Indeed, if K is an uncountable compact subspace of A, then K D K for some
< c and hence y 2 Q \ K B \ K which contradicts K A and A \ B D ;.
Analogously, if K B, then x 2 P \ K D A \ K which contradicts K B
and A \ B D ;. Therefore all properties (1)(3) are fulfilled for A and B and hence
Fact 4 is proved. t
u
Returning to our solution, observe that if nw.X / D !, then nw.Y / D ! for
every Y X and hence every Y X is a Lindelf -space by Problem 228.
Now assume that every subspace of X is a Lindelf -space. By Fact 4, there are
A; B X such that X D A [ B, and if K is a compact subset contained either in A
or in B, then K is countable. If we prove that nw.A/ D nw.B/ D !, then evidently,
nw.X / D !, so we can carry out our proof for X D A and X D B which means
that we can assume, without loss of generality, that all compact subspaces of X are
countable.
Let C be a compact cover of X such that there is a countable family F of closed
subsets of X which is a network with respect to C. For any x 2 X , let Fx D fF 2
F W x 2 F g; for any x; y 2 X , let x y if Fx D Fy . It is evident that
is an equivalence relationship on X ; denote by T the set of its equivalence classes.
Clearly, T 2 T if and only if, there is x 2 X such that T D Tx D fy 2 X W y xg.
Given any point x 2 X take any C 2 C with x 2 C and TobserveTthat the family
FC D fF 2 F W C F g is contained T in F x and hence Fx FC . Besides,
T
FC is a network at C and therefore FC D C which shows that Fx D Fx is
compact being a closed subset of a compact set C . Besides, Tx Fx for any x 2 X
and hence Tx is countable because all compact subsets of X are countable. Thus
we can choose an enumeration (possibly with repetitions) fxTn W n 2 !g of every
T 2T.
For any nS2 ! let Xn D fxTn W T 2 T g; it is clear that Xn \ T D fxTn g for every
T 2 T and fXn W n 2 !g D X . The family Fn D fF \ Xn W F 2 F g consists of
closed subsets of Xn for all n 2 !; we claim that Fn is T0 -separating.
Indeed, given distinct x; y 2 Xn , there are distinct T; S 2 T such that x D xTn
and y D xSn . Since T D Tx S D Ty , we have Fx Fy . Assume first that there
is F 2 Fx nFy and hence x 2 F while y F ; then F \fx; yg D fxg. Analogously,
if there is F 2 Fy nFx , then F \ fx; yg D fyg. This proves that Fn is a countable
T0 -separating family in Xn for all n 2 !. The space X is hereditarily Lindelf, so
we can apply Fact 2 to conclude that there is a countable T1 -separating family of
closed subsets of Xn . Since Xn is also Lindelf , we have nw.Xn / D ! by Fact 1.
It is an easy exercise that a countable union of spaces with a countable network also
has a countable network so nw.X / D ! and hence our solution is complete.
T.271. Prove that every subspace of X is a Lindelf p-space if and only if X is
second countable.
Solution. If X is second countable, then every subspace of X is second countable
and hence each Y X is a Lindelf p-space by Problem 221. Now assume that
every Y X is a Lindelf p-space. Since every Lindelf p-space is a Lindelf -
space (see Problem 223), every subspace of X is a Lindelf -space. Therefore we
can apply Problem 270 to conclude that nw.X / ! and hence w.X / D nw.X / D
! by Problem 244.
T.272. Observe that there exist hereditarily Cech-complete non-metrizable spaces.
Therefore a hereditarily p-space need not be metrizable. Prove that any hereditarily
Cech-complete space is scattered.
Solution. We denote by D the set f0; 1g with the discrete topology. As usual, we
will identify any ordinal with the set of its predecessors and, in particular, n D
f0; : : : ; n 1g for any n 2 !. Recall that D0 D f;g; if k 2 !; i 2 D and s 2 Dk ,
then s _ i 2 DkC1 is defined by .s _ i /.k/ D i and .s _ i /jk D s. If we have s 2
Dn ; t 2 Dk where n k and n; k 2 ! C 1, then s t says that tjn D s. For
any k 2 ! and s 2 Dk , let s D fx 2 D! W s xg. Given a point x 2 D! , it is
immediate that the family fxjn W n 2 !g is a local base at x in the space D! .
S
choice of i , so we have consistently defined a function f which will be denoted

theS
Fact 1. For any non-scattered Cech-complete space Z, there exists a dense-in-itself
compact K Z. In other words, if Z is Cech-complete and every compact subspace
of Z is scattered, then Z is itself scattered.
Proof. The fact that Z is not scattered implies that there is a dense-in-itself subspace
Y Z. It is clear that Y is also dense-in-itself; besides, it is Cech-complete by
Problem 260 of [TFS]. This shows that we can assume, without loss of generality,
that Z D Y , i.e., Z is a Cech-complete space without isolated points. There exists
aTfamily fOn W n 2 !g .Z/ such that O0 D Z; OnC1 On for all n 2 ! and
fOn W n 2 !g D Z. Denote by C the family of all non-empty closed subspaces of
Z which have no isolated points.
Let F; D Z and assume that, for some n > 0, we have constructed families
fFs W s 2 Dk g C for all k < n with the following properties:
(1) for any k < n and s 2 Dk , we have clZ .Fs / Ok ;
(2) for any k < n and any distinct s; t 2 Dk , we have Fs \ Ft D ;;
(3) if k < m < n; s 2 Dk ; t 2 Dm and s t then Ft Fs .
Fix any s 2 Dn1 ; since Fs is dense-in-itself, it is infinite so we can take distinct
points x; y 2 Fs . There exist U 2 .x; Z/; V 2 .y; Z/ such that clZ .U / [
clZ .V / On and clZ .U / \ clZ .V / D ;.
The sets Fs _ 0 D U \ Fs and Fs _ 1 D V \ Fs are closed in Z and disjoint
(the bar denotes the closure in Z). Since Fs has no isolated points, neither U \ Fs
nor V \ Fs have isolated points and hence Fs _ i is dense-in-itself for each i 2 D.
After we construct the pair fFs _ 0 ; Fs _ 1 g for all s 2 Dn1 , we obtain the family
fFs W s 2 Dn g for which the properties (1) and (3) hold. The condition (2) is satisfied
because the sets Fs _ 0 and Fs _ 1 are disjoint for any s 2 Dn1 .
Consequently, we can construct the family Fk D fFs W s 2 Dk g for every k 2 !
in such a way that the conditions (1)(3) are satisfied
S Tfor every n 2 !; let Pk D
Fk for every k 2 !. Observe that the set F D fPk W k 2 !g is compact.
Indeed, it follows from (1) that clZ .Pk / Ok for each k 2 ! and therefore
\ \ \ \
Pk D .clZ .Pk / \ Z/ D fclZ .Pk / W k 2 !g \ Z D fclZ .Pk / W k 2 !g
k2! k2!
T T
because fclZ .Pk / W k 2 !g k2! Ok D Z [see (1)]. Thus F is compact being
the intersection of compact subspaces of Z. Besides, F ; because the family
fclZS.Pk / W k 2 !g is decreasing by (3). Given any x 2 F and k 2 !, we have
x 2 Fk and hence there is a unique sk 2 Dk such that x 2 FS sk [see (2)]. Observe
that sk skC1 for each k 2 ! by (2) and (3) so the point y D fsi W i 2 !g 2 D!
is well-defined; let f .x/ D y. Observe that it follows from our definition of f .x/
that
(4) if k 2 !; s 2 Dk and x 2 Fs \ F , then s f .x/.
We claim that the map f W F ! D! is continuous. To prove it take any x 2 F

and W 2 .f .x/; D! /. There is n 2 ! such that f .x/jn W . Let s D f .x/jn;
then U D Fs \ F is an open neighborhood of the point x in the space F because
it follows from (2) that Fs \ F is open in F . Given any y 2 U , we have s f .y/
by (4) and hence f .y/jn D s D f .x/jn, i.e., y 2 f .x/jn . This proves that
f .U / f .x/jn W and therefore f is continuous at the point x.
The map f is surjective; indeed, given any y 2 D! , the family fFyjn W n 2 !g is
decreasing by (3) and therefore
!
\ \ \ \
Fyjn D .clZ .Fyjn / \ Z/ D clZ .Fyjn / \ Z D clZ .Fyjn /
n2! n2! n2! n2!
T T
because fclZ .Fyjn / W n 2 !g n2! On D Z by the property (1). Since the
familyTA D fclZ .Fyjn / W T
n 2 !g is decreasing and consists T of compact sets, we
have fFyjn W n 2 !g D A ;, and hence there is x 2 fFyjn W n 2 !g. It
is immediate that f .x/ D y; the point y was taken arbitrarily, so we proved that
f maps F continuously onto D! . If F is scattered, then every second countable
continuous image of F is countable by Problem 129 which is a contradiction. Thus
F is a non-scattered compact subspace of Z and hence there exists a closed dense-
in-itself subspace K F . It is clear that K is the desired compact subspace of Z,
u
Returning to our solution, observe that the space X D A.!1 / is a non-metrizable
compact space while every Z X is either compact (if it contains the unique non-
isolated point of X ) or discrete (if it does not contain the non-isolated point of X ).
Since both compact and discrete spaces are Cech-complete, every subspace of X is
Cech-complete. Every Cech-complete space is a p-space, so X is an example of a
non-metrizable hereditarily p-space.
Now assume that X is a non-scattered hereditarily Cech-complete space. By
Fact 1 there is a non-scattered compact K X . This implies that there is a
continuous onto map f W K ! I (see Problem 133). The map f being perfect,
there is a closed F K such that f .F / D I and g D f jF is irreducible (see
Problem 366 of [TFS]). Let A be a countable dense subset of I; for any a 2 A,
choose a point xa 2 g 1 .a/ and let Y D fxa W a 2 Ag.
Observe first that F has no isolated points. Indeed, if x is isolated in F , then F1 D
F nfxg is closed in F and hence g.F1 / is closed in I. The map g being irreducible,
we have g.F1 / I and therefore Ing.F1 / D fg.x/g. However, a complement
of any point of I is dense in I. Thus g.F1 / is closed and dense in I, so we have
g.F1 / D I which is a contradiction showing that F is dense-in-itself.
Observe also that A g.Y / and hence g.Y / is dense in I. The map g being
closed, we have g.Y / D g.Y / D I. Since g is irreducible, we have Y D F , i.e., Y
is dense in F and therefore Y does not have isolated points. This implies that fyg
is nowhere dense in Y for any y 2 Y . Since Y is countable, it turns out that Y is a
countable union of its nowhere dense sets, i.e., Y does not have the Baire property.
This implies that Y is a subspace of X that is not Cech-complete (see Problem 274
of [TFS]) which is a contradiction, making our solution complete.
T.273. Prove that !1 C 1 is a scattered compact space which is not hereditarily
Cech-complete.
Solution. We proved in Problem 314 of [TFS] that !1 C1 is compact. Every second
countable continuous image of !1 C 1 is countable by Fact 1 of S.319, so !1 C 1 is
scattered by Problem 133. The set L of limit ordinals of !1 is stationary by Problem
065 and hence there exist stationary sets A; B L such that L D A [ B and
A \ B D ; (see Problem 066). Denote by D the set of successor ordinals of !1 .
We claim that the subspace Y D .D [ A/ [ f!1 g of the space !1 C 1 is not Cech-
complete.
To prove it assume towards a contradiction that Y is Cech-complete. Since Y is
dense in .!1 C 1/, the set Z D .!1 C 1/nY D B is an F -set in .!1 C 1/ and
hence B is -compact. However, any compact K B is countable because the
stationary set A is disjoint from B and intersects any uncountable compact subset
of !1 C 1. This shows that B is a countable union of countable compact sets and
hence jBj D ! which is a contradiction (see Problem 065). Thus Y is a subspace of
!1 C 1 which is not Cech-complete.
T.274. Prove that every subspace of X is -compact if and only if X is countable.
Solution. If the space X is countable, then every subspace of X is -compact being
countable. Now assume that every subspace of X is -compact. Every -compact
space is Lindelf by Problem 226, so every subspace of X is a Lindelf -space
and hence nw.X / D ! by Problem 270. This implies hl.X / D ! and therefore
there exist A; B X such that A [ B D X and any compact K X is countable
if it is contained either in A or in B (see Fact 4 of T.270). By our assumption about
X , both sets A and B are -compact and hence countable. Therefore X is also
countable.
T.275. Prove that
(i) if an uncountable regular cardinal is a caliber of a space X , then is a
precaliber of X ;
(ii) if an infinite successor cardinal is a precaliber of a space X , then c.X / < .
In particular, if !1 is a precaliber of X , then c.X / D !.
Solution.
(i) Take a familyTU D fU W < g .X /. There exists a set A such that
jAj D and fU W 2 Ag ;. It is evident that the family fU W 2 Ag is
centered and hence is a precaliber of X .
(ii) Assume that D C and c.X / ; if jUj < for any disjoint U .X /,
then jUj for any such family U and hence c.X / , a contradiction.
Therefore there exists a disjoint U D fU W < g .X /. It is clear that no
non-empty subfamily of U is centered and therefore U witnesses that is not a

precaliber of X which is again a contradiction proving that c.X / < .
T.276. Let be an uncountable regular cardinal. Prove that if Sis a precaliber
(caliber) of Xn for every n 2 !, then is a precaliber (caliber) of fXn W n 2 !g.
Solution.
S Assume that is a precaliber (caliber) of Xn for every n 2 !. Let X D
fXn W n 2 !g and take any family U S D fU W < g .X /. If An D f < W
U \ Xn ;g for every n 2 !, then fAn W n 2 !g D and therefore jAn j D
for some n 2 !; let ' W ! An be a bijection.
By our choice of the set An , if V D U'./ \ Xn for each < , then the
family fV W < g consists of non-empty open subsets of Xn . Thus there exists
B such that jBj D and the family fV W 2 Bg D fV' 1 ./ W 2 '.B/g is
centered (has non-empty intersection). Since V' 1 ./ U for each 2 A D '.B/,
the family fU W 2 Ag is also centered (has non-empty intersection); besides,
jAj D jBj D which shows that is a precaliber (caliber) of X .
T.277. Let be an uncountable regular cardinal. Prove that if is a precaliber
(caliber) of X , then is a precaliber (caliber) of every continuous image of X .
Solution. Assume that is a precaliber (caliber) of the space X . Let f W X ! Y
be a continuous onto map. Given a family fU W < g .Y /, the family
ff 1 .U / W < g consists of non-empty open subsets of X and hence there is
A such that jAj D and the family ff 1 .U / W 2 Ag is centered (has non-
empty intersection). It is immediate that the family fU W 2 Ag is also centered (or
has non-empty intersection respectively) which proves that is a precaliber (caliber)
of the space Y .
T.278. Suppose that is an uncountable regular cardinal and Y is a dense subspace
of X . Prove that
(i) if is a caliber of Y , then it is a caliber of X ;
(ii) is a precaliber of Y if and only if it is a precaliber of X .
Solution.
(i) If fU W < g .X /, then fU \ Y W < g .Y / because the
T Y is dense in X . Therefore there is a set A such
subspace T that jAj D and
P D TfU \ Y W 2 Ag ;. It is evident that P fU W 2 Ag and
hence fU W 2 Ag ; which proves that is a caliber of X .
(ii) Assume first that is a precaliber of Y . If fU W < g .X /, then
fU \ Y W < g .Y / because Y is dense in X . Therefore there is a set
A such that jAj D and the family fU \ Y W 2 Ag is centered. It is
evident that fU W 2 Ag is also centered which proves that is a precaliber
of X .
Now let be a precaliber of X and take any family fV W < g .Y /. Take
U 2 .X/ such that U \Y D V for each < . Then fU W < g .X / and
therefore there exists A for which the family fU W 2 Ag is centered. Given
1 ; : : : ; n 2 A, the set U D U1 \

\ Un is non-empty and hence U \ Y ;
because Y is dense in X . Since V1 \

\ Vn D U \ Y ; and the ordinals
1 ; : : : n 2 A were chosen arbitrarily, we proved that the family fV W 2 Ag is
centered and hence is a precaliber of Y .
T.279. Show that an uncountable regular cardinal is a caliber of a compact space
X if and only if it is a precaliber of X .
Solution. Any caliber of X is also a precaliber of X by Problem 275, so assume
that is a precaliber of X and take a family fU W < g .X /. By regularity of
X , there is a family fV W < g .X / such that V U for all < . There
is A such that jAj D and the family fV W 2 Ag is centered. Therefore T
the family F D fV W 2 Ag is centered as well T and hence there is x 2 F
because X is compact. It is immediate that x 2 fU W 2 Ag and hence is a
caliber of X .
T.280. Let be an uncountable regular cardinal.
Q Prove that if is a precaliber of
Xt for every t 2 T , then is a precaliber of fXt W t 2 T g.
Q
Solution. If Xs D ; for some s 2 T , then X D t 2T Xt D ; and there is nothing
to prove, so we assume thatQ Xt ; for all t 2 T . A set U X is called a standard
open subset of X if U D t 2T Ut where Ut 2 .Xt / for all t 2 T and the set
supp.U / D ft 2 T W Ut Xt g is finite. Standard open subsets of X form a base of
X (see Problem 101 of [TFS]).
Fact 1. An uncountable regular cardinal is a precaliber of a space Z if and only
if for any family fUa W a 2 Ag .Z/ indexed by a set A with jAj D , there is a
set B A such that jBj D and the family fUa W a 2 Bg is centered.
Proof. Sufficiency is clear, so assume that is a precaliber of Z; given a set A with
jAj D and a family fUa W a 2 Ag .Z/ fix any bijection ' W ! A. If V D
U'./ for any < , then fV W < g .Z/, and hence there is a set C
such that jC j D and the family fV W 2 C g is centered. If B D '.C /, then
B A; jBj D and the family fUa W a 2 Bg D fU'./ W 2 C g D fV W 2 C g
is centered. Fact 1 is proved. t
u
To prove that precalibers are preserved by any products, we have to show first
that they are preserved by finite ones. To that end, let us prove that
(1) given spaces Y and Z if is a precaliber of both Y and Z, then is a precaliber
of Y Z.
Take any set A such that jAj D and a family U D fUa W a 2 Ag .Y Z/.
For any a 2 A there exist V 2 .Y / and W 2 .Z/ such that U V W .
Since is a precaliber of Y , by Fact 1, there is a set A1 A such that jA1 j D and
the family fVa W a 2 A1 g is centered; since is a precaliber of Z, we can choose a
set A2 A1 such that jA2 j D and the family W D fWa W a 2 A2 g is also centered
(here we used Fact 1 again).
We claim that the family U 0 D fUa W a 2 A2 g is centered as well. To see this,

take any Ua1 ; : : : ; Uan 2 U 0 ; since fVa W a 2 A2 g is centered, we can find a point
y 2 Va1 \

\Van . The family W being centered, there is a point z 2 Wa1 \

\Wan .
Hence w D .y; z/ 2 .Va1 Wa1 / \

\ .Van Wan / Ua1 \

\ Uan , so the
family U 0 is centered. This shows that is a precaliber of Y Z, i.e., the property
(1) is proved. A trivial induction shows that
(2) given spaces Y1 ; : : : ; Yn , if is a precaliber of Yi for all i D 1; : : : ; n, then is
a precaliber of Y1

Yn .
Now take any set A with jAj D and a family fUa W a 2 Ag .X /. For any
a 2 A there is a standard open set Va Ua ; let Pa D supp.Va / and fix a point va 2
Va . By the -lemma (see Problem 038), there is a finite set S D fs1 ; : : : ; sn g T
and B A such that jBj D and Pa \ Pb D S for any distinct a;Q b 2 B. The
family fpS .Va / W a 2 Bg consists of non-empty open subsets of XS D s2S Xs ; the
cardinal being a precaliber of XS by (2), there exists C B such that jC j D
and the family fPS .Va / W a 2 C g is centered (we again applied Fact 1 here). To
see that fUa W a 2 C g is centered take any a1 ; : : : ; am 2 C . There is a point
y 2 PS .Va1 / \

pS .Vam /. Since the family fPai nS W i D 1; : : : ; mg is disjoint, we
can choose a point x 2 X such that xjS D y and x.t/ D vai .t/ for any t 2 Pai nS
and i D 1; : : : ; m. It is immediate that x 2 Va1 \

\ Vam Ua1 \

\ Uam and
therefore Ua1 \

\ Uam ;. This proves that the family fUa W a 2 C g is centered
and hence is a precaliber of X .
T.281. Let be an uncountable regular Q cardinal. Prove that if is a caliber of Xt
for every t 2 T , then is a caliber of fXt W t 2 T g.
Q
Solution. If Xs D ; for some s 2 T , then X D t 2T Xt D ; and there is nothing
to prove so we assume that Q Xt ; for all t 2 T . A set U X is called a standard
open subset of X if U D t 2T Ut where Ut 2 .Xt / for all t 2 T and the set
supp.U / D ft 2 T W Ut Xt g is finite. Standard open subsets of X form a base of
X (see Problem 101 of [TFS]).
Fact 1. An uncountable regular cardinal is a caliber of a space Z if and only if
for any family fUa W a 2 Ag T
.Z/ indexed by a set A with jAj D , there is a
set B A such that jBj D and fUa W a 2 Bg ;.
Proof. Sufficiency is clear so assume that is a caliber of Z; given a set A with
jAj D and a family fUa W a 2 Ag .Z/, fix any bijection ' W ! A. If V D

T fV W < g .Z/, and hence there is a set C
U'./ for any < , then
suchTthat jC j D and T fV W 2 C g ;. If T
B D '.C /, then B A; jBj D
and fUa W a 2 Bg D fU'./ W 2 C g D fV W 2 C g ;, so Fact 1 is
proved. t
u
To prove that calibers are preserved by any products, we have to show first that
they are preserved by finite ones. To that end let us prove that
(1) given spaces Y and Z if is a caliber of both Y and Z, then is a caliber of
Y Z.
Take any set A such that jAj D and a family U D fUa W a 2 Ag .Y Z/.
For any a 2 A there exist V 2 .Y / and W 2 .Z/ such that U V W .
Since is a caliber of Y , by Fact 1, there is a set A1 A such that jA1 j D and
T
fVa W a 2 A1 g T;; since is a caliber of Z, we can choose a set A2 A1 such
that jA2 j D and fWa W a T 2 A2 g ; (here we used Fact T 1 again).
Choose T any points y 2 fV T
a W a 2 A 2 g and z 2 fWTa W a 2 A1 g. Then
.y; z/ 2 fVa Wa W a 2 A2 g fUa W a 2 A2 g and hence fUa W a 2 A2 g ;.
This shows that is a caliber of Y Z, i.e., the property (1) is proved. A trivial
induction shows that
(2) given spaces Y1 ; : : : ; Yn , if is a caliber of Yi for all i D 1; : : : ; n, then is a
caliber of Y1

Yn .
Now take any set A with jAj D and a family fUa W a 2 Ag .X /. For any
a 2 A there is a standard open set Va Ua ; let Pa D supp.Va / and fix a point va 2
Va . By the -lemma (see Problem 038), there is a finite set S D fs1 ; : : : ; sn g T
and B A such that jBj D and Pa \ Pb D S for any distinct a; b 2 QB. The
family fpS .Va / W a 2 Bg consists of non-empty open subsets of XS D s2S Xs ;
the cardinal
T being a caliber of XS by (2), there exists C B such that jC j D
and fPS .Va / W a 2 C g ; (we again T applied Fact 1 here). In fact, the set C
witnesses
T that is a caliber of X , i.e., fUa W a 2 C g ;. Indeed, take any point
y 2 fPS .Va / W a 2 C g. Since the family fPa nS W a 2 C g is disjoint, we can
choose a point x 2 X such that xjS T D y and x.t/ D Tva .t/ for any t 2 Pa nS and
a
T 2 C . It is immediate that x 2 fV a W a 2 C g fUa W a 2 C g and therefore
fUa W a 2 C g ; which proves that is a caliber of X .
T.282. Prove that any product X of separable spaces satisfies the Shanin condition,
i.e., every uncountable regular cardinal is a caliber of X .
Solution. Take any uncountable regular cardinal . By our hypothesis we have X D
Q
fXt W t 2 T g where each Xt is separable; fix a countable dense Dt Xt for
every t 2 T . Take any t 2 T and a family fU W < g .Xt /. Let Pd D
f <S W d 2 U g for every d 2 Dt . Since U \ Dt ; for each < , we
have fPd W d 2 Dt g D . The set Dt being countable,
T there is d 2 Dt such that
A D Pd has cardinality . As a consequence d 2 fU W 2 Ag, which, together
with jAj D , proves that is a caliber of Xt for all t 2 T . Finally apply Problem
281 to conclude that is a caliber of X .
T.283. Prove that any uncountable regular cardinal is a precaliber of Cp .X / for
any space X .
Solution. Any uncountable regular cardinal is a caliber of RX by Problem 282.
Since Cp .X / is dense in RX , we can apply Problems 275 and 278 to conclude that
is a precaliber of Cp .X /.
T.284. Prove that there exists a space X such that !1 is a precaliber of X while the
point-finite cellularity of X is uncountable. Observe that if !1 is a caliber of X , then
p.X / D !.
Solution. It is evident that !1 is a caliber of X if and only if any point-countable

family of non-empty open subsets of X is countable. This clearly implies that
any point-finite family of non-empty open subsets of X is also countable, i.e.,
p.X / !.
Now, if X D Cp .A.!1 //, then !1 is a precaliber of X by Problem 283.
Furthermore, A.!1 / embeds in Cp .Cp .A.!1 /// D Cp .X / which shows that
p.X / !1 by Problem 178 of [TFS].
T.285. Let X be a metrizable space. Prove that any regular uncountable cardinal is
a caliber of Cp .X /.
Solution. For any number n 2 N consider the set Mn D f1; : : : ; ng; we will also
need the family O D f.a; b/ W a; b 2 Q and a < bg. Given any points x1 ; : : : ; xn 2
X and sets O1 ; : : : ; On 2 O, let x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2
Oi for all i 2 Mn g. The family B D fx1 ; : : : ; xn I O1 ; : : : ; On W n 2 N; xi 2 X
and Oi 2 O for all i 2 Mn g is a base in the space Cp .X /. It is straightforward to
see that the family C D fx1 ; : : : ; xn I O1 ; : : : ; On 2 B W xi xj and Oi \ Oj D ;
for any distinct i; j 2 Mn g is a -base in Cp .X /, i.e., any non-empty open subset
of Cp .X / contains an element of C.
Fact 1. Let Z be a metrizable space. If is not a countably cofinal cardinal and
w.Z/ , then there is a closed discrete D Z such that jDj D .
S
Proof. Let U D fUn W n 2 !g .Z/ be a base in Z such that Un is discrete
for all n 2 !; it is clear that jUj . Since is not !-cofinal, we have jUn j
for some n 2 !. Now take a point z.U / 2 U for every U 2 Un ; then the set
D 0 D fz.U / W U 2 Un g is closed and discrete. Since jD 0 j , we can choose a set
D D 0 with jDj D . It is evident that D is as promised so Fact 1 is proved. u t
Fact 2. For any space Z if is a regular uncountable cardinal such that d.Z/ < ,
then is a caliber of Z.
Proof. Fix a set D Z such that D D Z and jDj D D d.Z/. If we are given a
family fU W < g .Z/, then
S let Ed D f < W d 2 U g for every d 2 D.
Since D is dense in Z, we have fEd W d 2 Dg D . TThe cardinal > being
regular there is d 2 D such that jEd j D . Since d 2 fU W 2 Ed g, the set Ed
witnesses that is a caliber of Z. Fact 2 is proved. t
u
Returning to our solution, fix a metric d on X which generates .X/ and take a
regular uncountable cardinal . It is an easy exercise that
. / if D is a discrete subset of X with jDj D , then there is E D such that E
is closed (and, obviously, discrete) in X and jEj D .
Given a family fUa W a 2 Ag .Cp .X // such that jAj D , we can choose
a set Va D x1a ; : : : ; xn.a/
a
I O1a ; : : : ; On.a/
a
2 C such that Va Ua ; let supp.Va / D
fx1 ; : : : ; xn.a/ g for all a 2 A. Since O is countable, there exists a set B A, a
a a
number n 2 N and O1 ; : : : ; On 2 O such that jBj D ; n.a/ D n and Oia D Oi for

any a 2 B and i 2 Mn .
Now we can apply the Delta-lemma (see Problem 038) to find a finite set
P D fp1 ; : : : ; pm g X and B1 B such that jB1 j D and supp.Va / \
supp.Vb / D P for any distinct a; b 2 B1 (it possible that m D 0, i.e.,
P D ;). Since the order of points and open sets in the definition of Va does
not influence the definition of Va and there are only countably many possibilities
of assigning an element of O to a point of P , we can consider that Va D
p1 ; : : : ; pm ; x1a ; : : : ; xnm
a
I G1 ; : : : ; Gm ; O1 ; : : : ; Onm for each a 2 B1 .
Given any E B1 we will need the set Ri .E/ D fxia W a 2 Eg for all i 2 Mnm ;
then the family fRi .E/ W i 2 Mnm g is disjoint for any E B1 . Consider the set
H D fi 2 Mnm W w.Ri .B1 // < g; it follows from Fact 1 and . / that for each
i 2 Mnm nH , we can find a set Di Ri .B1 / of cardinality which is discrete
and closed in X . Observe also that w.Y / D w.Y / for any Y X and hence
jRk .B1 / \ Di j < for any k 2 H and i 2 Mnm nH . Therefore we can choose a
set B2 B1 and " > 0 with the following properties:
(1) d.x; y/ > " for any x 2 P [ Rk .B2 / and y 2 Ri .B2 / whenever k 2 H and
i 2 Mnm nH ;
(2) jB2 j D and Ri .B2 / is closed and discrete in X for any i 2 Mnm nH .
S
Let Y D P [ . fRi .B2 / W i 2 H g/; then w.Y / < . Choose i1 ; : : : ; il 2 Mnm
such that H D fi1 ; : : : ; il g and consider the set Wa D ff 2 Cp .Y / W f .pi / 2 Gi
for all i 2 Mm and f .xiaj / 2 Oij for all j 2 Ml g for all a 2 B2 . It is evident
that fWa W a 2 B2 g .Cp .Y //. Since d.Cp .Y // nw.Cp .Y // D nw.Y / D
w.Y / < , the cardinal is a caliber of Cp .Y / by Fact T 2. Consequently, there is
B3 B2 and f 2 Cp .Y / such that jB3 j D and f 2 S fWa W a 2 B3 g.
The set Y is closedS in X and hence so is Y [ . fRi .B2 / W i 2 Mnm nH g/.
Since the set F D fRi .B2 / W i 2 Mnm nH g is closed, discrete and disjoint
from Y , there exists g 2 C.Y [ F / such that gjY D f and g.x/ 2 Oi for every
i 2 Mnm nH and x 2 Ri .B2 /. The set Y [ F closed and hence C -embedded in
the normal space X , T so there is a functionTh 2 Cp .X / with hj.Y [ F / D g. It is
immediate that h 2 fVa W a 2 B3 g fUa W a 2 B3 g which proves that is
a caliber of Cp .X /. Since an uncountable regular cardinal was taken arbitrarily,
we proved that every uncountable regular cardinal is a caliber of Cp .X /, i.e., our
T.286. Prove that an uncountable regular cardinal is a precaliber of X if and only
if it is a caliber of X .
Solution. If is a caliber of X , then it is a precaliber of X by Problem 278. Now,
if is a precaliber of X , then it is also a precaliber of X by Problem 278. Therefore
is a caliber of X by Problem 279.
T.287. Suppose that X is a compact space of countable tightness. Prove that if !1 is
a caliber of X , then X is separable. Give an example of a non-separable compact
space X such that !1 is a caliber of X .
Solution. It follows from t.X / D ! that the space X has a point-countable -base
B (see Problem 332 of [TFS]). Since !1 is a caliber of X , the family B must be
countable and therefore d.X / w.X / D !.
Now if D 2c , then Y D I is a compact space such that !1 is a caliber of Y
by Problem 282. However, Y is not separable because otherwise w.Y / 2d.Y / D
2! D c while w.Y / D 2c (see Fact 2 and Fact 3 of S.368).
T.288. Assuming MAC:CH, prove that !1 is a precaliber of any space which has
the Souslin property.
Solution. Let Z be an arbitrary space such that c.Z/ D ! and take any family
fU W < !1 g .Z/. Consider the set P D .Z/ where U V iff U V .
It is clear that is a partial order on .Z/; it is immediate that U; V 2 P are
compatible if and only if U \ V ;. An immediate consequence is that P is ccc
because c.Z/ D !.
If there is < !1 such thatTthe set A./ D f W U D U g is uncountable, then
for any x 2 U we have x 2 fU W 2 A./g and hence A./ witnesses that !1
is a caliber of Z. Now if A./ is countable for any 2 !1 , then we can choose an
uncountable B !1 such that U U for any distinct ; 2 B. Observe also
that a set E P is centered in the sense of Problem 049 if and only if the family
E is centered as a family of subsets of Z. Therefore we can apply Problem 049 to
conclude that the uncountable family U D fU W 2 Bg contains an uncountable
centered subfamily, i.e., there is an uncountable C B for which fU W 2 C g is
centered. This proves that !1 is a precaliber of Z.
T.289. Assume the axiom of Jensen (}). Prove that there exists a space X with
c.X / D ! while !1 is not a precaliber of X .
Solution. Under the axiom of Jensen, there exists a hereditarily Lindelf non-
separable compact space X (see Problem 073). Therefore c.X / hl.X / D !;
furthermore, t.X / .X / D .X / D !, so if !1 is a precaliber of X , then it is a
caliber of X by Problem 279 and hence X is separable by Problem 287 which is a
contradiction. Thus c.X / D ! while !1 is not a precaliber of X .
T.290. Prove that for any uncountable regular cardinal , the diagonal of Cp .X / is
-small if and only if is a caliber of X . In particular, !1 is a caliber of X if and
only if Cp .X / has a small diagonal.
Solution. For any n 2 N, denote by Mn the set f1; : : : ; ng; given x1 ; : : : ; xn 2 X
and " > 0, let O.x1 ; : : : ; xn ; "/ D ff 2 Cp .X / W jf .xi /j < " for all i 2 Mn g. The
function u 2 Cp .X / is defined by u.x/ D 0 for all x 2 X ; it is evident that the
family fO.x1 ; : : : ; xn ; "/ W n 2 N; x1 ; : : : ; xn 2 X and " > 0g is a local base of
Cp .X / at the point u. Let D f.f; f / W f 2 Cp .X /g Cp .X / Cp .X / be the
diagonal of the space Cp .X /.
To prove necessity assume, towards a contradiction, that the diagonal of
Cp .X / is -small while is not a caliber of the space X , i.e., there exists a family
U D fU W < g .X / such that
T
. / fU W 2 Ag D ; for any A with jAj D .
Take any point x 2 U and a function f 2 Cp .X; 0; 1 / such that f .x / D 1
and f .X nU / f0g for all < . Let P D f < W f .x / D 1g; it is
immediate from . / that jP j < for every < . Using regularity of , it is easy
to construct by transfinite induction a set A such that jAj D and f .x / 1
(which, evidently, implies f f ) for any distinct ; 2 A. For the set F D
ff W 2 Ag we claim that
(1) jf 2 A W f W gj < for any W 2 .u; Cp .X //.
Indeed, W1 D O.x1 ; : : : ; xn ; "/ W for some x1 ; : : : ; xn 2 X and " > 0. Each
xi belongs to strictly less than elements of U by . / and hence the cardinality of
the set B D f 2 A W f .xi / 0 for some i 2 Mn g is strictly less than . Since
f 2 A W f W g f 2 A W f W1 g B, the property (1) is proved.
It follows from the choice of the set A that F is faithfully indexed and hence
jF j D . Therefore the set H D f.f ; u/ W 2 Ag .Cp .X / Cp .X //n also
has cardinality . Given any set O 2 .; Cp .X / Cp .X //, we have .u; u/ 2 O
and hence there is W 2 .u; Cp .X // such that W W O. It follows from
(1) that there exists a set B A with jBj < such that f 2 W and hence
.f ; u/ 2 W W O for all 2 AnB. This proves that jH nOj jBj < ; since
the set O 2 .; Cp .X /Cp .X // was chosen arbitrarily, the set H witnesses that
is not -small which is a contradiction. Thus the family U D fU W < g .X /
with the property . / cannot exist and hence we established necessity.
Now assume that the cardinal is a caliber of X and the diagonal of Cp .X /
is not -small, i.e., there exists a set P D f.g ; h / W < g .Cp .X //2 n
such that .g ; h / .g ; h / for distinct ; < and jP nOj < for any set
O 2 .; Cp .X / Cp .X //. Let f D g h for every < ; it turns out that
(2) jf 2 W f W gj < for any W 2 .u; Cp .X //.
Indeed, O D f.f; g/ 2 .Cp .X //2 W f g 2 W g 2 .; Cp .X / Cp .X // and
hence jP nOj < which shows that jf < W g h D f W gj < and
therefore (2) holds for the set F D ff W < g.
Given any n 2 N, let Un D f1 .Rn n1 ; n1 / for all < . For each < ,
n./
we have f u, and hence there is n./ 2 N such that U ;. The cardinal
being regular, there is n 2 N and A such that jAj D and n./ D n for any
2 A. As a consequence, U D fUn W 2 Ag .X / and jAj D .
Furthermore, given x 2 X , we have W D O.x; n1 / 2 .u; Cp .X //, so it follows
from (2) that the set f 2 A W x 2 Un g f 2 A W jf .x/j n1 g f W f W g
has cardinality < . The point x 2 X was chosen arbitrarily, so any point of X
belongs to strictlyTless than -many elements of U. Thus for any B A with
jBj D , we have fUn W 2 Bg D ; and hence the family U witnesses that is
not a caliber of X which is a contradiction. This settles sufficiency and makes our
solution complete.
T.291. Prove that an uncountable regular cardinal is a caliber of X if and only if

it is a caliber of Cp .Cp .X //.
Solution. Say that a space Z has a property P if is a caliber of Z. It follows from
Problems 277, 281 and 282 that P is a complete property, i.e.,
(1) any metrizable compact space has P;
(2) if n 2 N and Zi has P for all i D 1; : : : ; n, then Z1

Zn has P;
(3) if Z has P, then every continuous image of Z has P.
Assume that is a caliber of X , i.e., X has P. For any x 2 X , let ex .f / D f .x/
for every f 2 Cp .X /. Then ex 2 Cp .Cp .X // for all x 2 X and the map x ! ex
is a homeomorphism of X onto the subspace E D fex W x 2 X g of the space
Cp .Cp .X // (see Problem 167 of [TFS]).
Since E is homeomorphic to X , the cardinal is a caliber of E. There exists an
algebra R.E/ in the space Cp .Cp .X // such that E R.E/ (see Fact 1 of S.312);
since P is a complete property, the space R.E/ has -P by Fact 2 of S.312 and
hence R.E/ has P because P is preserved by countable unions (see Problem 276).
It is easy to see that E (and hence R.E/) separates the points of Cp .X / and therefore
R.E/ is dense in Cp .Cp .X // by Problem 192 of [TFS]. Finally apply Problem 278
to conclude that Cp .Cp .X // has P, i.e., is a caliber of Cp .Cp .X //. This proves
necessity.
Now, if the cardinal is a caliber of Cp .Cp .X //, then the diagonal of the
space Cp .Cp .Cp .X /// is -small by Problem 290. It is immediate that having a
-small diagonal is a hereditary property, so Cp .X / has a -small diagonal being
homeomorphic to a subspace of Cp .Cp .Cp .X /// (see Problem 167 of [TFS]).
Finally, apply Problem 290 again to conclude that is a caliber of X ; this settles
sufficiency and makes our solution complete.
T.292. Suppose that an uncountable regular cardinal is a caliber of Cp .X /. Prove
that for any Y X the cardinal is a caliber of Cp .Y /.
Solution. Let W Cp .X / ! Cp .Y / be the restriction map, i.e., .f / D f jY for
all f 2 Cp .X /. Then is continuous and Z D .Cp .X // is dense in Cp .Y / (see
Problem 152 of [TFS]). Since Z is a continuous image of Cp .X /, the cardinal is a
caliber of Z (see Problem 277). Finally, is a caliber of Cp .Y / because Z is dense
in Cp .Y / (see Problem 278).
T.293. Let be an uncountable regular cardinal. Prove that if is a caliber of
Cp .X /, then the diagonal of X is -small. In particular, if !1 is a caliber of Cp .X /,
then the diagonal of X is small.
Solution. Since is a caliber of Cp .X /, the diagonal of the space Cp .Cp .X // is
-small by Problem 290. It is an easy exercise to see that having a -small diagonal
is a hereditary property, so X has a -small diagonal because it is homeomorphic to
a subspace of Cp .Cp .X // by Problem 167 of [TFS].
T.294. Let be an uncountable regular cardinal. Prove that if all finite powers of
X are Lindelf and X has a -small diagonal, then is a caliber of Cp .X /. As a
consequence, if l .X / D !, then X has a -small diagonal if and only if is a
caliber of Cp .X /. In particular, if X is compact, then the diagonal of X is small if
and only if !1 is a caliber of Cp .X /.
Solution. For any number n 2 N consider the set Mn D f1; : : : ; ng; we will also
need the family O D f.a; b/ W a; b 2 Q and a < bg. Given points x1 ; : : : ; xn 2 X
and sets O1 ; : : : ; On 2 O, let x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2
Oi for all i 2 Mn g. The family B D fx1 ; : : : ; xn I O1 ; : : : ; On W n 2 N; xi 2 X
and Oi 2 O for all i 2 Mn g is a base in the space Cp .X /. It is straightforward to
see that the family C D fx1 ; : : : ; xn I O1 ; : : : ; On 2 B W xi xj and Oi \ Oj D ;
for any distinct i; j 2 Mn g is a -base in Cp .X /, i.e., any non-empty open subset
of Cp .X / contains an element of C.
If Z is a space and n 2, let nij .Z/ S D fz D .z1 ; : : : ; zn / 2 Z n W zi D zj g for
any distinct i; j 2 Mn . The set n .Z/ D fnij .Z/ W 1 i < j ng is called the
n-diagonal of the space Z.
Given a space Z and a set F Z, say that F is -small if for any set A ZnF
with jAj D , there is U 2 .F; Z/ such that jAnU j D . Evidently, F is -small
if and only if for any A ZnF with jAj D , there is B A such that jBj D
and B \ F D ;.
Fact 1.
(i) given a space Z any finite union of -small subsets of Z is a -small subset of
Z;
(ii) if f W Z ! Y is a continuous mapping and F is a -small subset of Y , then
G D f 1 .F / is a -small subset of Z.
Proof. (i) Suppose that Fi isSa -small subset of Z for any i 2 Mn ; we must
prove that the set F D fFi W i 2 Mn g is -small. Take any A ZnF
with jAj D ; since F1 is -small, there is B1 A such that jB1 j D and
B 1 \ F1 D ;. Assume that 1 j < n and we have a set Bj A such that
(1) jBj j D and B j \ Fi D ; for all i j .
Since Fj C1 is -small, there is Bj C1 Bj such that jBj C1 j D and B j C1 \
Fj C1 D ;. It is evident that (1) still holds for the set Bj C1 , so we can continue
this inductive construction to obtain a set B D Bn A such that jBn j D and
B n \ Fi D ; for all i 2 Mn . This implies B \ F D ;, i.e., B witnesses that F
is -small and hence (i) is proved.
(ii) If A ZnG, then f .A/ Y nF . If jf .A/j < , then by regularity of ,
there is some y 2 f .A/ such that B D f 1 .y/ \ A has cardinality . Since
B f 1 .y/ ZnG, the set B witnesses that G is -small. If, on the other
hand, we have jf .A/j D , then there is C f .A/ such that jC j D and
C \ F D ;. It is immediate that for the set B D f 1 .C / \ A, we have jBj D
and B \ G D ;, so (ii) is settled and Fact 1 is proved.
t
u
Fact 2. If the diagonal of a space Z is -small, then the set n .Z/ is -small in Z n
for all n 2 N; n 2.
Proof. Let D f.z; z/ W z 2 Zg Z Z be the diagonal of the space Z. Let
qij W Z n ! Z Z be the natural projection onto the face defined by i and j , i.e.,
for any z D .z1 ; : : : ; zn / 2 Z n we have qij .z/ D .zi ; zj / 2 Z Z. It is clear that
nij .Z/ D qij1 ./ and therefore nij .Z/
S is n-small in Z for all distinct i; j 2 Mn
n
(see Fact 1). Consequently, n .Z/ D fij .Z/ W 1 i < j ng is a finite union
of -small sets, so it is -small by Fact 1. Fact 2 is proved. t
u
Returning to our solution, take any family fUa W a 2 Ag .Cp .X // such
that jAj D . We can choose a set Va D x1a ; : : : ; xn.a/ a
I O1a ; : : : ; On.a/
a
2 C such
that Va Ua for each a 2 A. Since O is countable, there exists a set B A, a
number n 2 N and O1 ; : : : ; On 2 O such that jBj D ; n.a/ D n and Oia D Oi
for any a 2 B and i 2 Mn . Let xa D .x1a ; : : : ; xna / for all a 2 B; it is evident that
P D fxa W a 2 Bg X n nn .X /. The set n .X / is -small by Fact 2 and therefore
there is D B such that jDj D and the closure of the set Q D fxa W a 2 Dg
does not intersect n .X /.
For any y D .y1 ; : : : ; yn / 2 Q fix a function hy 2 Cp .X / such that hy .yi / 2 Oi
for all i 2 Mn (such a function exists because yi yj whenever i; j 2 Mn ; i j
and hence we can apply Problem 034 of [TFS]). Let Wy D h1 1
y .O1 /

hy .On /;
it is clear that Wy 2 .y; X / for all y 2 Q. The space Q being Lindelf, there is
n
S
a countable R Q such that Q fWy W y 2 Rg. The cardinal being regular
there exists a set E D and y 2 R such that jEj D and xa 2 Wy for all
a 2 E. As a consequence, hy .xia / 2 Oi for all a 2 E and i 2 Mn which shows that
hy 2TVa Ua for all a 2 E. Thus we have found a set E A such that jEj D
and fUa W a 2 Eg ;. This proves that is a caliber of Cp .X /.
Finally, apply 293 together with our proved result to see that for a space X with
l .X / D ! the diagonal of X is -small if and only if is a caliber of Cp .X /.
T.295. Prove that any compact space of weight !1 with a small diagonal is
metrizable.
Solution. Let X be a compact space such that the diagonal of X is small and
w.X / !1 . If X is not metrizable, then the diagonal D f.x; x/ W x 2 X g
X X is a not a G -set of X X (see Problem 091). It is easy to see that
.; X X / D !1 and hence there exists an outer base U D fU W < !1 g
.; X X / of in X X .
For each < !1 it is easy to construct a sequence fUn W n 2 !g .; X X /
T Un D U and UT n U for all n 2 !. It is immediate that the set
0 nC1 n
such that
F D fU W n 2 !g D fU W n 2 !g is a closed G -subset of X X and
F UT for all < !1 .
If G D fF W g for all < !1 , then each G is also a closed G -set
of X X ; besides, G U and G G whenever < < !1 . Observe
also that
. / for any countable P .X X /n, there is < !1 such that G \ P D ;.
Indeed, since U is a base of in X X , for every p 2 P , there is p < !1 such

that p Up ; if > supfp W p 2 P g, then G \ P D ; which proves . /.
Choose any x0 2 G0 n; this is possible because G0 ; for otherwise,
is a G -set, a contradiction. Now, assume that < !1 and we have chosen points
fx W < g so that
(1) x 2 G n for every < ;
(2) x x
if ;
< and
.
The property . / shows that there is
such that G
\ fx W < g D ;;
if we take any x 2 G
n (this is possible again because is not a G -subset
of X X ) then x 2 G and the conditions (1) and (2) are satisfied for the set
fx W g. Thus our inductive construction can be continued to provide us a set
Y D fx W < !1 g .X X /n such that (1) and (2) hold for all < !1 .
It follows from (2) that jY j D !1 ; given any U 2 .; X X /, there is < !1
such that U U . Consequently, x 2 G G U U for all which
shows that Y nU fx W < g and therefore the set Y nU is countable for every
U 2 .; X X /. This contradiction with the fact that the diagonal of X is small
shows that X is metrizable and completes our solution.
T.296. Let X be a compact space with a small diagonal. Prove that if X is !-
monolithic and has countable tightness, then it is metrizable.
Solution. If X is not metrizable, then there is Y D fx W < !1 g X which is
not metrizable (see Problem 092). Let F D fx W < g for each < !1 ; observe
S
that the set F D fF W < !1 g is closed. Indeed, if x 2 F , then there is a
countable A F such that x 2 A. For any z 2 A, there is z 2 !1 such that
z 2 Fz ; if > supfz W z 2 Ag, then A F D F F and hence x 2 F .
Since Y F , the subspace F is not metrizable; by !-monolithity of the space
X , we have nw.F / D ! for every < !1 . It is an easy exercise that a union of
!1 -many spaces with a countable network has a network weight !1 which
proves that w.F / D nw.F / !1 (see Fact 4 of S.307). It is easy to see that
having a small diagonal is a hereditary property and therefore F is a space with a
small diagonal. Finally, apply Problem 295 to conclude that F is metrizable; this
contradiction shows that X is metrizable.
T.297. Let X be a compact space with a small diagonal. Prove that if X is
monolithic, then it is metrizable.
Solution. If X is not metrizable, then there is Y D fx W < !1 g X which
is not metrizable (see Problem 092). Since X is !1 -monolithic, we have w.Y / D
nw.Y / !1 (see Fact 4 of S.307). The property of having a small diagonal is
hereditary, so Y has a small diagonal which, together with Problem 095, implies
that Y Y is metrizable. This contradiction shows that X is metrizable.
T.298. Prove that, under CH, any compact space with a small diagonal is
metrizable.
Solution. For each n 2 N, let Mn D f1; : : : ; ng and denote by D the set f0; 1g with
the discrete topology; if f is a function, then dom.f / is its domain. If f and g are
functions, then f g says that dom.f / dom.g/ and gjdom.f / D f . Given a
set A let Fn.A/ be the set of functions f such that dom.f / is a non-empty finite
subset of A and f W dom.f / ! D. For any h 2 Fn.A/, let h D fs 2 DA W h sg.
It is easy to see that H D fh W h 2 Fn.A/g is a base in the space DA . A space Z
is called a convergent !1 -sequence if jZj D !1 and there is a point z 2 Z such that
jZnU j ! for any U 2 .z; Z/. Such a point z will be called !1 -small and hence
a convergent !1 -sequence is a space of cardinality !1 which has an !1 -small point.
Let f W X ! Y be a closed continuous onto map; then f is called irreducible if
for any closed F X with F X , we have f .F / Y . For any open non-empty
U X , we will need the sets f # .U / D Y nf .X nU / and U D f 1 .f # .U //. It is
straightforward that f # .U / f .U /; U U and f # .U / 2 .Y /; U 2 .X/.
Besides, if f is irreducible, then the set U is dense in U and hence f # .U / is dense
in f .U / (see Fact 1 of S.383).
Given an infinite cardinal , say that a family F D fFt0 ; Ft1 W t 2 T g of closed
subsets of a space X is -dyadic if we have jT j D and Ft0 \ Ft1 D ; for every
T h.t /
t 2 T while I.F ; h/ D fFt W t 2 dom.h/g ; for any h 2 Fn.T V /. In
particular, Fti ; for any t 2 T and i 2 D. If A is a family of sets, then A is
the family of all non-empty finite intersections of the elements of A.
Fact 1. If a space X has a small diagonal, then no convergent !1 -sequence is
embeddable in X .
Proof. Let D f.x; x/ W x 2 X g be the diagonal of X . Assume the contrary and
take any Z X for which jZj D !1 and there is z 2 Z such that jZnU j ! for
any U 2 .z; Z/ and hence for any U 2 .z; X /. Then A D f.t; z/ W t 2 Znfzgg
is contained in .X X /n and jAj D !1 ; given any O 2 .; X X /, we have
.z; z/ 2 O and therefore there exists U 2 .z; X / for which U U O. The point
z being !1 -small, the set ZnU is countable and therefore AnO An.U U /
.ZnU / fzg which shows that AnO is also countable and hence the set A witnesses
that the diagonal of X is not small, a contradiction. Fact 1 is proved. t
u
Fact 2. The space I is a continuous image of D for any infinite cardinal .
Proof. Fix a continuous onto map p W D! ! I (seeQ Problem 128 of [TFS]) and let
' D p for each < ; then the product ' D < ' maps .D! / continuously
onto I (see Fact 1 of S.271). The space .D! / is homeomorphic to D! ; the
cardinal being infinite, we have jj D j !j and hence D! is homeomorphic
to D . Thus ' maps D continuously onto I so Fact 2 is proved. t
u
Fact 3. Given a cardinal , any subspace of D is zero-dimensional, i.e., has a base
which consists of clopen subsets.
Proof. Given any Y D if B is a base in D , then BY D fU \ Y W U 2 Bg

is a base in Y ; it is straightforward that if all elements of B are clopen in D , then
all elements of BY are clopen in Y , so it suffices to show that D has a base whose
elements are clopen.
Given < and i 2 D, let Oi D fx 2 D W x./ D i g. It is evident that
S D fOi W < ; i 2 Dg is a subbase of D . Each Oi 2 S is clopen in D being
an inverse image of the clopen set fi g under the natural projection of D onto the
th
V factor. Since each finite intersection of clopen sets is a clopen set, the family
S is a base of D which consists of clopen sets so Fact 3 is proved. t
u
Fact 4. Given an infinite cardinal the following conditions are equivalent for any
compact space X :
(a) the space X can be mapped continuously onto I ;
(b) there exists a continuous onto map f W F ! D for some closed F X ;
(c) there exists a -dyadic family of closed subsets of X .
Proof. (a) H)(b). Let g W X ! I be a continuous onto map. It is clear that
D I , so the set F D g 1 .D / is closed in X and the map f D gjF is as
promised.
(b) H)(a). There is a continuous onto map ' W D ! I by Fact 2; clearly q D
' f maps F continuously onto I . For any < , let W I ! I be the
natural projection of I onto its th factor, i.e., .x/ D x./ for every x 2 I .
The map q D q is continuous and hence, by normality of X , there is a
continuous map r W X ! I such that r jF D q for all < . It is clear that
r D fr W < g maps X to I ; since rjF D q, we have r.X / q.F / D I
and hence X can be continuously mapped onto I .
(b) H)(c). Let Gi D fx 2 D W x./ D i g for each < and i 2 D.
It is immediate that fG0 ; G1 W < g is a -dyadic family in D , and it
is straightforward that if Fi D f 1 .Gi / for all < and i 2 D, then
fF0 ; F1 W < g is a -dyadic family of closed subsets of X .
(c) H)(b). Let F be a -dyadic family of closed subsets of X . After an evident
reindexation, we can assume that F D fF0 ; F1 W < g. If s 2 D , then
s./
consider the family Fs D fFT W < g. Since F is dyadic, the family Fs
is centered and hence Fs D Fs ;; let f .x/ D s for any x 2 Fs . If
s; t 2 D and s t, then s./ t./ for some < and hence Fs \ Ft
s./ t ./ S
F \ F D ;. This shows that the map f WTF D fFs W s 2 D g ! D
is well-defined. It is straightforward that F D fF0 [ F1 W < g, so the
subspace F is compact and hence closed in X .
It follows from the definition of f that f is surjective, so we must only prove
continuity of f . Take any x 2 F and let s D f .x/; given any O 2 .s; D /, there
is h 2 Fn./ such that s 2 h O. Let Wi D Fi \ F for all < and i 2 D. It is
evident that each Wi is closed in F ; besides, F D W0 [ W1 while W0 \ W1 D ;
which shows that each Wi is a complement of a closed subset of F and hence Wi
T h./
is also open in F . As a consequence, the set W D fW W 2 dom.h/g is
clopen in F . Observe also that for any y 2 W , if t D f .y/, then h t and hence
f .y/ h O which proves that f .W / O, i.e., W witnesses continuity of the
mapping f at the point x. Thus f is a continuous onto map and Fact 4 is proved.
t
u
Fact 5. If X is a compact space such that .x; X / > ! for each x 2 X , then there
is a closed subset P X for which there is a continuous onto map f W P ! D!1 .
Proof. Denote by C the family of all closed non-empty G -subsets of X . Each G 2
C has a countable outer base BG in X by Problem 327 of [TFS]. This shows that
G U 2 .X/ implies that there is V 2 BG with V U .
Suppose that C 0 C and jC 0 j !. Given a point x 2 X , if every U 2 .x; X/
contains some G S 2 C 0 , then it also contains some V 2 BG by the previous remark.
This shows that fBG W G 2 C 0 g is a countable local -base at x which is a
contradiction. This proves that
. / for any x 2 X and any countable C 0 C, there is W 2 .x; X/ such that
GnW ; for all G 2 C 0 .
To construct an !1 -dyadic family in X , we will also need the following
property:
. / for any countable family C 0 C, there are F; G 2 C such that F \ C ; and
G \ C ; for any C 2 C 0 but there exists B 2 C 0 such that F \ G \ B D ;.
To prove . /, apply . / for every x 2 X to obtain a set Wx 2 .x; X/ such that
C nWx ; for every C 2 C 0 . Taking a smaller Wx if necessary, we can assume that
Wx is an F -set and therefore Px D X nWx is a G -set for all x 2 X . Take a finite
subcover fWx1 ; : : : ; Wxn g of the cover fWx W x 2 X g; then Pxi \ C ; for any
C 2 C 0 and i 2 Mn while Px1 \

\ Pxn D ;. Let k 2 Mn be the minimal number
for which there exist Q1 ; : : : ; Qk 2 fPx1 ; : : : ; Pxn g such that Q1 \

\Qk \B D ;
for some B 2 C 0 . Then k 2 and the sets F D Q1 and G D Q2 \

\ Qk are as
promised so . / is proved.
Take any disjoint K00 ; K01 2 C and let C00 D K00 ; C01 D K01 . Assume that for
some < !1 , we have chosen the sets K0 ; K1 ; C0 ; C1 2 C for all < and a
function h 2 Fn./ for each 2 1; / D f
2 !1 W 1
< g so that
(1) if K D fC
0 ; C
1 W
< g, then I D I.K ; h / ; for all 2 1; /;
(2) C0 D I \ K0 ; C1 D I \ K1 for all 2 1; / and C0 \ C1 D ; for all
< ; V
(3) Ki \ H ; for any H 2 K whenever 0 < < and i 2 D.
V
The family K D fC0 ; C1 W < g is countable and hence so is C 0 D K .
V
K1 2 C such that Ki \H ; for any H 2 C 0
Thus . / is applicable to find K0 ;V
and i 2 D while there is F 2 C 0 such that K0 \ K1 \ F D ;. It follows
from (2) that there is h 2 Fn./ such that F D I D I.K ; h /. Finally, if we
let Ci D Ki \ F for every i 2 D, then the conditions (1)(3) are satisfied for
all . Thus our inductive construction can be continued to obtain a family
fK0 ; K1 W < !1 g and a set fh W 1 < !1 g such that (1)(3) hold for all
< !1 .
Given any 2 1; !1 /, let './ D maxfdom.h /g; it is clear that './ < for
all 2 1; !1 / and hence we can apply the pressing-down lemma (see Problem 067)
to find < !1 such that the set A0 D f < !1 W './ D g is uncountable. Since
dom.h / C 1 for every 2 A0 and the set Fn. C 1/ is countable, there is an
uncountable A00 A0 and h 2 Fn. C 1/ such that h D h for any 2 A00 ; let
A D A00 n. C 1/ and observe that
. / I D I
D I D I.KC1 ; h/ for any ;
2 A.
We claim that the family D D fC0 ; C1 W 2 Ag is !1 -dyadic. The property
(2) shows that, to prove it, we only have to establish that I.D; g/ ; for any
g 2 Fn.A/. Let dom.g/ D f1 ; : : : ; n g where 1 <

< n and ki D g.i / for
all i 2 Mn .
Observe (1)(3) imply that Ck11 ;; assume that 1 m n and we proved
V 0 1
that H D Ck11 \

\ Ckm1
m1 ;. Then H 2 fC ; C W < m g and hence
Km \ H ; which implies that
km
Ck11 \

\Ckm1
m1
\Ckmm D Ck11 \

\Ckm1
m1
\Kkmm \Im D Ck11 \

\Ckm1
m1
\Kkmm
because Ck11 \ Im D Kk11 \ I \ I D Ck11 (we used . /). Now it follows from (3)
that Ck11 \

\ Ckm1
m1 \ C km D K km \ H ; and hence we can go on inductively
m m
to finally establish that Ck11 \

\ Cknn ; and hence the family D is, indeed,
!1 -dyadic. Now apply Fact 4 to conclude that there is a closed P X which maps
continuously onto D!1 . Fact 5 is proved. t
u
Q
Fact 6. Let Mt be a second Q countable space for every t 2 T . If M D t 2T Mt ,
then let pS W M ! MS D t 2S Mt be the natural projection of M onto its face
MS (recall that pS is defined by pS .x/ D xjS for any x 2 M ). Then, for any
open U M , the set U depends on countably many coordinates, i.e., there exists a
countable S T such that pS1 .pS .U // D U .
Proof. The sets U and V D M nU are open and disjoint and hence separated in
M , i.e., U \ V D V \ U D ;. Thus we can apply Fact 3 of S.291 to find a
countable S T such that S .U / and S .V / are separated in MS . Therefore
S .V / \ clMS .S .U // D ;; by continuity of pS , we have S .U / clMS .pS .U //
and hence S .U / \ S .M nU / D ; which implies that pS1 .pS .U // D U so Fact 6
is proved. t
u
Fact 7. Suppose that X is a compact space such that there is a continuous onto map
f W X ! D!1 . Then X contains a convergent !1 -sequence.
Proof. The map f is perfect and hence there is a closed Y X such that
f .Y / D D!1 and f jY is irreducible (see Problem 366 of [TFS]). Since any
convergent !1 -sequence in Y is also a convergent !1 -sequence in X , we can assume,
without loss of generality, that Y D X , i.e., the map f is irreducible. Let us assume
first that X is zero-dimensional and denote by C the family of all non-empty clopen
subsets of X .
For any open non-empty U X we will need the sets f # .U / D D!1 nf .X nU /
and U D f 1 .f # .U //. It is straightforward that f # .U / f .U /; U U and
f # .U / 2 .D!1 /; U 2 .X/. Besides, the set U is dense in U and hence f # .U /
is dense in f .U / (see Fact 1 of S.383). An immediate consequence is
(4) Int.f .U // is dense in f .U / for any U 2 C
because f # .U / Int.f .U //. Another important property is
(5) Int.f .U \ V // D Int.f .U // \ Int.f .V // for any U; V 2 C.
To prove (5) observe that the inclusion Int.f .U \ V // Int.f .U // \ Int.f .V //
is straightforward, so assume that .Int.f .U //\Int.f .V ///nInt.f .U \V // ; and
therefore W D .Int.f .U // \ Int.f .V ///nf .U \ V / ;. Since W is a non-empty
open subset of f .U / and f # .U / is dense in f .U /, the set W1 D f # .U /\W is non-
empty, open and contained in f .V /. The set f # .V / being dense in f .V /, we have
W2 D f # .V / \ W1 2 .D!1 / and hence f 1 .W2 / f 1 .f # .U / \ f # .V // D
U \ V U \ V . Consequently, W2 f .U \ V / which contradicts W2 W
and W \ f .U \ V / D ;. This contradiction shows that (5) is true.
Let u 2 D!1 be defined by u./ D 0 for all < !1 ; furthermore, for all ; <
!1 , let u ./ D 0 if and u ./ D 1. The space K D fug [ fu W < !1 g
D!1 has the unique non-isolated point u and KnO is finite for any O 2 .u; D!1 /.
Pick any x 2 f 1 .u/ and let F D f 1 .u / for all < !1 .
For every < !1 denote by p W D!1 ! D , the natural projection of D!1
onto its face D , and in general, if S !1 , then pS W D!1 ! DS is the natural
projection of D!1 onto its face DS . Observe that if a set E D!1 depends on a set
S !1 (recall that this means that E D pS1 .pS .E//), then E depends on S 0 for
any S 0 S . Now it follows from (4) and Fact 6 that, for any U 2 C, the set f .U /
depends on some countable S !1 and therefore
(6) for any U 2 C there is < !1 such that f .U / depends on first coordinates.
Denote by C.x/ the family fU 2 C W x 2 U g and let C D fU 2 C S W x 2 U and
f .U / depends on first coordinatesg. We already saw that C.x/ D fC W <
!1 g. An easy consequence of (6) is that if t 2 D!1 ; U 2 C and p .t/ 2 p .f .U //,
then t 2 f .U /. In particular,
(7) if U 2 C , then u 2 f .U / for any
because u D f .x/ 2 f .U / and p .u/ D p .u /. Next we claim that
(8) if U; V 2 C , then W D U \ V 2 C .
To prove (8), let U 0 D Int.f .U // and observe that f .U / D p1 .p .f .U ///

implies that p1 .p .U 0 // f .U / and hence p1 .p .U 0 // Int.f .U // D
U 0 . Thus p1 .p .U 0 // D U 0 , and analogously, if V 0 D Int.f .V //, then
p1 .p .V 0 // D V 0 . Now let W 0 D Int.f .W // and apply (5) to conclude that
W 0 D U 0 \ V 0 and therefore p1 .p .W 0 // D W 0 . Finally, W 0 is dense in f .W /
which, together with the openness of the map p implies that
p1 .p .f .W /// p1 .p .W 0 // D p1 .p .W 0 // D W 0 D f .W /
which proves that p1 .p .f .W /// D f .W / and settles (8). Observe also that we
have C C if and therefore
T T
(9) C C whenever < !1 .
Given any < !1 it follows from (7) that U \ F ; for any U 2 C ; besides,
the family fU T \ F W U 2 C g is centered by (8) which shows that there exists
x 2 F \ . C / because F is compact.
We claim that Z D fx W < !1 g [ fxg is a convergent !1 -sequence.
In first place, jZj D !1 because the map f jZ condenses Z onto K. Fur-
thermore, for any U 2 .x; X/, there is < !1Tand V T 2 C such that
x 2 V U . Given any , we have x 2 C C V and
hence jZnU j jZnV j jfx W < gj !. Moreover, we proved that there is
a convergent !1 -sequence Z in X such that f jZ condenses Z onto the compact
space K D!1 .
This finishes the proof of our fact for compact zero-dimensional spaces. Now
assume that X is an arbitrary compact space; if w.X / D , then we can assume
that X I . Apply Fact 2 to take a continuous onto map ' W D ! I . Then
Y D ' 1 .X / is a zero-dimensional compact space by Fact 3 and '1 D 'jY maps
Y continuously onto X . Now, f '1 maps Y continuously onto D!1 , so we can
apply what we have proved for zero-dimensional compact spaces to conclude that
there is a convergent !1 -sequence Z1 Y such that f ' condenses Z1 onto
K. An immediate consequence is that ' condenses Z1 onto Z D '.Z1 / X . It is
immediate that a continuous image of a convergent !1 -sequence is also a convergent
!1 -sequence provided that it is uncountable. Since 'jZ1 W Z1 ! Z is a bijection,
we have jZj D jZ1 j D !1 and hence the space Z is also a convergent !1 -sequence
in X , i.e., Fact 7 is proved. t
u
Fact 8. Given an infinite cardinal suppose that X is a compact space and S X
is a free C -sequence. Then the set S can be continuously mapped onto C C 1. In
particular, if X is a compact space with t.X / > , then there is a closed Y X
which maps continuously onto C C 1.
Proof. If t.X / > , then there is a free sequence S D fx W < C g X (see
Problem 328 of [TFS]); let S D fx W < g for every C . To prove the
second part of our fact, it suffices to take Y D S , i.e., it is sufficient to show that S
maps continuously onto C C 1.
To do that, let f .xn / D n for any n 2 !; if 2 C n!, then let f .x / D C1. If

x 2 S nS , then the ordinal f .x/ D minf C W x 2 S g is well-defined; besides,
f .x/ is a limit ordinal for any x 2 S nS because if D
C 1, then x 2 S implies
x 2 S nS and therefore x 2 S nfx
g D S
. This defines a map f W S ! C C 1.
Observe that x is an isolated point of Y for any < C , so it suffices to
prove continuity of f at any x 2 Y nS . If f .x/ D and U 2 .; C C 1/,
then there is < such that .; D f
2 C W <
g U .
The set S being a free sequence, it follows from the definition of f .x/ that
the set W D X nSC1 [ .S nS / is an open neighborhood of x. Observe that
x
2 W implies
2 .; / and therefore f .x
/ 2 .; / U . Furthermore, if
y 2 .Y nS / \ W , then y S nS and therefore y 2 S whence f .y/ . Now,
it follows from y S C1 that f .y/ C 1 > , i.e., f .y/ 2 .; . This proves
that f .W \ Y / .; U and hence f is continuous at the point x. The point
x 2 Y nS was chosen arbitrarily, so f is a continuous map. The set f .S / is dense
in C C 1, so f .S / D C C 1 and hence Fact 8 is proved. t
u
Fact 9. If X is a compact space with t.X / > !, then X contains a convergent
!1 -sequence.
Proof. It follows from Fact 8 that there is a closed H X which maps continuously
onto !1 C 1. By Problem 366 of [TFS], there is closed H 0 H and a continuous
irreducible onto map f W H 0 ! .!1 C 1/. Since it suffices to find a convergent
!1 -sequence in H 0 , we can assume, without loss of generality, that X D H 0 , i.e.,
that there is a continuous irreducible onto map f W X ! .!1 C 1/.
Let X D fx 2 X W f .x/ D g for each !1 . Observe first that X!1
is nowhere dense in X because otherwise there is an open non-empty U X!1
S f .U / is a non-empty open subset of f!1 g which is a contradiction. Let
#
and hence
Y D fX W !1 g for every < !1 . If .x; X!1 / > ! for any x 2 X!1 ,
then there is a closed P X!1 that maps continuously onto D!1 by Fact 5 and
hence P contains a convergent !1 -sequence by Fact 7. Therefore we can assume,
without loss of generality, that there is x 2 X!1 such that .x; X!1 / D !.
Fix a family fOn W n 2 !g .X/ such that fOn \ X!1 W n 2 !g is a -base of
X!1 at the point x. Choose a set Un 2 .X/ such that Un \ X!1 ; and U n On
for all n 2 !.
We saw already that Y D X nX!1 is dense in X and hence Y \ Un is dense in
Un for all n 2 !. As a consequence, the set An D f < !1 W Un \ X ;g is
uncountable for each n 2 !. Furthermore, An D f .Un \ Y /; since the map f jY is
closed (see Fact 1 of S.261), the set BnTD f .U n \Y / is closed and unbounded in !1
for all n 2 !. Therefore the set B D fBn W n 2 !g is also closed and unbounded
by Problem 064.
Now let V D fV 2 .x; X/ W for any n 2 !; if U n \ X!1 V , then U n \ Y
V g for every < !1 ; observe that V1 ; V2 2 V implies V1 \ V2 2 V . Besides
V V whenever < < !1 .
T
Now let F D fV W V 2 V g for all < !1 . Observe T that for any set V 2
.x; X/, there is < !1 such that V 2 V . Indeed, X!1 D fY W < !1 g, so
T
U n \ X!1 V implies that fU n \ Y W < !1 g V and hence we can apply
Fact 1 of S.326 to conclude that there is n < !1 for which U n \ Yn V . Now, if
> supfn W U n \ X!1 V g, then V 2 V .
Consequently, we have a family fF W T< !1 g of closed subsets T of X such
that F F whenever < < !1 and fF W < !1 g D fV W V 2
.x; X /g D fxg.
Besides, for every ordinal < !1 and any V 2 V , there is n 2 ! such that
U n \ X!1 On \ X!1 V and hence U n \ Y V . This implies, however, that
Bn Bn n f .U n \ Y / f .V / which shows that f .V / Bn for each
V 2 V and hence Bn f .F /. As a consequence, F fxg for all < !1 .
Observe also that if Q X nfxg is a countable set, then for any y 2 Q there is
y < !1 such that y Fy ; if > supfy W y 2 Qg, then F \ Q D ;. This makes
it possible to construct, by an evident induction, a set Z 0 D fx W < !1 g such that
x 2 F nfxg and x x whenever ; < !1 and .
To see that Z D fx W < !1 g [ fxg is a convergent !1 -sequence observe first
that jZj D !1 because the enumeration of the set Z 0 is faithful. Furthermore, if
W 2 .x; X/, then we can apply Fact 1 of S.326 to conclude that there is < !1
for which F W . Therefore ZnW fx W < g and hence Z is a convergent
!1 -sequence. Fact 9 is proved. t
u
Now it is easy to finish our solution. Assume that CH holds and X is a compact
space with a small diagonal. It follows from Fact 1 and Fact 9 that t.X / !. Given
a countable A X , we have w.A/ 2jAj c by Fact 2 of S.368 and hence
w.A/ !1 by CH. Recall that having a small diagonal is hereditary, so A has a
small diagonal; thus we can apply Problem 295 to conclude that A is metrizable
and therefore X is !-monolithic. Finally, X is metrizable by Problem 296 so our
T.299. Assume that 2!1 D !2 . Prove that any compact X , with !1 and !2 calibers
of Cp .X /, is metrizable.
Solution. If the space X is not metrizable, then there is Y X such that jY j !1
and Y is not metrizable (see Problem 092). Thus Z D Y is not metrizable either
and w.Z/ 2jY j 2!1 D !2 (see Fact 2 of S.368). Furthermore, both cardinals
!1 and !2 are calibers of Cp .Z/ by Problem 292. If Cp .Z/ is separable, then Z is
metrizable (see Problems 174 and 140 of [TFS]) which is a contradiction with the
fact that Y Z is not metrizable.
As a consequence, ! < d.Cp .Z// nw.Cp .Z// D nw.Z/ w.Z/ !2 and
hence d.Cp .Z// D 2 f!1 ; !2 g. Let D D ff W < g be a dense subset of
Cp .Z/ and denote by D the set ff W < g for every < . Since jD j < for
each < , the set U D Cp .Z/nD is open and non-empty for all < .
Given any f 2 Cp .Z/, there is a countable A D with f 2 A (recall that
t.Cp .Z// D ! because Z is compact). Since D cf./ > !, there is < such
S
T A D and hence f 2 D . This proves that fD W < g D Z and hence
that
fU W < g D ;. This, together with the fact that U U whenever < ,
implies that every f 2 Cp .Z/ belongs to strictly less than -many elements of the
family U D fU W < g. Therefore U witnesses that is not a caliber of Cp .Z/.

This contradiction shows that a non-metrizable Y X with jY j !1 cannot exist
and hence X is metrizable.
T.300. Observe that any Lindelf -space with a diagonal G has a countable
network. Prove that, under CH, any Lindelf -space with a small diagonal has a
countable network.
V
Solution. For every n 2 N let Mn D f1; : : : ; ng; if A is a family of sets, then A
is the family of all non-empty finite intersections of the elements of A. If Z is a
space, then a family A exp.Z/ is called closed (compact) if all elements of A are
closed in Z (or compact, respectively). Given a space Z, a family F exp.Z/ is a
network with respect to a family C exp.Z/ if for any C 2 C and U 2 .C; Z/,
there is F 2 F such that C F U . The family F is a called a network at any
point of a set A Z if it is a network with respect to ffag W a 2 Ag. Say that distinct
points x; y 2 Z are T1 -separated by a family A exp.Z/ if there are A; B 2 A
such that x 2 A 63 y and y 2 B 63 x. Say that A is a T1 -separating family for the
points of Y Z if any distinct x; y 2 Y are T1 -separated by the family A.
A space Z is Lindelf if and only if there is a countable family F of closed
subsets of Z which is a network with respect to a compact cover C of the space Z
(see Problem 225 and Fact 1 of T.229).
Fact 1. Suppose that Z is a space in which there exists a countable closed network
F with respect to a compact cover C of the space Z such that every C 2 C is
metrizable. If, additionally, the space Z has a small diagonal, then nw.Z/ D !.
V
Proof. There is no loss of generality to consider that F D F , i.e., F is closed
under finite intersections. For every C 2 C, let FC D fF 2 F W C F g. It turns
out that
(1) givenVany C 2 C, there
V exists a countable family UC of cozero-sets of Z such
that UC D UC and .FC [ UC / is a network at all points of C .
To prove (1) take a countable base B of the space C and call a pair .U; V / 2 B
strongly disjoint if U \ V D ; (the bar denotes the closure in Z). The space Z
is Lindelf and hence normal, so if U; V 2 B are strongly disjoint, then we can
fix disjoint cozero-sets V GU ; GV of the space Z such that U GU and V GV .
The family UC D fGV U ; GV W the pair .U; V / 2 B B is strongly disjointg is,
evidently, countable and UC D UC . To see that UC is as promised take any x 2 C
and W V2 .x; Z/. Let fPn W n 2 !g be an enumeration V of all elements of the
family .FC [ UC / which contain x. Since the family .FC [ UC / is closed under
finite intersections,
T if Pn nW ; for every n 2 !, then we can choose a point
xn 2 . fPi W i ng/nW for every n 2 !; let S D fxn W n 2 !g. For any A Z
let S.A/ D fn 2 ! W xn 2 Ag.
Assume first that, for any point y 2 C , there exists Uy 2 .y; Z/ such that the
set S.Uy / is finite. The subspace C being compact we can choose a finite subcover
fUy1 ; : : : ; Uyk g of the open cover fUy W y 2 C g of the space C . Since FC is a
network with respect to C , there is n 2 ! such that Pn Uy1 [

[ Uyk and
hence S.Pn / S.Uy1 / [

[ S.Uyk / which shows that S.Pn / is a finite set, a
contradiction with S.Pn / fm 2 ! W m ng.
As a consequence, there is a point y 2 C such that S.G/ is infinite for any
G 2 .y; Z/. It is evident that y W and hence y x. The family B being a base
in C , there are U; V 2 B such that x 2 U; y 2 V and U \ V D ;. This implies
x 2 GU ; y 2 GV and GU \ GV D ;. Observe that GU D Pn for some n 2 !
and therefore S.Pn / fm 2 ! W m ng whence j!nS.Pn /j n. However, the
set S.GV / is infinite by the choice of y and S.GV / \ S.GU / D ;, i.e., the infinite
set S.GV / is contained in the finite set !nS.GU /. This contradiction shows that
x 2 Pn W for some n 2 ! and hence (1) is proved.
Denote by U the family of all cozero-sets of Z. Observe that each U 2 U is an
F -set of Z and the family U is a base in Z (see Fact 1 of T.252). We can assume
without loss of generality that
(2) if U 0 U is countable, then F [ U 0 does not T1 -separate the points of Z.
Indeed,Sif it does, then let AU be a countable family ofSclosed subsets of Z
such that AU D U for every U 2 U 0 . The family F [ . fAU W U 2 U 0 g/ is
also countable, consists of closed subsets of Z and T1 -separates the points of Z, so
Fact 1 of T.270 is applicable to conclude that ZThas a countable network.
Given a family A exp.Z/, let Az D fA 2 A W z 2 Ag for all z 2 Z.
Observe also that
(3) if x; y 2 Z and y 2 F x , then y 2 Cx
because if x 2 C 63 y for some C 2 C, then the family F being a network with
respect to C, there is F 2 F for which C F Znfyg and hence x 2 F 63 y
Since the family F does not T1 -separates the points of Z, there exist distinct
a0 ; b0 2 Z such that b0 2 F a0 . There is C0 2 C with a0 2 C0 ; it follows from (3)
that b0 2 C0 .
Assume that < !1 and we have chosen a family fC W < g C and points
fa ; b W < g Z such that
(4) a b and fa ; b g C forS all < ;
(5) For any < if V D F [ . fUC
W
< g/, then b 2 V a .
S
Applying (2) to the family V D F [ . fUC W < g/, we can find distinct
points a ; b 2 Z such that b 2 V a . There is C 2 C such that a 2 C ; it
follows from (3) that b 2 C . It is immediate that the conditions (4) and (5) are
satisfied for the family fC W g and the set fa ; b W g which shows that
our inductive construction can be continued to give us a family fC W < !1 g C
and a set fa ; b W < !1 g Z such that (4) and (5) are fulfilled for any < !1 .
Let D f.z; z/ W z 2 Zg be the diagonal of the space Z; since is small
and D D f.a ; b / W < !1 g .Z Z/n, there is an uncountable T !1
such that the closure R of the set f.a ; b / W 2 T g does not meet . For each
z D .x; y/ 2 R, take Uz 2 .x; Z/ and Vz 2 .y; Z/ such that U z \ V z D ;.
The space Z is Lindelf so Z Z isSLindelf (see Problem 256) and hence there
is a countable Q R such that R fUz Vz W z 2 Qg. Consequently, there is
an uncountable E !1 and z 2 R for which f.a ; b / W 2 Eg Uz Vz , i.e.,
fa W 2 Eg Uz and fb W 2V Eg Vz .
Given an 2 E,Vthe family .FV[ UC / is a network at all points of C by
(1); it follows from F D F and UC D UC that for every point x 2 C ,
there exists Ox 2 UC and Fx 2 FC such that Ox \ Fx misses either U z or V z .
The set C is compact, so there are finite subcollections fU1 ; : : : ; Un g UC and
fF1 ; : : : ; Fn g FC such that C U D U1 [

[ Un and Ui \ Fi misses
either U z or V z for every i 2 Mn . Since F is a network with respect to C, we can
choose F 2 F for which C F F1 \

\ Fn \ U for all 2 E.
The family F being countable, there exist ; 2 !1 such that < and F D
F D F . Therefore C [ C F U and hence a 2 Ui for some i 2 Mn .

The set Ui \ Fi misses either U z or V z while a 2 Ui \ Fi . Consequently,

Ui \ Fi misses V z which implies that b Ui \ Fi . However, Ui \ Fi 2 V
which shows that b V a which gives a contradiction with the property (5).
Therefore Z has a T1 -separating family F [ U 0 for some countable U 0 U which
shows that nw.Z/ D ! and hence Fact 1 is proved. t
u
Returning to our solution, observe that an evident consequence of Problem 235
is that every Lindelf -space with a G -diagonal has a countable network. Now, if
CH holds and X is a Lindelf -space with a small diagonal, then every compact
K X also has a small diagonal and hence K is metrizable by Problem 298. Take
a countable family F of closed subsets of X such that F is a network with respect
to a compact cover C of the space X (see Problem 225 and Fact 1 of T.229). Every
element of C is metrizable, so we can apply Fact 1 to conclude that nw.X / D ! and
finish our solution.
T.301. Let X be a zero-dimensional space. Prove that any subspace of X is also
zero-dimensional.
Solution. Fix a base B in the space X such that every B 2 B is clopen in X . Given
any Y X , the family BY D fB \ Y W B 2 Bg is a base in Y and it is immediate
that all elements of BY are clopen in Y . Therefore Y is zero-dimensional.
T.302. Prove that an arbitrary product of zero-dimensional spaces must be a zero-
dimensional space.
V Denote by Mn the set f1; : : : ; ng for all n 2 N. Given a family S, the

Solution.
family S is the collection of all finite intersections of the elements of S. Assume
that XtQis a zero-dimensional space for every t 2 T ; we must prove that the space
X D fXt W t 2 T g is also zero-dimensional. Let Bt be a base in Xt such that
every B 2 Bt is clopen in Xt . The map pt W X ! Xt is the natural projection for
all t 2 T .
The family S D fpt1 .B/ W B 2 Bt ; t 2 T g consists of clopen subsets of
X because each pt is a continuous map. Take any x 2 X and U 2 .x; X/. By
the definition of the product topology there exists a set T 0 D ft1 ; : : : ; tn g T and
Q Q
Oi 2 .Xti / for all i 2 Mn such that x 2 O D fOi W i 2 Mn g t 2T nT 0 Xt U .
Since each Xt is zero-dimensional, for every i 2 Mn , there is Vi 2 Bti such that
x.ti / 2 Vi Oi . V
Observe also that if Wi D pt1i V i 2 Mn , then V D W1 \

\Wn 2 S
.Vi / for all
and x 2 V O U which shows that S is a base in X , i.e., S is a V subbase of
X . Since every finite intersection of clopen sets is a clopen set, the base S of the
space X consists of clopen subsets of X and hence X is zero-dimensional.
T.303. Given a cardinal and an infinite space X with w.X / , prove that X is
zero-dimensional if and only if it can be embedded in D .
Solution. If X is a subspace of D , then X is zero-dimensional by Fact 3 of T.298.
Now assume that X is zero-dimensional and fix a base B 0 of the space X such that
every B 2 B 0 is a clopen subset of X . It follows from Fact 1 of T.102 that there is
B B 0 such that B is still a base of X and jBj .
For every B 2 B let B .x/ D 1 if x 2 B and B .x/ D 0 for all x 2 X nB. Then
B W X ! D is a continuous map because fB; X nBg is an open cover of X on the
elements of which B is constant (see Fact 1 of S.472).
Consequently, ' D fB W B 2 Bg W X ! DB is a continuous map; let
Y D '.X /. We claim that the map ' W X ! Y is a homeomorphism. In the first
place observe that for any distinct x; y 2 X , there is B 2 B such that x 2 B and
y B. Thus B .x/ D 1 while B .y/ D 0 which shows that '.x/ '.y/ and
therefore ' is a bijection. For any B 2 B let pB W DB ! D be the natural projection
of DB onto its Bth factor.
To see that ' 1 is continuous take any y 2 Y and U 2 .' 1 .y/; X /; let x D
1
' .y/. Since B is a base in X , there is B 2 B such that x 2 B U . The set
W D fz 2 Y W pB .z/ D 1g D pB1 .1/ \ Y is an open subset of Y because pB1 .1/
is open in DB . Now, if t 2 W and s D ' 1 .t/, then, by definition of the diagonal
product, 1 D pB .t/ D B .s/ and hence s 2 B which implies that s D ' 1 .t/ 2 U
for any t 2 W . Thus ' 1 .W / U , i.e., W witnesses continuity at the point y. The
point y has been chosen arbitrarily, so the map ' 1 is continuous and hence ' is an
embedding of X in DB . Since jBj , the space DB embeds in D (this is an easy
exercise that we leave to the reader) so X embeds in D as well.
T.304. Prove that any space X is a perfect image of a zero-dimensional space Y
such that w.Y / w.X /.
Solution. If D w.X / is finite, then X is zero-dimensional, so there is nothing to
prove. We will assume, therefore, that is an infinite cardinal; there is no loss of
generality to consider that X is a subspace of I (see Problem 209 of [TFS]).
There exists a continuous onto map ' W D ! I by Fact 2 of T.298; let Y D
1
' .X /. The map f D 'jY W Y ! X is perfect because so is ' (see Problem 122
of [TFS] and Fact 2 of S.261). The space Y is zero-dimensional by Problem 303 and
w.Y / w.D / which shows that X is a perfect image of a zero-dimensional
space Y such that w.Y / D w.X /.
T.305. Prove that any non-zero-dimensional space can be continuously mapped

onto I.
Solution. Let X be a non-zero-dimensional space. It was proved in Fact 4 of T.063
that there exists a continuous onto map f W X ! I D 0; 1 where I carries
the topology induced from R. If g.x/ D 2x 1 for all x 2 I , then g maps I
continuously onto I and hence g f maps X continuously onto I.
T.306. Prove that any Lindelf space is zero-dimensional if and only if it is
strongly zero-dimensional. In particular, compact zero-dimensional spaces and
second countable zero-dimensional spaces are strongly zero-dimensional.
Solution. Given a space Z let C.Z/ be the family of all clopen subsets of Z.
Fact 1. If a space Z is strongly zero-dimensional, then it is zero-dimensional.
Proof. To prove that C.Z/ is a base in Z, take any z 2 Z and U 2 .z; Z/. There
is V 2 .z; Z/ such that V U and hence the family U D fU; X nV g is a finite
open cover of Z. The space Z being strongly zero-dimensional, there is a disjoint
refinement V .X/S of the cover U. Observe first that every W 2 V is clopen in Z
because ZnW D .VnfW g/ is an open set. Take any W 2 V with z 2 W . Since V
is a refinement of U, we have either W U or W ZnV . The second inclusion is
impossible because z 2 W n.ZnV /. Therefore x 2 W U which implies, together
with W 2 C.Z/, that C.Z/ is a base of Z and hence Z is zero-dimensional. Fact 1
is proved. t
u
Returning to our solution observe that if X is strongly zero-dimensional, then it
is zero-dimensional by Fact 1; for this implication, there is no need to assume that
X is Lindelf.
Now let X be a Lindelf zero-dimensional space. To prove that X is strongly
zero-dimensional, take any open cover U of the space X (we should formally
assume that U is finite, but proving it without this assumption gives a stronger
statement). Since the space X is zero-dimensional, the family V D fV 2 C.X / W
there exists U.V / 2 U with V U.V /g is a cover of X .SThe space X being
Lindelf there is a family V 0 DSfVn W n 2 !g V such that V 0 D X . It is easy
to see that the set Wn D Vn n. i <n Vi / is clopen for all n 2 N; let W0 D V0 and
observe that W D fWn W n 2 !g is a disjoint open refinement of U. This proves that
every Lindelf zero-dimensional space is strongly zero-dimensional and finishes our
solution.
T.307. Let X be a space with jX j < c. Prove that X is zero-dimensional. In
particular, any countable space is strongly zero-dimensional.
Solution. If X is not zero-dimensional, then there exists a continuous onto map
f W X ! I by Problem 305 and hence jX j jIj D c which is a contradiction.
Therefore X is zero-dimensional. Furthermore, if X is countable, then X is Lindelf
and jX j < c. Therefore X is strongly zero-dimensional by Problem 306.
T.308. For an arbitrary space X , prove that X is strongly zero-dimensional if and

only if IndX D0. Observe that, as a consequence, any strongly zero-dimensional
space is normal.
Solution. Given a space Z let C.Z/ be the family of all clopen subsets of Z.
Suppose that Z is a space such that Ind.Z/ D 0. Given disjoint closed sets
F; G Z we have U D ZnG 2 .F; Z/ and hence there is O 2 C.Z/ such that
F O ZnG. As a consequence, O and ZnO are open disjoint neighborhoods
of F and G respectively which proves that
(1) for any space Z if Ind.Z/ D 0, then Z is normal.
To prove sufficiency assume that Ind.X / D 0 and take any open cover U D
fU1 ; : : : ; Un g of the space X . The set F D X n.U2 [

[ Un / is closed in X and
F U1 ; besides F [ U2 [

[ Un D X . Since Ind.X / D 0, there is V1 2 C.X /
such that F V1 U1 ; it is clear that V1 [ U2 [

[ Un D X .
Assume that 1 k < n and we have clopen sets V1 ; : : : ; Vk such that Vi Ui
for all i k and V1 [

[ Vi [ Ui C1 [

[ Un D X for any i k. The set
F D X n.V1 [

[ Vk [ UkC2 [

[ Un / is closed in X and F UkC1 . Since
Ind.X / D 0, there is VkC1 2 C.X / such that F VkC1 UkC1 ; it is clear that
V1 [

[ VkC1 [ UkC2 [

[ Un D X and hence our inductive construction can
be continued to obtain a family fV1 ; : : : ; Vn g C.X / such that Vi Ui for every
i n and V1 [

[ Vn D X . Finally, if W1 D V1 and Wk D Vk n.V1 [

[ Vk1 /
for all k 2 f2; : : : ; ng, then fWi W i ng is a disjoint open refinement of U which
proves that X is strongly zero-dimensional and settles sufficiency.
As to necessity, assume that X is strongly zero-dimensional and take any closed
F X and U 2 .F; X/. The family U D fU; X nF g is a finite open cover of X ,
so there exists a disjoint open S refinement V of the cover U. Observe S first that for
every V 0 V, the set W D V 0 is clopen in X because X nW D .VnV 0 / is an
open set.
Take any W 2 V 0 D fV 2 V W V \ F ;g. Since V is a refinement of U,
we have either W U or W X nF . The second inclusion is impossible because
0
there is x 2 W S \0 F and hence x 2 W n.X nF /. Thus W U for any W 2 V and
hence V D V is a clopen set such that F V U which implies, together
with V 2 C.X /, that Ind.X / D 0 and hence necessity is established. To finish our
solution observe that Ind.X / D 0 for any strongly zero-dimensional space X and
hence X is normal by (1).
T.309. Prove that any strongly zero-dimensional space is zero-dimensional. Give
an example of a normal zero-dimensional space which fails to be strongly zero-
dimensional.
Solution. Any strongly zero-dimensional space is zero-dimensional: this was
proved in Fact 1 of T.306. Denote by I the set 0; 1 with the topology inherited
from R. For any space Z let C.Z/ be the family of all clopen subsets of Z. When
we work with ordinals, then each ordinal considered as a set is identified with
the set of its predecessors, i.e., D f W < g. For any ; 2 !1 C 1, let

; / D f
W
< g.
Fact 1. Given any points a; b 2 R such that a < b, the interval a; b is
connected, i.e., C.a; b / D f;; a; b g. As a consequence, the space a; b is not
zero-dimensional.
Proof. Assume, towards a contradiction, that there is a clopen subset U of the space
a; b such that U ; and U a; b . Then the sets U and V D a; b nU are non-
empty, compact, disjoint and U [ V D a; b . Take any x 2 U and y 2 V ; then
x y because the sets U and V are disjoint. Interchanging U and V if necessary,
we can assume that x < y.
The sets P D U \ x; y and Q D V \ x; y are compact, disjoint, their union
is x; y and x 2 P; y 2 Q. Let D inf Q. The set Q x; y is bounded and
non-empty so 2 x; y . The point has to belong to P or to Q. Observe that
P and Q are complementary closed sets in x; y and hence they are both open in
x; y . Now, if 2 P , then < y, and by openness of P , there is " > 0 such that
; C "/ P , and hence there are no points of Q in ; C "/ which shows that
inf Q C ", a contradiction. If 2 Q, then x < , and, by openness of Q, we
have . "; 2 Q for some " > 0. Therefore all points of . "; belong to Q
and hence inf Q " and we again obtained a contradiction which shows that
a; b is connected.
Finally, observe that if a; b is zero-dimensional, then there is U 2 C.a; b /
such that a 2 U and b U . Therefore U is a clopen set in a; b such that U ;
and U a; b ; this contradiction completes the proof of Fact 1. t
u
Fact 2. If Z R, then Z is zero-dimensional if and only if Int.Z/ D ;.
Proof. Suppose that Z is zero-dimensional and Int.Z/ ;. Then there are a; b 2 R
such that a < b and a; b Z. Now, it follows from Problem 301 that a; b is
zero-dimensional which contradicts Fact 1. This proves necessity.
Now assume that Int.Z/ D ;; given any z 2 Z and U 2 .z; Z/ there are
a; b 2 R such that a < b and z 2 .a; b/ \ Z U . It follows from Int.Z/ D ; that
.a; z/nZ ; and .z; b/nZ ;. Pick any c 2 .a; z/nZ and d 2 .z; b/nZ. Then
z 2 W D .c; d / \ Z; it is evident that W 2 .Z/. Furthermore, W D c; d \ Z
and hence W is also closed in Z. Since z 2 W U and W 2 C.Z/, we showed
that C.Z/ is a base in Z, i.e., that Z is zero-dimensional. Fact 2 is proved. t
u
Fact 3. Let Q D Q \ I ; then there exists a disjoint family fQ W < !1 g of
countable dense subsets of I such that Q \ Q D ; for all < !1 .
Proof. Let Q1 D Q and suppose that for some ordinal < !1 , we have a disjoint
familySfQ W 1 < g of countable dense subsets of I . Observe that the set
R D fQ W 1 < g is countable and therefore I nR is dense in I . Since the
space I nR is second countable, there is a countable dense Q I nR. It is easy to
see that Q is also dense in I and hence our inductive construction can be continued
to provide a disjoint family fQ W 1 < !1 g of countable dense subsets of I

such that Q1 D Q. Clearly, the family fQ W 0 < !1 g is as promised, so
Fact 3 is proved. t
u
Fact 4. For any space Z, if K Z is compact, F Z is closed and K \ F D ;,
then K and F are open-separated, i.e., there are U 2 .K; Z/ and V 2 .F; Z/
such that U \ V D ;.
Proof. For every x 2 K, there is Ux 2 .x; Z/ such that U x \ F D ;. By
compactness of the space K, there are x1 ; : : : ; xn 2 K such that K Ux1 [

[Uxn .
For the set U D Ux1 [

[ Uxn , we have U D U x1 [

[ U xn ZnF
because U xi ZnF for all i n. As a consequence, U 2 .K; Z/ and
V D ZnU 2 .F; Z/; since the sets U and V are disjoint, the proof of Fact 4
is complete. t
u
Fact 5. If Z is a space which can be covered by its clopen zero-dimensional
subspaces, then Z is zero-dimensional.
S
Proof. Take any U C.Z/ such that each U 2 U is zero-dimensional and U D
Z. Given any z 2 Z and W 2 .z; Z/, there is U 2 U such that z 2 U . Evidently,
W \ U 2 .z; U /; since U is zero-dimensional, there exists O 2 C.U / such that
z 2 O U \ W . It is immediate that O 2 C.Z/ and z 2 O U which shows that
Z is zero-dimensional. Fact 5 is proved. t
u
Returning to our solutions let fQ W < !1 g be a disjoint family of countable
dense subsets of I such that Q \ Q D ; for S all < !1 , where Q D Q \ I
(see Fact 3). Observe that the set S D I n. fQ W g/ is dense in I for
every
S < !1 because Q S . Besides, S S whenever < < !1 and
fS W < !S1g D I .
Let X D ffg S W < !1 g .!1 C 1/ I . We consider X with the
topology inherited from the compact space .!1 C 1/ I so X is a Tychonoff space;
let X D . I / \ X for all < !1 . To see that X is normal, consider the space
Y D X [ .f!1 g I /. We show first that
(1) the space Y is normal.
Let F; G be closed disjoint subsets of Y . Then the sets F 0 D F \ .f!1 g I /
and G 0 D G \ .f!1 g I / are compact disjoint subsets of J D f!1 g I which
is homeomorphic to I . By Fact 4, there exist U 2 .F 0 ; Y / and V 2 .G 0 ; Y /
such that U \ V D ;. We claim that F nU X for some < !1 . Indeed,
suppose for a moment that for any < !1 , there is z D . ; t / 2 F nU for
some . The space I being compact, there is a point t 2 I such that the set
A" D f < !1 W jt t j < "g is uncountable for any " > 0. Let z D .!1 ; t/; if
W 2 .z; Y /, then there is " > 0 and
< !1 such that z 2 O D f.; u/ W
<
and ju tj < "g W . Since A" is uncountable, there is 2 A" with >
and
hence >
which shows that z 2 O \ F W \ F and hence z is in the
closure of F nU in Y . The set F nU being closed, we have z 2 .F nU / \ J which is
a contradiction with F 0 D F \ J U . The same reasoning applied to GnV shows
that there exists < !1 such that .F nU / [ .GnV / X . Since XC1 is clopen in
Y , the sets U 0 D U nXC1 and V 0 D V nXC1 are open in Y .
The space XC1 . C 1/ I is second countable and hence normal; since
F \ XC1 and G \ XC1 are disjoint closed subsets of XC1 , there are disjoint
U 00 ; V 00 2 .XC1 / such that F \ XC1 U 00 and F \ XC1 V 00 . It is evident
that U 00 ; V 00 2 .Y /; since U 0 [ U 00 and V 0 [ V 00 are disjoint open neighborhoods
of F and G respectively, we proved that Y is normal.
To establish that X is also normal we need to prove first that
(2) if F; G are disjoint closed subsets of X , then F \ G D ; (the bar denotes the
closure in Y ).
It is clear that F \G \X D ; so Sassume, towards a contradiction, that z 2 F \G
for some z D .!1 ; t/ 2 J . Since fS W < !1 g D I , there is < !1 such that
t 2 S for all and hence P D ; !1 / ftg Y . It is clear that P is
homeomorphic to ; !1 /; any two closed unbounded subsets of ; !1 / have non-
empty intersection (see Problem 064) and therefore both sets F \ P and G \ P
cannot be uncountable. Assume, for example, that F \ P is countable and hence
there is
< !1 for which F \ P
ftg.
It is easy to see that z 2 F implies that there exists a strictly increasing sequence
fn W n 2 !g !1 n
and ftn W n 2 !g I such that zn D .n ; tn / 2 F and
jtn tj < nC1 1
for all n 2 !. Letting D supfn W n 2 !g it is straightforward to
show that z0 D .; t/ 2 F D F which contradicts F \ P
ftg.
Analogously, if G \ P is countable, then z G; this final contradiction shows
that (2) is true.
Now it is easy to prove that X is normal. Indeed, take any disjoint closed F; G
X . The sets F and G are closed in Y and disjoint by (2). The property (1) guarantees
that there exist disjoint U 0 2 .F ; Y / and V 0 2 .G; Y /; it is clear that U D U 0 \X
and V D V 0 \ X are disjoint open (in X ) neighborhoods of F and G, respectively,
so X is normal.
Given any < !1 , the set S has empty interior in I (and hence in R) because
Q I nS is dense in I . Consequently, S is zero-dimensional by Fact 2. The
space is zero-dimensional being countable by Problem 307 and hence S is
zero-dimensional by Problem 302. Therefore X S is zero-dimensional by
Problem 301 for every < !1 . Thus fXC1 W < !1 g is a clopen cover of X by its
zero-dimensional subspaces which shows that X is zero-dimensional by Fact 5.
To finally see that X is not strongly zero-dimensional, assume the contrary; then
Ind.X / D 0 by Problem 308. Let F D !1 f0g and G D !1 f1g; it is evident that
F and G are closed disjoint subsets of X . Since Ind.X / D 0, there exists U 2 C.X /
such that F U X nG, i.e., U and V D X nU are disjoint clopen subsets of X .
Observe also that p D .!1 ; 0/ 2 F U and q D .!1 ; 1/ 2 G V which shows
that U 0 D U \ J and V 0 D V \ J are non-empty disjoint closed subsets of J . Since
U [ V D X and X is dense in Y , we have U [ V D Y and therefore U 0 [ V 0 D J .
The set J being homeomorphic to I , there exist non-empty complementary closed
subsets A; B I . Therefore A is clopen in I and A f;; I g which contradicts
Fact 1. Consequently, there is no U 2 C.X / such that F U X nG which shows

that X is not strongly zero-dimensional and finishes our solution.
T.310. Prove that a closed subspace of a strongly zero-dimensional space is strongly
zero-dimensional. Give an example of a strongly zero-dimensional space X such
that some Y X is not strongly zero-dimensional.
Solution. Given a space Z, let C.Z/ be the family of all clopen subsets of Z.
Suppose that Z is a strongly zero-dimensional space and Y is a closed subspace
of Z. It suffices to show that IndY D 0 by Problem 308, so take any closed subset
F of the space Y and let U 2 .F; Y /. There is V 2 .Z/ such that V \ Y D U ;
it is evident that F is also closed in Z and V 2 .F; Z/. Since IndZ D 0, there is
W 0 2 C.Z/ such that F W 0 V . It is immediate that W D W 0 \ Y 2 C.Y /;
since F W U , we established that IndY D 0 and hence Y is strongly zero-
dimensional.
It was proved in Problem 309 that there exists a zero-dimensional space Y such
that Y is not strongly zero-dimensional. There exists a cardinal such that Y
embeds in D by Problem 303, so we can consider that Y is a subspace of X D D .
The space X is compact and zero-dimensional by Problem 303, so it is strongly
zero-dimensional by Problem 306. Thus the strongly zero-dimensional space X
contains a subspace Y which is not strongly zero-dimensional.
S
T.311. Let X be a normal space such that X D fXn W n 2 !g, where each
Xn is strongly zero-dimensional and closed in X . Prove that X is strongly zero-
dimensional.
Solution. For every n 2 N, denote by Mn the set f1; : : : ; ng. Given a space Z
denote by C.Z/ the family of all clopen subsets of Z; if U D fU1 ; : : : ; Un g is
an open cover of Z, say that an open cover V of the space Z is a strong precise
refinement of U if V D fV1 ; : : : ; Vn g and V i Ui for all i 2 Mn . If A exp.Z/,
then A D fA W A 2 Ag; if Y Z, then AjY D fA \ Y W A 2 Ag.
Fact 1. Every open cover U D fU1 ; : : : ; Un g of a normal space Z has a strong
precise refinement.
Proof. It is evident that the set F D Zn.U2 [

[ Un / is closed in Z and F U1 ;
besides F [ U2 [

[ Un D Z. Since Z is normal, there is V1 2 .F; Z/ such that
F V1 V 1 U1 ; it is clear that V1 [ U2 [

[ Un D Z.
Assume that 1 k < n and we have open sets V1 ; : : : ; Vk such that V i Ui
and V1 [

[ Vi [ Ui C1 [

[ Un D Z for any i k. It is clear that the set
F D Zn.V1 [

[ Vk [ UkC2 [

[ Un / is closed in Z and F UkC1 . Since Z
is normal, there is VkC1 2 .Z/ such that F VkC1 V kC1 UkC1 ; it is evident
that V1 [

[VkC1 [UkC2 [

[Un D Z and hence our inductive construction can
be continued to obtain a family fV1 ; : : : ; Vn g .Z/ such that V i Ui for every
i n and V1 [

[ Vn D Z. Thus V D fV1 ; : : : ; Vn g is a strong precise refinement
of U and Fact 1 is proved. t
u
Fact 2. For every open cover U D fU1 ; : : : ; Un g of a strongly zero-dimensional

S
space Z, there is a disjoint family V D fV1 ; : : : ; Vn g C.Z/ such that V D Z
and Vi Ui for all i 2 Mn .
Proof. There is a Sdisjoint refinement W of the cover U. Observe S first that for any
W 0 W, the set W 0 is clopen in Z because its complement .WnW 0 / is open
in Z. For every WS2 W, there is O.W / 2 U such that W O.W /. For each
i 2 Mn , let Vi D fW 2 W W O.W S/ D Ui g. By our first observation, each Vi is
clopen in Z; since V1 [

[ Vn D W D Z and Vi Ui for every i 2 Mn , the
family V D fV1 ; : : : ; Vn g is as promised and Fact 2 is proved. t
u
Returning to our solution, take any open cover U D fU1 ; : : : ; Un g of the space
X . Since X0 is strongly zero-dimensional, we can apply Fact 2 to find a family
V D fV1 ; : : : ; Vn g C.X0 / such that V1 [

[ Vn D X0 and Vi Ui \ X0 for all
i 2 Mn . It is easy to see that the set Ui0 D .Ui nX0 / [ Vi is open in X for all i 2 Mn ;
besides, the family fUi0 \ X0 W i 2 Mn g is disjoint because so is V and Ui0 \ X0 D Vi
for all i 2 Mn .
Another easy observation is that the family U 0 D fUi0 W i 2 Mn g is a cover of
X and therefore Fact 1 can be applied to conclude that there exists a strong precise
refinement U0 D fU10 ; : : : ; Un0 g of the cover U 0 . Furthermore, Ui0 \X0 Ui0 \X0 D
Vi for all i 2 Mn and hence U 0 jX0 is disjoint.
Now assume that k 2 ! and we have a collection fUi W i 2 f0; : : : ; kgg of open
covers of X such that
(1) Ui D fU1i ; : : : ; Uni g for all i k;
(2) U0 is a strong precise refinement of U and Ui C1 is a strong precise refinement
of Ui for all i < k;
(3) U i jXi is disjoint for any i k.
Since the space XkC1 is strongly zero-dimensional and Uk jXkC1 is an open Scover
of XkC1 , there is a disjoint family V D fV1 : : : ; Vn g C.XkC1 / such that V D
XkC1 and Vi Uik \ XkC1 for all i n (see Fact 2). It is straightforward that
the set Ui0 D .Uik nXkC1 / [ Vi is open in X for all i 2 Mn ; besides, the family
fUi0 \ XkC1 W i 2 Mn g is disjoint because so is V and Ui0 \ XkC1 D Vi for all
i 2 Mn .
Another easy observation is that the family U 0 D fUi0 W i 2 Mn g is a cover of
X and therefore Fact 1 can be applied to conclude that there exists a strong precise
refinement UkC1 D fU1kC1 ; : : : ; UnkC1 g of the cover U 0 . It is immediate that the
properties (1) and (2) hold for the collection fU0 ; : : : ; UkC1 g; the property (3) holds
because UikC1 \ XkC1 Ui0 \ XkC1 D Vi for all i 2 Mn and hence U kC1 jXkC1 is
disjoint.
Therefore we can construct inductively a collection fUi W i 2 !g of open covers
T j
of X such that (1)(3) are satisfied for all k 2 !. Given i 2 Mn let Oi D fUi W
j
j 2 !g; it is clear that Oi is a closed subset of X and Oi Ui for all j 2 ! which,
together with (2), implies that Oi Ui . If x 2 X , then x 2 Xn for some n 2 !;
given distinct i; j 2 Mn , we have Oi \ Oj Uin \ Ujn and hence x Oi \ Oj
because Uin \ Ujn \ Xn D ; by (3). This shows that the family O D fO1 ; : : : ; On g
is disjoint. On the other hand, the family Ui is a cover of X and hence there is
m.i / 2 Mn such that x 2 Um.ii
/ for every i 2 !. There is an infinite A ! and
m 2 Mn such that m.i / D m for all i 2 A. It follows from (2) that x 2 Umi for all
T T
i 2 ! and hence x 2 fUmi W i 2 !g D fUmi W i 2 !g D Om .
Therefore O is a disjointScover of X ; observe that every O 2 O is open in
X because its complement .OnfOg/ is closed in X . Thus O is a disjoint open
refinement of U which proves that X is strongly zero-dimensional and makes our
solution complete.
T.312. Prove that there exists a space X which is not zero-dimensional while X D
S
fXn W n 2 !g, where each Xn is strongly zero-dimensional and closed in X .
Solution. For every n 2 N denote by Mn the set f1; : : : ; ng. Given a space Z let
C.Z/ be the family of all clopen subsets of Z. The space Z is called connected
if C.Z/ D f;; Zg. If A is a set, then Fin.A/ is the family of all non-empty finite
subsets of A.
Fact 1. (a) If Z is a connected space and jZj > 1, then Z is not zero-dimensional.
(b) If Z is a space such that some connected Y Z is dense in Z, then Z is
connected.
Proof. Assume that Z is connected and zero-dimensional; if a; b are distinct points
of Z, then there exists a set U 2 C.Z/ such that a 2 U and b U . It is evident that
U f;; Zg; this contradiction shows that (a) is true.
Now assume that Y is a connected dense subspace of Z. If U 2 C.Z/nf;; Zg,
then U 0 D U \Y ;. Moreover, ZnU 2 .Z/ and hence Y nU 0 D .ZnU /\Y
; which shows that U 0 2 C.Y /nf;; Y g; the obtained contradiction settles (b) and
completes the proof of Fact 1. t
u
Fact 2. The space .A/ D fx 2 RA W jfa 2 A W x.a/ 0gj < !g is connected for
any infinite set A.
Proof. Suppose that U 2 C..A//nf;; .A/g. Then V D .A/nU is a non-empty
clopen subset of .A/. Take any points x 2 U and y 2 V and consider the set
J D ftx C .1 t/y W t 2 0; 1 g RA . It is evident that J .A/ and hence
U \ J and V \ J are non-empty disjoint clopen subsets of J which shows that
U \ J ; and U \ J J , i.e., J is not connected.
Since RA can be identified with Cp .A/ where A carries the discrete topology,
we can apply Fact 1 of S.301 to conclude that J is homeomorphic to 0; 1 R.
It was proved in Fact 1 of T.309 that any interval p; q R is connected, so the
connected space 0; 1 is homeomorphic to the space J which is not connected. This
contradiction shows that .A/ is connected and proves Fact 2. t
u
S B note first that jR j D c for any finite non-empty

B
set B c and hence j fR W B 2 Fin.c/gjS D c. Therefore we can choose an
enumeration fs W < cg of the set E D fRB W B 2 Fin.c/g such that each s 2 E
occurs c-many times in this enumeration, i.e., jf < c W s D sgj D c for any s 2 E
(see Fact 3 of S.286). For each < c, let B 2 Fin.c/ be set for which s 2 RB .
S It is easy to construct a disjoint family T D fT W < cg exp c such cthat
T D c and jT j D c for every 2 c. Given < c, define a point x 2 R as
follows: x jB D s ; if 2 T nB , then x ./ D 1, and if 2 T
nB for some
, then x ./ D 0. We claim that the space X D fx W < cg (considered

with the topology induced from Rc ) is as promised. The proof is not easy and will
be done in several steps.
Let Xn D fx 2 X W jB jSD ng for any n 2 N; since the set B is finite for
every < c, we have X D fXn W n 2 Ng. Given any x D x 2 X , choose
any distinct 1 ; : : : ; nC1 2 T nB and let Ux D fx 2 X W x .i / > 0 for all
i 2 MnC1 g. It is evident that Ux 2 .x; X/. If and x 2 Xn , then x .
/ D 0
for all
2 T nB . Since jB j D n, there is i 2 MnC1 such that i B and hence
x .i / D 0.
This shows that x Ux if x 2 Xn and , i.e., Ux \ Xn fx g. Thus
every point of X has a neighborhood which contains at most one point of Xn and
hence Xn is closed and discrete in X for every n 2 N. It is evident that every
discrete space is strongly zero-dimensional, Sso Xn is strongly zero-dimensional for
all n 2 N. Letting X0 D X1 we have X D fXn W n 2 !g where Xn is closed in X
and strongly zero-dimensional for all n 2 !.
For any A c let A W X ! RA be the projection of X to the face RA defined
by A .x/ D xjA for any x 2 X . Observe first that X is dense in Rc . Indeed,
given a non-empty finite B c and s 2 RB , there is < c with B D B and
s D s . Recalling the definition of x , we can see that B .x / D s. This proves that
B .X / D RB for any B 2 Fin.c/; it is an easy exercise that this implies density of
X in Rc .
To prove that X is connected, assume that U 2 C.X /nf;; X g. Then V D X nU
is a non-empty clopen subset of X ; let f .x/ D 0 for all x 2 U and f .x/ D 1 for
every x 2 V . It is immediate that f W X ! D is a continuous onto map, so we
can apply Problem 299 of [TFS] to find a countable A c and a continuous map
g W A .X / ! D such that f D g A . Since f .X / D g.A .X // D D, the sets
U 0 D g 1 .0/ and V 0 D g 1 .1/ are non-empty disjoint and clopen in A .X /. An
immediate consequence is that A .X / is not connected.
Next we prove that A .X / .A/. Take any z 2 .A/ and a set B 2 Fin.c/
such that f < c W z./ 0g B. The set Q D f < c W T \ A ;g
is countable because the family T is disjoint. By the choice of our enumeration,
there is < c such that T \ A D ;; B D B and s D zjB. Then x jB D
s jB D zjB and x ./ D 0 for any 2 AnB because T . This proves that
A .x / D z and hence z 2 A .X /. The point z 2 .A/ was chosen arbitrarily,
so we proved that .A/ A .X /. The space .A/ is connected by Fact 2 and
dense in RA (this is an easy exercise) and hence .A/ is dense in A .X /. A space
which contains a dense connected subspace is itself connected by Fact 1, so A .X /
is connected which is a contradiction. Therefore X not zero-dimensional because
it is connected (see Fact 1). We finally proved that X is a connected (and hence
nonzero-dimensional) space which is a countable union of its closed discrete (and
hence strongly zero-dimensional) subspaces, so our solution is complete.
T.313. Prove that the space P of the irrationals is homeomorphic to ! ! and hence
P is zero-dimensional.
Solution. Given a function f the set dom.f / is its domain; if f and g are
functions, then f g says that dom.f / dom.g/ and gjdom.f / D f . Note
that if f is an empty function (in this case we write f D ;), then f g for any
function g.
Now suppose that we have a collection of functions ffi W i 2 I g such that
fi j.dom.fi / \ dom.fj // D fj j.dom.fi /S \ dom.fj // for any i; j 2 I . Then we
can define a function f with dom.f / D i 2I dom.fi / as follows: given any x 2
dom.f /, find any i 2 I with x 2 dom.fi / and let f .x/ D fi .x/. It is easy to check
that the value of f at x does not depend on the choice
S of i , so we have consistently
defined a function f which will be denoted by ffi W i 2 I g (this makes sense if
we identify each function with its graph). S
Let ! 0 D f;g S and denote by ! <! the set f! n W n 2 !g. We will also need
the set ! n D f! k W 0 k ng for every n 2 !. Given any s 2 ! <! , the set
s D ff 2 ! ! W f jn D sg is open in ! ! and the family ff jn W n 2 !g is a local
base at f in ! ! . Suppose that k 2 ! and s 2 ! k ; for any n 2 !, define the function
t D s _ n by tjk D s and t.k/ D n.
If .X; / is a metric space, call U .X / a ! <! -directed family if
(1) U D fUs W s 2 ! <! g;
(2) the family U.n/ D fUs W s 2 ! n g is disjoint for any n 2 !;
(3) for any s; t 2 ! <! if s t and s t, then U t Us ;
(4) diam.Us / n1 for any s 2 ! n and n 2 N.
Fact 1. If .X; / is aTcomplete
S metric space and U is a ! <! -directed family in
.X; /, then L.U/ D f U.n/ W n 2 !g is homeomorphic to ! ! .
Proof. For any f 2 ! ! consider the sequence fUf jn W n 2 !g. It follows from (4)
T T
that diam.Uf jn / ! 0 and hence Qf D fUf jn W n 2 !g D fU f jn W n 2 !g ;
(we applied (3) and Problem 236 of [TFS]). It follows from (4) that there is a point
xf 2 X such that Qf D fxf g; let '.f / D xf for any f 2 ! ! .
It is immediate from the definition
S of ' that '.! ! / L.U/. If x 2 L.U/, then
for any n 2 ! we have x 2 U.n/ and hence there is sn 2 ! n such that x 2 Usn ;
it follows from (2) that this sn is unique for every n 2 !. Furthermore, if n < m,
then xS 2 Usm jn \ Usn [see (3)] and hence sm jn D sn by (2). Thus the function
f D fsn W n 2 !g 2 ! ! is well-defined and sn D f jn for all n 2 ! which shows
that x D xf . As a consequence, L.U/ '.! ! / and hence L.U/ D '.! ! /.
To see that the mapping ' is continuous, take any f 2 ! ! and " > 0. Choose
n 2 N such that n1 < " and observe that '.f jn / Uf jn B .'.f /; "/ because
diam.Uf jn / n1 < " by (4). Thus the open neighborhood f jn of the point f in
! ! witnesses continuity of ' at the point f .
Now, if f; g are distinct points of ! ! , then there is n 2 ! such that f jn gjn; by

the definition of ', we have '.f / 2 Uf jn and '.g/ 2 Ugjn . Since Uf jn \Ugjn D ; by
(2), we have '.f / '.g/, i.e., ' is a bijection. To finally see that ' 1 is continuous
take any x 2 L.U/ and let f D ' 1 .x/. Given any O 2 .f; ! ! / there is n 2 !
such that f jn O. Furthermore, V D Uf jn \ L.U/ 2 .x; L.U//; if y 2 V and
g D ' 1 .y/, then y D '.g/ and hence y 2 Ugjn \ Uf jn . Applying property (2)
again we conclude that f jn D gjn and hence g 2 f jn O. The point y 2 V
was chosen arbitrarily, so we proved that ' 1 .V / f jn O, i.e., V witnesses
continuity of ' 1 at the point x. Thus ' is a homeomorphism and Fact 1 is proved.
t
u
Returning to our solution, take a faithful enumeration fOn W n 2 !g of the family
f.n; n C 1/ W n 2 !g [ f. n 1; n/ W n 2 !g; it is easy to choose an enumeration
fqn W n 2 !g of the set Q such that q0 D 0. Our first step is to define U; D R; for
any f 2 ! 1 , let Uf D Of .n/ . Assume that n 2 N and we have defined a family
fUs W s 2 ! n g such that
(5) the set Us is a non-empty open interval with rational endpoints for any S s ;;
(6) the family U.k/ D fUs W s 2 ! k g is disjoint and fq0 ; : : : ; qk g Rn. U.k//
Q whenever 1 k n;
(7) for any s; t 2 ! n if s t and s t then U t Us ;
(8) if 1 k n, then diam.Us / k1 for any s 2 ! k .
To construct the family U.nC1/, it suffices to define Us _ k for all s 2 ! n and k 2
!. Fix any s 2 ! n ; by the induction hypothesis, Us D .a; b/ where a; b 2 Q. Thus
b0 D aCb2
2 Q and the diameters of both intervals .a; b0 / and .b0 ; b/ do not exceed
2
n nC1 . Choose a strictly decreasing sequence fan W n 2 Ng Q \ .a; b0 /
1 1 1
and a strictly increasing sequence fbn W n 2 Ng Q \ .b0 ; b/ such that an ! a

and bn ! b. Let a0 D b0 and observe that if qnC1 2 Us , then we can choose our
sequences in such a way that qnC1 be listed either in As D fan W n 2 !g or in
Bs D fbn W n 2 !g. Choose a faithful enumeration fWn W n 2 !g of the family
f.anC1 ; an / W n 2 !g [ f.bn ; bnC1 / W n 2 !g and let Us _ n D Wn for each n 2 !.
Sall k 2 ! and s 2 ! , we obtain a family nfUs W s 2
n
After defining Us _ k for
! g. Observe that Us n. fWn W n 2 !g/ D As [ Bs Q for all s 2 ! ; besides,
nC1
if qnC1 2 Us , then qnC1 2 As [ Bs .

This guarantees that (6) is fulfilled for the family fUs W s 2 ! nC1 g; it is evident
that (5), (7) and (8) are also satisfied, so we can continue our inductive construction
to obtain a family U D fUs W s 2 ! <! g with the properties (5)(8). It is evident
that U is ! <! -directed and hence L.U/ is homeomorphic to ! ! by Fact 1. It easily
follows from (6) that L.U/ D P which proves that P is homeomorphic to ! ! .
Finally, observe that the discrete space ! is zero-dimensional and hence so is ! ! by
Problem 302. This shows that P is also zero-dimensional being homeomorphic to a
zero-dimensional space ! ! and completes our solution.
T.314. Let X be a paracompact space. Prove that X is strongly zero-dimensional if
and only if every open cover of X has a disjoint open refinement.
Solution. For sufficiency observe that it follows from the definition of a strongly
zero-dimensional space that if any cover of X has a disjoint open refinement, then
X is strongly zero-dimensional (paracompactness of X is not needed here). Now
assume that X is paracompact and strongly zero-dimensional and let C.X / be the
family of all clopen subsets of X .
If U is an open cover of X , then there exists a locally finite refinement V of
the cover U. Let us index V by some well-ordered set .S; </, i.e., consider that
V D fVs W s 2 S g. There exists a closed cover fFs W s 2 S g of the space X
such that Fs Vs for every s 2 S (see Fact 2 of S.226). Since IndX D 0 by
308, there exists Os 2 C.X / such that Fs Os Vs for all s 2 S . It is evident
S O D fOs W s 2 S g is a locally finite refinement of U. Now let
that the family
Ws SD Os n. fOt W t < sg/ for every s 2 S . The family O is closure-preserving,
so O0 is closed (and, evidently, open) for any O0 O. This proves that Ws is a
clopen set for any s 2 S . It is immediate that the family W D fWs W s 2 S g is a
disjoint open refinement of U so we settled necessity.
T.315. Let P be a strongly zero-dimensional paracompact space and suppose that
M is a completely metrizable space. Denote by CL.M / the set of all closed non-
empty subsets of M and let ' W P ! CL.M / be a lower semicontinuous map.
Prove that ' has a continuous selection, i.e., there exists a continuous map f W
P ! M such that f .x/ 2 '.x/ for any x 2 P .
Solution. Choose a complete metric d on the space M such that d.x; y/ 1 for
any x; y 2 M (such a choice is possible by Problem 206 of [TFS]). For any set
A M , let ' 1 .A/ D 'l1 .A/ D fx 2 P W '.x/ \ A ;g. Then ' 1 .U /
is open in P for any U 2 .M /. For each fixed non-empty A M , define the
map dA W M ! R by dA .x/ D inffd.x; y/ W y 2 Ag; then, for any closed set
A M , we have x 2 A if and only if dA .x/ D 0. If we are given functions
f; g 2 C.P; M / D C .P; M /, let .f; g/ D supfd.f .x/; g.x// W x 2 P g. Then
is a complete metric on C.P; M / (see Problem 248 of [TFS]).
Take a point u0 2 M and let f0 .x/ D u0 for all x 2 P . The function f0 W P !
M is constant and hence continuous. Now assume that k 2 ! and we have defined
continuous functions f0 ; : : : ; fk from P to M with the following properties:
(1) fn1 .u/ is a clopen (maybe empty) subset of P for any n k and u 2 M ;
(2) .fn ; fnC1 / 21n for any n < k;
(3) for all n k and x 2 P there is point y 2 '.x/ such that d.fn .x/; y// 21n .
Apply the property (3) to choose, for any x 2 P , a point yx 2 '.x/ such that
d.fk .x/; yx / 21k . Then Ox D ' 1 .Bd .yx ; 2kC1
1
// 2 .x; P / for any x 2 P .
Fix any w 2 M ; the family Vw D fOx \ fk .w/ W x 2 fk1 .w/g is an open
1
cover of the set fk1 .w/ which is clopen in P and hence paracompact and strongly
zero-dimensional (see Problem 310). Apply Problem 314 to find a disjoint open
refinement Uw of the cover Vw . For any U 2 Uw , there is x.U / 2 fk1 .w/ such that
U Ox.U / ; let fkC1
w
.x/ D yx.U / for any x 2 fk1 .w/. Observe that
w
. / d.fkC1 .x/; fk .x// D d.fkC1
w
.x/; w/ D d.yx.U / ; w/ D d.yx.U / ; fk .x.U ///
1
2k
for any x 2 fk1 .w/. After we have fkC1 w

for all w 2 M (observe that nothing has
1
to be done if fk .w/ D ;), we can define a map fkC1 W P ! M as follows:
for any x 2 P let w D fk .x/ and fkC1 .x/SD fkC1 w
.x/. It is immediate that the
function fkC1 jU is constant for any U 2 fUw W w 2 M g; therefore fkC1 is
1
continuous by Fact 1 of S.472 and it is easy to see that fkC1 .u/ is open for any
u 2 M . Consequently, (1) holds for n D k C 1. Furthermore, it follows from . /
that d.fkC1 .x/; fk .x// 21k for all x 2 P and hence (2) is fulfilled for k D n.
Given any x 2 P let w D fk .x/; then there is U 2 Uw such that x 2 U Ox.U /
and hence E D '.x/ \ Bd .yx.U / ; 2kC1 1
/ ;. If y 2 E, then d.fkC1 .x/; y/ D
d.yx.U / ; y/ < 2kC1 , so (3) is fulfilled as well for n D k C 1.
1
As a consequence, our inductive construction can be continued to provide us a

sequence S D ffn W n 2 !g for which (1)(3) are satisfied for all k 2 !. It is
an easy exercise to check that (2) implies that S is a Cauchy sequence in the space
.C.P; M /; / and hence there is a function f 2 C.P; M / such that, for all x 2 P ,
we have f .x/ D lim fn .x/ when n ! 1.
Fix any x 2 P and let A D '.x/. It follows from (3) that dA .fn .x// 21n for all
n 2 !; the function dA being continuous by Fact 1 of S.212, we have dA .f .x// D
0 and therefore f .x/ 2 A D '.x/. Thus f is as promised and our solution is
complete.
T.316. Let M be a strongly zero-dimensional completely metrizable space. Prove
that any closed non-empty F M is a closed retract of M , i.e., there exists a
closed continuous map f W M ! F such that f .x/ D x for any x 2 F .
Solution. It is easy to prove that there exists some retraction of M onto F . To
see it denote by CL.F / the family of all non-empty closed subsets of F ; define
a map ' W M ! CL.F / as follows: '.x/ D fxg if x 2 F and '.x/ D F
for any x 2 M nF . The map ' is lower semicontinuous for if U 2 .F /, then
'l1 .U / D U [ .M nF / is an open subset of M . The space M is paracompact by
Problem 218 of [TFS] and F is completely metrizable, so we can apply Problem
315 to conclude that there is a continuous g W M ! F such that g.x/ 2 '.x/ for
all x 2 X . In particular, g.x/ 2 '.x/ D fxg and hence g.x/ D x for any x 2 F ,
i.e., g is a retraction of M onto F . However, it takes a much harder proof to actually
find a closed retraction of M onto F .
Fact 1. If Y is a sequential space, then a continuous onto map f W Z ! Y is closed
if and only if f .D/ is not a nontrivial convergent sequence for any closed discrete
D Z.
Proof. If f is closed and D Z is closed and discrete, then f .D/ is also closed
and discrete, so it cannot be a nontrivial convergent sequence.
To prove sufficiency, assume that for any closed discrete D Z the set f .D/
is not a nontrivial convergent sequence. If f is not closed, then there is a closed
A Z such that the set f .A/ is not closed in the space Y and hence there is a
sequence S D fyn W n 2 !g f .A/ such that yn ! y for some y 2 Y nf .A/.

Take zn 2 f 1 .yn / \ A for every n 2 !; then the set D D fzn W n 2 !g is
closed and discrete in Z while f .D/ D S is a nontrivial convergent sequence. This
contradiction shows that f is a closed map and hence Fact 1 is proved. t
u
Returning to our solution, fix a complete metric d on the space M such that
.d / D .M / and d.x; y/ 1 for all x; y 2 M . Given functions p; q 2 C.M; F /,
let .p; q/ D supfd.p.x/; q.x// W x 2 M g. The function is a complete metric on
C.M; F / by Problem 248 of [TFS] (observe that C.M; F / D C .M; F / because
.F; d j.F F // is a bounded metric space).
Choose x0 2 F ; let U0 D fM g and f0 .x/ D x0 for all x 2 M . Assume that
k 2 ! and we have chosen maps f0 ; : : : ; fk and disjoint open covers U0 ; : : : ; Uk of
the space M with the following properties:
(1) fi W M ! F for all i k;
(2) for any i k and U 2 Ui , there is yUi 2 F such that fi .x/ D yUi for all x 2 U ;
(3) Ui C1 is a refinement of Ui for any i < k;
(4) For any i k, if U 2 Ui and U \ F ;, then yUi 2 U \ F .
(5) for any i < k, if U 2 Ui C1 and U \ F D ;, then yUi C1 D yVi where V is the
unique element of Ui for which U V ;
(6) diam.U / 21i for all U 2 Ui and i k.
Observe that the property (2) implies that each fi is continuous (see Fact 1 of
S.472) and fix any U 2 Uk ; since M is strongly zero-dimensional, there is a disjoint
open cover VU of the space U such that diam.V / 2kC1 1
for every V 2 VU . If
V \ F D ;, then let fkC1 .x/ D yU for all x 2 V ; if V \ F ;, then choose any
k
yVkC1 2 V \ F and let fkC1 .x/ D yVkC1 for all x 2 V . This defines fkC1 .x/ for all
x 2 U ; the set USwas chosen arbitrarily, so we can define fkC1 .x/ for all x 2 M .
Letting UkC1 D fVU W U 2 Uk g, we obtain the desired disjoint open cover of the
space M . It is evident that the properties (1)(6) still hold for the maps f0 ; : : : ; fkC1
and covers U0 ; : : : ; UkC1 . Thus our inductive construction can be continued to obtain
sequences ffi W i 2 !g and fUi W i 2 !g for which (1)(6) are fulfilled for all k 2 !.
Let us prove by induction on i that
(7) the set fi .M / is closed and discrete (in F and hence in M ) for all i 2 !.
This is true for i D 0 because f .M / D fx0 g. Now if n < ! and we proved that
fn .M / is closed Sand discrete, then consider the family V D fV 2 UnC1 W V \ F
;g and let O D V. It is clear that the set fnC1 .O/ D fyVnC1 W V 2 Vg is closed
and discrete because the family V is discrete and yVnC1 2 V for all V 2 V by (4).
The property (5) implies that fnC1 .M nO/ fn .M nO/; since fn .M nO/ is closed
and discrete by the induction hypothesis, the set fnC1 .M nO/ is closed and discrete
as well. Thus fnC1 .M / D fnC1 .O/ [ fnC1 .M nO/ is also closed and discrete so
(7) is proved. Next, let us show that
(8) .fi ; fi C1 / 1
2i
for all i 2 !.
Take any x 2 M ; there are U 2 Ui and V 2 Ui C1 such that x 2 U \V . It follows

from (3) that V U . If V \F ;, then fi C1 .x/ D yVi C1 2 V \F by (4) and hence
fi C1 .x/ 2 U . If V \ F D ;, then fi C1 .x/ D yUi D fi .x/. As a consequence, in
the second case, we have d.fi .x/; fi C1 .x// D 0 while in the first one yUi 2 U \ F
because U \F ; and hence d.fi .x/; fi C1 .x// D d.yVi C1 ; yUi / diam.U / 21i .
Thus d.fi .x/; fi C1 .x// 21i for all x 2 M and hence (8) is proved.
It is an easy consequence of (8) that the sequence T D ffn W n 2 !g C.M; F /
is fundamental in the complete metric space .C.M; F /; /. Therefore there exists
a function f 2 C.M; F / such that fn converges to f in .C.M; F /; / which is
equivalent to uniform convergence of T to f . It is an easy exercise to check, using
the property (8) that
(9) d.fn .x/; f .x// 1
2n1
for all x 2 M and n 2 N.
If x 2 F , then let Ui be the unique element of Ui which contains x; then Ui \F
; for all i 2 ! and hence fi .x/ D yUi i 2 Ui by the property (4). This implies
d.x; fi .x// diam.Ui / 21i and hence the sequence ffi .x/ W i 2 !g converges
to x. Therefore f .x/ D lim fi .x/ D x for any x 2 F and hence f is a retraction
i !1
of the space M onto F .
To see that f is a closed map assume that D D fdn W n 2 !g is a closed discrete
subset of M such that f .D/ is a nontrivial convergent sequence. We can assume that
di dj if i j . If D \ F is infinite, then it is closed and discrete and hence f .D/
contains a closed discrete subspace f .D \ F / D D \ F of the space F which is
a contradiction with the fact that f .D/ is a nontrivial convergent sequence. Thus
D \ F is finite, so if necessary, we can throw away finitely many point of D to
guarantee that D \ F D ;. Let z 2 F be the limit of f .D/; since K D f .D/ [ fzg
is a compact subspace disjoint from D,
(10) there is " > 0 such that d.x; y/ " for any x 2 K and y 2 D.
Pick n 2 ! with 21n < 4" and take any x 2 D; there is U 2 Un with x 2 U . If
U \ F ;, then fn .x/ D yUn 2 U and hence d.x; fn .x// diam.U / 21n . It
follows from (9) that d.x; f .x// d.x; fn .x// C d.fn .x/; f .x// 2n1
1
C 21n <
4 < " which is a contradiction with (10). As a consequence, U \ F D ; whenever
3"
U 2 Un and U \ D ;. If i n and V \ D ; for some V 2 Ui , then there is

U 2 Un with V U and hence V \ F U \ F D ;. This proves that
(11) for any i n if U 2 Ui and U \ D ;, then U \ F D ;.
It is easy to prove by an evident induction applying (11) and the property (5) that
fi .x/ D fn .x/ for all x 2 D and i n. Thus f .x/ D limk!1 fk .x/ D fn .x/ for
all x 2 D. As a consequence, f .D/ D fn .D/ is a closed discrete subset of F by
(7) which is a contradiction. Thus f .D/ cannot be a nontrivial convergent sequence
and hence the map f is closed by Fact 1. Our solution is complete.
T.317. Let F be a closed subspace of P and suppose that a space X is a continuous
image of F . Prove that X is also a continuous image of P.
Solution. Let g W F ! X be a continuous onto map. The space P is zero-

dimensional by Problem 313; being second countable, it is also strongly zero-
dimensional by Problem 306, so we can apply Problem 316 to conclude that there
is a continuous onto map f W P ! F (which is a retraction but we dont need that).
Then g f maps P continuously onto X .
T.318. Prove that for any second countable space X and every countable ordinal
, there exists a set U 2 0 .X K/ such that 0 .X / D fU y W y 2 Kg,
where U y D fx 2 X W .x; y/ 2 U g for any y 2 K. Observe that, as an easy
consequence, for any second countable space X and every countable ordinal ,
there exists a set V 2 0 .X K/ such that 0 .X / D fV y W y 2 Kg.
Solution. We consider that K D D! (see Problem 128 of [TFS]). Given an ordinal
< !1 and a space Z, a set U Z K will be called 0 .Z/-universal if U 2
0 .Z K/ and 0 .Z/ D fU y W y 2 Kg. Analogously, a set V Z K is called
0 .Z/-universal if V 2 0 .Z K/ and 0 .Z/ D fU y W y 2 Kg.
Fact 1. Given spaces Z and Y let f W Z ! Y be a continuous map. Then, for
any < !1 and A 2 0 .Y / (or A 2 0 .Y /), we have f 1 .A/ 2 0 .Z/ (or
f 1 .A/ 2 0 .Z/ respectively).
Proof. Since f is continuous, f 1 .U / is open (closed) in Z for any open (closed)
U Y . In terms of Borel subsets this means that A 2 00 .Y / (or A 2 00 .Y /)
implies f 1 .A/ 2 00 .Z/ (or f 1 .A/ 2 00 .Z/ respectively). Suppose that < !1
and, for any < , we proved that f 1 .A/ 2 0 .Z/ (f 1 .A/ 2 0 .Z/) for any
A 2 0 .Y / (or A 2 0 .Y / respectively).
Now, if A 2 0 .Y /, then there is aSsequence fn W n 2 !g and An 2
n .Y / for every n 2 ! such that A D fAn W n 2 !g. Thus f 1 .An / 2 0n .Z/
0
S
for all n 2 ! by the induction hypothesis so f 1 .A/ D ff 1 .An / W n 2 !g 2
0 .Z/. Thus f 1 .A/ 2 0 .Z/ for any set A 2 0 .Y /. Observe finally that if
A 2 0 .Y /, then B D Y nA 2 0 .Y / which implies f 1 .B/ 2 0 .Z/. As a
consequence, f 1 .A/ D Znf 1 .B/ 2 0 .Z/ and hence our inductive proof can
be carried out for all < !1 . Fact 1 is proved. t
u
Fact 2. Suppose that Z is a space and Y Z. If is a countable ordinal and
A 2 0 .Z/ (A 2 0 .Z/), then A \ Y 2 0 .Y / (or A \ Y 2 0 .Y / respectively).
Proof. If A is open (closed) in Z, then A \ Y is open (closed) in Y , so our fact is
evident for D 0. Assume that < !1 and we proved our fact for all < . If
A 2 0 .Z/, then there is a sequence fn W n 2 !g and An 2 0n .Z/ for all
S
n 2 ! such that A D fAn W n 2 !g. By the induction hypothesis,
S we have Bn D
An \ Y 2 0n .Y / for all n 2 ! and therefore A \ Y D fBn W n 2 !g 2 0 .Y /.
Now, if B 2 0 .Z/, then A D ZnB 2 0 .Z/ and hence C D A \ Y 2 0 .Y /.
Consequently, B \ Y D Y nC 2 0 .Y / and hence our fact holds for all .
Thus our inductive procedure shows that our fact is true for all < !1 and hence
Fact 2 is proved. t
u
Returning to our solution, suppose that for some ordinal < !1 , there exists a
0 .X /-universal set U X K. Then, for the set V D .X K/nU , we have
V 2 0 .X K/ and V y D X nU y for any y 2 K. Therefore
fV y W y 2 Kg D fX nU y W y 2 Kg D fX nW W W 2 0 .X /g D 0 .X /;
i.e., the set V is 0 .X /-universal. Consequently,

(1) if < !1 and there is a 0 .X /-universal set, then there exists a 0 .X /-
universal set as well.
Thus it suffices to construct a 0 .X /-universal set for all < !1 . We will do
this by induction on .
Fix a base B D fBn W n 2 !g in the space X ; let On D fy 2 K W y.n/ D 1g for
all n 2 !. Then On 2 .K/ and hence the set Un D S Bn On is open in the space
X K for every n 2 !. As a consequence, U D fUn W n 2 !g 2 00 .X K/.
For any point y 2 K the natural projection W X K ! X maps X fyg
homeomorphically onto X and hence U y D .U \ .X fyg// is an open subset
of X for any y 2 Y . Thus fU y W y 2 Kg 00 .X /.
Now, if W 2S.X/, then let A D fn 2 ! W Bn W g. Since B is a base of X ,
we have W D fBn W n 2 Ag; let y 2 D! be defined as follows: y.n/ D 1 for all
n 2 A and y.n/ D 0 whenever n 2 !nA. Observe that if x 2 X , then .x; y/ 2 U
if and only if .x; y/ 2 Un for some n 2 ! which takes Splace if and only if n 2 A
and x 2 Bn . Therefore .x; y/ 2 U if and only if x 2 fBn W n 2 Ag D W which
shows that W D U y and hence fU y W y 2 Kg D 00 .X /, i.e., the set U is
00 .X /-universal.
Now assume that < !1 and there exists a 0 .X /-universal set for each < .
Thus there also exists a 0 .X /-universal set for each < by (1). Let Ki be
a homeomorphic copy of the space K for all i 2 ! and choose an enumeration
fn W n 2 !g of the set D f W < g such that each < occurs infinitely
many times in this enumeration.
Un X Kn which is 0n .X /-universal for all n 2 !. In the
There is a set Q
space T D X fKn W n 2 !g let n W T ! X Kn be the natural projection
for every n 2 !. Then n1 .Un / 2 0n .T / for all n 2 ! by Fact 1 and therefore
S Q
U D fn1 .Un / W n 2 !g 2 0 .T / D 0 .X K/ where K D fKn W n 2 !g;
it is evident that K is homeomorphic to K. Let W X K ! X be the natural
projection. Given any y 2 K we have U y D .U \ .X fyg// and j.X fyg/
is a homeomorphism. Since U \ .X fyg/ 2 0 .X fyg/ by Fact 2, we have
U y 2 0 .X / for any y 2 K. Thus fU y W y 2 Kg 0 .X /.
To prove the inverse inclusion take any W 2 0 .X /. There exists a sequence
S
fn W n 2 !g and Wn 2 0n .X / for all n 2 ! such that W D fWn W n 2 !g.
It is easy to choose a sequence S D fm.n/ W n 2 !g ! such that m.n/ D n
for all n 2 !. Since Um.n/ is 0n .X /-universal, there is ym.n/ 2 Km.n/ such that
Um.n/ym.n/ D Wn for all n 2 !. Observe that the empty set belongs to 0 .X / for
any < !1 and hence we can take, for every n 2 !nS , a point yn 2 Kn such that
Un yn D ;. Now, if we let y.n/ D yn for every n 2 !, then y 2 K. Observe that

x 2 U y if and only if .x; y/ 2 n1 .Un / for some n 2 ! which takes place if
and only if n .x; y/ 2 Un , or, in other words, .x; yn / 2 Un , i.e., x 2 Un yn . Since
Un yn W for all n 2 !, we have U y W .
On the other hand, if x 2 W , then x 2 Wn D Um.n/ym.n/ for some n 2 ! and
1
therefore .x; ym.n/ / 2 Um.n/ which shows that .x; y/ 2 m.n/ .Um.n/ / U . Thus
x 2 U y for any x 2 W that x 2 U y and hence W U y , i.e., W D U y .
Thus fU y W y 2 Kg D 0 .X /, so identifying the spaces K and K, we can
conclude that U is 0 .X /-universal. Our inductive procedure shows that we proved
the existence of a 0 .X /-universal set for any ordinal < !1 which demonstrates,
together with (1), that our solution is complete.
T.319. Prove that for any uncountable Polish space X and every countable ordinal
, the classes 0 .X / and 0 .X / do not coincide.
Solution. If Z is a space, A exp.Z/ and Y Z, then AjY D fA \ Y W A 2 Ag.
Given an ordinal < !1 , a set U Z K will be called 0 .Z/-universal if
U 2 0 .Z K/ and 0 .Z/ D fU y W y 2 Kg. Here U y D fz 2 Z W .z; y/ 2 U g
for any y 2 K.
Fact 1. Given a countable ordinal , for an arbitrary space Z and any Y Z, we
have 0 .Z/jY D 0 .Y / and 0 .Z/jY D 0 .Y /.
Proof. The inclusions 0 .Z/jY 0 .Y / and 0 .Z/jY 0 .Y / were proved
in Fact 2 of T.318. It is an immediate consequence of the definition of subspace
topology that .Z/jY D .Y /, i.e., 00 .Z/jY D 00 .Y /. Furthermore, if F is a
closed subset of Y , then clZ .F / \ Y D F which shows that 00 .Z/jY D 00 .Y /.
Now assume that < !1 and we proved the equalities 0 .Z/jY D 0 .Y / and
.Z/jY D 0 .Y / for all < . Given any U 2 0 .Y /, there is a sequence
0
S
fn W n 2 !g and Un 2 0n .Y / for all n 2 ! such that U D fUn W n 2 !g.
By the induction hypothesis, there is Vn 2 0n .Z/ such that Vn \ Y D Un for all
S
n 2 !. Then V D fVn W n 2 !g 2 0 .Z/ and V \ Y D U which shows that
0 .Z/jY D 0 .Y /. Passing to complements, we obtain 0 .Z/jY D 0 .Y /, so
our inductive proof is valid for all < !1 . Fact 1 is proved. t
u
Returning to our solution T observe that the space X is a continuous image of P by
Fact 3 of T.132; since P D fRnfqg W q 2 Qg, the set P is G in R. Every open
subset of R is a countable union of compact sets, so P is a K -space, and hence X
is K-analytic. Therefore we can apply Fact 4 of T.250 to conclude that K embeds
in X . If 0 .X / D 0 .X / for some < !1 , then we can apply Fact 1 to conclude
that
(1) 0 .K/ D 0 .X /jK D 0 .X /jK D 0 .K/.
Let D f.y; y/ W y 2 Kg be the diagonal of the space K. Now apply Problem
318 to find a set U K K which is 0 .K/-universal. Then V D .K K/nU 2
0 .K K/ and hence V \ 2 0 ./ by Fact 1. Let W K K ! K be the
natural projection onto the first factor. It is evident that j is a homeomorphism
of onto K and therefore W D .V \ / 2 0 .K/. The property (1) shows that
W 2 0 .K/ and, by universality of U , there is y 2 K such that U y D W .
If y 2 W , then by the definition of W , we have .y; y/ U . On the other hand,
y 2 W D U y which shows that .y; y/ 2 U which is a contradiction. Now, if
y W , then .y; y/ V and hence .y; y/ 2 U . Therefore y 2 U y D W ;
this final contradiction shows that the equality 0 .K/ D 0 .K/ is impossible and
hence the equality 0 .X / D 0 .X / is impossible as well.
T.320. Let X be a second countable space. Given countable ordinals and > ,
prove that 0 .X / [ 0 .X / 0 .X /. Show that if X is an uncountable Polish
space, then 0 .X / [ 0 .X / 0 .X /.
Solution. If Z is a space and A exp.Z/, then AjY D fA \ Y W A 2 Ag for every
Y Z. The expression Z ' T says that the spaces Z and T are homeomorphic.
Fact 1. Let Z be a second countable space. Then, for any closed Y Z, we have
0 .Y / 0 .Z/ for all countable ordinals > 0 and 0 .Y / 0 .Z/ for all
< !1 . As a consequence, 0 .Z/jY D 0 .Y / for any countable ordinal > 0.
Proof. Any closed subset of Y is closed in Z, so any F -subset of Y is an F -
subset of Z as well. This proves that 00 .Y / 00 .Z/ and 10 .Y / 10 .Z/.
Furthermore, if A 2 10 .Y /, then Y nA is an F -set in Y and hence in X . Any open
set in Z is also an F -set because Z is second countable. Thus ZnA D .Y nA/ [
.ZnY / is an F -set, i.e., ZnA 2 10 .Z/. Consequently, A 2 10 .Z/ and therefore
10 .Y / 10 .Z/.
Assume that 1 < < !1 and we proved that 0 .Y / 0 .Z/ for all < and
.Y / 0 .Z/ whenever 0 < < . Given any A 2 0 .Y /, there is a sequence
0
S
fn W n 2 !g and An 2 0n .Y / for all n 2 ! such that A D fAn W n 2 !g.
By the induction hypothesis, we have An 2 0n .Z/ for all n 2 ! and therefore
A 2 0 .Z/. This proves that 0 .Y / 0 .Z/.
Now if B 2 0 .Y /, then A D Y nB 2 0 .Y / 0 .Z/ and hence A 2 0 .Z/.
ThereSis a sequence fn W n 2 !g and An 2 S n .Z/ for all n 2 ! such that
0
A D fAn W n 2 !g. Furthermore, we have ZnY D fPn W n 2 !g where each Pn

is closed in Z and hence Pn 2 00 .Z/. Choose an enumeration fQn W n 2 !g of the
family fAn W n 2 !g [ fPn W n 2 !g; it is evident that Qn 2 0n where n < for
S
all n 2 !. Thus ZnB D A [ .ZnY / D fQn W n 2 !g and hence ZnB 2 0 .Z/.
This shows that B 2 0 .Z/, i.e., 0 .Y / 0 .Z/, so our inductive procedure
completes the proof the first part of our fact.
It follows from Fact 1 of T.319 that 0 .Z/jY 0 .Y / for all < !1 . Now, if
> 0 and A 2 0 .Y /, then A 2 0 .Z/ and A 2 0 .Z/ by the first part of our
fact. Thus A 2 0 .Z/ and A D A \ Y 2 0 .Z/jY . Therefore 0 .Z/jY D 0 .Y /
u
Returning to our solution, observe that it follows from the inductive definition of
0 .X / that
(1) 0 .X / 0 .X / whenever < < !1 .

Given a family A exp.X /, let Ac D fX nA W A 2 Ag. It is evident that
A B exp.X / implies Ac B c , so it follows from (2) that
(2) 0 .X / 0 .X / whenever < < !1 ,
because 0 .X / D .0 .X //c and 0 .X / D .0 .X //c .

Since X is second countable, every U 2 .X/ is an F -set which shows that
00 .X / 10 .X /. Assume that < !1 and we established, for every ordinal
< , that 0 .X / C1
0
.X /. If U 2 0 , then we can find a sequence fn W n 2
S
!g and Fn 2 n .X / for all n 2 ! such that U D fFn W n 2 !g. It is evident
0
that fn W n 2 !g . C 1/, so U 2 C1 0 0
.X / by the definition of C1 .X /.
Therefore .X / C1 .X / and our inductive procedure shows that
0 0
(3) 0 .X / C1
0
.X / for any ordinal < !1 .
Therefore .0 .X //c .C1

0
.X //c so we obtain
(4) 0 .X / C1
0
.X / for any ordinal < !1 .
Assume first that D C1; since 0 .X / 0 .X / by (3) and 0 .X / 0 .X /

by (2), we have 0 .X / 0 .X /. Furthermore, it follows from the property (4)
that 0 .X / 0 .X /; besides, 0 .X / 0 .X / by (1) which implies 0 .X /
0 .X /. Thus 0 .X / [ 0 .X / 0 .X /.
Now, if C 1 < , then 0 .X / C1 0
.X / 0 .X / by (2) and (1). The
property (2) also implies that 0 .X / 0 .X /, so 0 .X / 0 .X /. Analogously,
0 .X / C10
.X / 0 .X / by (1) and (2). The property (1) also implies that
0 .X / 0 .X / so 0 .X / 0 .X /. Thus 0 .X / [ 0 .X / 0 .X / whenever
< < !1 .
To finish our solution, assume towards a contradiction that X is an uncountable
Polish space such that 0 .X / [ 0 .X / D 0 .X / for some ordinals and with
< .TThe space X is a continuous image of P by Fact 3 of T.132; it follows from
P D fRnfqg W q 2 Qg that P is a G -set in R. Every open subset of R is a
countable union of compact sets, so P is a K -space, and hence X is K-analytic.
Therefore we can apply Fact 4 of T.250 to conclude that K embeds in X .
Restricting the respective families to K and applying Fact 1 of this solution and
Fact 1 of T.319, we conclude that 0 .K/ [ 0 .K/ D 0 .K/. It is an easy exercise
that there are disjoint clopen U; V K such that U ' V ' K.
We have 0 .U / 0 .U / by Problem 319; besides, the inclusion 0 .U /
.U / implies 0 .U / 0 .U / (and vice versa) because each of the fam-
0
ilies 0 .U / and 0 .U / is obtained from the other by passing to the relevant

complements. Thus it follows from 0 .U / 0 .U / that 0 .U / D 0 .U / which

is a contradiction. Therefore neither of the families 0 .U / and 0 .U / is contained
in the other. Of course, the same is true for the families 0 .V / and 0 .V /. This
makes it possible to choose F 2 0 .U /n0 .U / and G 2 0 .V /n0 .V /. Now
Fact 1 of this solution and Fact 1 of T.319 imply that F 2 0 .K/ and G 2 0 .K/
(we must use the fact that U and V are open in K if D 0). Now, the first part of
our solution implies that F; G 2 0 .K/ and hence H D F [ G 2 0 .K/ (we leave
it as an easy exercise for the reader to prove that the union of any two elements of
0 .K/ belongs to 0 .K/). However, H 0 .K/ [ 0 .K/; indeed, if H 2 0 .K/,
then F D H \ U 2 0 .U / which is a contradiction. Analogously, if H 2 0 .K/,
then G D H \ V 2 0 .V / which contradicts the choice of G.
This contradiction shows that the equality 0 .K/ [ 0 .K/ D 0 .K/ is
impossible and hence 0 .X / [ 0 .X / 0 .X /, i.e., our solution is complete.
T.321. Suppose that X is aS
second countable space. Prove that for every countable
limit ordinal , we have f0 .X / W < g 0 .X /. Show that if X is
S
uncountable and Polish, then the inclusion is strict, i.e., f0 .X / W < g
0 .X /.
Solution. We have 0 .X / 0 .X / for all < by Problem 320; as a
S
consequence, f0 .X / W < g 0 .X /.
S assume, towards a contradiction, that X is uncountable and Polish, and we
Now
have f0 .X / W < g D 0 .X / for some countable limit ordinal . The space X
T
is a continuous image of P by Fact 3 of T.132; it follows from P D fRnfqg W q 2
Qg that P is a G -set in R. Every open subset of R is a countable union of compact
sets, so P is a K -space, and hence X is K-analytic. Therefore we can apply Fact 4
of T.250 to conclude that K embeds in X .
Restricting the relevantSfamilies to K and applying Fact 1 of T.320 and Fact 1 of
T.319, we conclude that f0 .K/ W < g D 0 .K/. We leave it to the reader
to verify that we can choose a disjoint family fUn W n 2 !g of clopen
S subsets of K
such that Un is homeomorphic to K for every n 2 !. Let U D fUn W n 2 !g and
F DS U nU . Since KnU is open in K, there is a sequence fFn W n W2 !g 00 .K/
with fFn W n 2 !g D KnU . Take a strictly increasing sequence fn W n 2 !g
such that D supfn W n 2 !g. It follows from Problem 320 that there exists a set
An 2 0nC1 .Un /n0n .Un / for all n 2 !. Apply Fact 1 of T.320 to conclude that
S
An 2 0nC1 .K/ for all n 2 !. Consequently, A D fAn W n 2 !g 2 0 .K/.
Observe also that Bn D Un nAn 2 0nC1 .K/ for all n 2 !. Furthermore, the
S S
set KnA D . fBn W n 2 !g/ [ F [ . fFn W n 2 !g/ is a union of the family
fBn W n 2 !g [ fF g [ fFn W n 2 !g which can be enumerated as fQn W n 2 !g
where Qn 2 0n .K/ and n < for all n 2 !. Thus KnA 2 0 .K/ which shows
that A 2 0 .K/ and hence A 2 0 .K/. Now, if < and A 2 0 .K/, then
An D A \ Un 2 0 .Un / for all n 2 ! (see Fact 1 of T.319). There is n 2 ! such
that < n and therefore An 2 0n .Un / by Problem 320 which is a contradiction
S
with the choice of An . Therefore the equality f0 .K/ W < g D 0 .K/ is false
S 0
and hence f .X / W < g 0 .X / so our solution is complete.
T.322. Prove that there exists a countable space which cannot be embedded into
Cp .B/ for any Borel set B.
Solution. Given a space Z let F .Z/ be the family of all closed subsets of Z; if
A exp.Z/, then AjY D fA \ Y W A 2 Ag for every Y Z. The expression
Z ' T says that the spaces Z and T are homeomorphic.
Fact 1. If Z is a space and w.Z/ D , then jF .Z/j D j.Z/j 2 . In particular,
jF .Z/j c for any second countable space Z.
Proof. It is evident that jF .Z/j D j.Z/j; furthermore, if B is a baseSin Z with
jBj , then for every U 2 .Z/ there is B 0 B such that U D B 0 . As a
consequence, j.Z/j jexp.B/j 2 so Fact 1 is proved.

t
u
Fact 2. There are at most c-many pairwise non-homeomorphic Polish spaces.
Proof. Every Polish space is homeomorphic to a closed subspace of R! by Problem
273 of [TFS]. Thus the cardinality of any family of pairwise non-homeomorphic
Polish spaces cannot be greater than jF .R! /j c (see Fact 1). Fact 2 is proved. u
t
Fact 3. For any second countable space X , we have jB.X /j c. An easy
consequence is that there are at most c-many pairwise non-homeomorphic Borel
sets.
Proof. If X is a second countable space, then j00 .X /j D j.X/j c by Fact 1.
Suppose that 0 < < !1 and we proved that j0 .X /j c for all < . Then
j0 .X
S/j c for all ordinals < and therefore the cardinality of the family
U D f0 .X / W < g does not exceed c as well. Since 0 .X / is the family of
unions of countable subfamilies of U, we have j0 .X /j jU ! j c! D c. Thus
our inductive procedure shows that j0 .X /j c for all < !1 . Consequently,
jB.X /j c
!1 D c.
S fX W < cg of
By Fact 2, there is an enumeration (with possible repetitions)
all Polish spaces. By the first part of our fact, the family V D fB.X / W < cg
has cardinality at most c
c D c; since any Borel set is an element of V, Fact 3 is
proved. t
u
Fact 4. In the family f! [ fg W 2 !n!g there exists a subfamily of 2c -many
pairwise non-homeomorphic spaces. Here ! [ fg is considered with the subspace
topology inherited from ! for all 2 !n!.
Proof. Denote by N the space ! [ fg and let E./ D f 2 !n! W N ' N g
for all 2 !n!. For any 2 E./ fix a homeomorphism f W N ! N . Then
f .!/ D ! and f ./ D ; besides, for a set A !, we have A [ fg 2 .; N / if
and only if f .A/ [ f g 2 . ; N /. If U D .; N /j! for every 2 !n!, then
f .U / D U for any 2 E./.
T
Observe that P D fA W A 2 U g D fg for any 2 !n! (the bar denotes
the closure in !). Indeed, it is clear that 2 P ; if , then there are disjoint
U; V 2 .!/ such that 2 U and 2 V . Then U \ ! 2 U ; it is straightforward
that U \ ! and hence P .
It follows from f g D P that the correspondence ! U D f .U / is an
injection on the set E./ and therefore the cardinality of E./ does not exceed the
cardinality of the set of all bijections from ! to !. Thus jE./j j! ! j D c and we
established that
(1) jE./j c for any 2 !n!.
Furthermore, if we let if and only if N ' N , then is an equivalence
relationship on !n! and E./ is the equivalence class of . Since j!n!j D 2c
and the cofinality of 2c is greater than c, the property (1) implies that there are 2c -
many distinct equivalence classes. If we enumerate them as fE W < 2c g, then
choosing 2 E for each < 2c , we obtain a family N D fN W < 2c g of
pairwise non-homeomorphic spaces with jN j D 2c . Fact 4 is proved. t
u
Returning to our solution, take a family S D fS W < 2c g of pairwise non-
homeomorphic countable spaces whose existence is guaranteed by Fact 4. Assume
that, for every < 2c , the space S embeds in Cp .B / for some Borel set B .
Since there are at most c-many Borel sets by Fact 3, there is a Borel set B such that
the set Q D f < 2c W B D Bg has cardinality 2c . The space B being second
countable, we have jCp .B/j D c because nw.Cp .B// D nw.B/ !. Therefore
there are at most c-many countable subspaces of Cp .B/ (because c! D c). As a
consequence, there are distinct ; 2 Q such that S and S are homeomorphic to
the same countable subspace of Cp .B/. Therefore S ' S which is a contradiction.
Consequently, some of the spaces from the family S does not embed in Cp .B/ for
any Borel set B, so our solution is complete.
T.323. Prove that a second countable space X is an absolute F if and only if X is
-compact.
S
Solution. Suppose that X D fKn W n 2 !g where Kn is compact for all n 2 !.
If X Y , then Kn is closed in Y (recall that all our spaces are Tychonoff and a
compact space is closed in any larger Hausdorff space). Thus X is an F -set in Y ,
i.e., X is an absolute F . This proves sufficiency.
Since X is second countable, we can consider that X I! ; if X is absolute F ,
then X is an F -set in I! . Since all closed subsets of I! are compact, the space X
has to be -compact; this settles necessity.
T.324. Prove that a second countable space X is an absolute G if and only if X is
Cech-complete.
Solution. We can consider that X I! ; if X is absolute G , then X is Cech-
complete being a G -set in the compact space I! (see Problem 260 of [TFS]). This
proves necessity.
Now, assume that X is Cech-complete and X M where M is a metrizable

space. Let Y D X (the bar denotes the closure in M ). Then Y is an extension of X
and hence X is a G -set in Y (see Problem 259 of [TFS]). Since M is metrizable,
the set Y is G in M . Finally, apply Fact 2 of S.358 to see that X is also a G -set in
M which shows that X is absolute G .
T.325. Suppose that X is a Polish space and f W X ! Y is a perfect map. Prove
that Y is Polish (remember that any perfect map is onto).
Solution. Being completely metrizable is equivalent to being metrizable and Cech-
complete by Problem 269 of [TFS], so to prove that Y is Polish, we must establish
that Y is Cech-complete and w.Y / D !.
A perfect image of a Cech-complete space is a Cech-complete space by Problem
261 of [TFS] so Y is Cech-complete. Furthermore, nw.Y / ! (see Problem 157 of
[TFS]) and hence we can apply Problem 270 of [TFS] to conclude that w.Y / D !.
Thus Y is a second countable Cech-complete space, i.e., Y is Polish.
T.326. Let X be a Polish space. Suppose that f W X ! Y is a continuous surjective
open map. Prove that there is a closed F X such that f .F / D Y and f jF is a
perfect map.
Solution. The space X is Cech-complete by Problem 269 of [TFS] and nw.Y /
nw.X / D ! by Problem 157 of [TFS], so the space Y is Lindelf and hence
paracompact. Therefore Fact 1 of S.491 is applicable to conclude that there exists a
closed F X such that f .F / D Y and f jF is a perfect map.
T.327. Prove that any open continuous image of a Polish space is a Polish space.
Solution. Suppose that X is Polish and f W X ! Y is an open continuous onto
map. We can apply Problem 326 to find a closed F X such that f .F / D Y
and f jF is a perfect map. The space F being Polish, we can apply Problem 325 to
conclude that Y is also Polish.
T.328. Prove that X is a Polish space if and only if it is an open continuous image
of P.
Solution. The space P is Cech-complete being a G -subspace of R (see Problem
260 of [TFS]). Thus P is Polish by Problem 269 of [TFS]. If X is an open continuous
image of P, then X is Polish by Problem 327.
Now assume that X is a Polish space and fix a complete bounded metric d on
the space X for which .d / D .X/. Given s 2 ! <! , let l.s/ D n if s 2 ! n for
some n 2 N; if s D ;, let l.s/ D 0. It was proved in Fact 3 of S.491 that there exists
an A-system in X , i.e., a family U D fUs W s 2 ! <! g .X/ with the following
properties:
S
(A1) U; D X and Us D fUt W t 2 ! nC1 and tjn D sg for any n 2 ! and s 2 ! n ;
(A2) if s 2 ! <! and l.s/ D n 2TN, then diam.Us / < n1 ;
(A3) for any f 2 ! ! , we have fUf jn W n 2 !g ;.
We will identify
T P with ! ! (see Problem 313). For any f 2 ! ! , observe that
the set Pf D fUf jn W n 2 !g consists of precisely one point because Pf ; by
(A3) and the diameters of the sets Uf jn approach zero. Let '.f / D x where x 2 X
is the point for which Pf D fxg. We claim that ' W ! ! ! X is a continuous open
onto map.
Fix m 2 ! and s 2 ! m and let Ws D ff 2 ! ! W f jm D sg. It is straightforward
from the definition of ' that '.Ws / Us . Take any x 2 Us and let sm D s. Assume
that m n 2 ! and we have constructed functions fsk W m k ng with the
(i) sk 2 ! k for each k 2 fm; : : : ; ng;
(ii) x 2 Usk for each k 2 fm; : : : ; ng;
(iii) skC1 jk D sk for all k 2 fm; : : : ; n 1g.
S
Since x 2 Usn D fUt W t 2 ! nC1 and tjn D sn g, there is snC1 2 ! nC1 such
that snC1 jn D sn and x 2 UsnC1 . It is evident that (i)(iii) hold for the sequence
fsk W k n C 1g, so our inductive construction can be continued, giving us a
sequence fsn W m n < !g with the properties (i)(iii). It follows from (iii) that
thereTexists f 2 ! ! such that f jn D sn for each n m. Condition (ii) implies that
x 2 fUf jn W n 2 !g and therefore '.f / D x, so '.Ws / D Us . Since s 2 ! <! has
been chosen arbitrarily, we proved that
. / '.Ws / D Us for any s 2 ! <! .
If we take s D ;, then we obtain '.! ! / D X , i.e., the map ' is onto. For
any point x 2 X take any f 2 ! ! such that '.f / D x. Observe that the family
Cx D fUf jn W n 2 !g is a local base of X at x. Indeed, all elements of Cx are open
in X and contain x. Given any " > 0, choose n 2 N with n1 < " and note that, for
any y 2 Uf jn , we have d.x; y/ diam.Uf jn / < n1 so x 2 Uf jn B.x; "/ which
proves that Cx is a local base at x. Besides, the family Bf D ffg 2 ! ! W gjn D
f jng W n 2 !g is a local base of ! ! at f and f'.U / W U 2 Bf g D Cx by . /, so we
can apply Fact 2 of S.491 to conclude that the map ' is continuous and open. We
already saw that ' is surjective, so X is an open continuous image of P and hence
T.329. Prove that a second countable space is Polish if and only if it is a closed
continuous image of P. Show that a closed continuous image of P is not necessarily
first countable.
Solution. Assume that X is a second countable space and f W P ! X is a closed
continuous onto map. For every x 2 X let Kx D f 1 .x/nInt.f 1 .x//. It turns out
that
(1) the set Kx is compact (maybe empty) for every x 2 X .
Indeed, if the set Kx is not compact for some point x 2 X , then there is a set
A D fan W n 2 !g Kx which is closed and discrete in P. Fix a local base
fUn W n 2 !g of X at x such that UnC1 Un for all n 2 !. There exists a discrete
family V D fVn W n 2 !g .P/ such that an 2 Vn for all n 2 !. By continuity

of f , making the sets Vn smaller if necessary, we can assume that f .Vn / Un
for all n 2 !. The set A is contained in the boundary of f 1 .x/ so we can take
yn 2 Vn nf 1 .x/ for each n 2 !.
The family V being discrete, the set D D fyn W n 2 !g is closed and discrete
in P; besides, f .D/ 63 x by our choice of D. Furthermore, f .yn / 2 Un for all
n 2 ! which shows that the sequence ff .yn / W n 2 !g converges to x which is a
contradiction because f is a closed map and hence f .D/ must be a closed set. This
proves (1).
Now, take any x 2 X ; if f 1 .x/ is open, then choose a.x/ 2 f 1 .x/ and
let Px D fa.x/g. If not, then let Px D KS x . It is immediate from (1) that Px is
a compact set for all x 2 X , so if Q D fPx W x 2 X g and g D f jQ, then
g.Q/ D S X and g 1 .x/ D Px is a compact set for all points x 2 X . Observe that
PnQ D ff 1 .x/nPx W x 2 X g is an open subset of P because f 1 .x/nPx is
open in P for each x 2 X . Thus Q is closed in P and therefore g is a perfect map.
The space Q is Polish being closed in P and hence we can apply Problem 325 to
conclude that X is Polish finishing the proof of sufficiency.
Now assume that X is a Polish space. Then there is a open map g W P ! X
by Problem 328. Apply Problem 326 to find a closed F P such that g.F / D X
and gjF is a perfect (and hence closed) map. The space P is Polish and strongly
zero-dimensional, so there is a closed retraction h W P ! F by Problem 316.
Consequently, f D .gjF / h W P ! X is a continuous closed onto map (it is
evident that a compositions of two closed maps is a closed p map).
To finally construct the desired example, let F D f 2 C n W n 2 !g P. Then
F is a closed discrete subset of P. Let us construct a space P=F as follows: for any
A P, let A D .AnF / [ fF g. Given z 2 P, let pF .z/ D z if z 2 PnF and
pF .z/ D F if z 2 F . It is clear that pF W P ! PF D fF g [ .PnF /. Then the family
F D fU 2 .P/ W U PnF g [ fU W U 2 .F; P/g is a topology on the set PF ;
the space P=F D .PF ; F / is T1 and normal (and hence Tychonoff) and the map
pF W P ! P=F is continuous, closed and onto (see Fact 2 of T.245).
To distinguish between the point F of the space P=F and the set F , let us denote
by zF the set F when it is considered as a point of the space Z D P=F . We
claim that Z is not first countable at the point zF . To prove it assume, towards a
contradiction, thatpthere is a family fOn W n 2 !g .zF ; Z/ which is a base of Z
at zF . Let xn D 2 C n for all n 2 !; it is easy to find Un 2 .xn ; P/ for every
n 2 ! such that the family fUn W n 2 !g is discrete.
Take a point yn 2 .On \ Un /nfxn g for each n 2 !. It is immediate that the set
E D fyn W n 2 !g is closed, discrete and E \ F D ;. Therefore PnE 2 .zF ; Z/
and hence there is n 2 ! for which On \ E D ;. However, yn 2 On \ E which is
a contradiction. Thus Z is a closed continuous image of P such that .Z/ > ! and
T.330. Prove that X is homeomorphic to a Borel subset of some Polish space if and
only if it is homeomorphic to a Borel subset of R! .
Solution. Suppose that X is a Borel subset of a Polish space M . We can assume

that M is a closed subspace of R! by Problem 273 of [TFS]. There is a ordinal
< !1 such that X 2 C1
0 0
.M /; since C1 .M / C1
0
.R! / by Fact 1 of T.320,
we have X 2 C1 .R / and therefore X is a Borel subset of R! .
0 !
T.331. Let X be a Borel set. Prove that every Y 2 B.X / is also a Borel set. In
particular, any closed and any open subspace of a Borel set is a Borel set.
Solution. We can assume that X is a Borel subspace of a Polish space M and
hence there is < !1 such that X 2 0 .M / and, consequently, X 2 C1 0
.M / by
Problem 320. Recall that a space Z is called perfect if every U 2 .Z/ is an F -set.
Fact 1. For any perfect space Z we have
(a) 0 .Z/ [ 0 .Z/ 0 .Z/ whenever < < !1 ;
S S S
(b) B.Z/ D f0 .Z/ W < !1 g D f0 .Z/ W < !1 g D f0 .Z/ W < !1 g;
(c) if A 2 B.Z/, then ZnA 2 B.Z/; S
(d) if Bi 2 B.Z/ for any i 2 !, then B D TfBi W i 2 !g 2 B.Z/;
(e) if Ci 2 B.Z/ for any i 2 !, then C D fCi W i 2 !g 2 B.Z/.
Proof. Given an ordinal with 0 < < !1 , it follows from the inductive definition
of 0 .Z/ that
(1) 0 .Z/ 0 .Z/ whenever < < !1 .

Given a family A exp.Z/, let Ac D fZnA W A 2 Ag. It is evident that
A B exp.Z/ implies Ac B c , so it follows from (2) that
(2) 0 .Z/ 0 .Z/ whenever < < !1 ,
because 0 .Z/ D .0 .Z//c and 0 .Z/ D .0 .Z//c .

Since the space Z is perfect, every U 2 .Z/ is an F -set which shows that
00 .Z/ 10 .Z/. Assume that < !1 and we established, for every ordinal < ,
that 0 .Z/ C1
0
.Z/. If U 2 0 , then we can find a sequence fn W n 2 !g
S
and Fn 2 0n .Z/ for all n 2 ! such that U D fFn W n 2 !g. It is evident that
fn W n 2 !g . C 1/, so U 2 C1
0 0
.Z/ by the definition of C1 .Z/. Therefore
.Z/ C1 .Z/ and our inductive procedure shows that
0 0
(3) 0 .Z/ C1
0
.Z/ for any ordinal < !1 .
Therefore .0 .Z//c .C1

0
.Z//c so we obtain
(4) 0 .Z/ C1
0
.Z/ for any ordinal < !1 .
Assume first that D C1; since 0 .Z/ 0 .Z/ by (3) and 0 .Z/ 0 .Z/
by (2), we have 0 .Z/ 0 .Z/. Furthermore, it follows from the property (4)
that 0 .Z/ 0 .Z/; besides, 0 .Z/ 0 .Z/ by (1) which implies 0 .Z/
0 .Z/. Thus 0 .Z/ [ 0 .Z/ 0 .Z/.
Now, if C 1 < , then 0 .Z/ C1 0

.Z/ 0 .Z/ by (2) and (1). The
property (2) also implies that 0 .Z/ 0 .Z/ so 0 .Z/ 0 .Z/. Analogously,
0 .Z/ C10
.Z/ 0 .Z/ by (1) and (2). The property (1) also implies that
0 .Z/ 0 .Z/ so 0 .Z/ 0 .Z/. Thus 0 .Z/ [ 0 .Z/ 0 .Z/ whenever
< < !1 and (a) is proved. The property (b) is an evident consequence of (a).
It follows from (b) that if A 2 B.Z/, then A 2 0 .Z/ for some < !1 and
therefore ZnA 2 0 .Z/ 2 B.Z/ which shows that ZnA 2 B.Z/, i.e., (c) is
proved.
As to (d), apply (b) to conclude that for every i 2 !, there is an ordinal i < !1
such that Bi 2 0i .Z/. If D supfi W i 2 !g C 1, then by definition of 0 .Z/
we have B 2 0 .Z/ B.Z/ and hence (d) is settled. Finally, the property (e) is an
immediate consequence of (c) and (d), so Fact 1 is proved. t
u
Returning to our solution, suppose first that F is a closed (open) subset of X ;
then there is a closed (open) subset G of the space M such that G \ X D F . Since
G 2 00 .M / [ 00 .M / 01 .M /, by Problem 320, we have G 2 C1 0
.M / and
therefore F D G \ X 2 C1 0
.M / (it is easy to see that the intersection of any two
elements of 0 .M / belongs to 0 .M / for every < !1 ) and hence F is a Borel
subset of M .
Now assume that < !1 and we established that for all ordinals < , if
F 2 0 .X / [ 0 .X /, then F is a Borel subset of the space M . If F 2 0 .X /,
then there is a sequence fn W n 2 !g and Fn 2 0n .X / for all n 2 ! such that
S
F D fFn W n 2 !g. Since Fn 2 B.M / for all n 2 ! by the induction hypothesis,
we have F 2 B.M / by Fact 1.
Now if F 2 0 .X /, then H D X nF 2 0 .X / and hence there S is a sequence
f
n W n 2 !g and Hn 2
0n .X / for all n 2 ! such that H D fHn W n 2 !g.
Since Hn 2 B.M / for all n 2 ! by the induction hypothesis, the sets H and M nH
are Borel subsets of M (see Fact 1). Therefore we can apply Fact 1 once more to
conclude that F D X nH D X \.M nH / is a Borel subset of M . Thus our inductive
procedure shows that for any < !1 , if Y 2 0 .X /, then Y is a Borel subset of M .
In other words, every Y 2 B.X / is a Borel subset of M so our solution is complete.
T.332. Given second countable spaces X and Y and a continuous map f W X ! Y ,
prove that for every Borel subset A of the space Y , the set f 1 .A/ is Borel in X .
Solution. If A is open (closed) in Y , then f 1 .A/ is also open (closed) in Y because
the map f is continuous. This proves that if A 2 00 .Y / [ 00 .Y /, then f 1 .A/ is
a Borel subset of X . Now assume that < !1 and we proved that for any < , if
A 2 0 .Y / [ 0 .Y /, then f 1 .A/ 2 B.X /.
If A 2 0 .Y /, then there is a sequence fn W n 2 !g and An 2 0n .Y / for
S
all n 2 ! such that A D fAn W n 2 !g. Since S f 1 .An / 2 B.X / for all n 2 !
by the induction hypothesis, we have f .A/ D ff 1 .An / W n 2 !g and hence
1
f 1 .A/ is a Borel subset of X by Fact 1 of T.331.

If, on the other hand, we have a set A 2 0 .Y /, then Y nA 2 0 .Y / and hence

X nf 1 .A/ D f 1 .Y nA/ is a Borel subset of X . Thus f 1 .A/ is also a Borel
subset of X by Fact 1 of T.331, so our inductive procedure shows that, for any
A 2 B.Y /, we have f 1 .A/ 2 B.X /.
T.333. Prove that any countable productSof Borel sets is a Borel set. Show that for
any second countable space X , if X D fXi W i 2 !g and each Xi is a Borel set,
then X is also a Borel set.
Solution. Given a space Z and Y Z, let idYZ W Y ! Z be the identity embedding,
i.e., idYZ .z/ D z for all z 2 Y .
Fact 1. Suppose that M is a complete metric space. Given any space Z and a
continuous map f W A ! M for some dense A Z, there is a G -set G of
the space Z such that A G and there is a continuous map g W G ! M with
gjA D f .
Proof. Take a complete metric d on the space M which generates the topology of
M . Given any n 2 N let Un D fz 2 Z W there exists W 2 .z; Z/ such that
diam.f .W \ A// n1 g. It is evident that Un is open in Z; the map f being
continuous, for any z 2 A, there is V 2 .z; A/ such that f .V / B.f .z/; 2n 1
/.
If W is an open subset of Z such that W \ A D V , then diam.f .W \ A// D
diam.f .V // diam.B.f .z/; 2n 1
// n1 which shows that z 2 Un . Consequently,
A Un for anyTn 2 N.
Thus G D fUn W n 2 Ng is a G -subset of Z and A G. To prove that the
map f can be continuously extended over the set G, take any z 2 G and consider
the family Fz D ff .W \ A/ W W 2 .z; Z/g. It is clear that Fz consists of closed
subsets of the space M ; if W1 ; : : : ; Wn 2 .z; Z/ and W D W1 \

\ Wn , then
f .W \ A/ f .W1 \ A/ \

\ f .Wn \ A/ which shows that the family Fz is
centered.
Given any " > 0 take n 2 N with n1 < ". Since z 2 Un , there is W 2 .z; Z/
such that diam.f .W \ A//T n1 < ". Therefore Fz contains elements of arbitrarily
smallTdiameter and hence Fz ; by Problem 236 of [TFS]. It is easy to see
that
T Fz cannot contain more than one point, so there is a point g.z/ such that
Fz D fg.z/g.
Once we have the function g W G ! M , observe first that gjA D f because
if z 2 A,T then f .z/ 2 f .W \ A/ f .W \ A/ for any W 2 .z; Z/ and hence
f .z/ 2 Fz , i.e., g.z/ D f .z/.
To see that g is continuous, take any z 2 G and " > 0. Choose n 2 N and
W 2 .z; Z/ such that n1 < " and diam.f .W \ A// D diam.f .W \ A// n1 . If
W 0 D W \ G, then W 0 2 .z; G/; given any t 2 W 0 , we have t 2 W 2 .t; Z/ and
hence f .W \ A/ 2 Ft \Fz . Consequently, fg.z/; g.t/g f .W \ A/ and therefore
d.g.z/; g.t// diam.f .W \ A// n1 < " which proves that g.W 0 / B.g.z/; "/,
i.e., g is continuous at the point z. Since z 2 G was taken arbitrarily, the function g
is continuous on G and Fact 1 is proved. t
u
Fact 2 (Lavrentieff theorem). Let M and N be complete metric spaces. If we are

given sets A M; B N and a homeomorphism f W A ! B, then there exist
A0 M; B 0 N such that A A0 ; B B 0 , the set A0 is G in M , the set B 0 is
G in N and there is a homeomorphism h W A0 ! B 0 such that hjA D f .
Proof. If we find the promised sets for the spaces M1 D A and N1 D B, then they
will be G -sets in M and N respectively by Fact 2 of S.358 (recall that every closed
subset in a metric space is G ). Thus we can assume that M1 D M and N1 D N ,
i.e., A is dense in M and B is dense in N .
Apply Fact 1 to find a G -set A1 in the space M and a G -set B1 in the space
N such that A A1 ; B B1 and there are continuous maps f1 W A1 ! N and
g1 W B1 ! M for which f1 jA D f and g1 jB D g D f 1 . Apply Fact 2 of S.358
once more to conclude that A0 D f11 .B1 / is a G -set in M and B 0 D g11 .A1 / is
a G -set in N . Since f1 .A/ D f .A/ D B B1 , we have A A0 ; it follows from
g1 .B/ D g.B/ D A A1 that B B 0 .
Let h D f1 jA0 ; then h.A0 / D f1 .A0 / B1 and therefore the function g1 h
is well-defined. Observe also that g1 h W A0 ! g1 .B1 / D A1 . Furthermore,
0
g1 .h.x// D g.f .x// D x for any x 2 A and hence the identity map i D idA A1 W
0 0
A ! A1 coincides with g1 h on a dense set A of the space A which shows that
g1 h D i (see Fact 0 of S.351), i.e.,
(1) g1 .h.x// D x for all x 2 A0 .
In particular, g1 .h.x// 2 A0 A1 and hence g1 .h.x// 2 A1 , i.e., h.x/ 2 B 0 for
all x 2 A0 ; thus h.A0 / B 0 .
Analogously, for the map u D g1 jB 0 , we have u.B 0 / D g1 .B 0 / A1 and hence
the function f1 u is well-defined. Furthermore, f1 .u.y// D f .g.y// D y for every
0
y 2 B and therefore f1 u coincides with the identity map j D idBB1 on a dense set
0
B of the space B which shows that f1 u D j (see Fact 0 of S.351), i.e.,
(2) f1 .u.y// D y for all y 2 B 0 .
An immediate consequence is that f1 .u.y// 2 B1 and hence u.y/ 2 A0 for each
y 2 B 0 ; thus u.B 0 / A0 .
Given any y 2 B 0 , we have y D f1 .u.y// D h.x/ where x D u.y/ 2 A0 ;
consequently, h.A0 / D B 0 . Analogously, x D g1 .h.x// D u.y/ for y D h.x/ 2 B 0
and therefore u.B 0 / D A0 . Finally, the properties (1) and (2) show that we have
u.h.x// D g1 .h.x// D x for all x 2 A0 and h.u.y// D f1 .u.y// D y for all
y 2 B 0 , i.e., the map h is the promised homeomorphism and u D h1 . Fact 2 is
proved. t
u
Fact 3. Suppose that M is a Polish space and Z 2 0 .M / (or Z 2 0 .M /) for
some countable ordinal 2. Then, for any Polish space N , if Z embeds in N ,
then Z 2 0 .N / (or Z 2 0 .N / respectively). In particular, if Z embeds as a
Borel subset in some Polish space M , then for any Polish space N , if Z embeds in
N , then Z 2 B.N /.
Proof. Denote by Y the subspace of N homeomorphic to Z; let f W Z ! Y be the

respective homeomorphism. Apply Fact 2 to find a G -set Z 0 in the space M and
a G -set Y 0 in the space N such that there is a homeomorphism h W Z 0 ! Y 0 for
which hjZ D f .
Since Z 2 0 .M / (or Z 2 0 .M / respectively), we have Z D Z \ Z 0 2
.Z 0 / (or Z D Z \ Z 0 2 0 .Z 0 / respectively) by Fact 1 of T.319. Therefore Y 2
0
0 .Y 0 / (or Y 2 0 .Y 0 / respectively) (it is evident that homeomorphisms preserve

both additive and multiplicative classes). Now apply Fact 1 of T.319 once more to
see that there is A 2 0 .N / (or A 2 0 .N / respectively) such that Y D A \ Y 0 .
Since Y 0 is a G -subset of N , we have Y 0 2 0 .N / and hence Y 2 0 .N / (or
Y 2 0 .N / respectively) by Fact 1 of T.341 and hence Fact 3 is proved. t
u
Returning to our solution, assume that Yi is a Borel set for everyQi 2 ! and fix a
Polish space Mi such that Yi 2 B.Mi / for all i 2 !. Let i Q W M D n2! Mn ! Mi
be the natural projection for all i 2 !. To prove that Y D i 2! Yi M is a Borel
subset of M observe that T Yi0 D i1 .Yi / is a Borel subset of M for every i 2 ! by
Problem 332. Since Y D fYi0 W i 2 !g, we can apply Fact 1 of T.331 to conclude
that Y 2 B.M /, i.e., Y is a Borel set. This proves that any countable product of
Borel sets is a Borel set.
Now, consider any embedding of X in a Polish space N . There is an embedding
of Xi into a Polish space Ni such that Xi 2 B.Ni / for all i 2 !. Since Xi is also
embedded in N , we have Xi 2 B.N / by Fact 3. Thus X 2 B.N / by Fact 1 of T.331
which shows that X is a Borel set and hence our solution is complete.
T.334. Prove that every Borel set is an analytic space.
Solution. Let X be a Borel set; we can assume that X M for some Polish space
M and hence there is < !1 such that X 2 0 .M /. Let us prove, by induction on
, that X is analytic.
If X is closed or open in M, then X is Polish and hence it is a continuous image
of P by Problem 328, so X is analytic. Now assume that 0 < < !1 and we have
proved, for all < , that if X 2 0 .M / [ 0 .M /, then X is analytic.
Now, if X 2 0 .M /, then there exists a sequence fn W n 2 !g and
S
Xn 2 0n .ML / for all n 2 ! such that X D fXn W n 2 !g. It is an easy exercise to
see that P D fPn W n 2 !g where Pn is homeomorphic to P for all n 2 !. By the
induction hypothesis, there is a continuous onto map fn W Pn ! Xn for all n 2 !.
Define f W P ! X as follows; if x 2 P, then there is a unique n 2 ! with x 2 Pn ;
let f .x/ D fn .x/. It is easy to prove, using Fact 1 of S.472, that f is a continuous
onto map and hence X is analytic.
Finally, if X 2 0 .M /, then M nX 2 0 .M / and therefore there exist a
sequence of ordinals S fn W n 2 !g and a set T Yn 2 0n .M / for all n 2 !
such that M nX D fYn W n 2 !g. Therefore X D fM nYn W n 2 !g. The space
M nYn 2 0n .M / is analytic for all n 2 ! by the induction hypothesis and hence X
is analytic by Fact 3 of T.132. Thus our inductive procedure shows that all elements
of B.M / are analytic and hence X is analytic.
T.335. Prove that

(i) any closed subspace of an analytic space is an analytic space;
(ii) any open subspace of an analytic space is an analytic space;
(iii) any countable product of analytic spaces is an analytic space.
Solution. The items (i) and (iii) were proved in Fact 3 of T.132. To prove that
(ii) also holds, assume that X is an analytic space and U 2 .X/. Since X has a
S network, there is a family fFn W n 2 !g of closed subsets of X such that
countable
U D n2! Fn . Each Fn is analytic by (i) and a countable union of analytic spaces
is analytic: this was also proved in Fact 3 of T.132. Thus U is analytic and the proof
of (ii) is complete.
T Y is a space and Xi Y is an analytic space for all i 2 !.

T.336. Assume that
Prove that X D fXi W i 2 !g is also an analytic space.
Solution. This was proved in Fact 3 of T.132.
S
T.337. Assume that X D fXi W i 2 !g and Xi is an analytic space for every
i 2 !. Prove that X is an analytic space.
T.338. Let X and Y be Polish spaces. Suppose that f W X ! Y is a continuous
map. Prove that for any analytic set B Y , the set f 1 .B/ is also analytic.
Solution. Every Polish space is analytic by Problem 334 and hence the space X Y
is analytic by Problem 335. The graph G.f / D f.x; f .x// W x 2 X g of the map f
is a closed subspace of X Y (see Fact 4 of S.390) and hence G.f / is also analytic
by Problem 335. The space X B is analytic as well being a product of two analytic
spaces (see Problem 335) and therefore the space F D .X B/ \ G.f / is analytic
by Problem 336. Now denote by W X Y ! X the relevant natural projection and
observe that .F / D f 1 .B/. It is evident that a continuous image of an analytic
space is an analytic space so f 1 .B/ is analytic.
T.339. Let A and B be two disjoint analytic subsets of a Polish space M . Prove that
there exist Borel subsets A0 and B 0 of the space M such that A A0 ; B B 0 and
A0 \ B 0 D ;.
Solution. Given a Polish space N , call sets C; D N Borel-separated in N if
there exist disjoint C 0 ; D 0 2 B.N / such that C C 0 and D D 0 . We identify
P with the space ! ! (see Problem 313). If f is a function, then dom.f / is its
domain; given a function g, the expression f g says that dom.f / dom.g/ and
gjdom.f / D f . Given any s 2 ! <! , let Os D fx 2 P W s xg.
If k; n 2 ! and s 2 ! k , then s _ n 2 ! kC1 is defined by .s _ n/.k/ D n and
.s n/jk D s. Suppose that we have a set of functions ffi W i 2 I g such that
_
fi j.dom.fi / \ dom.fj // D fj j.dom.fi /S \ dom.fj // for any i; j 2 I . Then we

can define a function f with dom.f / D i 2I dom.fi / as follows: given any x 2
dom.f /, find any i 2 I with x 2 dom.fi / and let f .x/ D fi .x/. It is easy to check
that the value of f at x does not depend on the S

choice of i , so we have consistently
defined a function f which will be denoted by ffi W i 2 I g.
Fact 1. Suppose that N is a Polish space and Cn ; Dn N for all n 2 !. IfSthe sets
Cn and DmSare Borel-separated in N for all m; n 2 !, then the sets C D n2! Cn
and D D n2! Dn are also Borel-separated in N .
Proof. Take disjoint Borel sets U.m; n/ and V .m; n/ in the space N such that
Cn SU.m;T n/ and Dm V .m; n/ for Tn 2 ! and observe that the sets
S all m;
U D n2! . m2! U.m; n// and V D m2! . n2! V .m; n// are Borel in N (see
Fact 1 of T.331) and disjoint while C U and D V . Fact 1 is proved. t
u
Returning to our solution, fix continuous onto maps f W P ! A and g W P ! B.
S towards a contradiction, that A and B are not Borel-separated in M . Since
Assume,
P D fOs W s 2 ! 1 g, we can apply Fact 1 to conclude that there exist s1 ; t1 2 ! 1
such that f .Os1 / is not Borel-separated in M from g.Ot1 /.
Suppose that n 2 N and we have constructed s1 ; : : : ; sn ; t1 ; : : : ; tn 2 ! <! with
(1) si ; ti 2 ! i for every i n;
(2) si 1 si and ti 1 ti if 2 i n;
(3) the sets f .Osi / and g.Oti / are not Borel-separated in M for all i n.
S S
Since Osn D fOsn _ k W k 2 !g and Otn D fOtn _ k W k 2 !g, we can apply
Fact 1 to conclude that there are k; l 2 ! such that f .Osn _ k / is not Borel-separated
in M from g.Otn _ l /. Letting snC1 D sn _ k and tnC1 D tn _ l, we obtain a sequence
s1 ; : : : ; sn ; snC1 ; t1 ; : : : ; tn ; tnC1 2 ! <! for which the conditions (1)(3) hold for all
i n C 1.
Thus our inductive construction can go on to provide sequences fsi W i 2 Ng and
fti W i 2 Ng such S that (1)(3) are fulfilled for Sall i 2 N. It follows from (2) that
the points s D fsi W i 2 Ng 2 P and t D fti W i 2 Ng 2 P are well-defined.
The sets A and B being disjoint, we have f .s/ g.t/ and hence there are disjoint
U; V 2 .M / such that f .s/ 2 U and g.t/ 2 V . The families fOsn W n 2 Ng and
fOtn W n 2 Ng are local bases at the points s and t respectively, so, by continuity of
f and g, there is n 2 N such that f .Osn / U and g.Otn / V . The sets U and
V being Borel in M , we conclude that f .Osn / and g.Otn / are Borel-separated in
M which contradicts (3). Therefore the sets A and B are Borel-separated in M and
T.340. Let X be a subspace of a Polish space M . Prove that X is Borel if and only
if X and M nX are analytic.
Solution. If X and M nX are analytic, then there are disjoint A; B 2 B.M / such
that X A and M nX B (see Problem 339). It is immediate that X D A and
M nX D B, so X is a Borel subset of M .
Now, if X 2 B.M /, then M nX 2 B.M / by Fact 1 of T.331, so both X and
M nX are analytic by Problem 334.
T.341. Prove that X is a Borel set if and only if there exists a closed subspace of P
which condenses onto X . As a consequence, if a Borel set X can be condensed onto
a second countable space Y , then Y is a also Borel set.
Solution. The expression Z ' T says that the spaces Z and T are homeomorphic.
A family of sets A is closed under
T finite (countable) intersections if for any finite
(countable) A0 A, we have A0 2 A. Analogously, A is closed S under finite
(countable) unions if for any finite (countable) A0 A, we have A0 2 A. Say
that a second countable space Z is P-representable if there is a closed F P
which condenses onto Z. We must prove that Z is a Borel set if and only if it is
P-representable.
We identify P with the space ! ! (see Problem 313). If f is a function, then
dom.f / is its domain; given a function g, the expression f g says that
dom.f / dom.g/ and gjdom.f / D f . Given any s 2 ! <! let Os D fx 2
P W s xg. If k; n 2 ! and s 2 ! k , then s _ n 2 ! kC1 is defined by .s _ n/.k/ D n
and .s _ n/jk D s.
Fact 1. If Z is an arbitrary space and is a countable ordinal, then
(a) the class 0 .Z/ is closed under countable unions and finite intersections;
(b) the class 0 .Z/ is closed under finite unions and countable intersections.
Proof. It is evident that a finite intersection of open subsets of Z is open in Z.
Besides, any union of open subsets of Z is open in Z which shows that (a) is proved
for D 0. Analogously, (b) is true for D 0 because any finite union of closed
subsets of Z is closed in Z and any intersection of closed subsets of Z is again
closed in Z.
S that < !1 and we proved the properties (a) and (b) for all S< . Let
Assume
D f0 .Z/ W < g; if An 2 0 .Z/ for all n 2 !, then An D m2! Am
S S m n
where Am n 2 for all m; n 2 ! and hence A D n2! An D fAn W m; n 2 !g 2
0 .Z/ which shows that 0 .Z/ is closed under countable unions.
To prove that 0 .Z/ is closed under finite intersections, it suffices to establish it
for intersections of any two elements of 0 .Z/, so take any A; B 2 0 .Z/.SChoose
families fASn W n 2 !g and fBn W n 2 !g such that A D n2! An
and B D n2! Bn . By the inductionS hypothesis, we have An \ Bm 2 for all
m; n 2 ! and hence A \ B D fAn \ Bm W n; m 2 !g 2 0 .Z/ which proves (a)
for D .
Now assume that Bn 2 0 .Z/ and hence An D ZnBn 2 0 .Z/ for all n 2 !.
For every n 2 !, there is S
S fAm
a collection S n W m 2 !g such that An D
T m2! A m
n and hence A D n2! A n D fA n W n; m 2 !g 2 .Z/. Thus B D
m 0
n2! Bn D ZnA 2 .Z/ which shows that .Z/ is closed under countable
0 0
intersections.
Finally, if A; B 2 0 .Z/, then A0 D ZnA 2 0 .Z/ and B 0 D ZnB 2 0 .Z/
which implies that A0 \ B 0 2 0 .Z/ because we proved that 0 .Z/ is closed under
finite intersections. Therefore A [ B D Zn.A0 \ B 0 / 2 0 .Z/ and hence 0 .Z/
is closed under finite unions. This settles (a) and (b) for all , so our inductive
procedure shows that (a) and (b) hold for all < !1 . Fact 1 is proved. t
u
Fact 2. Suppose that Z is a second countable space. If fAn W n 2 !g is a disjoint
family of analytic subspaces of Z, then there is a disjoint family fBn W n 2 !g
B.Z/ such that An Bn for all n 2 !.
Proof. We can consider that there Sis a Polish space M such that Z M . For every
n 2 !, the sets An and A0n D fAk W k 2 !nfngg are analytic and disjoint, so
there are disjoint Bn0 ; CS
0 0 0 0
n 2 B.M / such that An Bn and An Cn by Problem
0 0
339. Then Dn D Bn n. i <n Bi / 2 B.M / (see Fact 1 of T.331) for every n 2 !. It
follows from the choice of the sets Bn0 that An Dn for all n 2 !; it is immediate
that the family fDn W n 2 !g is disjoint. If Bn D Dn \ Z, then Bn 2 B.Z/ for all
n 2 ! and the family fBn W n 2 !g is as promised. Fact 2 is proved. t
u
Our proof of necessity will be accomplished in several steps:
Step 1. The space R is P-representable. L
It is easy to see that we can represent P as i 2! Pi where Pi is homeomorphic
to P for all i 2 !. Let fqi W i 2 Ng be a faithful enumeration of Q; choose a point
zi 2 Pi for all i 2 N and fix a homeomorphism f0 W P0 ! P.
It is evident that F D fzi W i 2 Ng [ P0 is a closed subset of P. Let f .zi / D qi
for all i 2 N; if z 2 P0 , then let f .z/ D f0 .z/. It is immediate that f W F ! R
is a condensation.
Step 2. Any countable product of P-representable spaces is P-representable and,
in particular, the space R! is P-representable.
If Zi is P-representable for all i 2 !, then there is a closed Q Fi P and a
condensation
Q f i W F !Q Z i . It is easy to see that the map g D i 2! fi condenses
F D i 2! Fi onto i 2! Zi . The set F is closed in P! and the latter space
is homeomorphic to P. Thus there is a closed F P which condenses onto
Q
i 2! Zi . Finally, recall that R is P-representable (see Step 1) and hence R is
!
also P-representable.
Step 3. If a second countable space Z is P-representable and A is closed in Z,
then A is also P-representable.
Indeed, if there is a closed G P and a condensation f W G ! Z, then
F D f 1 .A/ is a closed subset of G and hence of P which condenses onto A.
Step 4. Any Polish space is P-representable.
Indeed, if M is a Polish space, then we can consider that M is a closed subspace
of R! (see Problem 273 of [TFS]). Now apply the results of StepsS2 and 3.
Step 5. Suppose that Z is a second countable space such that Z D fZi W i 2 !g
and Zi \ Zj D ; whenever i j . If Zi is P-representable for all i 2 !, then
Z is also P-representable.
For each i 2 !, let Fi be a closed subspace of a space Pi ' P for which L there
exists a condensation fi W Fi ! Zi . It is immediate that theLspace F D i 2! Fi
condenses onto Z; since F is a closed subspace of P D i 2! Pi and P ' P,
the space Z is P-representable.
Step 6. Suppose that Z is a second countable T space and a subspace Zi Z is

P-representable for all i 2 !. Then A D fZi W i 2 !g is also
Q P-representable.
Indeed, the space A is homeomorphic to a closed subspace of i 2! Zi (see Fact 7
of S.271), so A is P-representable (see Steps 2 and 3).
Now let X be a Borel subset of a Polish space M . To start an inductive proof
assume that X 2 00 .M / [ 00 .M /; then X is either closed or open in M and
hence Polish, so it is P-representable by Step 4.
Suppose that we have an ordinal < !1 such that any A 2 0 .M / [ 0 .M / is
P-representable for each < . Let D fA \ B W A 2 0 .Z/ and B 2 0 .Z/g
for all < . Then every element of is P-representable (see Step 6) for all
< .
If X 2 0 .M /, then there is a sequence fn W n 2 !g and An 2 n .M /
S
for all n 2 ! such that X D n2! An . We can assume that n nC1 and hence
A0 ; : : : ; An 2 0n .M / for all n 2 ! (see Problem 320). Thus A0n D A0 [

[ An 2
0n .M / for all n 2 ! (see Fact 1). Let X0 D A0 and XnC1 D AnC1 nA0n for all
n 2 !. It is easy to see that Xn 2 n for all n 2 !; besides, S each Xn is P-
representable, the family fXn W n 2 !g is disjoint and X D n2! Xn . Therefore X
is P-representable by Step 5.
Finally, assume that X 2 0 .M /; then M nX 2 0 .M / and hence there is a
sequence of ordinals
S fn W n 2 !g and Bn 2 n .M / for every n 2 ! such
that M nX D n2! Bn . Therefore Cn D M nBn 2 0n .M / is P-representable for
T
all n 2 ! by the induction hypothesis and X D n2! Cn which shows that X is
P-representable by Step 6. Thus our inductive procedure can be continued to prove
that every X 2 B.M / is P-representable. Since every Borel set X belongs to B.M /
for some Polish space M , we proved that every Borel set X is P-representable, i.e.,
there is a closed subspace F P which condenses onto X . This settles necessity.
Now suppose that F P is closed and f W F ! X is a condensation; we can
assume that X M for some Polish space M . Consider the set Bs D f .Os \ F /
for any s 2 ! <! . The set Os is analytic being closed in P; therefore Os \ F is also
analytic and hence Bs is analytic as well (maybe empty) for all s 2 ! <! .
The family fBs W s 2 ! 1 g is countable and disjoint (because f is a condensation),
so we can apply Fact 2 to find a disjoint family fBs0 W s 2 ! 1 g B.M / such that
Bs Bs0 for all s 2 ! 1 ; let Cs D Bs0 \ B s for all s 2 ! 1 (the bar denotes the closure
in M ); then the family fCs W s 2 ! 1 g B.M / is still disjoint and Bs Cs for
every s 2 ! 1 .
Assume that 1 n < ! and we have a family fCs W s 2 ! i g B.M / for every
i n with the following properties:
(1) fCs W s 2 ! i g is disjoint for all i n;
(2) Bs Cs B s for every s 2 ! i and i n;
(3) Cs Csjj whenever s 2 ! i and 1 j < i n.
Fix any s 2 ! n ; since the family fBs _ k W k 2 !g is disjoint and consists of
analytic sets, we can find a disjoint family fAk W k 2 !g B.M / such that Bs _ k
Ak for all k 2 ! (see Fact 2). If Cs _ k D Ak \ Cs \ B s , then Cs _ k 2 B.M / and
Bs _ k Cs _ k for all k 2 !. After we construct Cs _ k for all s 2 ! n and k 2 !,

we will have a family fCs W s 2 ! nC1 g and it is immediate that (1)(3) still hold for
all i n C 1. Thus our inductive construction can be continued to provide families
fCs W s 2 ! i g for all i 2 N such that the conditions (1)(3) are satisfied for all
n 2 !. T S
We claim that X D Q D f fCs W s 2 ! i g W i 2 Ng; of course, S if we prove
this, then X is a Borel subset of M by Fact 1 of T.331. Since X fBs W s 2 ! i g
for al i 2 N, the condition (2) implies that X Q. T
Now take any z 2 Q; there is s 2 P such that z 2 fCsjn W n 2 Ng. As a
T
consequence, z 2 fB sjn W n 2 Ng by (2) and hence Osjn \ F ; for all n 2 N.
The family fOsjn W n 2 Ng is a local base of P at s, so s 2 F D F (recall that
F is closed in P). If f .s/ z, then, by continuity of f , there is n 2 N such that
z f .Osjn \ F / D B sjn which is a contradiction. Therefore z D f .s/ 2 X and
hence Q X which shows that Q D X and hence the space X is a Borel set. This
finishes the proof of sufficiency.
Finally, assume that X is a Borel set and g W X ! Y is a condensation of X onto
a second countable space Y . We proved that there exists a closed F P for which
there is a condensation f W F ! X . Therefore g f is a condensation of F onto
Y and hence Y is a Borel set. Our solution is complete.
T.342. Show that there exists a subspace X R which is not analytic (and hence
not Borel).
Solution. In Fact 5 of T.250, we constructed a space Y K R which is not
K-analytic. Observe that P is a G -subset of R; since every open subset of R is K ,
the space P is K and hence any analytic space is K-analytic. This shows that Y
is not analytic. Since every Borel set in analytic by Problem 334, the space Y is not
Borel either.
T.343. Prove that
(i) any closed subset of a K-analytic space is a K-analytic space;
(ii) any countable product of K-analytic spaces is a K-analytic space.
T Y is a space and Xi Y is a K-analytic space for all i 2 !.

T.344. Assume that
Prove that X D fXi W i 2 !g is also a K-analytic space.
Q
Solution. The space X embeds in Z D fXi W i 2 !g as a closed subspace by
Fact 7 of S.271. The space Z is K-analytic by Fact 3 of T.250 and hence X is also
K-analytic by Problem 343 (or by Fact 3 of T.250).
S
T.345. Assume that X D fXi W i 2 !g and Xi is a K-analytic space for all i 2 !.
Prove that X is a K-analytic space.
T.346. Observe that there exist K-analytic non-analytic spaces. Show that any
analytic space is a K-analytic space. Prove that for any space X with a countable
network, X is analytic if and only if it is K-analytic.
Solution. Given a space Z denote by K.Z/ the family of all compact subsets of Z.
We identify P with the space ! ! (see Problem 313). If f is a function, then dom.f /
is its domain; given a function g the expression f g says that dom.f / dom.g/
and gjdom.f / D f . Given any s 2 ! <! , let O Ss D fx 2 P W s xg. If Z and Y
are spaces and ' W Z ! exp.Y /, then '.A/ D f'.z/ W z 2 Ag for any A Z and
' 1 .B/ DS fz 2 Z W '.z/ Bg for every B Y . Note that ' is called onto map if
'.Z/ D f'.z/ W z 2 Zg D Y which need not imply f'.z/ W z 2 Zg D exp.Y /.
Fact 1. Suppose that Z and T are spaces and we have a map ' W Z ! exp.T /.
Then ' is upper semicontinuous if and only if it is upper semicontinuous at any
z 2 Z, i.e., for any z 2 Z and U 2 .'.z/; T / there is V 2 .z; Z/ such that
'.V / U .
Proof. If the mapping ' is upper semicontinuous and U 2 .'.z/; T /, then the set
V D ' 1 .U / is open in Z and z 2 V . It is evident that '.V / U , so necessity is
proved.
Now assume that ' is upper semicontinuous at any z 2 Z and take any set
U 2 .T /. If z 2 ' 1 .U /, then '.z/ U and hence there isSVz 2 .z; Z/ such that
'.Vz / U which implies Vz ' 1 .U /. Thus ' 1 .U / D fVz W z 2 ' 1 .U /g is
an open set, so ' is upper semicontinuous. Fact 1 is proved. t
u
Fact 2. For any K-analytic space Z there is an upper semicontinuous onto map
' W P ! K.Z/.
Proof. Fix a K -space Y and a continuous onto map f W Y ! Z. There is a space
T such that TY T and a family fYn W n 2 !gSof -compact subspaces of T for
which Y D n2! Yn . We can assume that Yn D m2! Knm where Knm KnmC1 and
T s.n/
Knm is compact for all m; n 2 !. Given any s 2 P, let L.s/ D fKn W n 2 !g.
Then L.s/ Y is compact and therefore K.s/ D f .L.s// is a compact subspace
of Z for every s 2 P. Let us prove that the map ' W P ! K.X / defined by '.s/ D
K.s/, is upper semicontinuous.
Take any s 2 P and U 2 .K.s/; Z/; the set W D f 1 .U / is open in Y , so
there is W 0 2 .T / such that W 0 \ Y D W . Since K.s/ D f .L.s// U , we have
T s.n/
L.s/ W 0 and hence there is k 2 ! such that K 0 D fKn W n < kg W 0
(see Fact 1 of S.326). Then V D Osjk 2 .s; P/; given any point t 2 V , we have
T t .n/ T s.n/
tjk D sjk and hence L.t/ fKn W n < kg D fKn W n < kg D K 0
which shows that '.t/ D f .L.t// f .K 0 \ Y / f .W / D U . Thus ' is upper
semicontinuous at any point of P so it is upper semicontinuous by Fact 1. To finally
see that '.P/ D Z, take any z 2 Z and y 2 Y with f .y/ D z. For every n 2 !,
s.n/
we have y 2 Yn and hence there is s.n/ 2 ! such that y 2 Kn . This defines a
function s 2 P such that y 2 L.s/ and therefore x 2 K.s/ D f .L.s// D '.s/ so
Fact 2 is proved. t
u
Fact 3.T Suppose that Z is a space and K0 ; : : : ; Kn are compact subspaces of Z. If

K D i n TKi and U 2 .K; Z/, then there exists UiT2 .Ki ; Z/ for every i n
such that i n Ui U . In particular, if K D ;, then i n Ui D ;.
Proof. Fix a set V 2 .Z/ such that V \ Z D U ; then K V . Consider the
family F D fV 0 \

\ V n W Vi 2 .Ki ; Z/ for all i ng (the bar denotes
the closure in Z). If z 2 ZnK, then z Ki for some i n, and hence, by
regularity of Z, there is Vi 2 .Ki ; Z/ such that z V i . Letting Vj D Z for
all j 2 .n C 1/nfi g, we obtain a set F D V 0 \

T \ V n D V i 2 F with z F .
This proves that F is a family of compacts sets with F D K V .
Now weTcan apply Fact 1 of S.326 to find a finite Ui .K T i ; Z/ for each i n
such that fW W W 2 U0 [

[ Un g V . If Vi D Ui for all i n, then
V
T 0 \

\ V n V and hence, if U i D V i \ Z, then U i 2 .K i ; Z/ for all i n and
i n Ui U . Fact 3 is proved. t
u
Fact 4. Given spaces Z and Y let ' W Z ! exp.Y / be an upper semicontinuous
map. If F is a closed subset of Y and G D fz 2 Z W '.z/ \ F ;g, then G is a
closed subset of Z and the map '1 W G ! exp.Y / defined by '1 .z/ D '.z/ \ F is
also upper semicontinuous.
Proof. The set G is closed because G D Zn' 1 .Y nF / and ' 1 .Y nF / is open by
upper semicontinuity of '. Now take any z 2 G and U 2 .'1 .z/; Y /. Then U1 D
U [ .Y nF / 2 .'.z/; Y / and hence there is V1 2 .z; Z/ such that '.V1 / U1 .
Evidently, V D V1 \ G 2 .z; G/ and '1 .V / '.V / \ F U1 \ F U and
we proved that '1 is upper semicontinuous at the point z. This, together with Fact 1,
shows that Fact 4 is proved. t
u
Returning to our solution observe that any non-metrizable compact space Z is
K-analytic but not analytic because nw.Z/ D w.Z/ > ! while nw.T / ! for any
analytic space T . Thus I!1 is K-analytic but not analytic.
T
Furthermore, P is a G -subset of R because P D fRnfqg W q 2 Qg; since every
open subset of R is -compact, the space P is K and hence any analytic space is
K-analytic.
Now assume that X is a K-analytic space with a countable network. Then there
is a weaker second countable (Tychonoff!) topology on the space X (see Problem
156 of [TFS]); take a base B D fBn W n 2 !g of the space .X; / and let Fn0 D
X nBn ; Fn1 D B n for all n 2 ! (the bar denotes the closure in X ).
By Fact 2, there is an upper semicontinuous onto map ' W P ! K.X /. The sets
Pn0 D fs 2 P W '.s/ \ Fn0 ;g and Pn1 D fs 2 P W '.s/ \ Fn1 ;g are closed
in P by Fact 4. Define a map 'n W Pn0 Pn1 ! X as follows: if s 2 Pn0 , then
'n .s/ D '.s/ \ Fn0 ; if s 2 Pn1 , then 'n .s/ D '.s/ \ Fn1 . Observe that in the space
P, the sets Pn0 and Pn1 may intersect while they are considered disjoint in the space
Pn0 Pn1 . It is an easy consequence of Fact 4 that 'n is upper semicontinuous for
all n 2 !. Evidently, 'n is compact-valued, and it follows from '.P/ D X that
'n .Pn0 Pn1 / D X for every n 2 !.
The space Qn D Pn0Q Pn1 is homeomorphic to a closed subset of P for every

n 2 !, and hence Q DT n2! Qn is also homeomorphic to a closed subset of P. For
any s 2 Q, let .s/ D n2! 'n .s.n//; we claim that
(1) .s/ contains at most one element of X for every s 2 Q.
To prove (1) suppose, towards a contradiction, that there are distinct points x; y 2
.s/. Since B is a base in .X; /, there is n 2 ! such that x 2 Bn and y B n ; thus
x Fn0 and y Fn1 . If s.n/ 2 Pn0 , then .s/ '.s.n// \ Fn0 and hence x .s/,
a contradiction. If s.n/ 2 Pn1 , then .s/ '.s.n// \ Fn1 and hence y .s/; this
final contradiction proves (1).
Let F D fs 2 Q W .s/ ;g; it follows from (1) that for any s 2 F there is
h.s/ 2 X such that .s/ D fh.s/g and therefore we have a map h W F ! X . Let us
prove first that
(2) the set F is closed in Q.
T
Take any s 2 QnF ; then n2!T'n .s.n// D ;. Since all sets 'n .s.n// are
compact, there is k 2 ! such that n<k 'n .s.n//TD ;. By Fact 3 there exists a
set Wn 2 .'n .s.n//; X / for each n < k such that n<k Wn D ;. Since the map 'n
is upper semicontinuous, there is Vn 2 .s.n/; Qn / such that 'n .Vn / Wn for all
n < k. The set O D ftT2 Q W t.n/ 2 VnTfor all n < kg T is open in Q and s 2 O;
if t 2 O, then .t/ n<k 'n .t.n// n<k 'n .Vn / n<k Wn D ; and hence
.t/ D ;. Thus every s 2 QnF has a neighborhood O QnF which proves that
QnF is open in Q, so F is closed in Q and (2) is proved. Our next step is to show
that
(3) the map h is continuous.
T any s 2 F and U 2 .h.s/; X /. Since .s/ U , there is k 2 ! such

Take
that n<k 'n .s.n// U (see Fact 1 of S.326). T Apply Fact 3 again to find, for each
n < k, a set Un 2 .'n .s.n//; X / such that n<k Un U . The map 'n being upper
semicontinuous, there is Vn 2 .s.n/; Qn / such that 'n .Vn / Un for all n < k.
The set W 0 D ft 2 Q W t.n/ 2 Vn for all n < kg is open in Q and s 2 W 0 ;
thus W T D W 0 \ F 2 .s;T F /. If t 2 W , then
T t.n/ 2 Vn for all n < k and hence
.t/ n<k 'n .t.n// n<k 'n .Vn / n<k Un U , i.e., h.W / U which
proves that h is continuous at the point s.
Furthermore, h is surjective; to see it take any x 2 X and t 2 P with x 2 '.t/.
i.n/ i.n/
For every n 2 !, there is i.n/ 2 f0; 1g such that x 2 Fn and hence '.t/ \ Fn
i.n/
; which implies tn D t 2 Pn for all Tn 2 !. Thus, for a point s 2 Q defined by
s.n/ D tn for all n 2 !, we have x 2 n2! 'n .s.n// and therefore h.s/ D x.
This proves that h.F / D X , i.e., X is a continuous image of F . We already saw
that Q is analytic being homeomorphic to a closed subspace of P. Therefore F is
analytic by Problem 335 and hence X is also analytic which shows that our solution
is complete.
T.347. Prove that a non-empty Polish space X is homeomorphic to P if and only if
X is zero-dimensional and any compact subspace of X has empty interior.
Solution. If k; n 2 ! and s 2 ! k , then s _ n 2 ! kC1 is defined by .s _ n/.k/ D n

and .s _ n/jk D s.
We proved in Problem 313 that P is zero-dimensional. Considering P as a
subspace of R assume that K P is compact. Since K \ Q D ;, the interior
of K in R is empty and hence U D RnK is a dense open subspace of R. Since P is
dense in R, the set P \ U is also dense in R and hence P \ U D PnK is dense in P.
Consequently, the interior of K in P has to be empty. This settles necessity.
Now assume that X is a zero-dimensional Polish space such that Int.K/ D ; for
any compact K X . Fix a complete metric d on X such that .d / D .X/ and
d.x; y/ 1 for any x; y 2 X . Denote by C.X / the family of all non-empty clopen
subsets of X . It is evident that
(0) any U 2 C.X / is a complete metric zero-dimensional space such that every
compact K U has empty interior.
Let U; D X and U.0/ D fU; g; assume that n 2 ! and we have a family
U.k/ D fUs W s 2 ! k g C.X / for all k n with the following properties:
S
(1) U.k/ is disjoint and U.k/ D X forSall k n;
(2) if 0 k < n and s 2 ! k , then Us D fUs _ i W i 2 !g;
(3) for any k n and s 2 ! k , we have diam.Us / 21k .
Fix any s 2 ! n ; since Us is not compact by (0), there is an infinite closed discrete
D Us . Using paracompactness of Us it is easy to construct a disjoint open cover
Ws C.X / of Us such that diam.W / 2nC1 1
and jW \ Dj 1 for each W 2 Ws .
As a consequence, Ws is infinite and hence we can choose a faithful enumeration
fWn W n 2 !g of the family Ws . Let Us _ k D Wk forS all k 2 !. After we construct
Ws for all s 2 ! n , we obtain a family U.n C 1/ D fWs W s 2 ! n g D fUs W s 2
! nC1 g and it is evident that the conditions (1)(3) are satisfied for all k n C 1.
Therefore our inductive construction can be continued to give us families U.k/ such
that (1)(3) hold for all k 2 !. S
Observe that our family U D k2! U.k/ D fUs W s 2 ! <! g is ! <! -directed, i.e.,
has the properties (1)(4)T fromS T.313, so Fact 1 of T.313 is applicable to conclude
that
S the space L.U/ D f U.k/ W k 2 !g is homeomorphic to P. However,
U.k/ D X for all k 2 ! by the property (1). Therefore L.U/ D X and hence X
is homeomorphic to P which proves sufficiency and makes our solution complete.
T.348. Prove that a metrizable compact X is homeomorphic to the Cantor set if and
only if X is zero-dimensional and has no isolated points.
Solution. We identify the Cantor set K with S then space D (see Problem 128 of
!
[TFS]). As usual, D D f;g and D

0 <!
D fD W n 2 !g. If f is a function,
then dom.f / is its domain; given a function g the expression f g says that
dom.f / dom.g/ and gjdom.f / D f . For any s 2 D<! , let Os D fx 2 D! W s
xg. Given k 2 !; i 2 D and s 2 Dk , let s _ i 2 DkC1 be defined by .s _ i /.k/ D i
and .s _ i /jk D s. If U is a family of sets in a metric space M , then diam.U/ D
supfdiam.U / W U 2 Ug.
If we have a set of functions ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D

S of i , so we have consistently defined a function f which
Fact 1. Let .M; / be a complete metric space. Suppose that we have a family U D
fUs W s 2 D<! g .M / with the following properties:
(a) the family U.n/ D fUs W s 2 Dn g is disjoint for all n 2 !;
(b) if 0 k < n < !; s 2 Dk ; t 2 Dn and s t, then U t Us ;
(c) diam.U.n// ! 0 when n ! 1.
T S
Then the space L.U/ D f U.n/ W n 2 !g is homeomorphic to K.
Proof. Given any s 2 D! , the family Vs D fUsjn W n 2 !g is decreasing by (b)
and the diameters of its elements approachT zero Tby (c). Therefore we can apply
Problem 236 of [TFS] to conclude
T that V s D fU sjn W n 2 !g ;. An evident
consequence of (c) isTthat Vs can have at most one element and therefore there is
x.s/ 2 M such that Vs D fx.s/g; let '.s/ D x.s/ for any s 2 D! .
If s; t 2 D! and s t, then there is n 2 ! for which sjn tjn; then '.s/ 2 Usjn
and '.t/ 2 Ut jn while Usjn \ Ut jn D ; by (a) and therefore '.s/ '.t/, i.e.,
the map ' is injective. If x 2 L.U/, then there is sn 2 Dn such that x 2 Usn
for all n 2S!. It follows from (a) and (b) that sn snC1 for every n 2 ! and
hence s D n2! sn 2 D! is well-defined; it is immediate that '.s/ D x and hence
' W D! ! L.U/ is a bijection.
Finally, if we have an arbitrary s 2 D! and U 2 .'.s/; M /, then there is " > 0
such that B .'.s/; "/ U . Take n 2 ! with diam.U.n// < ". The set Osjn is an
open neighborhood of the point s in D! and, for any t 2 Osjn , we have tjn D sjn
and therefore '.t/ 2 Ut jn D Usjn . Furthermore, if y 2 Usjn , then .'.s/; y/
diam.Usjn / < " which shows that Usjn B .'.s/; "/ U . Thus '.Osjn / U
which proves continuity of ' at the point s. We established that ' is a continuous
bijection and hence homeomorphism because D! is compact. Therefore K D D! is
homeomorphic to L.U/ and Fact 1 is proved. t
u
Returning to our solution observe that we proved in Problem 128 of [TFS] that
the Cantor set K is homeomorphic to D! . This, together with Problem 302, implies
that the space K is zero-dimensional. If s 2 D<! , then Os is an infinite open
subset of D! ; since the family fOs W s 2 D<! g is a base in D! , the space D! has
no isolated points. Thus K is a zero-dimensional second countable compact space
without isolated points.
Now assume that X is a metrizable compact zero-dimensional space without
isolated points and denote by C.X / the family of all non-empty clopen subsets of
X . Let d be a metric on X such that .d / D .X/ and d.x; y/ 1 for all x; y 2 X .
It is easy to construct a sequence fWn W n 2 !g of finite disjoint open covers of X
such that WnC1 is a refinement of Wn and diam.Wn / 21n for all n 2 !.
If A X , let p.n; A/ be the minimal number of elements of Wn needed to

cover A. Since Wn is a finite cover of the space X , the number p.n; A/ is well-
defined and p.n; A/ jWn j for every A X and n 2 !. For any A exp.X /,
let P .n; A/ D maxfp.n; A/ W A 2 Ag. It is evident that P .n; A/ 1 says that
A is inscribed in Wn , i.e., for any A 2 A there is W 2 Wn such that A W .
Consequently, if P .n; A/ D 1, then diam.A/ diam.Wn / 21n .
Now, if we have a non-empty family A .X /, then every A 2 A is infinite
because X has no isolated points. Fix A 2 A and pick distinct x; y 2 A; then
r D d.x; y/ > 0. There is m 2 ! such that 21m < r and hence diam.W / < r for
any W 2 Wm . This shows that no element of Wm can contain A and hence the set
N.A/ D fn 2 ! W P .n; A/ > 1g is non-empty. As a consequence, the number
i.A/ D min N.A/ is well-defined for any non-empty A .X /.
Let U; D X and U.0/ D fU; g. Suppose that n 2 ! and have families U.k/ D
fUs W s 2 Dk g C.X / for all k n with the following properties:
S
(1) U.k/ is disjoint and U.k/ D X for all k n;
(2) if 0 m < k n; s 2 Dm ; t 2 Dk and s t, then Ut Us ;
(3) for each k < n if m D i.U.k//, then P .m; U.k C 1// < P .m; U.k//.
To construct the family U.n C 1/ take any s 2 Dn and let m D i.U.n//. If there
is W 2 Wm such that Us W , then split Us into two disjoint clopen sets arbitrarily,
e.g., take distinct x; y 2 Us (which is possible because X has no isolated points),
find V 2 C.X / with x 2 V X nfyg and let V0 D V \ Us ; V1 D Us nV0 . If we
let Us _ j D Vj for all j 2 D, then we get a family Vs D fUs _ 0 ; Us _ 1 g such that
P .m; Vs / D 1 < P .m; U.n//.
If k D p.m; Us / > 1, then take W1 ; : : : ; Wk 2 Wm such that Us W1 [

[ Wk
and let V0 D W1 \ Us and V1 D Us nV0 (observe that V0 ; and V1 ; for
otherwise Us can be covered by less than k-many elements of Wm ). If we let Us _ j D
Vj for all j 2 D, then p.m; V0 / D 1 and p.m; V1 / D k 1, i.e., for the family
Vs D fUs _ 0 ; Us _ 1 g the inequality P .m; Vs / < P .m; U.n// still holds.
After we construct
S the sets Us _ j for all s 2 Dn and j 2 D, we obtain the family
U.n C 1/ D fVs W s 2 Dn g D fUs W s 2 DnC1 g; it is evident that the conditions (1)
and (2) are now satisfied for all k n C 1. The property (3) also holds for k D n
because P .m; U.n C 1// D maxfP .m; Vs / W s 2 Dn g < P .m; U.n// (recall that
our construction guarantees that P .m; Vs / < P .m; U.n// for every s 2 Dn ). Thus
our inductive procedure can be continued to obtain a sequence fU.k/ W k 2 !g such
that (1)(3) are fulfilled for all n 2 !.
Let jWj j D nj and mj D n0 C

C nj for all j 2 !; observe that we have the
following property:
(4) for any j 2 ! and n mj , we have the equality P .j; U.n// D 1. As a
consequence, diam.U.n// 21j for all n mj and hence diam.U.n// ! 0
when n ! 1.
Indeed, it follows from (2) that if P .j; U.n// D 1, then P .j; U.k// D 1 for all
k n. This, together with (3), implies that if j D i.U.n//, then P .j; U.nCnj // D
1 and hence (4) is true.
Thus our family U D fUs W s 2 D<! g satisfies (1)(4) and hence the conditions
(a)(c) of Fact 1 are fulfilled for U. Therefore Fact 1 can be applied to conclude that
L.U/ is homeomorphic to K D D! . The property (1) implies that L.U/ D X , so X
is homeomorphic to K and hence our solution is complete.
T.349. Prove that a countable metrizable space X is homeomorphic to Q if and only
if X has no isolated points.
Solution. We Sidentify P with the space ! ! (see Problem 313). As usual, ! 0 D f;g
and ! D f! W n 2 !g. If f is a function, then dom.f / is its domain; given a
<! n
function g the expression f g says that dom.f / dom.g/ and gjdom.f / D f .

If we have a set of functions ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D
Given a space Z denote by C.Z/ the family of all non-empty clopen subsets of
Z. If Z is a metric space and we have a family A exp.Z/, then diam.A/ D
supfdiam.A/ W A 2 Ag. The expression Y ' Z says that the spaces Y and Z are
homeomorphic. If k; n 2 ! and s 2 ! k , then s _ n 2 ! kC1 is defined by .s _ n/.k/ D
n and .s _ n/jk D s. For any s 2 ! <! , let Os D fx 2 ! ! W s xg.
Fact 1. If P is a countable dense subspace of R, then there exists a homeomorphism
f W Q ! P such that x; y 2 Q and x < y implies f .x/ < f .y/. In particular, if
P and Q are countable dense subspaces of R, then P ' Q.
Proof. Choose some faithful enumerations fqn W n 2 !g and fpn W n 2 !g of the
sets Q and P respectively. If x; y 2 R and x y, then the expression ord.x; y/ D
R says that xRy where R is one of the symbols < " or > ".
Letting r0 D p0 assume that n 2 ! and we have defined a point ri 2 P for all
i n in such a way that
(1) for any i < n if Rk D ord.qi C1 ; qk / for each k i , then ri C1 D pj where
j D minfl 2 ! W ord.pl ; rk / D Rk for all k i g.
Let Rk D ord.qnC1 ; qk / for all k n. Then U D fp 2 P W ord.p; pk / D Rk
for all k ng is a non-empty open subset of P (it is evident that we have three
possibilities for U : either U D .pk ; C1/ \ P for some k n or U D . 1; pk / \
P for some k n or else U D .pk ; pm / \ P for some k; m n). Therefore
the number j D minfl 2 ! W pl 2 U g is well-defined; let rnC1 D pj . It is
straightforward that (1) is fulfilled for all i n which shows that we can continue
our inductive construction to obtain a set fri W i 2 !g with the property (1) fulfilled
for all i 2 !. Observe that it follows from (1) that
(2) if i j , then ord.qi ; qj / D ord.ri ; rj / and, in particular, qi < qj implies
ri < rj .
Let f .qi / D ri for all i 2 !; then f W Q ! P . An immediate consequence of (2)

is that p; q 2 Q and p < q implies f .p/ < f .q/. To see that P 0 D fri W i 2 !g D
P suppose that it is false and consider the numbers n D minfi 2 ! W pi 2 P nP 0 g
and k D minfi 2 ! W fp0 ; : : : ; pn1 g fr0 ; : : : ; ri gg. For the set A D fr0 ; : : : ; rk g
we have fp0 ; : : : ; pn1 g A; it is evident that the sets L D fr 2 A W r < pn g
and R D fr 2 A W r > pn g are disjoint and x < y for any x 2 L and y 2 R
(note that it is possible that some of the sets L; R is empty). For the sets A0 D
fq0 ; : : : ; qk g; L0 D fqi 2 A0 W ri 2 Lg and R0 D fqi 2 A0 W ri 2 Rg we have
f .A0 / D A; f .L0 / D L and f .R0 / D R. The property (2) shows that a < b for
any a 2 L0 and b 2 R0 and therefore J D fx 2 R W a < x < b for any a 2 L0 and
b 2 R0 g is a non-empty open subset of R. We have J \ A0 D ;; the set Q being
dense in R, the number m D minfi 2 ! W qi 2 J g is well-defined and k < m.
For any i k, if qm < qi , then qi 2 R0 and hence ri 2 R which implies pn < ri .
Analogously, if i k and qi < qm , then ri 2 L and hence ri < pn .
Now, if k < i < m and qi < qm (or qi > qm ), then it follows from qi J that
we have two possibilities: either there is a 2 L0 with qi a or there is b 2 R0 such
that b qi (in fact, the inequalities have to be strict because our enumeration of Q is
faithful, but that does not matter). Observe that the second (first) possibility cannot
take place because then qm < b qi (qi a < qm ) which is a contradiction.
Thus qi a (qi b) for some a 2 L0 (for some b 2 R0 respectively) and hence
ri f .a/ < pn (or pn < f .b/ ri respectively). This proves that
(3) ord.qm ; qi / D ord.pn ; ri / for all i < m,
which implies, together with (2), that rm D pn because the numbers p0 ; : : : ; pn1
belong to the set fri W i < mg and hence n is a minimal natural number for which
(3) holds. As a consequence, pn 2 P 0 ; this contradiction shows that f .Q/ D P and
hence f is a bijection.
Now let B D f.p; q/ \ Q W p; q 2 Q and p < qg. It follows from density of
Q in R that B is a base in Q. An immediate consequence of (2) is that ff .U / W
U 2 Bg D C D f.a; b/ \ P W a; b 2 P and a < bg. The set P is also dense in
R and therefore C is a base in P . Thus we can apply Fact 2 of S.491 to see that f
is continuous. The map f being a bijection, we have ff 1 .V / W V 2 Cg D B, so
Fact 2 of S.491 can be applied again to conclude that f 1 is continuous and hence
f is a homeomorphism. Fact 1 is proved. t
u
Returning to our solution, observe that it is evident that if X is homeomorphic to
Q, then X is a countable metrizable space without isolated points. Now assume that
X is a countable metrizable space without isolated points and fix a metric d on X
such that .d / D .X/ and d.x; y/ 1 for all x; y 2 X .
Note first that
(4) if U 2 .X /, then U is not compact and hence there is an infinite D U
which is closed and discrete in X .
Indeed, if V 2 .X / and V is compact, then it must have an isolated point x
(see Problem 133); the set V being dense in V , we have fxg \ V ;, i.e., x 2 V .
Thus fxg 2 .V / which implies fxg 2 .X/, i.e., the point x is isolated in X ; this
contradiction shows that V is not compact. Since X is zero-dimensional by Problem
307, there is a clopen V U . We proved above that V D V is not compact and
hence there is an infinite D V which is closed and discrete in V . The set V is
closed, so D V U is closed and discrete in X and hence (4) is proved. As an
easy consequence we obtain the following property:
(5) For any W 2 C.X / and " > 0, there
S is a disjoint W D fWn W n 2 !g C.X /
such that diam.W/ < " and W D W.
Indeed, let D W be an infinite closed and discrete subset of X which exists by
(4). Using paracompactness and zero-dimensionality of W , we can find a disjoint
cover W of the space W such that diam.W/ < " and jV \ Dj 1 for any V 2 W
(see Problem 314). It is evident that W cannot be finite, so we can choose a faithful
enumeration fWn W n 2 !g of the family W finishing the proof of (5).
Letting U; D X assume that n 2 ! and we have, for all k n, a family
U.k/ D fUs W s 2 ! k g C.X / such that
S
(6) U.k/ is disjoint and U.k/ D XSfor all k n;
(7) if k < n and s 2 ! , then Us D fUs _ i W i 2 !g;
k
(8) diam.U.k// 21k for all k n.

To construct a family U.n C 1/ take any s 2 ! n ; the property (5) guarantees
S
existence of a disjoint family W D fWi W i 2 !g C.X / such that Us D W and
diam.W/ 2nC1 1
. Let Us _ i D Wi for all i 2 !. After the family fUs _ i W i 2 !g
is constructed for all s 2 ! n , we obtain U.n C 1/ D fUs W s 2 ! nC1 g and it is
straightforward that the conditions (6)(8) are fulfilled for all k n C 1. Thus our
inductive construction can be continued to obtain a sequence fU.k/ W k 2 !g for
which (6)(8) hold for all n 2 !. T
Consider the set S D fs 2 P W Fs D fUsjn W n 2 !g ;g. Since for every
s 2 P we have diam.Usjn / ! 0 when n ! 1, the set Fs can have at most one point.
Therefore, for any s 2 S , there is x.s/ 2 X such that Fs D fx.s/g; let '.s/ D x.s/.
This gives us a map ' W S ! X . Given any x 2 X , there is sn 2 ! n such that
x 2 Usn for all n 2 !.S It follows from (6) and (7) that sn snC1 for all n 2 ! and
hence the point s 2 n2! sn is well-defined. It is immediate that x D '.s/ which
proves that ' is an onto map.
If s and t are distinct points of S , then sjn tjn for some n 2 ! and therefore
'.s/ 2 Usjn while '.t/ 2 Ut jn . The sets Usjn and Ut jn being disjoint by (6), we have
'.s/ '.t/ and hence ' is a bijection.
Now take a point s 2 S and " > 0; there is n 2 ! such that 21n < ". The set
V D Osjn \ S is an open neighborhood of s in S . If t 2 V , then '.t/ 2 Ut jn D Usjn ;
besides, for any y 2 Usjn , we have d.'.s/; y/ diam.Usjn / < 21n < " and therefore
'.V / Usjn Bd .'.s/; "/ which proves continuity of ' at the point s.
Now assume that x 2 X and ' 1 .x/ D s; given any V 2 .s; S/, there is n 2 !
such that Osjn \ S V . Observe that x 2 Usjn and ' 1 .Usjn / Osjn which shows
that ' 1 is continuous at the point x. Thus ' is a homeomorphism. Observe also
that fOs W s 2 ! <! g is a base in P; for any s 2 ! <! , the set Us is non-empty, and if
x 2 Us , then ' 1 .x/ 2 Os \ S which proves that S is dense in P.
Thus we established that X is homeomorphic to a dense subspace of P; recalling
that P is dense in R, we conclude that X is homeomorphic to a dense subspace of R
and hence X ' Q by Fact 1. Our solution is complete.
T.350. Prove that every countable metrizable space is homeomorphic to a closed
subspace of Q.
Solution. Let J D . 1; 1/ I and take any countable metrizable space X . Pick
a countable dense subspace D of the space J ! I! . It is evident that D is also
dense in I! and hence D D is a dense countable subspace of I! I! . Since X
is homeomorphic to a subspace of I! (see Problem 209 of [TFS]), we can consider
that X I0 D f 1g I! I! I! . Since E D D D J ! J ! , we have
E \ I0 D ; and hence E \ X D ;.
Furthermore, E is dense-in-itself because I! I! does not have isolated points
and therefore the space F D E[X is also dense-in-itself. The set I0 is closed in I!
I! and hence X D I0 \ F is closed in F . Finally, observe that F is homeomorphic
to Q because it is a countable metrizable space without isolated points (see Problem
349). Thus X can be embedded in Q as a closed subspace.
T.351. Let X be a second countable -compact space. Prove that X is not Polish if
and only if it contains a closed subspace homeomorphic to Q.
Solution. The expression Y ' Z says that the spaces Y and Z are homeomorphic.
If U is a family of non-empty subsets of Z and z 2 Z, say that U converges to z
(denoting this by U ! z) if, for any V 2 .z; Z/, the family fU 2 U W U nV ;g
is finite. Note that any finite U exp.Z/nf;g converges to any point of Z.
Fact 1.S Suppose S that Z is a space, U exp.Z/nf;g and U ! z for some z 2 Z.
Then U fU W U 2 Ug [ fzg.
S
Proof. If y 2 Znfzg and y 2 U, then take W 2 .z; Z/ such that y W .
Then V D ZnW 2 .y; Z/ and V intersects only finitely many elements of U say,
S y 2 U 0 [

[ U n which shows that y 2 U i for
U0 ; : : : ; Un . As a consequence,
some i n and hence y 2 fU W U 2 Ug so Fact 1 is proved. t
u
Fact 2. Suppose that Z is a first countable space; if z 2 Z; O 2 .Z/ and z 2
OnO, then there is family U D fUn W n 2 !g .O/ such that U n \ U m D ; if
n m and U ! z.
Proof. Fix a local base B D fBn W n 2 !g of Z at the point z such that B nC1 Bn
for every n 2 !. We can choose a sequence fzn W n 2 !g O such that zn ! z
when n ! 1. Let j.0/ D 0; there is i.0/ 2 ! such that zi.0/ 2 B0 .
Suppose that n 2 ! and we have i.0/ <

< i.n/ and j.0/ <

< j.n/
such that zi.k/ 2 Bj.k/ for all k n while zi.k/ Bj.kC1/ for all k < n. Since
zi.n/ z, there exists j.n C 1/ > j.n/ such that zi.n/ Bj.nC1/ ; since zn ! z, there
is i.n C 1/ > i.n/ for which zi.nC1/ 2 Bj.nC1/ . Thus we can construct sequences
fi.n/ W n 2 !g ! and fj.n/ W n 2 !g ! such that
(1) i.n/ < i.n C 1/ and j.n/ < j.n C 1/ for all n 2 !;
(2) zi.n/ 2 Bj.n/ nBj.nC1/ for all n 2 !.
Let Vn D .Bj.n/ nBj.nC1/ / \ O for all n 2 !. It is easy to check that the family
fVn W n 2 !g .O/ is disjoint and converges to z. If we take a set Un 2
.zi.n/ ; Z/ such that U n Vn for all n 2 !, then the family U D fUn W n 2 !g is as
promised so Fact 2 is proved. t
u
Fact 3. Let Z be a first countable space. Suppose that P is a nowhere dense subset
of Z and A Z is finite. Then, for any O 2 .Z/ with A OnO, there is
H 2 .O/ such that A H nH and H \ P A.
Proof. We can find Wa 2 .a; Z/ for every a 2 A such that the family fW a W a 2
Ag is disjoint; let Va D .Wa \ O/nP for all a 2 A. It is clear that a 2 V a nVa and
therefore there is a family Ga D fGan W n 2 !g .Va / such that Gan \ Gam D ; if
n m and Ga ! a (see Fact 2).
Choose Han 2 .Z/ such that Han Gan for all a 2 A and n 2 !. It is clear
that Ha D fHan W n 2 !g still converges
S to a and Han \ PSD ; for all a 2 A and
S2 !.
n Therefore, for the set Ha D Ha , we have H a fHan W n 2 !g [ fag
fGaSW n 2 !g [ fag (see Fact 1) whence H a \ P fag. Consequently, the set
n
H D a2A Ha is as promised and Fact 3 is proved. t

u
Fact 4. Let Z be a second countable space without isolated points. If Z is of first
category in itself, then some closed Y Z is homeomorphic to Q.
S
Proof. We have Z D fZn W n 2 !g where Zn is closed, nowhere dense in Z
and Zn ZnC1 for all n 2 !. Let d be a metric on Z such that .d / D .Z/ and
d.x; y/ 1 for all x; y 2 Z.
Take z0 2 Z arbitrarily and let A0 D fz0 g. By Fact 3 applied to the sets P D
Z0 ; A D A0 and O D Znfz0 g, there is V0 2 .O/ such that A0 V 0 nV0 and
V 0 \ Z0 A0 . If U0 D ZnV 0 then Z0 nA0 U0 and U0 \ V0 D ;.
Assume that n 2 ! and we have chosen non-empty finite sets Ak Z and
Uk ; Vk 2 .Z/ for all k n with the following properties:
(3) the family fA0 ; : : : ; An g is disjoint;
(4) if k < n, then for any z 2 Bk D A0 [

[ Ak , there is y 2 AkC1 such that
d.z; y/ 21k ;
(5) Zk nBk Uk ; Bk V k nVk and Uk \ Vk D ; for all k n;
(6) VkC1 Vk for all k < n.
Since Bn V n nVn , for any z 2 Bn , we can choose y.z/ 2 Vn such that
d.z; y.z// 21n ; let AnC1 D fy.z/ W z 2 Bn g. It is immediate that the properties (3)
and (4) hold for k D n. It is easy to see that for the set O D Vn nAnC1 , we have
BnC1 D Bn [ AnC1 OnO, so we can apply Fact 3 to find VnC1 2 .O/ such
that BnC1 V nC1 nVnC1 and V nC1 \ ZnC1 BnC1 . If UnC1 D ZnV nC1 then
ZnC1 nBnC1 UnC1 ; VnC1 Vn and UnC1 \ VnC1 D ;. It is clear that (3)(6) are
fulfilled for all k n C 1 and therefore our inductive construction can be continued
to give us families fAi W i 2 !g; fUi W i 2 !g and fVi W i 2 !g with the conditions
S for all n 2 !.
(3)(6) satisfied
Let Y D i 2! Ai ; given any x 2 Y and m 2 !, we have x 2 An for some
n 2 ! and hence there is xm 2 AnCmC1 such that d.x; xm / 2nCm 1
. It follows from
(3) that fxm W m 2 !g Y nfxg; since xm ! x, the point x is not isolated in Y and
hence Y has no isolated points.
Observe also that Y \ Ui D ; for all i 2 !. Indeed, if x 2 Y \ Ui for some
i 2 !, then there is n 2 ! such that x 2 An . The property (5) implies that x 2
V i CnC1 V i [the last inclusion holds by (6)]. However, Vi \SUi D ; by (5) and
hence V i \ Ui D ; which contradicts x 2 V i \ Ui . Thus Y \ . i 2! Ui / D ;. Now,
if z 2 ZnY , then z 2 Zn for some n 2 ! and therefore z 2 Zn nBn . It follows S from
(5) that z 2 Un ; the pointSz 2 ZnY was chosen arbitrarily, so ZnY i 2! Ui .
This proves that ZnY D i 2! Ui and hence Y is a closed subset of X . Since Y is
a countable metrizable space without isolated points, we have Y ' Q by Problem
349 and Fact 4 is proved. t
u
Returning to our solution suppose that a closed F X is homeomorphic to Q.
If X is Polish, then so is F (see Problem 260 of [TFS]) and hence Q is Polish which
is false because Q is of first category in itself (see Problem 274 of [TFS]). This
contradiction shows that X is not Polish and proves sufficiency.
Now assume that X is not Polish; we can consider that X is a dense subspace of
a metrizable compact space K. Since X is not a G -set in K (see Problem 259 of
[TFS]), the space KnX is not -compact. However, KnX is a G -set in K because
X is -compact. Consequently, KnX is a Borel set and hence it is analytic by
Problem 334.
Therefore we can apply Fact 4 of T.132 to conclude that there is a subspace
P KnX which is closed in KnX and homeomorphic to P. The space Z D P
(the bar denotes the closure in K) is compact and hence F D Z \ X is closed
in X . Besides, Z \ .KnX / D P because P is closed in KnX . The space P is
not -compact (see Fact 1 of T.132) and hence not locally compact (it is an easy
exercise that any second countable locally compact space is -compact). Besides,
P is homeomorphic to ! ! and hence any non-empty open U P contains a closed
subspace homeomorphic to P (this also follows from Fact 4 of T.132). This proves
that clP .U / is not compact for any U 2 .P /; an easy consequence is that F is
dense in Z. The space Z does not have isolated points because it is a closure of P
which is dense-in-itself. Since F is also dense in Z, it does not have isolated points
either.
Furthermore, F is -compact being closed
S in X . Let fFn W n 2 !g be a family
of compact subsets of F such that F D n2! Fn . If W D IntF .Fn / ; for some
n 2 !, then there is W 0 2 .Z/ such that W 0 \ F D W . The density of F in Z
implies that clZ .W 0 / D clZ .W / clZ .Fn / D Fn (the last equality holds because
Fn is compact). Thus W 0 Fn and hence W 0 \ P D ; which contradicts density
of P in Z.
This contradiction shows that Fn is nowhere dense in F for all n 2 !, i.e., F is a

dense-in-itself second countable space of first category in itself. Thus we can apply
Fact 4 to find a set Y F which is closed in F and Y ' Q. It is immediate that
Y is also closed in X and hence X contains a closed set homeomorphic to Q. This
settles necessity and makes our solution complete.
T.352. Let X be an analytic non--compact space. Prove that some closed subspace
of X is homeomorphic to P.
Solution. If Z is a space, call a family A D fAs W s 2 S g exp.Z/ indexed
discrete if for any z 2 Z there is W 2 .z; Z/ such that the set fs 2 S W W \ As
;g is finite. As usual, we identify the space P with ! ! (see Problem 313). The
expression X ' Y says that the space X is homeomorphic to the space Y . As
usual, we identify any ordinal with the set of its predecessors and, in particular,
n D f0; : : : ; n 1g for any n 2 !. If k; n 2 ! and s 2 ! k , then s _ n 2 ! kC1 is
defined by .s _ n/.k/ D n and .s _ n/jk D s. If we have s 2 ! n ; t 2 ! k where
n k and n; k 2 ! C 1, then s t says that tjn D s. For any s 2 ! <! , let
Os D fx 2 ! ! W s xg.
If d is a metric on a set M , then .d / is the topology generated by the metric
d . If we work with a metric space .M; d /, then for any x 2 X and r > 0 the set
Bd .x; r/ D fy 2 M W d.x; y/ < rg is the ball of radius r centered at x. If f
is a function, then dom.f / is its domain. Suppose that we have a set of functions
ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D fj j.dom.fi / \S dom.fj // for
any i; j 2 I . Then we can define a function f with dom.f / D i 2I dom.fi /
as follows: given any x 2 dom.f /, find any i 2 I with x 2 dom.fi / and let
f .x/ D fi .x/. It is easy to check that the value of f at x does not depend on the
choice
S of i so we have consistently defined a function f which will be denoted by
ffi W i 2 I g
Since X is analytic, there is a continuous onto map f W P ! X . Fix a complete
metric on P such that ./ D .P/. We will need the family S D fA X W there
is a -compact P X such that A P g. Let A; D P; assume that n > 0, and
for each k < n, we have a family fAs W s 2 ! k g of closed subsets of P with the
(1) if 0 < k < n, then diam .As / < 2k for all s 2 ! k ;
(2) if 0 k < n, then fAs W s 2 ! k g is an indexed discrete family in P;
(3) if 0 k < n, then ff .As / W s 2 ! k g is an indexed discrete family in X ;
(4) if 0 k < l < n; s 2 ! k ; t 2 ! l and s t, then At As ;
(5) if 0 k < n, then f .As / S for any s 2 ! k .
Fix s 2 ! n1 and consider the set Bs D fx 2 As W f .U / S for any U 2
.x; As /g. For each x 2 As nBs there is Wx 2 .x; As / such that f .Wx / 2 S. The
open cover W D fWx W x 2 As nB Ss g of the Lindelf space As nBs has a countable
subcover W 0 ; it is evident that f . W 0 / D f .As nBs / 2 S. Since f .As / S, we
have f .Bs / S and hence there exists an infinite set D f .Bs / which is discrete
and closed in X . It is easy to find, for every y 2 D, a set Vy 2 .y; X/ such that
the family fVy W y 2 Dg is indexed discrete in X . The family ff 1 .Vy / W y 2 Dg is

indexed discrete in P; pick a point x.y/ 2 f 1 .y/ \ Bs for each y 2 D. It is clear
that x.y/ 2 f 1 .Vy /, so we can find Uy 2 .x.y/; P/ such that Uy f 1 .Vy / and
diam .Uy / < 2n for every y 2 D.
Take a faithful enumeration fyn W n 2 !g of the set D and let As _ n D Uyn \ Bs
for each n 2 !. It follows from Uyn \ Bs ; that f .As _ n / S for each n 2 !.
It is clear that diam .As _ n / diam .Uyn / < 2n which shows that (1) is fulfilled
for the sets As _ n for all n 2 !. The properties (2) and (3) hold because the families
U D fUyn W n 2 !g and V D fVyn W n 2 !g are indexed discrete and the families
fAs _ n W n 2 !g and ff .As _ n / W n 2 !g are obtained by shrinking U and V
respectively. After we define the family fAs _ n W n 2 !g for each s 2 ! n1 , we have
the collection fAt W t 2 ! n g. This method of construction guarantees the property
(4). We also assured that f .As _ n / S for each s 2 ! n1 and n 2 ! which shows
that (5) is also true for the family fAt W t 2 ! n g.
Thus our inductive construction can be completed for all natural n to give us
families fAs W s 2 ! n g with (1)(5) for all n 2 !. Now we are ready to describe a
closed subset of X which is homeomorphic to P. Given any x 2 P let sn D xjn for
each n 2 !; it follows from (1) and (4) that fAsn W n 2 !g is a decreasing family of
closed subsetsT of P such that diam .Asn / ! 0. As a consequence there is a unique
point g.x/ 2 fAsn W n 2 !g (see Problem 236 of [TFS]). Let h.x/ D f .g.x//;
this gives us a map h W P ! X . We must check that F D h.P/ is closed in X and
the map h W P ! F is a homeomorphism.
Take any point x 2 P and an arbitrary W 2 .h.x/; X /; since h.x/ D f .g.x//
and f is a continuous
T map, there is " > 0 such that f .B .g.x/; "// W . We saw
that fg.x/g D fAxjn W n 2 !g; since diam .Axjn / ! 0 when n ! 1, there exists
m 2 ! such that Axjm B .g.x/; "/. The set U D Oxjm is an open neighborhood of
x in P and, for any y 2 U , we have yjm D xjm which implies g.y/ 2 Ayjm D Axjm
and therefore h.y/ D f .g.y// 2 f .Axjm / f .B .g.x/; "// W which shows
that h.U / W and hence h is continuous at the point x.
If x and y are distinct points of P, then there is n 2 ! such that xjn yjn.
The property (3) implies that f .Axjn / \ f .Ayjn / D ;; since h.x/ 2 f .Axjn / and
h.y/ 2 f .Ayjn /, we have h.y/ h.x/ and hence h is a condensation.
To show that h1 W F ! P is continuous, take any y 2 T F and x 2 P with
h.x/ D y; then h1 .y/ D x. If si D xji for all i 2 !, then y 2 ff .Asi / W i 2 !g.
Observe that the family fOxjn W n 2 !g is a local base at x in P, so if we take any
W 2 .x; P/, then there is n 2 ! such that Oxjn W . The set U D f .Asn /\F 3 y
is open in F by (3). If z 2 U , then h1 .z/jn D sn by (2) and (3); this implies
h1 .z/jn D xjn, i.e., h1 .z/ 2 Oxjn W . The point z 2 U was chosen arbitrarily,
so h1 .U / Oxjn W . Thus we proved that for any W 2 .x; P/, there exists
U 2 .y; F / such that h1 .U / W . Therefore h1 is continuous at every y 2 F
and h is a homeomorphism. S
To see that F is closed in T X take any point y 2 F . If Fn D ff .As / W s 2 ! n g
for each n 2 !, then F fFn W n 2 !g. Therefore y 2 F n for each n 2 !;
the family ff .As / W s 2 ! n g being discrete, there is sn 2 ! n such that y 2 f .Asn /
for all n 2 !. The properties (3)S and (4) imply that sn snC1 for any n 2 !
and
T hence we can define x D fsn W n 2 !g 2 P. Now observe that h.x/ 2
ff .Asn / W n 2 !g; if y h.x/, then there is W 2 .h.x/; X / such that y W .
We have h.x/ D f .g.x//, so continuity of f implies that there is " > 0Tsuch that
f .B .g.x/; "// W ; take m 2 ! such that 2m < ". Now, fg.x/g D n2! Axjn
and hence g.x/ 2 Axjm which shows that .g.x/; z/ diam .Axjm / < 2m < " for
every z 2 Axjm and hence Asm D Axjm B .g.x/; "/.
As a consequence, f .Asm / f .B .g.x/; "// W and therefore f .Asm / W
which contradicts y W and y 2 f .Asm /. This contradiction shows that y D h.x/
and hence y 2 F . The point y 2 F was chosen arbitrarily, so F F which shows
that F is closed in X . Thus there exists a closed F X with F ' P and our
T.353. Prove that any uncountable analytic space contains a closed subspace which
is homeomorphic to the Cantor set.
Let X Sbe an uncountable analytic space. Suppose first that X is -compact, i.e.,
X D fKn W n 2 !g where Kn is compact for all n 2 !; it follows from
jX j > ! that there is n 2 ! such that Kn is uncountable. Every analytic space has
a countable network, so w.Kn / D nw.Kn / D ! (see Fact 4 of S.307) which shows
that Kn is an uncountable second countable K-analytic space. It was proved in
Fact 4 of T.250 that every second countable uncountable K-analytic space contains
a subspace homeomorphic to the Cantor set, so K embeds in Kn and hence in X .
Now, if X is not -compact, then P embeds in X by Problem 352. Since P ' ! !
(see Problem 313) and K ' D! ! ! , the Cantor set K embeds in P and hence
in X .
T.354. Prove that any non--compact Borel set can be condensed onto I! as well
as onto R! .
We denote by J any of the spaces I or R, so if a statement is proved for the symbol J ,
then we actually give two identical proofs: one for I and another one for R. Suppose
that we have a set of functions ffi W i 2 I g such that fi j.dom.fi / \ dom.fj // D
fj j.dom.fi /S
\ dom.fj // for any i; j 2 I . Then we can define a function f with
dom.f / D i 2I dom.fi / as follows: given any x 2 dom.f /, find i 2 I with
S countable space. Suppose that Mi M; Mi Mi C1

Fact 1. Let M be a second
for all i 2 ! and M D i 2! Mi . Then, for every i 2 !, we can choose a continuous
function hi W Mi ! J so that the family fhi W i 2 !g separates the points of M , i.e.,
for any distinct x; y 2 M there is i 2 ! such that x; y 2 Mi and hi .x/ hi .y/.
Proof. The space M embeds in J ! , so let e W M ! J ! be the respective

embedding. Define ei W M ! J by ei .x/ D e.x/.i / for all x 2 M . It is clear
that the set E D fei W i 2 !g separates the points of M . Choose an enumeration
fh0i W i 2 !g of the set E in which every e 2 E occurs !-many times and let
hi D h0i jMi for all i 2 !.
To see that the sequence fhi W i 2 !g is as promised, take two distinct x; y 2 M .
There are i 2 ! and e 2 E such that fx; yg Mi and e.x/ e.y/. There is j > i
such that h0j D e and hence hj .x/ D e.x/ e.y/ D hj .y/ so Fact 1 is proved. u t
Fact 2. Given spaces Z and Y , assume that F D fFt W t 2 T g is a closed locally
finite cover of Z and we have a family fft W t 2 T g of functions such that ft W Ft !
Y is continuous for all tS2 T and ft j.Ft \ Fs / D fs j.Ft \ Fs / for any s; t 2 T .
Then the function f D t 2T ft W Z ! Y is also continuous.
S
Proof. If H is a closed subset of Y , then f 1 .H / D 1
t 2T ft .H / is closed
1
because the family fft .H / W t 2 T g is locally finite (and hence closure-preserving
by Fact 2 of S.221) being a shrinking of the family F . Therefore f is continuous
u
Returning to our solution, assume that we are given a non--compact Borel set
X . We will construct a condensation of X onto J ! . Given a point a 2 J ! , let
.a/ D fx 2 J ! W jfn 2 ! W x.n/ a.n/gj < !g. It is evident that if a; b 2 J ! and
the set fn 2 ! W a.n/ b.n/g is infinite, then .a/ \ .b/ D ;. This observation
makes it possible to find a set fai W i 2 !g J ! such that .ai / \ .aj / D ;
for any distinct i; j 2 !. It is an easy exercise that .ai / is a countable union of
! !
nowhere dense compact S subspaces of J! ; thus .ai!/ is of first category in J for
all i 2 !, and hence i 2! .ai / J because J is Cech-complete and hence
has the Baire property.
S
Let Sn D J ! n. f.ai / W i ng/ for all n 2 !. Since the set .aSi / is -compact
for every i 2 !, the set Sn Borel in J ! for all n 2 !. Observe that n2! Sn D J ! ;
besides, Sn SnC1 and SnC1 nSn D .an / for all n 2 !.
Any Borel set is analytic by Problem 334, so we can apply Problem 352 to find
a closed P X with P ' P. It is easy to Sfind a disjoint family P D fPi W i 2 !g
of clopen subsets of P such that P D i 2! Pi and Pi ' P for each i 2 !. It
is immediate that P is a discrete family of closed subsets of X . Define a function
W P ! R by letting .x/ D i for all x 2 Pi and i 2 !; it is evident that is
continuous.
Since X is normal, there is a continuous function W X ! R such that jP D .
Let F0 D 1 .. 1; 12 / and Fn D 1 .n 12 ; n C 12 / for every n 2 N; we will
also need the sets O0 D 1 .. 1; 12 // and On D 1 ..n 12 ; n C 12 // for every
n 2 N. It is clear
Sthat F D fFn W n 2 !g is a locally finite family of closed subsets
of X such that F D X while the family O D fOn W n 2 !g .X/ is disjoint
and Pn On Fn for all n 2 !. Let Gn D F0 [

[ Fn for all n 2 !.
Next we will inductively construct, for every n 2 !, a continuous mapping 'n W
Gn ! J ! J ! such that
(1) 'n is injective for all n 2 !;

(2) 'nC1 jGn D 'n for all n 2 !;
(3) J ! Sn 'n .Gn / J ! SnC1 for all n 2 !.
The set J ! S0 is Borel (see Problem 333), so there is closed E0 P0 and a
condensation f0 W E0 ! J ! S0 (see Problem 341). There is aSfamily fHi W i 2 !g
of closed subsets of X such that Hi Hi C1 for all i 2 ! and i 2! Hi D F0 nE0 .
In the product J ! J ! let p W J ! J ! ! J ! and q W J ! J ! ! J ! be the
natural projections onto the first and the second factor respectively. Apply Fact 1
to find a sequence H D fhi W i 2 !g of functions such that hi 2 C.Hi ; J / for
all i 2 ! and H separates the points of F0 nE0 . For each i 2 !, define a function
ui W E0 [ Hi ! J as follows: ui .x/ D p.f0 .x//.i / if x 2 E0 and ui .x/ D hi .x/
if x 2 Hi ; let vi 2 C.F0 ; J / be any continuous extension of ui for every i 2 !
(which exists because the space F0 is normal, the set E0 [ Hi is closed in F0 and
ui 2 C.E0 [ Hi ; J /).
Furthermore, for each i 2 !, define si W E0 [ Hi ! J as follows: si .x/ D
q.f0 .x//.i / if x 2 E0 and si .x/ D a0 .i / for all x 2 Hi . Let ti 2 C.F0 ; J / be a
continuous extension of si for all i 2 !. Now let '0 D .i 2! vi /.i 2! ti /; it is
immediate that '0 W F0 ! J ! J ! and '0 .x/ D .v.x/; t.x// where v.x/.i / D vi .x/
and t.x/.i / D ti .x/ for all x 2 F0 and i 2 !. Since any diagonal product of
continuous maps is continuous, the map '0 is continuous. It is immediate from the
definition that '0 jE0 D f0 and hence '0 .G0 / D '0 .F0 / f0 .E0 / D J ! S0 .
Now, if x 2 F0 nE0 , then there is m 2 ! such that x 2 Hi for all i m and hence
t.x/.i / D a0 .i / for all i m. Therefore t.x/ 2 .a0 / for all x 2 F0 nE0 . Since
t.x/ 2 S0 for all x 2 E0 , we have '0 .x/ D .v.x/; t.x// 2 J ! .S0 [ .a0 // D
J ! S1 , so the condition (3) is satisfied for n D 0. Since (2) is vacuous for n D 0,
let us check that (1) is fulfilled, i.e., '0 is injective.
Given two distinct x; y 2 F0 assume first that x; y 2 E0 . The map f0 being
injective, we have '0 .x/ D f0 .x/ f0 .y/ D '0 .y/. If x 2 E0 and y 2 F0 nE0 ,
then '0 .x/ 2 J ! S0 while '0 .y/ 2 J ! .a0 / which shows that '0 .x/ '0 .y/.
Finally, if x; y 2 F0 nE0 , then by our choice of the sequence fhi W i 2 !g there is
i 2 ! such that x; y 2 Hi and hi .x/ hi .y/. Consequently, vi .x/ vi .y/ which
shows that v.x/ v.y/ and hence '0 .x/ '0 .y/. Thus the conditions (1)(3) are
satisfied for n D 0.
Now assume that we have '0 ; : : : ; 'k1 for some k 2 N with the properties
(1)(3) for all n k 1. To construct 'k , observe that 'k1 .Gk1 / is a Borel subset
of J ! J ! because 'k1 is a condensation (see Problem 314 and Fact 3 of T.333).
Therefore A D .J ! Sk /n'k1 .Gk1 / is a Borel subset of J ! J ! and hence
there exists a closed Ek Pk and a condensation fk W Ek ! A (see Problem 341;
note that it is possible that A D Ek D ;).
There is a family
S fHi W i 2 !g of closed subsets of X such that Hi Hi C1 for
all i 2 ! and i 2! Hi D Fk n.Gk1 [ Ek /.
Apply Fact 1 to find a sequence H D fhi W i 2 !g of functions such that
hi 2 C.Hi ; J / for all i 2 ! and H separates the points of Fk n.Gk1 [Ek /. For each
i 2 ! define a function ui W Gk1 [ Ek [ Hi ! J as follows: ui .x/ D p.fk .x//.i /
if x 2 Ek ; if x 2 Gk1 , then ui .x/ D p.'k1 .x//.i /; and ui .x/ D hi .x/ if x 2 Hi ;

let vi 2 C.Gk ; J / be any continuous extension of ui for every i 2 ! (which exists
because the space Gk is normal, the set Gk1 [Ek [Hi is closed in Fk and, evidently,
ui 2 C.Gk1 [ Ek [ Hi ; J /).
Furthermore, for each i 2 !, define si W Gk1 [Ek [Hi ! J as follows: si .x/ D
q.fk .x//.i / if x 2 Ek ; if x 2 Gk1 , then si .x/ D q.'k1 .x//.i / and si .x/ D ak .i /
for all x 2 Hi . Let ti 2 C.Gk ; J / be a continuous extension of si for all i 2 !.
Now let 'k D .i 2! vi /.i 2! ti /; it is immediate that 'k W Gk ! J ! J ! and
'k .x/ D .v.x/; t.x// where v.x/.i / D vi .x/ and t.x/.i / D ti .x/ for all x 2 Gk and
i 2 !. Since any diagonal product of continuous maps is continuous, the map 'k is
continuous. It is immediate from the definition that 'k jGk1 D 'k1 ; 'k jEk D fk
and hence 'k .Gk / 'k1 .Gk1 / [ fk .Ek / D J ! Sk .
Now, if x 2 Fk n.Gk1 [Ek /, then there exists m 2 ! such that x 2 Hi and hence
t.x/.i / D ak .i / for all i m. Therefore t.x/ 2 .ak / for all x 2 Fk n.Gk1 [ Ek /.
Since t.x/ 2 Sk for all x 2 Ek , we have 'k .x/ D .v.x/; t.x// 2 J ! .Sk [
.ak // D J ! SkC1 , so the condition (3) is satisfied for n D k. The condition (2)
is satisfied for n D k 1 by our construction; let us check that (1) is fulfilled, i.e.,
'k is injective.
Given two distinct points x; y 2 Gk , assume first that x; y 2 Gk1 ; then 'k .x/ D
'k1 .x/ 'k1 .y/ D 'k .y/. If x 2 Gk1 and y 2 Fk nGk1 , then 'k .x/ D
'k1 .x/ 2 'k1 .Gk1 / while 'k .y/ 'k1 .Gk1 / because 'k .y/ D fk .y/
'k1 .Gk1 / if y 2 Ek and 'k .y/ 2 J ! .ak / .J ! J ! /n'k1 .Gk1 / for all
y 2 Fk n.Gk1 [ Ek /.
Next observe that if x; y 2 Ek , then 'k .x/ D fk .x/ fk .y/ D 'k .y/ because
fk is a condensation. If x 2 Ek and y 2 Fk n.Gk1 [ Ek /, then 'k .x/ 2 J ! Sk
while we have 'k .y/ 2 J ! .ak / which shows that 'k .x/ 'k .y/. Finally, if
x; y 2 Fk n.Gk1 [ Ek /, then by our choice of the sequence fhi W i 2 !g there is
i 2 ! such that x; y 2 Hi and hi .x/ hi .y/. Consequently, vi .x/ vi .y/ which
shows that v.x/ v.y/ and hence 'k .x/ 'k .y/.
Thus the conditions (1)(3) are satisfied for n D k and hence our inductive
construction gives us a sequence of mappingsS f'n W n 2 !g with the properties (1)
follows from (2) that a map ' D i 2! 'n W X ! J ! J ! is well-defined.
(3). It S
Since n2! Sn D J ! , the condition (3) guarantees that '.X / D J ! J ! ' J ! .
The conditions (1) and (2) imply that ' is injective. Finally observe that it follows
from (2) thatS the maps f n D 'n jFn W n 2 !g satisfy the assumptions of Fact 2 and
hence ' D n2! n is continuous. Thus ' condenses X onto J ! and hence our
T.355. Give an example of a non--compact subspace of R which cannot be
condensed onto a compact space.
Solution. If CH holds, then there exists a Luzin space X R (see Fact 1
of T.046). If some compact K X is uncountable, then K is a Luzin space
which is a contradiction because a Luzin space cannot even condense onto a
compact space by Problem 045. Thus every compact subspace of X is countable;
since X is uncountable, it is not -compact. Consequently, X is an example of a

non--compact subspace of R which cannot be condensed onto a compact space by
Problem 045.
Now, if CH is false, then take any X R with jX j D !1 . Since CH does not
hold, we have jX j < c. If a compact K X is uncountable, then K embeds in K
by Problem 353 and hence jKj jKj D c > !1 D jX j, a contradiction. Thus every
compact K X is countable and hence X is not -compact. If X condenses onto a
compact space Y , then w.Y / D nw.Y / nw.X / D ! (see Fact 4 of S.307), i.e., Y
is a second countable uncountable compact space. Therefore we can apply Problem
353 again to find a subspace Z Y which is homeomorphic to the Cantor set K. As
a consequence, jX j D jY j jZj D jKj D c > !1 ; this final contradiction shows
that X is a non--compact subspace of R which does not condense onto a compact
space.
T.356. Prove that Q cannot be condensed onto a compact space.
Solution. If there is a condensation f W Q ! K of Q onto a compact space K,
then K is countable and hence scattered because otherwise it can be continuously
mapped onto the space I which is uncountable (see Problem 133). Thus there is a
point x 2 K which is isolated in K. Since f is a condensation, there is q 2 Q such
that fqg D f 1 .x/; since fxg 2 .K/, the set fqg is open in Q, i.e., q is an isolated
point of Q which is a contradiction.
T.357. Prove that for any metrizable compact X , the space Cp .X / condenses onto
a compact space.
Solution. Given a space Z let Z be a metric on the set C .Z/ defined by the
formula Z .f; g/ D supfjf .z/ g.z/j W z 2 Zg for all f; g 2 C .Z/. It was proved
in Problem 248 of [TFS] that .C .Z/; Z / is a complete metric space. Given any
set A C.Z/ let clu .A/ D ff 2 C.Z/ W there is a sequence ffn W n 2 !g A
such that fn !
!f g. It was proved in Problem 084 of [TFS] that there exists a unique
topology Zu on C.Z/ such that clu .A/ is the closure of A in .C.Z/; Zu / for any
A C.Z/. The space .C.Z/; Zu / is denoted by Cu .Z/; the space Cu .Z/ is the set
C .Z/ with the topology induced from Cu .Z/.
Fact 1. For any space Z, the identity map i W C .Z/ ! C .Z/ is a homeomor-
phism between Cu .Z/ and .C .Z/; Z / or, in other words, .Cu .Z// D .Z /.
Proof. Let B.f; "/ D fg 2 C .Z/ W Z .f; g/ < "g for all f 2 C .Z/ and
" > 0, i.e., B.f; "/ is the "-ball (with respect to Z ) centered at f . Suppose that
F is a closed subset of Cu .Z/. If F is not closed in .C .Z/; Z /, then take any
f F which is in the .Z /-closure of F . For every n 2 N there exists fn 2
B.f; n1 / \ F ; it is immediate that fn !
!f and hence f is in the closure of F in the
space Cu .Z/ which is a contradiction. Thus every closed subset of Cu .Z/ is also
closed in .C .Z/; Z / and therefore .Cu .Z// .Z /.
Now assume that a set F is closed in .C .Z/; Z / and there is a function f 2
clu .F /nF , i.e., there exists a sequence S D ffn W n 2 !g F with fn ! !f .
For any " > 0, there is m 2 ! such that jfm .z/ f .z/j < 2" for all z 2 Z and
hence Z .fm ; f / 2" < ". Therefore fm 2 B.f; "/ \ S which shows that f is
in the closure of S in the space .C .Z/; Z / and hence f is in the closure of F
in .C .Z/; Z /; this contradiction proves that any closed subset of .C .Z/; Z / is
closed in Cu .Z/, i.e., .Z / .Cu .Z//. Thus .Cu .Z// D .Z / and Fact 1 is
proved. t
u
Fact 2. If K is a compact space, then w.Cu .K// D w.K/. In particular, if K is
metrizable, then Cu .K/ is a Polish space.
Proof. Observe first that Cu .K/ D Cu .K/ and hence Cu .K/ is a completely
metrizable space by Fact 1. As a consequence, w.Cu .K// D d.Cu .K//; choose
a dense A Cu .K/ with jAj D d.Cu .K//. Since the identity map i W Cu .K/ !
Cp .K/ is continuous (see Problem 086 of [TFS]), the set A has to be dense in Cp .K/
as well and hence w.K/ D i w.K/ D d.Cp .K// jAj D w.Cu .K//. Therefore
w.K/ w.Cu .K//.
Now if w.K/ D , then there is a set B Cp .K/ with jBj which is
dense in Cp .K/. It is evident that B separates the points of K. Let P .B/ D ff1
: : :
fk W k 2 N and fi 2 B for all i kg. We will need the sets R.B/ D
f0 C 1 g1 C

C n gn W n 2 N; gi 2 P .B/ and i 2 R for all i ng and
Q.B/ D fq0 C q1 g1 C

C qn gn W n 2 N; gi 2 P .B/ and qi 2 Q for all i ng.
It is easy to check that R.B/ is an algebra in C.K/; it follows from B R.B/ that
R.B/ separates the points of K and hence R.B/ is dense in Cu .K/ (see Problem
191 of [TFS]).
Observe that jQ.B/j D jBj , so it suffices to show that Q.B/ is dense in
Cu .K/. We will do it in several steps.
(1) If frn W n 2 !g R and rn ! r then rn f !
!rf for any f 2 C.K/.
Indeed, there is A > 0 such that jf .x/j < A for all x 2 K. Given " > 0,
there is m 2 ! such that jrn rj < A" for all n m. Then jrn f .x/ rf .x/j D
jf .x/jjrn rj < A
A" D " for all n m and x 2 K which shows that (1) is true.
(2) f 2 clu .Q.B// for any f 2 R.B/.
There are g1 ; : : : ; gn 2 P .B/ such that f D 0 C 1 g1 C

C n gn . For
each i D 0; : : : ; n, choose a sequence frki W k 2 !g Q such that rki ! i .
Now, rki gi 2 Q.B/ for all k 2 ! and i n, so we can apply (1) to conclude that
rki gi !
!i gi for all i n; therefore fk D rk C rk g1 C

C rk gn belongs to Q.B/
0 1 n
!
for all k 2 ! and fk !f by Problem 035 of [TFS]. Thus f 2 clu .Q.B//, i.e., (2)
is proved.
It is an immediate consequence of (2) that we have R.B/ clu .Q.B// and
therefore Cu .K/ D clu .R.B// clu .Q.B//, i.e., the set Q.B/ is dense in the
space Cu .K/. Thus d.Cu .K// jQ.B/j jBj D D w.K/ so Fact 2 is proved.
t
u
Fact 3. Every locally compact space condenses onto a compact space.
Proof. Let Z be a locally compact space; if Z is compact, then there is nothing to
prove, so we assume that Z is not compact. Then the one-point compactification K
of the space Z is a compact space such that jKnZj D 1; let a 2 K be the point for
which Z [ fag D K. Fix any point b 2 Z. In the space K collapse the closed set
F D fa; bg to a point to obtain the space K=F . Recall that K=F D .KnF / [ fF g
(i.e., the set F is considered to be a point in the space K=F ; denote this point by
zF ). Define a map pF W K ! K=F as follows: p.x/ D x for all x 2 KnF and
p.x/ D zF for all x 2 F . It was proved in Fact 2 of T.245 that K=F is a Tychonoff
space and the map pF is continuous. Thus K=F is compact; it is immediate that
pF jZ is a condensation of Z onto the compact space K=F , so Fact 3 is proved. u t
Returning to our solution observe first that if X is finite, then Cp .X / D RX is a
locally compact space; therefore Cp .X / condenses onto a compact space by Fact 3.
Now, if X is infinite, then fix a nontrivial convergent sequence S X (we
consider that S contains its limit and hence S is a countably infinite compact
subspace of X ). If Y Z X , let YZ W Cp .Z/ ! Cp .Y / be the restriction
map, i.e., YZ .f / D f jY for all f 2 Cp .Z/; if Z D X , then YZ will be denoted
by Y . Let Cp .Y jX / D Y .Cp .X // for any Y X .
Let A be a countable dense subset of X such that S A. Observe that
Cp .S jX / D Cp .S / because S is compact (see Fact 1 of T.218). It follows from
S D SA A that SA maps Cp .AjX / onto Cp .S /; the space Cp .S / is not -
compact by Problem 186 of [TFS], so the space Cp .AjX / is not -compact either.
Furthermore, the Polish space Cu .X / condenses onto Cp .X / (see Fact 2 and
Problem 086 of [TFS]). Since the map A condenses Cp .X / onto Cp .AjX /, there
exists a condensation of Cu .X / onto Cp .AjX /. Therefore Cp .AjX / is a non--
compact Borel set (see Problem 341) and we can apply Problem 354 to conclude that
there is a condensation ' W Cp .AjX / ! K of Cp .AjX / onto a compact space K.
Then D ' A condenses Cp .X / onto the compact space K and hence our
T.358. Prove that a Polish space X is dense-in-itself if and only if P condenses
onto X .
Solution. For the sake of brevity, we say that a space is crowded if it is non-empty
and dense-in-itself, i.e., has no isolated points. We identify P with the space ! ! (see
Problem 313). Given any s 2 ! <! let Os D ft 2 P W s tg. If A is a family of sets
in a metric space .Z; /, then diam .A/ D supfdiam .A/ W A 2 Ag.
Fact 1. If F and G are closed subsets of a crowded Polish space .Z; d / and F nG
family fFn W n 2 !g of closed subsets
is crowded, then for any " > 0, there exists a S
of Z such Sthat diamd .Fn / < ", the set Fn n. fFi W i < ng/ is crowded for every
n 2 ! and fFn W n 2 !g D F nG.
Proof. For each x 2 F nG fix a set Ux 2 .x; Z/ such that diamd .Ux / < " and
U x \ G D ; (the bar denotes the closure in Z). Choose a countable subcover fVi W
i 2 !g of the cover fUx W x 2 F nGg of the space F nG. The set Wi D Vi \ .F nG/
is crowded being a non-empty open subset of a crowded set F nG for every i 2 !.
Therefore Hi D W i is also crowded;
S it is clear that Hi is a closed subset of Z and
Hi F nG for all i 2 !. Besides, i 2! Hi D F nG.
If no finite subfamily of the family H D fHi W i 2 !g covers the space F nG,

then start off with the set F0 D H0 , and for any n 2 !, let FnC1 D Hk where
k D minfi 2 ! W Hi is not contained in F0 [

[ FS
S n g. It is easy to see that
0
n2! Fn D F nG; every F n is crowded and F n D Fn n. i <n Fi / is a non-empty
open subset of Fn by our definition of the family F D fFn W n 2 !g. Thus Fn0 is
crowded for every n 2 ! and hence the family F is as promised.
Now assume that F nG D H0 [

[ Hn for some n 2 !. Throwing S away the
empty elements of this union, we can consider that Hk0 D Hk n. i <k Hi / ; for
all k n. The set Hn0 is crowded and has diameter < ", so it suffices to split it into
countably many pieces as required in our fact without caring about the diameter.
To see that it is possible, observe that the space Hn0 being crowded, we can choose
a faithfully enumerated convergent sequence A D fan W n 2 !g Hn0 such that
an ! a 2 Hn0 nA. It is an easy exercise that it is possible to construct a family
O D fOi W i 2 !g .Hn0 / such that O i Hn0 nfag and ai 2 Oi for all i 2 !
while the family fO i W i 2 !g is disjoint and O ! a in the sense that for any
O 2 .a; Z/, there are only finitely many elements of O which are not contained
in O. It is easy to see thatSO0 ! a for any O0 O.
Observe that if E D fO 2k W k 2 !g, then E D E [ fag by Fact 1 of T.351.
S
Since a 2 fO2kC1 W k 2 !g Hn0 nE, we have P0 D Hn0 n.E [ fag/ D Hn0 nE
and hence P0 is crowded being a closure of a non-empty open subset of a crowded
space Hn0 . The set Pi D O 2i 2 is also crowded for the same reason for all i 2 N.
Finally, let Fi D Hi for all i < n and FnCi D Pi for all i 2 !. It is immediate that
the sequence fFi W i 2 !g witnesses that Fact 1 is proved. t
u
Returning to our solution, assume that f W P ! X is a condensation and some
x 2 X is isolated in X . Then there is s 2 P such that fsg D f 1 .x/ is open in P,
i.e., the point s is isolated in P. However P has no isolated points by Problem 347;
this contradiction proves that X has no isolated points either.
Now suppose that X is a dense-in-itself Polish space and fix a complete metric
d on X such that .d / D .X/ and d.x; y/ 1 for all x; y 2 X . Let Q; D X ;
it is evident that we can apply Fact 1 to the sets F D X and G D ; to find a
sequence fP Sn W n 2 !g of closed subsets of X such that diamd .Pn / 2 and
1
0
Pn D Pn n. i <n Pi / is crowded for all n 2 !. Now let Qs D Ps.0/ for each s 2 ! 1 .
Assume that k > 1 and we have, for every i < k, a family Qi D fQs W s 2 ! i g
S
(1) Q; D X and Qs D fQs _ i W i 2 !g for all s 2 ! m whenever m C 1 < k;
(2) the set Qs is crowded for every s 2 ! i whenever i < k;
(3) the family Qi is disjoint and diam.Qi / 21i for all i < k;
(4) Qs _ 0 [

[ Qs _ j is closed in X for any j 2 ! and s 2 ! m with m C 1 < k.
To construct Qk fix any t 2 ! k1 ; let s D tj.k 2/ and m D t.k 2/. It follows
from (4) that the sets G D Qs _ 0 [

[Qs _ .m1/ and F D Qs _ 0 [

[Qs _ .m1/ [
Qt are closed in X and F nG D Qt is a crowded set.S Apply Fact 1 to find a sequence
P D fPn W n 2 !g S of closed subsets of X such that i 2! Pi D Qt ; diam.P/ 2k
1
0 0
and Pn D Pn n. i <n Pi / is crowded for all n 2 !. Let Qt _ i D Pi for every
i 2 !; after we have the set Qt _ i for all t 2 ! k1 and i 2 !, we obtain a family
Qk D fQs W s 2 ! k g.
It is straightforward that the conditions (1)(4) are still satisfied for the families
Q0 ; : : : ; Qk and hence we can construct inductively
T a sequence fQi W i 2 !g with
the properties (1)(4). Given s 2 P, let F D fQsjn W n 2 !g. It follows from
(1) that the family fQsjn W n 2 !g is decreasing. Since .X; d / is complete and
diamd .Qsjn / ! 0 by (3), the set F consists of a single point x.s/; let f .s/ D x.s/.
This defines a function f W P ! X .
Take any s 2 P and " > 0; there is n 2 ! such that diamd .Qsjn / < ". If
y 2 Qsjn , then d.f .s/; y/ diamd .Qsjn / < " and hence y 2 Bd .f .s/; "/ which
shows that Qsjn Bd .f .s/; "/. The set W D Osjn is an open neighborhood of s in
P. If t 2 W , then tjn D sjn and hence f .t/ 2 Qt jn D Qsjn Bd .f .s/; "/, i.e.,
f .W / Bd .f .s/; "/ which proves that f is continuous at the point s.
If s; t 2 P and s t, then there is n 2 ! such that sjn tjn; we can assume
without loss of generality that s.n 1/ < t.n 1/. It follows from the properties (3)
and (4) that Qsjn \Qt jn D ;; besides, Qt j.nC1/ Qt jn and hence Qsjn \Qt j.nC1/ D
;. Since f .s/ 2 Qsjn and f .t/ 2 Qt j.nC1/ , we have f .s/ f .t/, i.e., f is an
injection.
Finally, if x 2 X , then it follows from the property (1) that there is a sequence
fsn W n 2 !g ! <! such that sn 2 ! n and x 2 Qsn for S all n 2 !. The property (3)
showsTthat sn snC1 T for all n 2 !
T and hence s D n2! sn is well-defined. Thus
x 2 n2! Qsn D n2! Qsjn n2! Qsjn and therefore f .s/ D x, i.e., f is a
condensation of P onto X . Our solution is complete.
T.359. Prove that for any metrizable compact X , the space P condenses onto
Cp .X /.
Solution. Since X is compact and metrizable, the space Cu .X / is Polish (see Fact 2
of T.357). It is evident that Cu .X / has no isolated points, so there is a condensation
f W P ! Cu .X / by Problem 358. If i W Cu .X / ! Cp .X / is the identity map, then i
condenses Cu .X / onto Cp .X / and hence i f is a condensation of P onto the space
Cp .X /.
T.360. Prove that Cp .X / is analytic if and only if R! maps continuously onto
Cp .X /. Observe that not every analytic space is a continuous image of R! .
Solution. The space R! is Polish and hence analytic (see Problem 334). If R!
maps continuously onto Cp .X /, then Cp .X / is analytic because, evidently, any
continuous image of an analytic space is an analytic space.
Now assume that the space Cp .X / is analytic and fix a continuous onto map
f W P ! Cp .X /. We can consider that P is a closed subset of R! (see Problem
273 of [TFS]). Since Cp .X / is a locally convex space, we can apply Problem 104
to find a continuous map g W R! ! Cp .X / such that gjP D f . It is clear that
g.R! / g.P/ D f .P/ D Cp .X /, so Cp .X / is a continuous image of R! .
Recall that a space Z is called connected if every clopen subset of Z belongs

to the family f;; Zg. Call a clopen U Z nontrivial if U f;; Zg. In this
terminology, a space Z is connected if and only if it has no nontrivial clopen sets.
Let .!/ D fx 2 R! W jfn 2 ! W x.n/ 0gj < !g. The set .!/ is connected
by Fact 2 of T.312; besides, .!/ is dense in R! , so we can apply Fact 1 of T.312
to conclude that R! is connected as well. The two-point space D is analytic being
compact and metrizable; however it is impossible to map R! continuously onto D.
Indeed, if h W R! ! D is a continuous onto map, then h1 .0/ is a nontrivial clopen
subset of R! which is a contradiction because R! is connected. Thus, an analytic
space D is not a continuous image of R! .
T.361. Suppose that X is an infinite space such that Cp .X / is analytic. Prove that
Cp .X / can be continuously mapped onto R! . Deduce from this fact that if X and Y
are infinite spaces such that Cp .X / and Cp .Y / are analytic, then each of the spaces
Cp .X / and Cp .Y / maps continuously onto the other one.
Solution. For every i 2 ! let i W R! ! R be the natural projection of R! onto its
i th factor. The space Cp .X / is not -compact by Problem 186 of [TFS], so we can
apply 352 to conclude that there is a closed F Cp .X / such that F ' P. Since
R! is analytic, there is a continuous onto map f W F ! R! ; let fi D i f for
all i 2 !. The space Cp .X / is normal because it has a countable network; therefore
there is a continuous function gi W Cp .X / ! R such that gi jF D fi for all i 2 !.
Then g D i 2! gi W Cp .X / ! R! . It is easy to see that g extends f and therefore
g.Cp .X // g.F / D f .F / D R! and hence g maps Cp .X / continuously onto R! .
Now if Cp .Y / is analytic, then R! maps continuously onto Cp .Y / by Problem
360. We proved that Cp .X / maps continuously onto R! , so the composition of
the relevant functions maps Cp .X / continuously onto Cp .Y /. Analogously, if Y is
infinite and Cp .Y / is analytic, then Cp .Y / maps continuously onto Cp .X /.
T.362. Prove that for any second countable -compact space X , the space Cp .X /
is K .
S
Solution. We have X D fXn W n 2 Ng where Xn is compact for all n 2 N. Let d
be a metric on X such that .d / D .X/ and d.x; y/ 1 for all x; y 2 X . For any
k; l; n 2 N, let Mkln D ff 2 IX W jf .x/ f .y/j k1 whenever x and y are points
of X such that d.x; z/ < n1 and d.y; z/ < n1 for some z 2 Xl g. Our first observation
is that
(1) the set Mkln is compact for any k; l; n 2 N.
Indeed, if f 2 IX nMkln , then there exist points x; y 2 X such that d.x; z/ < n1 and
d.y; z/ < n1 for some z 2 Xl while jf .x/ f .y/j > k1 ; then r D jf .x/ f .y/j
k > 0. The set Of D fg 2 I W jg.x/ f .x/j < 2 and jg.y/ f .y/j < 2 g is open
1 X r r
in the space I and f 2 Of . For any function g 2 Of , we have jf .x/ f .y/j

X
jf .x/ g.x/j C jg.x/ g.y/j C jg.y/ f .y/j and therefore

1 r r 1
jg.x/ g.y/j jf .x/ f .y/j jf .x/ g.x/j jf .y/ g.y/j > r C D
k 2 2 k
which shows that g 2 IX nMkln . Since a function g 2 Of was chosen arbitrarily, we

have Of IX nMkln . Thus, for every f 2 IX nM Skln , there OfX2 .I / such that
X
Of I nMkln . As a consequence, I nMkln D fOf W f 2 I nMkln g is open in

X X
IX and hence Mkln is compact being closed in IS X

.
The property (1) implies
T that the set P kl D fM kln W n 2 Ng is -compact; we
claim that Cp .X; I/ T D fP kl W k; l 2 Ng and hence C p .X; I/ is K .
Take any f 2 fPkl W k; l 2 Ng; given a point z 2 Z and " > 0, there is l 2 N
with z 2 Xl . Pick any k 2 N such that k1 < ". Since f 2 Pkl , there is n 2 N
such that jf .x/ f .y/j k1 < " whenever d.x; t/ < n1 and d.y; t/ < n1 for some
t 2 Xl . In particular, jf .x/ f .z/j < " if d.x; z/T< n1 which proves continuity of
f at the point z. Thus f is continuous and hence fPkl W k; l 2 Ng Cp .X; I/.
Now if f 2 Cp .X; I/, fix any k; l 2 N. If f Pkl , then for any n 2 N, there exist
xn ; yn 2 X and zn 2 Xl such that jf .xn / f .yn /j > k1 while d.xn ; zn / < n1 and
d.yn ; zn / < n1 . Since Xl is compact, the sequence fzn W n 2 Ng has a subsequence
which converges to a point z 2 Xl . Passing to that subsequence and reindexing the
sequences fxn W n 2 Ng and fyn W n 2 Ng if necessary, we can assume, without loss
of generality, that zn ! z. We also have xn ! z and yn ! z because d.xn ; zn / < n1
and d.yn ; zn / < n1 for all n 2 N. Since f is continuous at the point z, there is
W 2 .z; X / such that jf .w/ f .z/j < 2k 1
for all w 2 W . There is m 2 N such that
xm ; ym 2 W and hence jf .xm / f .ym /j jf .xm / f .z/j C jf .z/ f .ym /j <
1
2k
CT 1
2k
D k1 which is a contradiction. Thus T f 2 Pkl for all k; l 2 N and therefore
f 2 fPkl W k; l 2 Ng. This proves that fPkl W k; l 2 Ng D Cp .X; I/.
Let In D 1SC nC1 1
; 1 nC11
; it is evident that In I; In Int.InC1 / for
all n 2 N and fIn W n 2 Ng D J D . 1; 1/. It is easy to see that the set
Qln D ff 2 IX S W f .Xl / In g is closed in IX and hence compact for all l; n 2 N.
ThereforeTEl D fQln W n 2 Ng is -compact for all l 2 N which shows that the
set E D fEl W l 2 Ng is K .
If f 2 Cp .X; J / and l 2 N, then f .Xl / is a compact subset of J which
implies that there is n 2 N such that f .Xl / In ; thus f 2 El for all l 2 N
and therefore f 2 E. Consequently, Cp .X; T J / E and hence Cp .X; J / D
Cp .X; J / \ Cp .X; I/ E \ P where P D fPS kl W k; l 2 Ng.
On the other hand, if f 2 E, then f .X / D ff .Xl / W l 2 Ng J because
f .Xl / J for every l 2 N. Therefore f 2 J X . If, additionally, f 2 P , then
f has to be continuous because so are all functions from P . As a consequence,
E \ P Cp .X; J / which shows that E \ P D Cp .X; J /. We proved that the
space Cp .X; J / is the intersection of two K -subsets of IX and hence Cp .X; J / is
K . Since Cp .X / is homeomorphic to Cp .X; J / (see Problem 025 of [TFS]), we
proved that Cp .X / is also a K -space.
T.363. Let X be a space with a countable network. Prove that X is analytic if and
only if every second countable continuous image of X is analytic.
Solution. In this solution we will need some (possibly) discontinuous maps. Given
spaces Y and Z call a map f W Y ! Z measurable if f 1 .B/ 2 B.Y / for any
B 2 B.Z/. A space is called cosmic if it has a countable network. A space Z is
perfect if every U 2 .Z/ is an F -set in Z.
Fact 1. Given perfect spaces Y and Z a map f W Y ! Z is measurable if and only
if f 1 .U / 2 B.Y / for any U 2 .Z/.
S 0
Proof. We have B.Z/ D f .Z/ W < !1 g by Fact 1 of T.331. By our
assumption, f 1 .U / 2 B.Y / for any U 2 00 .Z/; if F 2 00 .Z/, then U D
ZnF 2 00 .Z/ and hence U 0 D f 1 .U / 2 B.Y /. Since F 0 D f 1 .F / D Y nU 0 ,
we have F 0 2 B.Y / by Fact 1 of T.331.
Now assume that < !1 and, for any < we have f 1 .U / 2 B.Y / for any
U 2 0 .Z/ [ 0 .Z/. If U 2 0 .Z/, then there is a sequence fn W n 2 !g
S
and Un 2 0n .Z/ for every n 2 ! such that U D n2! Un . Therefore f 1 .U / D
S 1
ff .Un / W n 2 !g; since f 1 .Un / 2 B.Y / for every n 2 ! by the induction
hypothesis, we have f 1 .U / 2 B.Y / by Fact 1 of T.331.
Finally, if F 2 0 .Z/, then U D ZnF 2 0 .Z/ and hence U 0 D f 1 .U / 2
B.Y /. Since F 0 D f 1 .F / D Y nU 0 , we have F 0 2 B.Y / by Fact 1 of T.331. Thus
our inductive procedure can be carried out to establish that f 1 .U / 2 B.Y / for any
U 2 0 .Z/ [ 0 .Z/ for all < !1 . Therefore f is a measurable map and Fact 1
is proved. t
u
Fact 2. Given a space Z for any A Z, denote by ZA the space with the
underlying set Z and the topology generated by .Z/ [QfA; ZnAg. Suppose that
A exp.Z/ is a non-empty family. In the product P D fZA W A 2 Ag consider
the set T .A/ D ff 2 P W f .A/ D f .B/ for any A; B 2 Ag. Then
(a) the space ZA is Tychonoff for any A 2 A and hence both spaces P and T .A/
are Tychonoff as well.
(b) T .A/ is a closed subspace of P homeomorphic to the space Z.A/ whose
underlying set is Z and whose topology is generated (as a subbase) by the
family S D .Z/ [ A [ fZnA W A 2 Ag.
Proof. The Tychonoff property of every ZA follows from ZA ' A .ZnA/
(we leave it to the reader to verify that any free union of Tychonoff spaces is a
Tychonoff space). Since products and subspaces of Tychonoff spaces are Tychonoff,
both spaces P and T .A/ are Tychonoff.
To see that T .A/ is closed in P take any f 2 P nT .A/. Then for some distinct
A; B 2 A, we have x D f .A/ y D f .B/. Pick disjoint U; V 2 .Z/ such that
x 2 U and y 2 V . The set W D ff 2 P W f .A/ 2 U and f .B/ 2 V g is an open
neighborhood of f in P such that W \ T .A/ D ;. This shows that P nT .A/ is
open in P and hence T .A/ is a closed subset of P .
Let qA W P ! ZA be the natural projection for every A 2 A. For each z 2 Z,
let fz .A/ D z for all A 2 A and observe that T .A/ D ffz W z 2 Zg. Furthermore,
qA .fz / D z for every z 2 Z and A 2 A. If i.z/ D fz for every z 2 Z, then
i W Z ! T .A/ is a bijection. Let id W Z ! Z be the identity map.
To check that the map i W .Z; / ! T .A/ is continuous, it suffices to show that
qA i is continuous for every A 2 A. It is clear that, qA i D id; besides, the map
id W Z.A/ ! ZA is continuous. Indeed, B D .Z/ [ fAg [ fZnAg is a subbase
of ZA , so it suffices to show that id1 .U / 2 for all U 2 B. We have B S, so
id1 .U / D U 2 S which shows that qA i is continuous for each A 2 A and
hence the map i W .Z; / ! T .A/ is continuous.
Furthermore, observe that j D i 1 D rA D qA jT .A/ for every A 2 A. It is
sufficient to prove that j 1 .U / is open in T .A/ for every U 2 S. Now, if U 2 S,
then there is A 2 A such that U 2 .ZA / and hence j 1 .U / D rA1 .U / is open
in T .A/ because the map rA is continuous. Therefore j D i 1 is also a continuous
map and hence i W .Z; / ! T .A/ is a homeomorphism. Fact 2 is proved. t
u
Fact 3. Let Y and Z be second countable spaces. If f W Y ! Z is a measurable
onto map and Y is analytic, then the space Z is also analytic.
Proof. Fix a continuous onto map h W P ! Y ; if B 2 B.Z/, then B 0 D f 1 .B/ 2
B.Y / and hence B 00 D h1 .B 0 / 2 B.P/ by Problem 332. Therefore .f h/1 .B/ is
a Borel subset of P for any Borel set B Z, i.e., the map g D f h is measurable.
Let B D fBn W n 2 !g be a base in Z; then, by our observation, the set Cn D
g 1 .Bn / is Borel in P for all n 2 !. Therefore Cn is a Borel set (as a space) for
every n 2 !. Thus PnCn is also a Borel set (see Fact 1 of T.331) and therefore the
space Tn D Cn .PnCn / is a Borel set as well for all n 2 ! (see Problem 333).
The family .P/ [ fCn W n 2 !g [ fPnCn W n 2 !g generates a topology
on P as a subbase. The Qspace P D .P; / is homeomorphic to a closed subspace
F of the space T D n2! Tn by Fact 2. Each Tn is analytic by Problem 334, so
F (and hence P ) is an analytic space by 335. By our construction of , we have
g 1 .Bn / D Cn 2 for all n 2 ! and therefore the map g W P ! Z is continuous.
Every continuous image of an analytic space is an analytic space, so Z is analytic
u
Fact 4. If Z is a cosmic space, then there exist second countable spaces M and N
and condensations f W M ! Z and g W Z ! N such that the mappings f 1 and
g 1 are measurable.
Proof. Take a countable network F in the space Z such that all elements of F are
closed in Z and any finite intersection of the elements of F belongs to F . Let
be a topology generated on Z by the family S D .Z/ [ F . Observe that ZnF 2
.Z/ for every F 2 F , and hence we can apply Fact 2 to conclude that .Z; /
is a Tychonoff space.
The family F is a base in the space M D .Z; /. To see it take any U 2
and x 2 U . The family S is a subbase of , so there are V1 ; : : : ; Vn 2 .Z/ and
F1 ; : : : ; Fk 2 F such that x 2 V \ F U where V D V1 \

\ Vn 2 .Z/ and
F D F1 \

\ Fn 2 F . The family F being a network in Z, there is G 2 F for
which x 2 G V . Then H D G \ F 2 F and x 2 H U which shows that F
is a base in .Z; /.
Let f W M ! Z be the identity map. Since the topology of M is stronger than
the topology of Z, the map f is a condensation. Given any U 2 .M / there is
S
F 0 F such that F 0 D U . All elements of F 0 are closed in Z, so U is an F -set
of Z. Thus U D .f 1 /1 .U / is an F -set in Z for every U 2 D .M /. Since
every F -set of Z is a Borel subset of Z, the map f 1 is measurable by Fact 1.
For any F 2 F fix a continuous function pF W Z ! I D 0; 1 such that
pF1 .0/ D F . This is possible because Z is cosmic and hence perfectly normal.
The diagonal product g D fpF W F 2 F g W Z ! I F is a continuous map; let
N D g.Z/. For every F 2 F , let qF W I F ! I be the natural projection.
If x; y 2 Z and x y, then there is F 2 F such that x 2 F and y F . Then
pF .x/ D 0 pF .y/ and hence g.x/ g.y/. Thus g W Z ! N is a condensation.
For every F 2 F , let F 0 D g.F /. We have F 0 D qF1 .0/ \ N which shows that F 0
is closed in N for every F 2 F . S 0
For anyS set U 2 .Z/ there is a family F 0 F such that U D F . Then
g.U / D fg.F / W F 2 F 0 g is an F -set in N because g.F / D F 0 is closed in N
for each F 2 F 0 . This proves that .g 1 /1 .U / D g.U / is an F -set in N for every
U 2 .Z/, so we can apply Fact 1 again to conclude that g 1 is also a measurable
map. Fact 4 is proved. t
u
Returning to our solution observe that any continuous image of an analytic space
is analytic, so if X is analytic, then every second countable continuous image of
X is analytic. Now assume that every second countable continuous image of X
is analytic and choose condensations f W M ! X and g W X ! N such that
w.M / D w.N / D ! and the maps h D g1 and u D f 1 are measurable (see
Fact 4). Since every second countable continuous image of X is analytic, the space
N is analytic. It is an easy exercise that a composition of measurable maps is a
measurable map, so u h W N ! M is a measurable map. Therefore M is analytic
by Fact 3 and hence X is analytic being a continuous image of an analytic space M .
Our solution is complete.
T.364. Let X be a space with a countable network. Prove that X is -compact if
and only if every second countable continuous image of X is -compact.
Solution. Necessity is evident so assume that every second countable continuous
image of X is -compact. Then X is analytic by Problem 363 because every -
compact second countable space is analytic. If X is not -compact, then there is a
closed F X with F ' P by Problem 352. Since P is a subset of R, there exists a
mapping h W F ! R which is a homeomorphism of F onto h.F /. The space X is
normal, so there is a continuous function h1 W X ! R such that h1 jF D h.
By perfect normality of X there is a continuous function f W X ! R such
that F D f 1 .0/. If g D h1 f and M D g.X /, then M R R is second
countable. For any z 2 F , we have g.z/ D .h1 .z/; f .z// D .h.z/; 0/; besides, if
g.y/ D .t; 0/ 2 M for some y 2 X , then f .y/ D 0 and hence y 2 F . This shows
that g.F / D M \ .R f0g/ is a closed subspace of M ; since g.F / D h.F / f0g,
we have g.F / ' h.F / ' P. It turns out that g.F / is a closed non--compact
subspace of M which is a second countable continuous image of X . Therefore M
is not -compact; this contradiction settles sufficiency.
T.365. Prove that a second countable space M is Polish if and only if there exists a
map f W P ! exp.M / with the following properties:
(a) f .s/ is compact for any s 2 P;
(b) for any s; t 2 P, if s t, then f .s/ f .t/;
(c) for any compact K M , there is s 2 P such that K f .s/.
Solution. TheS space P is identified with ! ! (see Problem 313). For any n 2 N we
let ! D f! W k < ng. For any s 2 ! <! there is n 2 ! such that s 2 ! n ; let
<n k
l.s/ D n. If n; i 2 ! and s 2 ! n , then s _ i 2 ! nC1 is defined by .s _ i /jn D s and

.s _ i /.n/ D i .
If M is Polish, then there exists an open continuous onto map h W P ! M by
Problem 328. By Fact 1 of S.491, there is a closed F PQsuch that h.F / D M and
h1 D hjF is a perfect map. For any s 2 P let K.s/ D n2! .s.n/ C 1/ (here we
consider the integer .s.n/C1/ to be the set f0; : : : ; s.n/g). Then K.s/ is compact for
any s 2 P and it is immediate that K.s/ K.t/ is s t. The set f .s/ D h.K.s// is
compact and it is clear that f .s/ f .t/ if s t. Now, if K is a compact subset of
M , then L D h1 1 .K/ is a compact subset of F by Fact 2 of S.259. If i W P ! !
is the natural projection onto the i th factor, then the set i .L/ is bounded for every
i 2 ! because any compact subspace of ! is finite.
As a consequence there is s.i / 2 ! such that i .L/ s.i / for all i 2 !. This
defines s 2 ! ! D P for which L K.s/ and hence f .s/ D h.K.s// h.L/ D K
which implies K f .s/, so we proved that the function f has the properties
(a)(c). This settles necessity.
Now assume that f W P ! exp.M / is a function with the properties (a)(c) and
fix a metric d on M such that .d / D .M /. As usual, for any x 2 M and " > 0,
the set B.x; "/ D fy 2 M W d.x; y/ < "g is the ball of radius " centered at x. We are
going to construct a map W ! <! ! ! <! and a family fUs W s 2 ! <! g .M /
S
(1) U; D M and Us D fU s _ i W i 2 !g for every s 2 ! <! ;
(2) if s 2 ! <! and l.s/ D i 1, then diam.Us / 1i ;
(3) .! n / ! n for all n 2 ! and if s; t 2 ! <! and s t, then .s/ .t/;
(4) if s 2 ! <! and K Us is compact, then there is t 2 P such that .s/ t and
K f .t/.
To start our inductive construction let .;/ D ; and U; D M . Assume that
n 2 N and we have a family fUs W s 2 ! <n g and a map W ! <n ! ! <n such that (1)
is fulfilled for all s 2 ! <.n1/ and the statements (2)(4) are true for all s 2 ! <n .
Fix any s 2 ! n1 ; we claim that for any x 2 Us we can find Vx 2 .x; M / and
k 2 ! such that V x Us ; diam.Vx / n1 and for any compact K Vx there is
t 2 P such that .s/_ k t and K f .t/. Indeed, if this were not true, then we
can choose a sequence fVi W i 2 !g .x; M / such that for all i 2 ! we have
V i Us \ B.x; 2nCi 1
/ and there is a compact Ki Vi for which
S Ki 6 f .t/ for any
t 2 P with .s/ i t. It is easy to see that the set K D . i 2! Ki / [ fxg Us is
_
compact and hence it follows from the induction hypothesis that there is t 2 P such
that .s/ t and K f .t/. If i D t.n 1/, then .s/_ i t and Ki K f .t/
Thus we canS construct sequences fVi W i 2 !g .M / and fki W i 2 !g !

such that Us D fV i W i 2 !g; diam.Vi / n1 for every i 2 ! and, for any compact
K Vi , there is t 2 P such that .s/_ ki t and K f .t/. Now let Us _ i D Vi
and .s _ i / D .s/_ ki for all i 2 !. After we accomplish this construction for all
s 2 ! n1 , we obtain a family fUs W s 2 ! n g and a map W ! <.nC1/ ! ! <.nC1/ such
that (1) holds for all s 2 ! <n and the assertions (2)(4) are true for all s 2 ! <.nC1/.
Therefore our inductive construction can be carried out for all n 2 ! to give us a
map W ! <! ! ! <! and a family fUTs W s 2 ! <! g with the properties (1)(4).
Take any t 2 P and assume that fUt jn W n 2 !g D ;. Let sn D tjn and pick
a point xn 2 Usn for every n 2 !. The property (4) implies that there exists tn 2 P
such that .sn / tn and xn 2 f .tn / for all n 2 !. Observe that .sn / .snC1 / by
(3) and therefore tm jn D .sn / D tn jn for all m n. Thus if we define a function
u 2 ! ! by u.i / D t0 .i / C

C ti .i / for all i 2 !, then tn u for all n 2 !.
Therefore xn 2 f .tn / f .u/ for all n 2 !, i.e., S D fxn W n 2 !g f .u/.
Observe that xn 2 U sn for all n 2 ! which implies that Fn D U sn \ f .u/ ; for
all n 2 !. Since
T the sequence
T T sets fFn W n 2 !g is decreasing by (1),
of compact
there is x 2 n2! Fn n2! U t jn D n2! Ut jn . This contradiction proves that
T
.20 / for any t 2 P, we have n2! Ut jn ;,
and hence the family U D fUs W s 2 ! <! g is an A-system in M (which means
precisely that the properties (1),(2) and .20 / hold for U). It was proved in Fact 3 of
S.491 that any second countable space with an A-system is Cech-complete, so M is
Polish and hence our solution is complete.
T.366. Prove that if Cp .X / is analytic, then X is -compact.
Solution. The space P is identified with ! ! (see Problem 313). If Z is a set and
f; g 2 RZ , then f g if f .z/ g.z/ for all z 2 Z.
Assume, towards a contradiction, that Cp .X / is analytic while the space X is
not -compact. We have nw.X / D nw.Cp .X // D ! and hence there exists a
continuous onto map ' W X ! M such that w.M / D ! and M is not -compact
(see Problem 364). Define the dual map ' W RM ! RX by ' .f / D f ' for all
f 2 RM . Let X 0 and M 0 be the discrete spaces with the underlying sets X and M
respectively. Then Cp .X 0 / D RX and Cp .M 0 / D RM ; since the map ' W X 0 ! M 0
is open, the set ' .RM / is closed in RX (see Problem 163 of [TFS]). Observe that
' .Cp .M // P D ' .RM / \ Cp .X /. The set P is analytic being closed in the
analytic space Cp .X /. Since ' is an embedding, we proved that
. / there is an analytic space Q such that Cp .M / Q RM
because it suffices to take Q D .' /1 .P /. Let RC D .0; C1/ R and fix a
homeomorphism W R ! RC (we can take, e.g., .t/ D et for all t 2 R). The map
W RM ! RC M
defined by
.f / D f for all f 2 RM is a homeomorphism (see
Problem 091 of [TFS]) and
.Cp .M // D Cp .M; RC /. An immediate consequence
of . / is that
. / the space Q0 D
.Q/ is analytic and Cp .M; RC / Q0 RC
M
.
We can consider that M is a metric subspace of a metric compact Q space .C; d /.

Fix a continuous onto map h W P ! Q0 . For any s 2 P let K.s/ D n2! .s.n/ C 1/
(here we consider the integer .s.n/ C 1/ to be the set f0; : : : ; s.n/g). Then K.s/ is
compact for any s 2 P and it is immediate that K.s/ K.t/ if s t. The set
w.s/ D h.K.s// is a compact subspace of Q0 and it is clear that w.s/ w.t/ if
s t.
For any x 2 M let x W RC M
! RC be the natural projection onto the factor
determined by x. Given any s 2 P, the set x .w.s// is a compact subspace of RC
and therefore rsx D inf.x .w.s// 2 RC . We have rs f .x/ for any f 2 w.s/; let
bs .x/ D rsx for every x 2 M . This gives us a set fbs W s 2 Pg RC M
such that
(1) if s; t 2 P and s t, then bt bs ;
(2) for any f 2 Cp .M; RC /, there is s 2 P such that bs .x/ f .x/ for all x 2 M .
The property (1) follows from the inclusion w.s/ w.t/ and the definition of bs
and bt . If f D h.s/ for some s 2 P, then f 2 w.s/ and hence bs .x/ x .f / D
f .x/ for all x 2 M , i.e., (2) is proved. As usual, given a point x 2 C and " > 0 the
set B.x; "/SD fy 2 C W d.x; y/ < "g is the "-ball in C centered at x. For any s 2 P,
let Ws D fB.x; bs .x// W x 2 M g and K.s/ D C nWs . Then K.s/ is a compact
subset of C nM ; besides, s; t 2 P and s t implies K.s/ K.t/ because it follows
from (1) that Ws Wt .
Now, if K C nM is compact, let f .x/ D d.x; K/ D inffd.x; y/ W y 2 Kg for
every x 2 M . Then f 2 Cp .M; RC / (see Fact 1 of S.212), so there exists s 2 P
such that bs f [see (2)]. As a consequence, Ws \ K D ; and hence K K.s/.
We proved that the map K W P ! exp.C nM / satisfies the conditions (a)(c) of
Problem 365. Thus C nM is a Polish space which implies that C nM is a G -set
of E D clC .C nM / (see Problem 259 of [TFS]). Any closed subset of a metric
space is a G -set, so E is a G -set in C and therefore C nM is a G -set in C (see
Fact 2 of S.358). Consequently, M is an F -subset of C , i.e., M is -compact. This
contradiction shows that X is also -compact and finishes our solution.
T.367. Prove that the following are equivalent for an arbitrary second countable
space X :
(i) Cp .X / is analytic;
(ii) Cp .X / is a K -space;
Observe that, as a consequence, the spaces Cp .P/ and Cp .R! / are not analytic.
Solution. The implication (i)H)(iii) was proved in Problem 366 while (iii)H)(ii)
was established in Problem 362. Since every second countable K -space is analytic
by Problem 346, we also have the implication (ii)H)(i).
Finally observe that P is not -compact by Fact 1 of T.132 and the space R! is
not -compact because P embeds in R! as a closed subspace by Problem 273 of
[TFS]. Therefore neither of the spaces Cp .R! / and Cp .P/ is analytic.
T.368. For a second countable space X let A be a countable dense subspace of X .

Prove that the following conditions are equivalent:
(i) Cp .AjX / is a Borel set;
(ii) Cp .AjX / is analytic;
Solution. Given spaces Y and Z, a map f W Y ! Z V is called measurable if
f 1 .U / 2 B.Y / for any U 2 B.Z/. If A exp.Z/, then A is the family of all
finite intersections of elements of A. A space Z is perfect if every U 2 .Z/ is an
F -subset of Z. The spaces with a countable network are called cosmic.
Fact 1. If Y is a perfect space and hl.Z/ D !, then a map f W Y ! Z is
measurable if and only if there is a subbase S in the space Z such that f 1 .U / 2
B.Y / for any U 2 S.
Proof. If f is measurable, then we can let S D .Z/ so necessity is clear. Now,
1
suppose that there is a subbase V S in the space Z such that f .U / 2 B.Y / for any
U 2 S. The family B D T S is a base of Z and U T2 B implies that there are
U1 ; : : : ; Un 2 S with U D i n Ui . Thus f 1 .U / D i n f 1 .Ui / 2 B.X / (see
Fact 1 of T.331) which proves that f 1 .U / 2 B.Y / for any U 2 B.
Now if U 2 .Z/, then it follows S from l.U / hl.Z/ DS! that there is a
countable B 0 B such that U D B 0 . Therefore f 1 .U / D ff 1 .V / W V 2
B 0 g, i.e., f 1 .U / is a Borel subset of Y being a countable union of Borel subsets of
Y (here we applied again Fact 1 of T.331).
We proved that f 1 .U / 2 B.Y / for any U 2 .Z/, so we can apply Fact 1 of
T.363 to conclude that f is a measurable map. This settles sufficiency and finishes
the proof of Fact 1. t
u
Fact 2. Given a cosmic space Z, suppose that B is a dense subset of Z and, for
any point z 2 Z, there is a sequence fbn W n 2 !g B such that bn ! z. If
W Cp .Z/ ! Cp .BjZ/ is the restriction map, then e D 1 W Cp .BjZ/ ! Cp .Z/
is measurable. In particular, if Z is second countable and B is dense in Z, then the
map e is measurable.
Proof. For any O 2 .R/ and z 2 Z let z; O D ff 2 Cp .Z/ W f .z/ 2 Og; then
the family S D fz; O W z 2 Z and O 2 .R/g is a subbase in Cp .Z/. Take any
U D z; O 2 S.
It is easySto find a family fOn W n 2 !g .R/ such that O n OnC1 for every
n 2 ! and n2! On D O. S T fbn W n 2 !g B such that bn ! z. We
Fix a sequence
S
claim that .U / D HU D n2! k2! i k ff 2 Cp .BjZ/ W f .bi / 2 On g.
Indeed, if f D .g/ for some g 2 U , then g.z/ 2 O and hence g.z/ 2 On for
some n 2 !. Furthermore, f .bi / D g.bi / for all i 2 !; by continuity of g, we have
g.bi / ! g.z/ and therefore f .bi / ! g.z/ which shows that there is k 2 ! such
that f .bi / 2 On for all i k. Thus f 2 HU and we established that .U / HU .
Conversely, if f 2 HU and f D .g/ for some g 2 Cp .Z/, then there are
n; k 2 ! such that for all i k we have g.bi / D f .bi / 2 On . Since g is a
continuous function, g.z/ D lim g.bi / when i ! 1 and therefore g.z/ 2 O n

OnC1 O which shows that g.z/ 2 O, i.e., g 2 U . Thus f 2 .U / and we proved
that HU .U /, i.e., .U / D HU .
Finally observe that the set Sni D ff 2 Cp .BjZ/ S SW f .bTi / 2 On g is open
in Cp .BjZ/ for all n; i 2 !, so the set HU D n2! k2! i k Sni is G in
Cp .BjZ/. Consequently, e 1 .U / D .U / D HU is a Borel subset of Cp .BjZ/ for
any U 2 S; since both spaces Cp .Z/ and Cp .BjZ/ are cosmic, we can apply Fact 1
to conclude that the map e is measurable. Fact 2 is proved. t
u
Fact 3. Given any space Z, suppose that .M; d / is a complete metric space and
f W A ! M is a continuous map for some dense A Z. For a point z 2 Z say that
osc.f; z/ D 0 if for any " > 0, there is W 2 .z; Z/ such that diam.f .W \A// < ".
Then G D fz 2 Z W osc.f; z/ D 0g is a G -subset of Z such that A G and there
is a continuous map g W G ! M with gjA D f .
Proof. As usual, for any x 2 M and " > 0, the set B.x; "/ D fy 2 M W d.x; y/ <
"g is the "-ball centered at x. Let Un D fz 2 Z W there exists W 2 .z; Z/ such
that diam.f .W \ A// n1 g; it is evident that the set Un is open in Z for any
n 2 N. The map f being continuous, for any z 2 A, there is V 2 .z; A/ such
that f .V / B.f .z/; 2n 1
/. If W is an open subset of Z such that W \ A D V ,
then diam.f .W \ A// D diam.f .V // diam.B.f .z/; 2n 1
// n1 which shows that
z 2 Un . Consequently, A UTn for any n 2 N.
It is immediate that G D fUn W n 2 Ng, so G is a G -subset of Z and A G.
To prove that the map f can be continuously extended over the set G take any
z 2 G and consider the family Fz D ff .W \ A/ W W 2 .z; Z/g. It is clear
that Fz consists of closed subsets of the space M ; if W1 ; : : : ; Wn 2 .z; Z/ and
W D W1 \

\ Wn , then f .W \ A/ f .W1 \ A/ \ : : : \ f .Wn \ A/ which
shows that the family Fz is centered.
Given any " > 0, take n 2 N with n1 < ". Since z 2 Un , there is W 2 .z; Z/
such that diam.f .W \ A//T n1 < ". Therefore Fz contains elements of arbitrarily
smallTdiameter and hence Fz ; by Problem 236 of [TFS]. It is easy to see
that
T Fz cannot contain more than one point, so there is a point g.z/ such that
Fz D fg.z/g. This defines a map g W G ! M .
Observe first that gjA D f because if z 2 A, T then f .z/ 2 f .W \ A/
f .W \ A/ for any W 2 .z; Z/ and hence f .z/ 2 Fz , i.e., g.z/ D f .z/.
To see that g is continuous, take any z 2 G and " > 0. Choose n 2 N and
W 2 .z; Z/ such that n1 < " and diam.f .W \ A// D diam.f .W \ A// n1 . If
W 0 D W \ G, then W 0 2 .z; G/; given any t 2 W 0 we have t 2 W 2 .t; Z/ and
hence f .W \ A/ 2 Ft \Fz . Consequently, fg.z/; g.t/g f .W \ A/ and therefore
d.g.z/; g.t// diam.f .W \ A// n1 < " which proves that g.W 0 / B.g.z/; "/,
i.e., g is continuous at the point z. Since z 2 G was taken arbitrarily, the function g
is continuous on G and Fact 3 is proved. t
u
Returning to our solution, observe that the implication (i)H)(ii) is obvious.
Now, assume that Cp .AjX / is analytic and let W Cp .X / ! Cp .AjX / be the
restriction map. It follows from Fact 2 that the map e D 1 is measurable, so
Cp .X / is an image of an analytic space Cp .AjX / under a measurable map. Thus

Cp .X / is analytic by Fact 3 of T.363, and hence X is -compact by Problem 366,
so we settled (ii)H)(iii). S
Now, if (iii) holds, then X D fXn W n 2 Ng where Xn is compact for all n 2 N.
Let d be a metric on X such that .d / D .X/ and d.x; y/ 1 for all x; y 2 X .
For any k; l; n 2 N, let Mkln D ff 2 RA W jf .x/ f .y/j k1 whenever x and y
are points of A such that d.x; z/ < n1 and d.y; z/ < n1 for some z 2 Xl g. Our first
observation is that
(1) the set Mkln is closed in RA for any k; l; n 2 N.
Indeed, if f 2 RA nMkln , then there exist points x; y 2 A such that d.x; z/ < n1 and
d.y; z/ < n1 for some z 2 Xl while jf .x/ f .y/j > k1 ; then r D jf .x/ f .y/j
k > 0. The set Of D fg 2 R W jg.x/ f .x/j < 2 and jg.y/ f .y/j < 2 g is open
1 A r r
in the space R and f 2 Of . For any function g 2 Of we have jf .x/ f .y/j

A
jf .x/ g.x/j C jg.x/ g.y/j C jg.y/ f .y/j and therefore
1 r r 1
jg.x/ g.y/j jf .x/ f .y/j jf .x/ g.x/j jf .y/ g.y/j > r C D
k 2 2 k
which shows that g 2 RA nMkln . Since a function g 2 Of was chosen arbitrarily,

we have Of RA nMkln . Thus, for every f 2 RA nMkln S, there Of 2 .R
A
/ such
that Of R nMkln . As a consequence, R nMkln D fOf W f 2 R nMkln g is
A A A
open in RA and hence Mkln is closed in R S.

A
The property (1) impliesTthat Pkl D fMkln W n 2 Ng is an F -subset of RA .

We claim that Cp .AjX / D fPkl W k; l 2 Ng and hence Cp .AjX / is an F -subset
of RA . T
Take any f 2 fPkl W k; l 2 Ng; given a point z 2 X and " > 0, there is l 2 N
with z 2 Xl . Pick any k 2 N such that k1 < ". Since f 2 Pkl , there is n 2 N such
that jf .x/ f .y/j k1 < " whenever d.x; t/ < n1 and d.y; t/ < n1 for some t 2 Xl .
As a consequence, for the set W D fx 2 X W d.z; x/ < n1 g we have d.x; z/ < n1 and
d.y; z/ < n1 for all x; y 2 W and hence jf .x/ f .y/j k1 whenever x; y 2 W \A.
This shows that diam.f .W \ A// k1 < " and therefore osc.f; z/ D 0. The point
z 2 X was chosen arbitrarily so osc.f; z/ D 0 for all z 2 X and hence we can
apply Fact 3 to conclude that there is a continuous T function g W X ! R such that
gjA D f . Therefore f 2 Cp .AjX / and hence fPkl W k; l 2 Ng Cp .AjX /.
Now if f 2 Cp .AjX /, fix g 2 Cp .X / such that gjA D f and take any k; l 2 N.
If f Pkl , then, for any n 2 N, there exist xn ; yn 2 A and zn 2 Xl such that
jf .xn / f .yn /j > k1 while d.xn ; zn / < n1 and d.yn ; zn / < n1 . Since Xl is compact,
the sequence fzn W n 2 Ng has a subsequence which converges to a point z 2 Xl .
Passing to that subsequence and reindexing the sequences fxn W n 2 Ng and fyn W
n 2 Ng if necessary, we can assume, without loss of generality, that zn ! z. We
also have xn ! z and yn ! z because d.xn ; zn / < n1 and d.yn ; zn / < n1 for
all n 2 N. Since g is continuous at the point z, there is W 2 .z; X / such that
jg.w/ g.z/j < 2k 1
for all w 2 W . There is m 2 N such that xm ; ym 2 W and
hence jf .xm / f .ym /j D jg.xm / g.ym /j jg.xm / g.z/j C jg.z/ g.ym /j <
2k CT 2k D k which is a contradiction. ThusT f 2 Pkl for all k; l 2 N and therefore
1 1 1
f 2 fPkl W k; l 2 Ng. This proves that fPkl W k; l 2 Ng D Cp .AjX / and hence

Cp .AjX / is an F -subset of RA . Consequently, Cp .AjX / is a Borel set (see Fact 1
of T.331), which finishes the proof of (iii)H)(i) and makes our solution complete.
T.369. Given a countable space X prove that Cp .X / is analytic if and only if
Cp .X; I/ is analytic.
Solution. The subspace Cp .X; I/ is closed in Cp .X /, so if Cp .X / is analytic, then
Cp .X; I/ is also analytic by Problem 335. This proves necessity and shows that
countability of X is not needed for this part of the proof.
Now assume that the space Cp .X; I/ is analytic. For any point x 2 X the set
Px D ff 2 Cp .X; I/ W T f .x/ 2 . 1; 1/g is open in Cp .X; I/ and it is immediate
that Cp .X; . 1; 1// D fPx W x 2 X g. Thus Cp .X; . 1; 1// is a G -subset of
Cp .X; I/. Any open subspace of an analytic space is analytic by Problem 335, so
Cp .X; . 1; 1// is analytic being a G -subset of an analytic space Cp .X; I/ (see
336). Finally, Cp .X / is analytic because it is homeomorphic to an analytic space
Cp .X; . 1; 1//.
T.370. Prove that a countable space X embeds into Cp .P/ if and only if Cp .X / is
analytic.
Solution. If the space Cp .X / is analytic, then there is a continuous onto map ' W
P ! Cp .X /. The dual map ' W Cp .Cp .X // ! Cp .P/ embeds Cp .Cp .X // in
Cp .P/ (see Problem 163 of [TFS]). Since X embeds in Cp .Cp .X // by Problem 167
of [TFS], it also embeds in Cp .P/ so sufficiency is proved.
Now assume that X is a countable subspace of Cp .P/. For any x 2 X and
m; n 2 N, let Q.x; m; n/ D f' 2 IX W there is a point s D .s1 ; : : : ; sn / 2 Pn for
which j'.y/ '.x/j m1 whenever jy.si / x.si /j < n1 for all i ng. We will
also need the set R.x; m; n/ D f.s; '/ 2 Pn IX W s D .s1 ; : : : ; sn / 2 Pn and
j'.y/ '.x/j m1 whenever jy.si / x.si /j < n1 for all i ng. It is immediate
that
(1) if m; n 2 N and the mapping p W Pn IX ! IX is the natural projection, then
p.R.x; m; n// D Q.x; m; n/ for any x 2 X .
Our next step is to prove that
(2) the set R.x; m; n/ is closed in Pn IX for all x 2 X and m; n 2 N.
Indeed, if w D .s; '/ 2 .Pn IX /nR.x; m; n/, then there exists y 2 X such
that jy.si / x.si /j < n1 for all i n while j'.y/ '.x/j > m1 . Since y and x are
continuous functions on P, the set Os D ft D .t1 ; : : : ; tn / 2 Pn W jy.ti / x.ti /j < n1
for all i ng is open in Pn ; it is clear that s 2 Os .
Furthermore, the set O' D f 2 IX W j.y/ .x/j > m1 g is open in IX (the proof
is an easy exercise left to the reader). Therefore the set O D Os O' is an open
neighborhood of w in Pn IX such that O \ R.x; m; n/ D ;. Thus every point of
A D .Pn IX /nR.x; m; n/ has a neighborhood which is contained in A. Therefore

A is open in Pn IX which shows that R.x; m; n/ is closed in Pn IX , i.e., (2) is
proved.
For any x 2 X let CSx D f' 2 I
X
W ' is continuous at the point xg; for any
m 2 N, let Q.x; m/ D fQ.x; m; n/ W n 2 Ng. The following property is a key
fact for our proof.
T
(3) Cx D fQ.x; m/ W m 2 Ng for any x 2 X .
To prove (3) assume that ' 2 Cx and take any m 2 N. By the definition of the
topology of pointwise convergence on X and continuity of the function ' at x, there
exist k 2 N; " > 0 and .s1 ; : : : ; sk / 2 Pk for which j'.y/ '.x/j < m1 for any y 2 X
such that jy.si / x.si /j < " for all i k. Pick n 2 N such that k n and n1 < "
and let skC1 D : : : D sn D sk ; it is clear that, for the point s D .s1 ; : : : ; sn / 2 Pn ,
we have j'.y/ '.x/j < m1 for any y 2 X such that jy.si / x.si /j < n1 for all
i n. This
T proves that ' 2 Q.x; m; n/; since T m 2 N was chosen arbitrarily, we
have ' 2 fQ.x; T m/ W m 2 Ng, i.e., C x fQ.x; m/ W m 2 Ng.
Now, if ' 2 fQ.x; m/ W m 2 Ng, then take any " > 0 and choose m 2 N such
that m1 < ". Since ' 2 Q.x; m/, there exists n 2 N for which ' 2 Q.x; m; n/ and
hence there is s D .s1 ; : : : ; sn / 2 Pn such that j'.y/ '.x/j m1 < " for any y 2 X
with jy.si / x.si /j < n1 for all i n. The set W D fy 2 X W jy.si / x.si /j < n1 for
all i ng is an open neighborhood of x in X such T that '.W / .'.x/ "; '.x/C"/
and hence ' is continuous at the point x. Thus fQ.x; m/ W m 2 Ng Cx and (3)
is proved.
Observe that the space Pn IX is analytic and hence (2) implies that so is
R.x; m; n/ for any x 2 X and m; n 2 N (see Problem 335). The property (1)
shows that for any x 2 X and m; n 2 N, the space Q.x; m; n/ is analytic being
a continuous image of R.x; m; n/. This implies that Q.x; m/ is analytic for any
x 2 X and m 2 N (see Problem 337). Consequently, for any x 2 X , the space
Cx is analyticTby (3) and Problem 336. Furthermore, the space X is countable, so
Cp .X; I/ D fCx W x 2 X g is also analytic by Problem 336; apply Problem 369
to conclude that Cp .X / is analytic as well. This settles necessity and completes our
solution.
T.371. Take any 2 !n! and consider the space X D ! [ fg with the topology
inherited from !. Prove that neither Cp .X / nor RX nCp .X / is analytic. As a
consequence, X cannot be embedded into Cp .P/.
Solution. Given a space Z and A; B Z let A B if AB D .AnB/ [ .BnA/
is a set of first category in Z. Say that A Z is a B-set if there is an open U Z
such that A U .
Fact 1. If Z is a space and A 2 B.Z/, then A is a B-set.
Proof. If A is open, then U D A witnesses that A is a B-set. If A is closed, then,
for U D Int.A/, the set AU D AnU is nowhere dense and hence A U .
Now assume that < !1 and we proved for any < and A 2 0 .Z/[0 .Z/
that A is a B-set. If A 2 0 .Z/, then thereSis a sequence fn W n 2 !g and
An 2 0n .Z/ for all n 2 ! such that A D n2! An . By the induction hypothesis,
S
there is Un 2 .Z/
S such that An Un for all n 2 !. For the set U D n2! Un , we
have AU n2! .An Un / which shows that AU is of first category and hence
A U , i.e., A is a B-set.
Now if A 2 0 .Z/, then A0 D ZnA 2 0 .Z/ and hence there is U 0 2 .Z/
such that A0 U 0 . Observe that for the set F D ZnU 0 we have AF D A0 U 0
and therefore A F . If U D Int.F /, then the set AU .AF / [ .FU /
is of first category because so is AF and FU D F nU is nowhere dense. Thus
A U and hence A is a B-set. Therefore our inductive procedure can be carried out
for all < !1 which shows that every A 2 B.Z/ is a B-set. Fact 1 is proved. t
u
Fact 2. Let D be an infinite discrete space. Then
(a) for any A D, the set A is clopen in D (the bar denotes the closure in the
space D);
(b) if z 2 DnD and Dz D fU \ D W U 2 .z; Z/g, then a set A D belongs to
Dz if and only if z 2 A;
(c) the family Dz is an ultrafilter on D for any z 2 DnD.
Proof. For an arbitrary set A D, we have A \ DnA D ; by Fact 1 of S.382. It
is evident that A [ DnA D D, so both sets U D A and V D DnA are clopen in
D; this proves (a).
It is clear that z 2 A for any A 2 Dz . Now if A D and z 2 A, then U D A
is a clopen subset of D by (a); since z 2 U and U \ D D A, the item (b) is also
settled.
To prove (c) observe that Dz is a centered family and take any B D. The sets
U D B and V D DnB are clopen in D by (a) and only one of them contains z. If
z 2 U , then U \ D D B 2 Dz and z 2 V implies that V \ D D .DnB/ 2 Dz . Thus
B 2 Dz or DnB 2 Dz for any B D and hence Dz is an ultrafilter by Problem 117
of [TFS]. Fact 2 is proved. t
u
Fact 3. Let Z be a homogeneous space, i.e., for every x; y 2 Z, there is a
homeomorphism h W Z ! Z such that h.x/ D y. If Z is of second category in
itself, then it is a Baire space.
Proof. Assume that some U 2 .Z/ is of first category in Z. Consider a family
U D fV 2 .Z/ W V there is a homeomorphism h W Z ! Z such that h.V / U g.
It is clear that every V 2 U is of first Scategory in Z. Let V be a maximal disjoint
subfamily of U. We claim that W D V is dense in Z.
Indeed, if this is not the case, then there is a point z 2 ZnW . Take a point x 2 U
and fix a homeomorphism h W Z ! Z such that h.z/ D x. By continuity of h, there
is V 2 .z; Z/ such that V ZnW and h.V / U . Then the family V 0 D V [ fV g
is still disjoint and V 0 U which contradicts maximality of V.
This proves that W is dense in Z and hence F D ZnW is nowhere dense in Z.
Since every V 2 V is of first category in Z, there is a family UV D fPVn W n 2 !g of
S S
nowhere dense subsets of Z such that V D UV . The set Pn D fPVn W V 2 Vg
is nowhere dense Sin Z for every n 2 ! (we leave
S an easy verification of this fact to
the reader) and n2! Pn D W whence Z D . n2! Pn / [ F is of first category in
itself which is a contradiction.
This contradiction shows that every U 2 .Z/ is of second category in Z and
hence Z is a Baire space. Fact 3 is proved. t
u
Returning to our solution, let W DX ! D! be the restriction map and consider
the set D D ff 2 DX W f ./ D 0g. It is evident that 1 D jD W D ! D! is a
homeomorphism. The family D D fU \ ! W U 2 .; X/g is an ultrafilter on ! by
Fact 2.
Now assume that Cp .X / is analytic; since Cp .X; D/ is closed in Cp .X /, the
space Cp .X; D/ is analytic too. Therefore F D ff 2 Cp .X; D/ W f ./ D 0g
is also analytic. For any f 2 F the set f 1 .0/ is open in X and contains , so
f 1 .0/ \ ! 2 D for any f 2 F.
On the other hand, if we have f 2 D and W D f 1 .0/ \ ! 2 D , then
f 1 .0/ D W [ fg 2 .; X/ because 2 V D cl! .W / by Fact 2 while the set
V is open in ! (here we used Fact 2 again) and V \ X D W [ fg. Therefore
f 2 F if and only if f 2 D and f 1 .0/ \ ! 2 D . This shows that F1 D ff 2
D! W f 1 .0/ 2 D g is an analytic space because F1 D 1 .F /.
Given any f 2 D! , let f 0 .n/ D 1 f .n/ for all n 2 !. If e.f / D f 0 for each
f 2 D! , then e W D! ! D! is a homeomorphism; it is easy to see that F2 D e.F1 /
is disjoint from F1 and F1 [ F2 D D! .
Observe that G D .RX nCp .X //\D coincides with the set ff 2 DX W f ./ D 0
and f is discontinuous at g, so G D ff 2 D W f 1 .0/ \ ! D g (here we used
Fact 2 which guarantees that D is an ultrafilter). It is evident that G is a closed
subset of RX nCp .X / and .G/ D F2 . As a consequence, we proved that
. / if Cp .X / is analytic, then F1 is analytic; if RX nCp .X / is analytic, then F2
is analytic. Since F1 and F2 are homeomorphic, it follows from analyticity of
either of spaces Cp .X / or RX nCp .X / that F1 is analytic.
It is easy to see that F1 and F2 are both dense in D! , so if i 2 f1; 2g, then A Fi
S dense in Fi if and only if it is nowhere dense in D . Consequently, if
!
is nowhere
F1 D n2! Pn where Pn is nowhere dense in F1 for each n 2 !, then Qn D e.Pn /
S in F2 for each n 2 ! and therefore the space D D F1 [ F2 D
!
isSnowhere dense
. n2! Pn / [ . n2! Qn / is of first category in itself which is false by Problem 274
of [TFS]. This contradiction shows that
(1) both spaces F1 and F2 are of second category in themselves and in D! .
We next prove that F1 is homogeneous. Given any f; g 2 F1 define h D f g
by the equalities h.n/ D 1 if f .n/ g.n/ and h.n/ D 0 if f .n/ D g.n/. For any
f 2 D! , let Lf .g/ D f g for every g 2 F1 . It is easy to see that f g 2 F1 for
any f; g 2 F1 . Furthermore, the map Lf W F1 ! F1 is a homeomorphism for any
f 2 F1 ; given any f; g 2 F1 , if h D f g, then Lh .f / D g which shows that F1
is a homogeneous space. Now, Fact 3 and (1) imply that
(2) both spaces F1 and F2 have the Baire property.

Observe that if F1 is analytic, then F2 D D! nF1 is also analytic, so we can apply
Problem 340 to conclude that F1 is a Borel subset of D! . Therefore F1 is a B-set by
Fact 1; pick O 2 .D! / such that OF1 is of first category in D! . If O D ;, then
F1 D F1 O is of first category in D! which is impossible by (1). Thus O ; and
hence O \ F2 D OnF1 OF1 is a non-empty open subset of F2 which is of first
category in D! and hence in itself. However, F2 has the Baire property by (2), so all
non-empty open subsets of F2 are of second category; this contradiction shows that
F1 is not analytic. Finally, apply . / to conclude that neither of the spaces Cp .X /
and RX nCp .X / is analytic. Since X is countable, it cannot be embedded in Cp .P/
by Problem 370 and hence our solution is complete.
T.372. Prove that if < !1 , then there exists a countable
S space X with a unique
non-isolated point such that Cp .X / 2 B.RX /n. < 0 .RX //, i.e., the space
Cp .X / can have an arbitrarily high Borel complexity for a countable space X with
a unique non-isolated point.
Solution. If Y and Z are spaces, the expression Y ' Z says that Y is homeomor-
phic to Z; a map f W Y ! Z is called measurable if f 1 .A/ 2VB.Y W / for any
A 2 B.Z/. Given a set Z and a family A exp.Z/, we denote by A ( A) the
family of all finite intersections (unions) of the elements of A.
For any space Z, let F .Z/ be the family of Sall closed subsets of Z; if n 2 !, then
LnZ D fY Z W jY j ng and Fin.Z/ D fLnZ W n 2 !g. For each U 2 .Z/,
let I.U / D fF 2 F .Z/ W F U g and M.U / D fF 2 F .Z/ W F \ U ;g.
Denote by V.Z/ the Vietoris topology on the set F .Z/, i.e., the topology generated
by the family S.Z/ S D fI.U / W U 2 .Z/g [ fM.U / W U 2 .Z/g as a subbase
(it is immediate that S.Z/ D F .Z/, so the family S.Z/ can, indeed, generate a
topology on F .Z/ as a subbase).
Fact 1. If Z is a compact Hausdorff space, then TZ D .F .Z/; V.Z// is also a
compact Hausdorff (and hence Tychonoff) space.
Proof. It follows from Problem 118 of [TFS] that to prove compactness of the space
.F .Z/; V.Z//, it suffices to show that any cover of F .Z/ with the elements of S.Z/
has a finite subcover. LetSSI D fI.U / W U 2 .Z/g and SM D fM.U / W U 2
.Z/g; if U SS.Z/ and U D F .Z/, then let UI D U \ SI and UM D U \ SM .
The set O D fU W M.U / 2 Ug is open in Z and S hence H D S ZnO 2 F .Z/.
It is immediate from the definition of O that H UM , so H 2 UI , i.e., there
is V 2S.Z/ such that I.V / 2 U and H V . The set G D ZnV is compact and
G fU W M.U / 2 Ug which shows that there are U1 ; : : : ; Un 2 .Z/ such that
M.Ui / 2 U for all i n and G U1 [

[ Un .
We claim that U 0 D fI.V /; M.U1 /; : : : ; M.Un /g is a (finite)
S subcover of U.
Indeed, take any F 2 F .Z/; if F V , then F 2 I.V / U 0 . If not, then
F \ G ; and Shence there is i n for which F \ Ui ;, i.e., F 2 M.Ui / and
therefore F 2 U 0 . Thus U 0 is a finite subcover of U, so we proved that the space
TZ is compact.
To see that TZ is Hausdorff take distinct F; G 2 F .Z/. We can assume, without

loss of generality, that F nG ;; pick any x 2 F nG. The space Z is regular (see
Problem 124 of [TFS]), so there are disjoint U; V 2 .Z/ such that x 2 U and
G V . It is straightforward that F 2 M.U /; G 2 I.V / and M.U / \ I.V / D ;
which shows that TZ is Hausdorff (and hence Tychonoff by Problem 124 of [TFS]).
Fact 1 is proved. t
u
Fact 2. If Z is a compact space, w.Z/ D ! and TZ D .F .Z/; V.Z//, then
w.TZ / .
W
Proof. Take a base B in Z such that jBj D and let C D B; then jCj D . We
claim that U D fI.U / W U 2 Cg [ fM.U / W U 2 Bg is a subbase in TZ . Indeed,
take any F 2 F .Z/ and O 2 .F; TZ /. Since S.Z/ is a subbase of TZ , there exist
U1 ; : : : ; Un ; V1 ; : : : ; Vk 2 .Z/ such that
F 2 O 0 D I.U1 / \

\ I.Un / \ M.V1 / \

\ M.Vk / O:
This implies that F U D U1 \

\ Un and we can choose xi 2 F \ Vi for
each i k. There are W1 : : : ; Wk 2 B such that xi 2 Wi Vi for all i k; besides,
the family fW 2 B W W U g is, evidently, an open cover of the Scompact space F .
Consequently, there are H1 ; : : : ; Hm 2 B such that F H D i m Hi U . It is
clearVthat H 2 C and F 2 I.H / \ M.W1 / \

\ M.Wk / O 0 V O which proves
that U is a base in TZ , i.e., U is a subbase of TZ . Since jUj D j Uj jCj D ,
we have w.TZ / so Fact 2 is proved. t
u
Fact 3. For any space Z the set LnZ D fF 2 Fin.Z/ W jF j ng is closed in the
space TZ D .F .Z/; V.Z//. As a consequence, if Z is compact and we consider
Fin.Z/ as a subspace of TZ , then Fin.Z/ is -compact.
Proof. If F 2 F .Z/nLnZ , then there exists distinct points z1 ; : : : ; znC1 2 F . Pick
Ui 2 .zi ; Z/ for all i n C 1Tso that the family fUi W i n C 1g is disjoint.
It is immediate that F W D i nC1 M.Ui / while W \ LnZ D ;. This proves
that LnZ is closed in TZ . If Z is compact, then TZSis also compact by Fact 1 so LnZ
is compact for all n 2 ! and hence Fin.Z/ D n2! LnZ is -compact. Fact 3 is
proved. t
u
Fact 4. Let T be an infinite set and z T . For an arbitrary filter C on T with
T
C D ;, consider the family .C/ D exp.T / [ ffzg [ A W A 2 Cg. Then .C/ is
a topology on T C D T [ fzg, the space ET; F ; z D .T C ; .C// is normal and T1
and z is the unique non-isolated point of ET; F ; z .
Proof. Since ; 2 exp.T /, we have ; 2 .C/ and it follows from T 2 C that
T C 2 .C/. If A; B 2 .C/ and z A \ B, then A \ B 2 .C/; if z 2 A \ B,
then A \ T 2 C and B \ T 2 C and hence C D A \ B \ T 2 C which shows that
A \ B D C [ fzg 2 .C/. S S
Finally,
S if U .C/ and U T , then U 2 .C/; if, on the other S hand,
z 2 U, then there is A 2 C such that A [ fzg 2 U and hence A A0 D . U/ \ T
S
which implies A0 2 C proving that U D A0 [ fzg 2 .C/. Thus .C/ is indeed a
topology on T C .
Since ftg 2 .C/ for each t 2 T , all point of T are isolated in ET; F ; z ; every
neighborhood of z is an element of F , so the point z is not isolated in ET; F ; z .
It is straightforward to see that all points of ET; F ; z are closed in ET; F ; z .
Furthermore, every open subset which contains z is also closed in ET; F ; z and
fftgg is a clopen base at any t 2 T . This proves that ET; F ; z is zero-dimensional
and hence Tychonoff by Fact 1 of S.232. Consequently, ET; F ; z is normal by
Claim 2 of S.018, so Fact 4 is proved. t
u
The notions of the following construction will be used all through the rest of this
solution.
Fact 5S(Basic construction and its properties). For each n 2 ! let n D Dn and
D n2! n . We fix a point w and let D [ fwg. For any s 2 D! , let
A.s/ D fsjn W n 2 !g and define .s/ 2 D by .s/.p/ D 1 for all p 2 A.s/
and .s/.p/ D 0 if p 2 nA.s/. In other words, .s/ is the characteristic function
of the set A.s/ in . The correspondence s ! .s/ defines a map W D! ! D .
Then
(1) the map is continuous and injective and hence W D! ! .D! / is a
homeomorphism;
(2) for any distinct s; t 2 D! the set A.s/ \ A.t/ isS
finite;
(3) for any S D! the family FS D fnA W A fA.s/ W s 2 S 0 g [ P for some
S 0 2 Fin.S / and P 2 Fin./g is a filter on ;
(4) if X S D E; FS ; w (see Fact 4), then S embeds in Cp .X S / as a closed
subspace.
Proof. (1) Fix any s0 2 D! , let t0 D .s0 / and take a set U 2 .t0 ; D /. There
exists a finite B such that U 0 D ft 2 D S W tjB D t0 jBg U . As a
consequence, there is n 2 ! for which B \ i n i . It is clear that the
set V D fs 2 D! W sjn D s0 jng is an open neighborhood of s0 in D! ; it is
S also sji D s0 ji for all
evident that S i n, so, by definition of the map , we have
A.s/ \ . i n i / D A.s0 / \ . i n i / which implies A.s0 / \ B D A.s/ \ B
and therefore .s/jB D .s0 /jB for all s 2 V . This proves that .V / U 0 U
and hence is continuous at the point s0 .
To see that is injective take distinct s; t 2 D! and fix n 2 ! such that a D
sjn tjn D b. Then a; b 2 n and .s/.a/ D 1 0 D .t/.a/ (because from
all points of n , the function .t/ only equals 1 at the point b) which shows that
.t/ .s/, so is injective and (1) is settled.
To see S take n 2 ! such that sjn tjn and observe that A.s/ \
Sthat (2) is true
A.t/ i n i and i n i is a finite set.
As to (3), letting S 0 D ; and P D ;, we can see that 2 FS , i.e., FS ;.
Now, observe that jA.s/S \ i j D 1 for any s 2 D! and i 2 !; thus, if S 0 2 Fin.S /
and jS j D n, then i n.S fA.s/ W s 2 S 0 g/ has at least 2i n elements for all i > n
0
which proves that n. fA.s/ W s 2 S 0 g/ is an infinite set and hence F ; for

any F 2 FS . It is straightforward that F 2 FS and F F 0 implies F 0 2 F and

F; G 2 FS implies F \ G 2 FS , so FS is, indeed, a filter and (3) is proved.
Given any s 2 S , the set U D nA.s/ is a clopen neighborhood of w; since
.s/.U / D f0g, the function .s/ is continuous at the point w and therefore .s/ 2
Cp .X S / for any s 2 S . On the other hand, if f 2 Cp .X S / \ .D! /, then
f 2 D ; f .w/ D 0 and f is continuous at the point w. Consequently, there exists
U 2 .w; X S / such that f .U / . 12 ; 12 /, i.e., f .U / D f0g.
By the definition of the topology
S of X S , there is a finite P and a set
S 0 2 Fin.S / such that nU fA.s/ S W s 2 S 0 g [ P . We have f D .t/ for some
t 2 D and hence f .1/ D A.t/ fA.s/ W s 2 S 0 g [ P which shows that
! 1
A.t/ \ A.s/ is infinite for some s 2 S 0 and therefore t D s [see (2)], i.e., t 2 S .
Consequently, f 2 .S / and we established that Cp .X S / \ .D! / D .S /; since
.D! / is compact by (1), the set .S / is a closed subspace of Cp .X S / which is a
homeomorphic copy of S by (1). This settles (4) so Fact 5 is proved. t
u
Fact 6. If S is a Borel subspace of D! and X S D E; FS ; w (see Fact 4 and
Fact 5), then Cp .X S / is a Borel subset of the space R .
Proof. By Fact 3 of T.333 being a Borel set as a space is equivalent to being a Borel
subset of every larger Polish space, so we will use both terms interchangeably when
proving that, in some spaces, certain subsets are Borel.
If A , then A 2 D is the characteristic function of A, i.e., A .p/ D
1 for all p 2 A and A .p/ D 0 if p 2 nA. Consider the set C 0 D ff 2
Cp .X S / W f .w/ D 0g; it follows from Fact 1 of S.409 that Cp .X S / ' C 0 R,
so it suffices to show that C 0 is a Borel set (see Problem 333 and Fact 3 of T.333).
Let W R ! R be the restriction map; if D D ff 2 R W f .w/ D 0g, then
1 D jD W D ! R is a homeomorphism; it is evident that 1 .D/ D R , so it
suffices to show that C D 1 .C 0 / is a Borel subset of R .
1
For any f 2 R and m 2 N, let H.f; S m/ D f 0 .Rn. m ; m //; we0 will also
1 1
need the family HS D fA W A fA.s/ W s 2 S g [ P for some S 2 Fin.S /

and P 2 Fin./g which, evidently, consists of the complements of the elements of
FS ; let H S D fA W A 2 HS g.
For every m 2 N define a map m W R ! D by letting m .f / to be the
characteristic function of the set H.f; m/, i.e., m .f /.p/ D 1 if p 2 H.f; m/ and
m .f /.p/ D 0 for all p 2 nH.f; m/. It is evident that m .f / D f for any
f 2 D , so m .R / D D for each m 2 N.
Observe that f 2 D belongs to C 0 if and only if f 1 .. m1 ; m1 //nfwg 2 FS for all
m 2 N, or, equivalently, if and only if f 1 .Rn. m1 ; m1 //nfwg 2 HS for all m 2 N.
Therefore a function f 2 R belongs to the set C if and only if H.f; m/ 2 HS or,
equivalently, m .f / 2 H S for all m 2 N. Thus, we established that
T
(5) C D f m1 .H S / W m 2 Ng.
We will show next that
(6) m is a measurable map for all m 2 N.

To do this, for any p 2 and i 2 D, let O.p; i / D ff 2 D W f .p/ D i g;
it is evident that the family S D fO.p; i / W p 2 and i 2 Dg is a subbase of
D . Therefore it suffices to show that m1 .O.p; i // is a Borel subset of R for all
m 2 N; p 2 and i 2 D (see Fact 1 of T.368).
Now, m1 .O.p; 0// D ff 2 R W m .f /.p/ D 0g D ff 2 R W p
H.f; m/g, i.e., m1 .O.p; 0// D ff 2 R W jf .p/j < m1 g is an open (and hence
Borel) subset of R for every p 2 .
Analogously, the equality m1 .O.p; 1// D ff 2 R W p 2 H.f; m/g shows
that m1 .O.p; 1// D ff 2 R W jf .p/j m1 g is a closed (and hence Borel) subset
of R for each p 2 so (6) is proved.
Our plan is to establish that H S is a Borel subset of D . Since this is not easy,
we will need several intermediate steps. The first one is to prove that
S
(7) the space H D H D! D fA W A fA.s/ W s 2 S 0 g [ P for some
S 0 2 Fin.D! / and P 2 Fin./g is -compact.
Fix an arbitrary number n 2 N and a finite set P and consider a
subset HnP D f.f; g1 ; : : : ; gn / 2 D ..D! //n W f 1 .1/ g11 .1/ [

[
gn1 .1/ [ P g of the compact space D ..D! //n . The set HnP is closed in
D ..D! //nS ; indeed, given a D .f; g1 ; : : : ; gn / 2 .D ..D! //n /nHnP , pick
p 2 f .1/n.. i n gi1 .1// [ P / and observe that W D f.h; u1 ; : : : ; un / W h.p/ D
1
1 and ui .p/ D 0 for all i ng is an open neighborhood of a in D ..D! //n with

W \ HnP D ;. Thus the space HnP is compact being closed in D ..D! //n .
Let n W D ..D! //nS! D be the natural projection for all n 2 N. It is
straightforward that H D fn .HnP / W n 2 N and P 2 Fin./g so H is -
compact and (7) is proved.
Denote by VC the set of all closed subsets of .D! / endowed with the Vietoris
topology; then VC is a metrizable compact space (see Fact 1 and Fact 2). Now let
be the set Fin..D! // with the topology induced from VC ; then is a metrizable
-compact space by Fact 3. Let S D Fin..S // considered to be a subspace
of . Then
(8) S is a Borel subset of .
Let D .D! /; n D fF 2 W jF j ng and n S D n \ S for every
n 2 !. Given any n 2 N, we have an order forgetting map 'n W n ! n defined
by 'n .x/ D fx.0/; : : : ; x.n 1/g for any x 2 n . Recall that we identify any n 2 N
with the set f0; : : : ; n 1g and consider the product n to be the set of functions
from f0; : : : ; n 1g to .
The sets I.U / D fF 2 W F U g and M.U / D fF 2 W F \ U ;g
form a subbase S of when U runs through all open subsets of . Fix U 2 ./
and x 2 'n1 .I.U //. Then F D fx.0/; : : : ; x.n 1/g U and therefore V D U n
is an open neighborhood of x in n such that 'n .V / I.U /. This proves that
'n1 .I.U // is open in n .
Furthermore, if x 2 'n1 .M.U //, then F \ U ; where F D fx.0/; : : : ; x.n

1/g. Pick j < n such that x.j / 2 U and let V D fy 2 n W y.j / 2 U g. Then
V 2 .x; n / and y.j / 2 'n .y/\U , i.e., 'n .y/ 2 M.U / for any y 2 V . Therefore
'n .V / M.U / and hence 'n1 .M.U // is also open in n for every U 2 ./.
Now that we proved that 'n1 .O/ is open in n for any O 2 S, we can conclude
that the map 'n is continuous for every n 2 N.
Consider the set Gn D fx 2 n W x.i / x.j / if i j g 2 .n /. It follows
from Fact 3 that the set Gn0 D n nn1 is open in n ; given any element F D
fx0 ; : : : ; xn1 g 2 Gn0 , the point x D .x0 ; : : : ; xn1 / belongs to Gn and 'n .x/ D F .
Choose fU0 ; : : : ; Un1 g ./ such that the family fU 0 ; : : : ; U n1 g is disjoint and
xi 2 Ui for all i < n. The set K D U 0

U n1 Gn is compact and for
any distinct y; z 2 K we have 'n .y/ 'n .z/. Therefore the map 'n0 D 'n jK is an
injection and hence 'n0 W K ! 'S 0
n .K/ is a homeomorphism. Besides, for the sets
U D U0

T Un1 and V D i <n Ui , we have F 2 'n .U / D W \ Gn0 where
W D I.V / \ . i <n M.Ui //. Thus, every F 2 n nn1 has an open neighborhood
(in n nn1 ) which is homeomorphic to an open subset of n .
Since S1 D .S / is a Borel set, the space S1n a Borel set as well by Problem 333.
Therefore S1n \ Gn is a Borel set too and it is immediate that n S nn1 S D
'n .S1n \ Gn /. We proved in the previous paragraph that any point F 2 n nn1
has a neighborhood WF 2 .F; n nn1 / such that for some UF 2 .Gn /,
the map 'n jUF W UF ! WF is a homeomorphism. As a consequence, if F 2
n S nn1 S , then 'n j.UF \ S1n / W .UF \ S1n / ! WF0 D WF \ .n S nn1 S /
is a homeomorphism. This proves that any point of n S D n S nn1 S has a
neighborhood which is a Borel set being homeomorphic to a Borel set UF \S1n . This
gives us an open cover C of n S which consists of Borel sets; choosing a countable
subcover of C, we represent n S as a countable union of Borel S sets. Therefore
n S is a Borel set by Problem 333 and hence S D f;g [ . fn S W n 2 Ng/
is also a Borel set (here we usedS Problem 333 again) so (8) is proved.
Recall that H D fA W A fA.s/ W s 2 S 0 g [ P for some S 0 2 Fin.D! / and
P 2 Fin./g; for any h 2 H , let .h/ D ff 2 W h1 .1/ \ f 1 .1/ is infiniteg.
Then
(9) .h/ 2 for any h 2 H and the map W H ! is measurable.
Take any h 2 H ; by definition there are s1 ; : : : ; sn 2 D! and a finite P
such that h1 .1/ A.s1 / [

[ A.sn / [ P . If h1 .1/ if finite, then .h/ D ; 2 .
If h1 .1/ is infinite and f 2 .h/, then f 1 .1/ \ h1 .1/ is infinite and therefore
f 1 .1/ \ A.si / is infinite for some i n. This implies f D .si / by (2) and hence
we have the inclusion .h/ f.s1 /; : : : ; .sn /g 2 which proves that .h/ 2
for any h 2 H .
To prove that the function is measurable, we will use again the subbase S
constructed in the proof of (8). Assume that U 2 ./ and consider the set Q D
1 .I.U // D fh 2 H W .h/ U g. Since is compact and metrizable and U
is open in , it is -compact. Fix any n 2 N and a finite set P ; the space
H being also -compact by (7), the product H U n is -compact as well. The set
S
Q.n; P / D f.h; f0 ; : : : ; fn1 / 2 H U n W h1 .1/ . i <n fi1 .1//[P g is closed
in H U n because for Sany a 1 D .h; f0 ; : : : ; fn1 / 2 .H U n /nQ.n; P /, we can
1
choose p 2 h .1/n.. i <n fi .1// [ P / and observe that G D f.k; g0 ; : : : ; gn / 2
H U n W k.p/ D 1 and gi .p/ D 0 for all i < ng is an open neighborhood of a in
H U n and G \ Q.n; P / D ;.
Consequently, the set Q.n; P / is -compact for all n S 2 N and P 2 Fin./; if
n W H U n ! H is the natural projection, then Q D fn .Q.n; P // W n 2 N
and P 2 Fin./g. Thus
(9a) Q D 1 .I.U // is -compact and hence Borel for any U 2 ./.
Now assume that U is a clopen subset of ; then G D 1 .M.U // D fh 2
H W .h/ \ U ;g and therefore G 0 D H nG D fh 2 H W .h/ nU g D
1 .I.nU // is a Borel set by (9a). Consequently, G D H nG 0 is also a Borel set,
so we proved that
(9b) for any compact U 2 ./, the set 1 .M.U // is a Borel subset of H .
Finally, take an arbitrary set U 2 ./; it is evident that we can represent it
as U D fUn W S n 2 !g where Un is a clopen subset of for all n 2 !. Then
1 .M.U // D f1 .M.Un// W n 2 !g is a Borel set because 1 .M.Un // is
Borel for all n 2 ! by (9b). Thus we proved that 1 .U / is a Borel subset of H for
any U 2 S whence is measurable so (9) is proved.
Our last step is to show that
(10) 1 .S / D H S .
To do this take any h 2 H S ; by definition there are s1 ; : : : ; sn 2 S and a finite
set P such that h1 .1/ A.s1 / [

[ A.sn / [ P . If h1 .1/ if finite, then
.h/ D ; 2 S . If h1 .1/ is infinite and f 2 .h/, then f 1 .1/ \ h1 .1/ is
infinite and therefore f 1 .1/ \ A.si / is infinite for some i n. This implies f D
.si / by (2) and hence we have the inclusion .h/ f.s1 /; : : : ; .sn /g 2 S
which proves that .h/ 2 S for any h 2 H S .
To prove the remaining inclusion suppose that h 2 H and S .h/ 2 S . By
definition of H there is a finite E D! such that h1 .1/ fA.t/ W t 2 Eg [ P
for some finite P . Observe that if h1 .1/ is Infinite, then h1 .1/ \ A.t/ is
infinite for some t 2 E and hence .t/ 2 .h/ .S /. This shows that .h/ D ;
implies that h1 .1/ is finite and hence h 2 H S .
Now if we have fs1 ; : : : ; sn g S and .h/ D f.s1 /; : : : ; .sn /g while the set
P D h1 .1/n.A.s1 /[

[A.sn // is infinite, then P \A.t/ is infinite for some t 2 E
which shows that .t/ 2 .h/ and hence t 2 fs1 ; : : : ; sn g which is a contradiction.
Thus the set P is finite and therefore h1 .1/ A.s1 /[

[A.sn /[P which shows
that h 2 H S and (10) is proved.
Observe finally that H S is a Borel subset of H because the map W H !
is measurable by (9), the set S is Borel in by (8) and H S D 1 .S / by
(10). Since H is -compact by (7), it is Borel in D , so the set H S is also Borel
in D (see Fact 1 of T.319 and Fact 1 of T.331).
1
The map m W R ! D is measurable T by1(6), so m .H S / is Borel in R

for all m 2 N which implies that C D f m .H S / W m 2 Ng is also a Borel

subset of R . We already noted in the first paragraph of the proof of this fact that
this implies that Cp .X S / is a Borel subset of R so Fact 6 is proved. t
u
Returning to our solution, take a set S 2 B.D! /nC2 0
.D! / which exists by
Problem 321 and let X D X S ; then X is a space with a unique non-isolated point.
By Fact 5 the space S is homeomorphic to a closed subspace T of the space Cp .X /.
The set Cp .X / is Borel in RX by Fact 6. Now, if Cp .X / 2 C1 0
.RX /, then the
set T D T \ Cp .X / (the bar denotes the closure in R ) also belongs to C1
X 0
.RX /
because T 2 0 .R / C1 .R / and finite intersections preserve Borel classes
0 X 0 X
(see Problem 320 and Fact 1 of T.341).

Apply Fact 2 of T.333 to find a G -subset S 0 of the space D! and a G -subset T 0
of RX such that S S 0 ; T T 0 and there is a homeomorphism h W S 0 ! T 0 . We
have T 2 C1 0
.T 0 / by Fact 1 of T.319 and therefore S 2 C1 0
.S 0 / (it is evident
0
that homeomorphisms preserve Borel classes). Since S is a G -set in D! , we have
S 0 2 10 .D! / C2
0
.D! / by Problem 320. Apply Fact 1 of T.319 again to find a
set E 2 C1 .D / such that E \S 0 D S . Since Borel classes are preserved by finite
0 !
intersections (see Fact 1 of T.341), we have S 2 C2 0

.D! /Swhich contradicts the
choice of S . Thus Cp .X / 2 B.R /nC1 .R / B.RX /n. f0 .RX / W < g/
X 0 X
(see Problem 321) and hence our solution is complete.

T.373. Prove that the following are equivalent for any metrizable space X :
(i) X is an absolute F ;
(ii) there is a completely metrizable space M such that X is an F -subset of M ;
(iii) X has a complete sequence of -discrete closed covers.
Solution. If Z is a space and A1 ; : : : ; An are families of subsets of Z, then let
A1 ^

^ An D fA Z W A ; and A D A1 \

\ An W Ai 2 Ai for all
i ng. If .Z; d / is a metric space, call a family A exp.Z/ uniformly discrete if
there is > 0 such that the ball Bd .z; / meets at most one element of A for any
z 2 Z. We will say that witnesses uniform discreteness of A. It is evident that
any uniformly
S 0 discrete
S family is discrete. A family A0 is a refinement of the family
A if A D A, and for any A0 2 A0 , there is A 2 A such that A0 A. If
A exp.Z/ and Y Z, then AjY D fA \ Y W A 2 A and A \ Y ;g.
S Then every open cover U D fUs W s 2 S g of

Fact 1. Let .Z; d / be a metric space.
the space Z has a refinement B D fBi W i 2 !g such that every Bi is uniformly
discrete for all i 2 !.
S
Proof. Fix a well-order < on the set S and let Hs D Us n. fUt W t < sg/ for each
s 2 S . Given i 2 ! and s 2 S , let Bs;i D fc 2 Hs W B.c; 3=2i / Us g. Next we
Sinduction on i 2 ! the sets Vs;i for all s 2 S . The first step is to define
define by
Vs;0 D fB.c; 1/ W c 2 Bs;0 g for all s 2 S S . If we have constructed SVs;j for each
j < i and s 2 S , consider the sets Vs;i D fB.c; 1=2i / W c 2 Bs;i n. fVs;j W s 2
S; j < i g/g for all s 2 S . Observe that Vs;i Us is an open set for all s 2 S and
S
i 2 !. Let Bi D fVs;i W s 2 S g and B D fBi W i 2 !g. For any z 2 Z there is a
minimal s 2 S with z 2 Us . This implies z 2SHs . Pick i 2 ! such that B.z; 3=2Si
/
Us ; for this i we have z 2 Bs;i . Now, if z 2 SfVs;j W s 2 S; j < i g,Sthen z 2 B.
If not, then B.z; 1=2i / Vs;i and again z 2 B. This yields Z D B and hence
B is a refinement of U.
To see that every Bi is uniformly discrete, observe that
(1) if x 2 Vs;i ; y 2 Vt;i where s < t, then d.x; y/ > 1=2i .
Indeed, there is a point c 2 Bs;i such that x 2 B.c; 1=2i / and c 0 2 Bt;i with
y 2 B.c 0 ; 1=2i /. We have B.c; 3=2i / Us while c 0 Us . Thus d.c; c 0 / 3=2i ,
and if d.x; y/ 1=2i , then d.c; c 0 / d.c; x/ C d.x; y/ C d.y; c 0 / < 1=2i C
1=2i C 1=2i D 3=2i ; this is a contradiction which proves (1).
Now take any z 2 Z and Uz D B.z; 1=2i C1 / 2 .z; Z/. If there exist s; t 2 S
such that s < t and Uz \ Vs;i ; Uz \ Vt;i , then pick x 2 Uz \ Vs;i and
y 2 Uz \ Vt;i ; our claim implies d.x; y/ > 1=2i . However
d.x; y/ d.x; z/ C d.z; y/ 1=2i C1 C 1=2i C1 D 1=2i
which is a contradiction. Thus, for D 2i 1 the ball Bd .z; / meets at most one
element of Bi for all z 2 Z. Therefore each Bi is uniformly discrete and Fact 1 is
proved. t
u
Fact 2. Let .Z; d / be a metric space. Given Y Z and family A exp.Y / which
is uniformly discrete in .Y; d j.Y Y //, the family A is also uniformly discrete (and
hence discrete) in .Z; d /.
Proof. There is 0 > 0 which witnesses uniform discreteness of A in .Y; d j.Y Y //.
Then D 20 witnesses uniform discreteness of A in .Z; d /. Indeed, assume that
z 2 Z and there are distinct A; B 2 A such that Bd .z; / meets both A and B.
Then there are a 2 A and b 2 B such that d.a; z/ < and d.b; z/ < which
implies d.a; b/ d.a; z/ C d.z; b/ < 2 D 0 which is a contradiction because the
0 -ball of the point a in Y cannot intersect both sets A and B. Thus the family A is
uniformly discrete (and hence discrete) in .Z; d /, so Fact 2 is proved. t
u
Fact 3. If .Z; d / is a metric space and a family Ai is uniformly discrete in .Z; d /
for all i < n 2 !, then the family A D A0 ^

^ An1 is also uniformly discrete.
Proof. Let i > 0 be the number which witnesses uniform discreteness of Ai for
all i < n. If D minf0 ; : : : ; n1 g, then take any z 2 Z and assume that there
are distinct A; B 2 A such that W DTBd .z; / meets bothTA and B. For all i < n
choose Ai ; Bi 2 Ai such that A D i <n Ai and B D i <n Bi . Since A B,
there is i < n such that Ai Bi . It follows from W \ A ; and W \ B ;
that W \ Ai ; and W \ Bi ; which contradicts the choice of i . Thus
witnesses uniform discreteness of A and Fact 3 is proved. t
u
Returning to our solution, assume that fFn W n 2 !g is a complete sequence
of -discrete closed covers of X and X is a subspace of a metric space .Y; d /; let
d1 D d j.X X /. Every F -subset of clY .X / is an F -subset of Y , so in proving

that X is an F -subset of Y , there is no loss of generality to assume that X is dense
in Y .
It is easy to see that
(2) if S 0 D fFn0 W n 2 !g is a sequence of covers of X and Fn0 is a refinement of Fn
for each n 2 !, then S 0 is also complete.
Another evident property is
(3) if fPt W t 2 T g is a uniformly discrete family in .Y; d / and Qt Pt for all
t 2 T , then the family fQt W t 2 T g is uniformly discrete.
S
By our assumption Fn D fFnm W m 2 !g where Fnm is a discrete family of X ,
so we can choose an open cover Unm of the space X such that every U 2 Unm has
n
diameter S 2 m and intersects at most one element of Fnm . There is a refinement
Vn D k2! Vn .k/ of the cover Un such that each Vnm .k/ is uniformly discrete in
m m
the space .X; d1 / (see Fact 1).

For all n; m; k 2 ! the family G.n; m; k/ D Vnm .k/ ^ Fnm is uniformly discrete
by (3) because every S set V 2 Vnm .k/ contains at most one element of G.n; m; k/.
It is clear
S that m Gn D k2! G.n; m; k/ is a refinement of the family Fnm and hence
m
0
Gn D m2! Gn is a refinement of the cover Fn for every n 2 !. Since Fn consists
of closed subsets of X , the family Gn D fclX .G/ W G 2 Gn0 g is also a refinement of
Fn for all n 2 !.
The property (2) shows that the sequence fGn W n 2 !g is complete; observe that
we also have
n
S/ 2 for every n 2 ! and V 2 Gn ;
(4) diam.V
(5) Gn D fGn .k/ W k 2 !g where Gn .k/ is uniformly discrete for every k 2 !.
Furthermore, the family H.n; k/ D fP W P 2 Gn .k/g is uniformly discrete in
.Y; d / by Fact 2 and an evident observation that the closures of the elements of
a uniformly discrete family form a uniformly discrete family (the bar denotes the
closure in Y ).
Now if m 2 N and we are given distinct numbers n1 ; : : : ; nm 2 ! and arbitrary
k1 : : : ; km 2 !, then let
Q.n1 ; : : : ; nm ; k1 ; : : : ; km / D fQ 2 H.n1 ; k1 / ^

^ H.nm ; km / W Q \ X D ;g:
It follows from Fact 3 that Q.n1 ; : : : ; nm S ; k1 ; : : : ; km / is uniformly discrete and

hence the set Q.n1 ; : : : ; nm ; k1 ; : : : ; km / D Q.n1 ; : : : ; nm ; k1 ; : : : ; km / is closed
in Y for Sany distinct n1 ; : : : ; nm 2 ! and arbitrary k1 : : : ; km 2 !. Consequently,
D D fQ.n1 ; : : : ; nm ; k1 ; : : : ; km / W n1 ; : : : ; nm are distinct elements of ! and
k1 ; : : : ; km 2 !g is an F -subset of Y whence R D Y nD is a G -subset of Y which
contains X .
Every open subset of Y is an F -set, S so R is an F -subset of Y . Since H.n; k/
is uniformly discrete, the set Hnk D H.n; k/ is closed in Y for any n; k 2 !.
S T
Consequently, Hn D k2! Hnk is an F -subset of Y . Therefore H D . n2! Hn / \
R is an F -subset of Y so it suffices to show that X D H .
The inclusion X H is evident, so assume that there is a point y 2 H nX .
There exists a sequence fkn W n 2 !g ! such that y 2 Hnkn for all n 2 !.
The family H.n; kn / being uniformly discrete, we can choose Pn 2 Gn .kn /, so
that y 2 P n for every n 2 !. Now, for any distinct n1 ; : : : ; nm 2 !, we have
y 2 Q D P n1 \

\ P nm and hence Q \ R ;, i.e., Q 6 D which shows that
Pn1 \

\ Pnm D Q \ X ; because D contains all elements of Gn1 .kn1 / ^

^
Gnm .knm / which do not meet X .
This proves that the family P D fPn W n 2 !g is centered and hence we can
choose a filter F P (see Problem 117 of [TFS]). Since T Pn 2 F \ Gn and the
sequenceT n fG W n 2 !g is complete, we have P D n2! clX .Pn / ; because
P fclX .F / W F 2 F g ;. It follows from (4) that there is x 2 X such
that fxg D P and therefore x 2 P n for all n 2 !. However, y 2 P n and hence
d.x; y/ diam.P n / D diam.Pn / D 2n for all n 2 ! which implies x D y, a
contradiction with x 2 X and y 2 Y nX . This contradiction shows that H D X ,
i.e., X is an F -subset of Y and proves that (iii)H)(i).
The implication (i)H)(ii) follows from the fact that any metrizable space X can
be embedded into a completely metrizable space (see Problem 237 of [TFS]).
NowT assume that there is a complete metric space .M; d / such that X M and
X D n2! Mn where each MnS is an F -subset of M .
Fix n 2 !; we have Mn D k2! Mnk where Mnk is closed in M for S all k 2 !.
It is easy to find, for every k 2 !, a family Gnk .Mnk / such that Gnk D Mnk ,
S Gn kis -discrete
the family k
in Mnk and diam.U / 2n for any U 2 Gnk . Now, if
0
Gn D n2! Gn and Gn D fG W G 2 Gn g (the bar denotes the closure in M ), then
S 0
(6) Gn X , the family Fn D Gn0 jX is -discrete in X and consists of closed
subsets of X and for any F 2 Fn we have diam.F / 2n and F Mn .
We claim that fFn W n 2 !g is a complete sequence in X . Indeed, if F is a filter
on X and F \ Fn ;, then fix Pn 2 F \ Fn for every n 2 !. It is clear that
G D fF W F 2 F g is a centered family of closed subsets of M ; besides, if " > 0 and
2n < ", then G D P n 2 G while diam.G/ D diam.Pn / n
T 2 < ". We proved
that G has elements of arbitrarily small diameter and hence TG ; (see Problem
236 of [TFS]). It is easy to see that there is p 2TM such that G D fpg. Since p 2
P n T Mn for all n 2 ! [see (6)],
T we have p 2 n2! MT n D X and therefore p 2 X .
Thus fclX .F / W T F 2 F g D fF \ X W F 2 F g D . G/ \ X D fpg \ X D fpg
which shows that fclX .F / W F 2 F g ; and hence fFn W n 2 !g is a complete
sequence of closed -discrete covers of X . We settled the implication (ii)H)(iii)
so our solution is complete.
T.374. Prove that Cp .X / is an absolute F for any countable metrizable X .
Solution. Let d be a metric on X such that .d / D .X/. We first prove that
(1) given n; k 2 N and x 2 X the set Pnk .x/ D ff 2 RX W jf .x/ f .y/j 1
n
whenever d.y; x/ < k1 g is closed in RX .
To see that (1) holds, take any f 2 RX nPnk .x/; there exists y 2 X with
d.y; x/ < k1 while jf .x/ f .y/j > n1 . If ex and ey are the natural projections of RX
onto the factors determined by x and y respectively, then they are continuous and
hence so is the map ' W RX ! R defined by '.g/ D jex .g/ ey .g/j D jg.x/ g.y/j
for every g 2 RX (it is immediate that ' D jex ey j). Now it follows from continuity
of ' and the equality Wf D fg 2 RX W jg.x/ g.y/j > n1 g D ' 1 .Rn n1 ; n1 / that
Wf is an open neighborhood of f which does not intersect Pnk .x/. Thus Pnk .x/ is
closed in RX and (1) is proved. S
It follows from (1) that Pnx D k2! Pnk .x/ is an F -subset of RX for every
n 2 N and x 2 X . Let us show next that
T
(2) Cp .X / D P D fPnx W n 2 N and x 2 X g.
Pick any f 2 Cp .X /; if x 2 X , then f is continuous at the point x, so, for any
n 2 N, there exists > 0 such that d.x; y/ < implies jf .x/ f .y/j < n1 . Now,
if k 2 N and k1 < , then T f 2 Pnk .x/ Pn ; thus, for any n 2 N and x 2 X , we
x
have f 2 Pnx , i.e., f 2 fPnx W n 2 N and x 2 X g D P . The function f 2 Cp .X /

was chosen arbitrarily, so we proved that Cp .X / P .
Now if f 2 P , then take any x 2 X and " > 0. There exists a number n 2 N
such that n1 < "; since f 2 Pnx , there is k 2 N such that d.y; x/ < k1 implies
jf .x/ f .y/j n1 < " which shows that f is continuous at x. The point x was
chosen arbitrarily, so f 2 Cp .X / which establishes that P Cp .X / and therefore
P D Cp .X /, i.e., (2) is proved.
It immediate from (1) and (2) that Cp .X / is an F -set in the completely
metrizable space RX . Therefore we can apply Problem 373 to conclude that Cp .X /
is an absolute F .
T.375. Let K be a compact space. Given a countable X Cp .K/, prove that Cp .X /
is an absolute F .
Solution. For any m; n; 2 N and f 2 X consider the set Pmn .f / D f' 2 RX W
there exist x1 ; : : : ; xn 2 K for which j'.f / '.g/j m1 for any g 2 X such that
jf .xi / g.xi /j < n1 for all i ng. Let us first prove that
(1) the set Pmn .f / is closed in RX for any f 2 X and m; n 2 N.
We will need the set Qmn .f / D f.'; x/ 2 RX K n W j'.f / '.g/j m1
whenever x D .x1 ; : : : ; xn / and jg.xi / f .xi /j < n1 for all i ng. If W RX
K n ! RX is the natural projection, then .Qmn .f // D Pmn .f /.
To see that the set Qmn .f / is closed in the space RX K n , take any point w D
.'; x/ 2 .RX K n /nQmn .f / where x D .x1 ; : : : xn /; there exists g 2 X such that
jg.xi / f .xi /j < n1 for all i n while j'.f / '.g/j > m1 .
Since the functions f and g are continuous on the space K, the set W D
f.y1 ; : : : ; yn / 2 K n W jg.yi / f .yi /j < n1 for all i ng is open in K n ; clearly,
x 2 W . It is an easy exercise to see that the set V D f 2 RX W j.f / .g/j > m1 g
is open in RX and contains '. Therefore U D V W is an open neighborhood
of w in RX K n such that U \ Qmn .f / D ;. This shows that Qmn .f / is closed
in RX K n ; since K n is compact, the projection is a closed map (see Fact 3 of

S.288) so Pmn .f /SD .Qmn .f // is a closed subset of RX and (1) is proved.
Let Rm .f / D n2N Pmn .f / for all m 2 N and f 2 X . Our last step is to show
that
T
(2) Cp .X / D P D fRm .f / W m 2 N and f 2 X g.
Take any ' 2 Cp .X /; for any f 2 X , the map ' is continuous at the point f , so
for any m 2 N there exists y1 ; : : : ; yk 2 K and " > 0 such that j'.g/ '.f /j < m1
for any g 2 X with jg.yi / f .yi /j < " for all i k. It is easy to find n 2 N such
that k n and n1 < "; if x D .x1 ; : : : ; xn / 2 K and fy1 ; : : : ; yk g fx1 ; : : : ; xn g,
then j'.f / '.g/j < m1 for any g 2 X such that jg.xi / f .xi /j < n1 for all
i n. Consequently, f 2 Pmn .f / Rm .f /; we have chosen f 2 X and m 2 N
arbitrarily so f 2 P and hence Cp .X / P .
Now, if ' 2 P , take any f 2 X and " > 0; there is m 2 N such that m1 < ".
Since ' 2 Rm .f /, there is n 2 N such that ' 2 Pmn .f / and hence there exists
.x1 ; : : : ; xn / 2 K n such that j'.f / '.g/j m1 < " for any point g 2 X such that
jg.xi / f .xi /j < n1 for all i n. If O D fg 2 X W jg.xi / f .xi /j < n1 for all
i ng, then O 2 .f; X/ and '.O/ .'.f / "; '.f / C "/, i.e., O witnesses
continuity of ' at the point f . Thus ' is continuous at every f 2 X and therefore
' 2 Cp .X /. This shows that P Cp .X / and hence P D Cp .X /.
Finally, observe that each Rm .f / is an F -subset of RX by (1) so Cp .X / is an
F -subset of a completely metrizable space RX by (2). Thus Cp .X / is an absolute
F by Problem 373.
T.376. Prove that any second countable space embeds into Cp .K/, where K is the
Cantor set.
Solution. It suffices to embed I! in Cp .K/ by Problem 209 of [TFS]. Let a D 0 and
an D n1 for all n 2 N. Then S D fag [ fan W n 2 Ng is a convergent sequence and
hence a compact metrizable space. There exists a continuous onto map r W K ! S
by Problem 128 of [TFS]; the dual map r W Cp .S / ! Cp .K/ embeds Cp .S / in
Cp .K/ (see Problem 163 of [TFS]), so it suffices to embed I! in Cp .S /.
Let F D ff 2 IS W f .a/ D 0 and f .an / 2 0; n1 for all n 2 Ng. It is immediate
Q
that F is homeomorphic to the product f0g f0; n1 W n 2 Ng which is a compact
subspace of IS . Besides, F Cp .S / because f .an / ! 0 when Q n ! 1 for every
f 2 F . Another easy observation is that F is homeomorphic to f0; n1 W n 2 Ng
which in its turn is homeomorphic to I! because 0; n1 is homeomorphic to I for all
n 2 N. Thus F is a subspace of Cp .S / homeomorphic to I! , so I! also embeds in
Cp .K/ and hence any second countable space embeds in Cp .K/.
T.377. Give an example of a second countable X such that for any compact K, the
space X cannot be embedded in Cp .K/ as a closed subspace.
Solution. The existence of our X can be easily deduced from the following fact.
Fact 1. If K is a compact space and M is a second countable closed subspace of

Cp .K/, then M is a K -space.
Proof. Let '.x/.f / D f .x/ for any x 2 K and f 2 M . Then '.x/ 2 Cp .M /
for any x 2 X and the map ' W K ! Cp .M / is continuous (see Problem 166 of
[TFS]). If L D '.K/, then L is compact and w.L/ D nw.L/ nw.Cp .M // D
nw.M / w.M / D ! which shows that L is a metrizable compact space.
The dual mapping ' W Cp .L/ ! Cp .K/ defined by ' .g/ D g ' for every
g 2 Cp .L/ is an embedding by Problem 163 of [TFS] and it is immediate that
M ' .Cp .L//. Since M is closed in a larger space Cp .K/, it is also closed
in ' .Cp .L//. This proves that M embeds as a closed subspace in Cp .L/. Since
Cp .L/ is a K -space by Problem 362, so is M by V1.338 and Fact 1 is proved. u
t
Returning to our solution, take a second countable space X which is not K
(such a space exists by Problems 334, 342 and 346) and observe that X cannot be
homeomorphic to a closed subspace of Cp .K/ for any compact K by Fact 1.
T.378. Prove that any countable second countable space embeds into Cp .K/ as a
closed subspace.
Solution. Let a D 0 and an D n1 for all n 2 N. Then S D fag [ fan W n 2 Ng is
a convergent sequence and hence it is a compact countable (and hence metrizable)
space. There exists a continuous onto map r W K ! S by Problem 128 of [TFS]; the
dual map r W Cp .S / ! Cp .K/ embeds Cp .S / in Cp .K/ as a closed subspace (see
Problem 163 of [TFS]). The space Cp .S / is dense-in-itself, second countable and of
first category in itself (see Problem 278 and 284 of [TFS]), so we can apply Fact 4 of
T.351 to conclude that there is a closed F Cp .S / which is homeomorphic to Q.
Since Cp .S / embeds in Cp .K/ as a closed subspace, there is a closed G Cp .K/
which is homeomorphic to Q.
Now, if M is a countable metrizable space, then there is a closed N G which
is homeomorphic to M (see Problem 350). It is evident that N is also a closed
subspace of Cp .K/, so M embeds in Cp .K/ as a closed subspace.
T.379. Given a space X and a function f W X ! R, consider the following
conditions:
(i) for any open U R, the set f 1 .U / is an F -set in X ;
(ii) there exists a sequence ffn W n 2 !g Cp .X / which converges to f .
Prove that (ii)H)(i) for any space X . Show that if X is second countable, then
also (i)H)(ii) and hence (i) (ii).
Solution. Given a space Z let B .Z/ D ff 2 RZ W f is bounded on Zg. It is easy
to see that B .Z/ D C .Z 0 / where Z 0 is the set Z with the discrete topology. For a
second countable space Z and A Z the statements A is an F -subset of Z and
A 2 10 .Z/ are identical, so we will use them alternatively to avoid repetitions
and shorten our expressions.
Fact 1. Let Z be a second countable space; given an n 2 ! assume that Ai 2

10 .Z/ for all i < n. Then there exists S family fBi W i < ng 1 .Z/
0
S a disjoint
such that Bi Ai for all i < n and i <n Ai D i <n Bi .
Proof. It is an easy exercise that
. / the intersection of two F -sets of Z is an F -set of Z and if F and G are
closed in Z then F nG is an F -subset of Z.
In this proof we will use . / many times without explicitly referring to it. Our fact
is evident for n 2 f0; 1g; to prove it for n D 2, take any sets P; Q 2 10 .Z/. There
existSfamilies P D S fP .i / W i 2 !g; Q D fQ.i / W i 2 !g of closed subsets of Z such
that P D P; Q D Q and P .i / P .i C 1/; Q.i / Q.i C 1/ for all i 2 !.
Let P0 D P .0/; Q0 D Q.0/ and Pi C1 D P .i C 1/nP .i /; Qi C1 D Q.i C 1/nQ.i /
for all i 2 !. Then fPi W i 2 !g [ fQi W i 2 !g 10 .Z/.
Furthermore, the sets Pi0 D P .i /nQ.i 0
S / and0 Qi D Q.i /nP
S .i / are F in Z for all
i 2 !; it is straightforward that P D i 2! Pi and Q0 D i 2! Qi0 are disjoint and
0
F while P nQ S P 0 and QnP Q0 . Now, the set R D .P [Q/n.P 0 [Q0 / is also

F because R D i 2! Ri where Ri D Pi \ Qi for all i 2 !. As a consequence,
P 00 D P 0 [ R and Q0 are disjoint F -subsets of Z for which P P 00 ; Q Q0 and
P 00 [ Q0 D P [ Q. This, evidently, settles our fact for n D 2.
Now assume that n D k C 1 > 2 and we have proved our fact for all numbers
n k. Given a family fAi W i < k C 1g of F -sets of Z, we can apply the
0 0 0
S sets B0 ; : : : ; Bk1 2 1 .Z/ such that Bi Ai
0
induction hypothesis S to find0 disjoint
for all i < k and i <k Bi D S i <k Ai . Since we proved our fact for n D 2, we
can apply it to the F -sets P D i <k Ai and Q D Ak to obtain disjoint F -sets
P 0 P; Q0 Q such that P [Q D P 0 [Q0 . If we let Bk D Q0 and Bi D Bi0 \P 0
for all i < k, then it is immediate that we obtain the desired sets B0 ; : : : ; Bk ; this
verifies our fact for n D k C 1 and shows that our inductive procedure guarantees
that it is fulfilled for all n 2 !. Thus Fact 1 is proved. t
u
Fact 2. For any space Z the set B1 .Z/ is closed in the uniform topology on RZ . In
other words, if fn 2 B1 .Z/ for all n 2 ! and fn f , then f 2 B1 .Z/.
Proof. For any g 2 B .Z/ let jjgjj D supfg.z/ W z 2 Zg. If Z 0 is the set Z with
the discrete topology, then B .Z/ D C .Z 0 / and the function .g; h/ D jjg hjj
is a metric on B .Z/ (see Problem 248 of [TFS]). If 0 2 RZ is identically zero on
Z, then jjgjj D .g; 0/ for any g 2 B .Z/. As a consequence,
(1) jju C vjj jjujj C jjvjj for any u; v 2 B .Z/.
It follows from fn f that jjfn f jj ! 0, so choosing an appropriate
subsequence of the sequence ffn gn2! , we can assume, without loss of generality,
that jjfn f jj 2n2 for all n 2 !, i.e.,
(2) jfn .x/ f .x/j 2n2 for any n 2 ! and x 2 Z.
Let p0 D f0 and pnC1 D fnC1 fn for all n 2 !. Then pn 2 B1 .Z/ and

jjpnC1 jj jjfnC1 f jj C jjf fn jj 2n1 for all n 2 ! (we applied the property
(1) here). Let un .x/ D 2n and vn .x/ D 2n for all n 2 N and x 2 Z; then
(3) vn .x/ pn .x/ un .x/ for all n 2 N and x 2 Z.
For every n 2 ! there is a sequence frnm W m 2 Ng C.Z/ such that rnm ! pn
as m ! 1. It follows from (2) that if qnm D max.rnm ; vn / and pnm D min.qnm ; un /
for all m 2 N, then
(4) for all n 2 ! and m 2 N, we have pnm 2 C.Z/ and jjpnm jj 2n ; besides,
pnm ! pn when m ! 1 for all n 2 !.
P
Now let snm D niD0 pim for all n; m 2 N. Apply (3) and Problem 030 of [TFS]
to conclude that for any m 2 N, there is a function sm such that snm sm (when
n ! 1). Therefore sm 2 C.Z/ for every m 2 N by Problem 029 of [TFS]; let us
prove that sm ! f .
Observe first that for any x 2 Z and n 2 N, we have the equality jsm .x/
snm .x/j D limk!1 jskm .x/ snm .x/j. Now,
X
k X
k 1
X
jjskm .x/ snm .x/jj D jj pim jj jjpim jj 2i D 2n
i DnC1 i DnC1 i DnC1
for any x 2 Z and k n which shows that

(5) jsm .x/ snm .x/j 2n for any n; m 2 N and x 2 Z.
n
Fix a point z 2 Z and an arbitrary " > 0; there is n 2 N such "
Pnthat 2 < 3 and
n
hence jsm .z/ sn .z/j 2 < 3 for all m 2 N. Observe that i D0 pi D fn , so it
m "
follows from pim !Ppi (for each i n as m ! 1) that there is k 2 N such that
jsnm .z/ fn .z/j D j niD0 pim .z/ fn .z/j < 3" for all m k. Therefore
"
jsm .z/ f .z/j jsm .z/ snm .z/jCjsnm .z/ fn .z/jCjfn .z/ f .z/j 2n C C2n D "
3
for any m k [we used the properties (2) and (5)] which proves that the sequence
fsm .z/ W m 2 Ng Cp .Z/ converges to f .z/ for every z 2 Z. Now apply Problem
143 of [TFS] to conclude that sm ! f and hence f 2 B1 .Z/. Fact 2 is proved. u t
Returning to our solution assume that w.X / D ! and f 1 .U / is an F -set in X
for any open U R. Let ' W R ! .0; 1/ be a homeomorphism; we will also need
the maps un ; vn 2 Cp .X / defined by un .x/ D nC1 1
and vn .x/ D 1 nC1 1
for all
x 2 X and n 2 N. If we prove that g D ' f 2 B1 .X /, then there is fpn W n 2
!g Cp .X / with pn ! g. Letting qn D max.pn ; un / and rn D min.qn ; vn / for
all n 2 N, we obtain a sequence frn W n 2 Ng Cp .X; .0; 1// with rn ! g. It is
straightforward that if fn D ' 1 rn for each n 2 N, then ffn W n 2 Ng Cp .X /
and fn ! f .
.X / .0; 1/. Call a

Therefore we can assume, without loss of generality, that f S
finite family A exp.X / a partition of X if A is disjoint and A D X . Our next
step is to prove the following property.
(6) If A is a partition of X and each A 2 A is an F -subset of X , then every
g 2 RX which is constant on all elements of A, belongs to B1 .X /.
To see that (6) is true let A D fA1 ; : : : ; An g and assume that r1 ; : : : ; rn 2 R are
chosen so that for every i n we have g.x/ D ri for any S x 2 Ai . There is a family
j j j C1
Ai D fAi W j 2 !g of closed subsets of X such that Ai D Ai and Ai Ai
for all j 2 ! and i n.
j
For any j 2 !, the family Bj D fAi W i ng is finite, disjoint and consists of
closed subsets ofSX , so by normality of X , there is a continuous function gj W X !
R such that gj j. Bj / D g. Given x 2 X it is immediate that gj .x/ D g.x/ for all
but finitely many j 2 ! and hence the sequence fgj W j 2 !g Cp .X / converges
to g. This settles (6).
For any n 2 N it is easy to choose an open finite cover fU1 ; : : : ; Uk g of the space
.0; 1/ in such a way that diam.Ui / n1 for all i k. The set Vi D f 1 .Ui / is an
F -set in X for every i k and V1 [

[ Vk D X . By Fact 1 there is a partition
P D fP1 ; : : : ; Pk g of the space X such that Pi is an F -subset of X and Pi Ui for
all i n. Choose ri 2 Ui for all i k and given x 2 X let fn .x/ D ri if x 2 Pi .
Since P is a disjoint family, this consistently defines a function fn W X ! R which
is constant on all elements of P. Therefore fn 2 B1 .X / by (6); furthermore, if
x 2 X , then x 2 Pi for some i n and hence f .x/ 2 Ui which shows that
jf .x/ fn .x/j D jf .x/ ri j diam.Ui / n1 . Thus jfn .x/ f .x/j n1 for
every x 2 X and n 2 N which implies that the sequence ffn W n 2 Ng converges
uniformly to f . Now apply Fact 2 to conclude that f 2 B1 .X /, i.e., we established
that (i)H)(ii) for any second countable space X .
To prove the remaining implication assume that fn 2 Cp .X / for all n 2 ! and
the sequence ffn W n 2 !g converges to f . Given an open set U R,Schoose a
family V D fVn W n 2 !g .R/ such that V n VnC1 for all n 2 ! and V D U .
It turns out that
S T
(7) f 1 .U / D H D n2! kn fk1 .V n /.
Indeed, if x 2 H , then there is n 2 ! such that fk .x/ 2 V n for all k n. Since
the sequence ffk .x/ W k ng converges to f .x/, we have f .x/ 2 V n U , i.e.,
f .x/ 2 U which shows that x 2 f 1 .U / and therefore H f 1 .U /.
On the other hand, if x 2 f 1 .U /, then f .x/ 2 U and hence f .x/ 2 Vm
for some m 2 !. The sequence ffk .x/ W k 2 !g converges to f .x/, so there is
n 2 ! such that n m and fk .x/ 2 Vm for all k n. We have Vm Vn , so
fk .x/ 2 Vn V n for all k n and therefore x 2 H . Thus f 1 .U / H , i.e.,
f 1 .U / D H so (7) is proved.
Finally, apply (7) to conclude that f 1 .U / is an F -subset of X for any open
U R; this proves the implication (ii)H)(i) for an arbitrary space X and makes
T.380. Prove that if X D R, then B1 .X / RX .

Solution. Let f .x/ D 0 if x 2 Q and f .x/ D 1 for all x 2 P. Then f W R ! R;
the sets U D . 12 ; 12 / and V D . 12 ; 32 / are open in R while Q D f 1 .U / and
P D f 1 .V /. Since Q is not Cech-complete (see Problem 274 of [TFS]), it is not
a G -subset of R (see Problem 260 of [TFS]). As a consequence P D f 1 .V /
is not an F -subset of R which shows that f B1 .R/ (see Problem 379), i.e.,
f 2 RX nB1 .X /.
T.381. Prove that a compact space X is countable if and only if B1 .X / D RX .
Solution. If X is a compact countable space, then X is second countable by Fact 4
of S.307. If f 2 RX , then f 1 .U / is an F -subset of X (because all subsets of X
are F in X ) for any U R and hence f 2 B1 .X / by Problem 379. This proves
necessity.
Now assume that X is a compact space such that B1 .X / D RX . Given any
A X , let f be the characteristic function of A in X , i.e., f .x/ D 1 if x 2 A and
f .x/ D 0 for all x 2 X nA. Then A D f 1 .. 12 ; 12 //; since f 2 B1 .X / by our
assumption, the set A is F in X by Problem 379. The space X being compact we
proved that every A X is -compact, so we can apply Problem 274 to conclude
that X is countable. This proves sufficiency and completes our solution.
T.382. Prove that, under MAC:CH, there exists an uncountable X R such that
B1 .X / D RX .
Solution. Assume MAC:CH and let X be a subset of R of cardinality !1 < c.
It follows from MAC:CH that every A X is a G -subset of X (see Problem
055). Thus every subset of X is F being a complement of a G -subset of X . As a
consequence, if f 2 RX , then f 1 .U / is an F -subset of X for any U R. This
shows that f 2 B1 .X / (see Problem 379) and therefore B1 .X / D RX .
T.383. Prove that the two arrows space is Rosenthal compact.
Solution. For any t 2 .0; 1 let at0 .x/ D 0 if 0 x < t and at0 .x/ D 1 for
all x 2 t; 1 ; this defines a function at0 W 0; 1 ! R. Now, if t 2 0; 1/, then
consider the function at1 W 0; 1 ! R defined by at1 .x/ D 0 for all x 2 0; t and
at1 .x/ D 1 if x 2 .t; 1 . If A0 D fat0 W t 2 .0; 1 g and A1 D fat1 W t 2 0; 1/g,
then A0 [ A1 B1 .0; 1 / because f 1 .U / is an F -subset of 0; 1 for any (not
necessarily open) U R and f 2 D D A0 [ A1 (see Problem 379).
Recall that the underlying set T of the two arrows space .T; / is defined by
T D ..0; 1 f0g/ [ .0; 1/ f1g/ R2 while the topology is generated on T by
families fBz W z 2 T g as local bases where the collection B D fBz W z 2 T g is defined
as follows. Given z D .t; 0/ 2 T , let Bz D f.."; t f0g/[.."; t/f1g/ W 0 < " < tg;
if z D .t; 1/ 2 T , then Bz D f.t; "/ f1g/ [ ..t; "/ f0g/ W t < " < 1g. Observe
that for any x 2 .0; 1/, the sets Ox0 D f.t; i / 2 T W i 2 f0; 1g and t < xg [ f.x; 0/g
and Ox1 D f.t; i / 2 T W i 2 f0; 1g and t > xg [ f.x; 1/g are disjoint, clopen in T
and T D Ox0 [ Ox1 . For technical reasons it is convenient to consider also the sets
O00 D ;; O01 D T and O10 D T; O11 D ;.
Let '..t; 0// D at0 for all t 2 .0; 1 and '..t; 1// D at1 for all t 2 0; 1/. It is
evident that the map ' W T ! D is a bijection. Besides, ' maps T into the product
space R0;1 ; for any x 2 0; 1 , let x W R0;1 ! R be the natural projection onto
the factor determined by x. If z D .t; i /, then .x '/.z/ D ati .x/ D 1 if t < x and
.x '/.z/ D ati .x/ D 0 if t > x.
Since also .x '/..x; 0// D 1 and .x '/..x; 1// D 0 for any x 2 .0; 1/, we
have the equalities .x '/1 .0/ D Ox1 and .x '/1 .1/ D Ox0 for all x 2 0; 1
which show that the sets .x '/1 .0/ and .x '/1 .1/ are clopen in T . This
implies that x ' is continuous for every x 2 0; 1 and therefore the map '
is continuous by Problem 102 of [TFS]; being a bijection, it is a homeomorphism
because T is compact by Problem 384 of [TFS]. Thus T is homeomorphic to the
subspace D of B1 .0; 1 /, i.e., the two arrows space T is Rosenthal compact.
T.384. Prove that every Rosenthal compact space is FrchetUrysohn.
Solution. If Z is a space and A Z, then the expressions A is an F -subset of Z
and A 2 10 .Z/ say the same thing, so we will use them interchangeably to avoid
repetitions and save the space. Given f W Z ! R let osc.f; U / D diam.f .U // for
any U Z. If z 2 Z, then osc.f; z/ D inffosc.f; U / W U 2 .z; Z/g. It is easy to
see that f is continuous at a point z if and only if osc.f; z/ D 0. If we have infinite
sets A and B, say that A B if AnB is finite. If S D fzn W n 2 !g Z, then
z 2 Z is accumulation point of the sequence S if the set fn 2 ! W zn 2 U g is infinite
for any U 2 .z; Z/. Observe that if z S , then z is an accumulation point for the
set S if and only if it is an accumulation point for S considered as a sequence (the
difference is that S can be finite as a set, so it has no accumulation points while it
has accumulation points as a sequence).
Fact 1. If Z is a set and we have a sequence fAi W i 2 !g exp.Z/ of infinite
subsets of Z such that Ai C1 Ai for all i 2 !, then there exists an infinite A Z
such that A Ai for all i 2 !.
Proof. Let B0 D A0 and suppose that we have infinite sets B0

Bk such that
Bi Ai and Ai nBi is finite for all i k. Since C D AkC1 nBk .AkC1 nAk / [
.Ak nBk / is finite, for the set BkC1 D AkC1 nC our assumption still holds. Thus we
can construct a sequence fBi W i 2 !g of infinite subsets of Z such that Bi C1 Bi
and Bi Ai for all i 2 !. Take a0 2 B0 arbitrarily and, if we have distinct points
a0 ; : : : ; ak such that ai 2 Bi for each i k, then we can choose a point akC1 2
BkC1 nfa0 ; : : : ; ak g which shows that we can construct a set A D fai W i 2 !g such
that ai aj if i j and ai 2 Bi for all i 2 !. It is straightforward that the set A
is as promised so Fact 1 is proved. t
u
Fact 2. Suppose that we have sequences fAi W i 2 !g and fBi W i 2 !g of infinite
subsets of ! such that Ai C1 Ai for all i 2 ! and Bj Ai for all i; j 2 !.
Then there is an infinite A ! such that Bj A Ai for all i; j 2 !.
Proof. It is easy to see that for any infinite sets P1 ; : : : ; Pn ; Q !, if Pi Q
for all i n, then P1 [

[ Pn Q. Furthermore, if Q Pi for all i n,
T T
then QS fPi W i ng. These observations prove that if Di D ki Ak and
Ei D ki Bi for all i 2 !, then Ei Dj for all i; j 2 !. It is easy to choose a
sequence fmi W i 2 !g ! such that miS< mi C1 and Ei Di \ .mi ; C1/ for all
i 2 !. Finally, observe that the set A D fEi \ .mi ; mi C1 W i 2 !g is as promised
u
Fact 3. Given a Polish space M , a function f W M ! R belongs to B1 .M / if and
only if for any closed F M , the function f jF has a point of continuity in the
space F .
Proof. Assume first that f 2 B1 .M / and fix a sequence ffn W n 2 !g Cp .M /
with fn ! f . Let g D f jF and gn D fn jF for every n 2 !. Since the restriction
map is continuous, we have gn ! g and hence g 2 B1 .F /. This shows that, to
prove necessity, we can assume, without loss of generality, that F D M , i.e., it
suffices to show that every f 2 B1 .M / is continuous at some point of M .
Let B D fOn W n 2 !g be a some base in R. The set f 1 .On / is F in M for
each n 2 ! (see Problem 379) S so fix a family fPmn W m 2 !g of closed subsets
of M such that f 1 .On / D fPmn W m 2 !g. The set Emn D Pmn nInt.Pmn / is
nowhere dense
S in M for all m; n 2 !; since M is complete, we can choose a point
x 2 M n. fEmn W m; n 2 !g/.
To see that f is continuous at x take any " > 0; there is n 2 ! such that f .x/ 2
On .f .x/ "; f .x/ C "/. Since x 2 f 1 .On /, there is m 2 ! such that x 2 Pmn
and hence x 2 Int.Pmn /. Now, the open set V D Int.Pmn / witnesses continuity of
f at the point x because x 2 V and f .V / On .f .x/ "; f .x/ C "/. This
proves necessity.
To prove sufficiency, assume that f 2 RM and, for every closed non-empty
F M , the function f jF has a point of continuity in the space F . It suffices to
show that f 1 .O/ is an F -subset of M for any O 2 .R/ (see Problem 379). To do
this we will prove an auxiliary statement. Say that sets A; B M are F -separated
if there are disjoint A0 ; B 0 M such that A A0 ; B B 0 and A0 ; B 0 2 10 .M /.
Then
(1) if A; B M are not F -separated, then there is a non-empty closed F M
such that A \ F D B \ F D F .
Let F0 D A \ B; then AnF0 A00 D AnF0 , BnF0 B00 D BnF0 , the sets
A00 ; B00
are disjoint and we have A00 [ B00 M nF0 while A00 ; B00 2 10 .M /. Now
assume that 0 < < !1 and we have constructed a family fF W < g of
closed subsets of M and a family fA0 ; B0 W < g 10 .M / with the following
properties:
(2) if 0
< < , then F F
; A0
A0 and B
0 B0 ;
(3) for every ordinal < , we have A0 \ B0 D ; and A0 ; B0 2 10 .M / while
AnF A0 ; BnF B0 and A0 [ B0 M nF ;
(4) if D
C 1 < , then F D A \ F
\ B \ F
.
T S
If is a limit ordinal, then the sets F D fF W < g; A0 D fA0 W
S
< g and B0 D fB0 W < g witness that (2)(4) hold for all . If
D
C 1 then the sets F D A \ F
\ B \ F
; A0 D .A \ F
nF / [ A0
and
B0 D .B \ F
nF / [ B
0 show that we also have (2)(4) for all .
Consequently, we can continue our inductive construction to obtain families
fF W < !1 g and fA0 ; B0 W < !1 g such that (2)(4) hold for all < !1 .
Since M is a second countable space, there is < !1 such that F D F for all
. If F D F D ;, then (3) applied for D implies that A and B are F -
separated which is a contradiction. Thus the set F is non-empty and the property
(4) applied for D C 1, implies that F D A \ F D B \ F , i.e., (1) is proved.
S To finish the proof of our fact take any open O R and represent it as O D
fRn W n 2 !g where Rn is closed in R and Rn RnC1 for all n 2 !. Let
U D f 1 .O/ and Qn D f 1 .Rn / for every n 2 !. Assume first that the sets
M nU and Qn are F -separated for all n 2 !. Then for each n 2 !, there is
Sn 2 S1 .M / such that Qn Sn and Sn \ .M nU / D ;, i.e., Sn U . Therefore
0
U D fSn W n 2 !g is an F -subset of M , i.e., we proved that f 2 B1 .M /.

Now if Qn and M nU are not F -separated for some n 2 !, then we can apply (1)
to find a non-empty closed F M such that Qn \ F D .M nU / \ F D F . By
our assumption, there is a point x 2 F such that f jF is continuous at x. Choose
sequences, fqi W i 2 !g Qn \ F and fri W i 2 !g .M nU / \ F such that
qi ! x and ri ! x. By continuity of f jF , we have f .qi / ! f .x/ and hence
f .x/ 2 Rn because ff .qi / W i 2 !g Rn O and Rn is closed in R.
On the other hand, ff .ri / W i 2 !g RnO and RnO is closed in R which
implies that f .x/ D limi !1 f .ri / 2 RnO; this contradiction shows that f 1 .O/
is an F -subset of M for every open O R, i.e., f 2 B1 .M / so Fact 3 is proved.
t
u
Fact 4. Given a Polish space M , a function f W M ! R belongs to RM nB1 .M / if
and only if there is a countable A M and > 0 such that osc.f; U / > for any
U 2 .A/.
Proof. If such a set A exists, then let F D A and g D f jF . It is straightforward
that osc.g; x/ for any x 2 F and hence g is discontinuous at x. Therefore
f B1 .M / by Fact 3 and hence we proved sufficiency.
Now assume that f 2 RM nB1 .M /. By Fact 3 there exists a closed non-empty
F M such that g D f jF is discontinuous at all points of F . Thus osc.g; x/ > 0
for any x 2 F . IfSFn D fx 2 F W osc.g; x/ nC1 1
g, then Fn is closed in F for
each n 2 ! and n2! Fn D F . The space F being Polish, there is U 2 .F /
with U Fn for some n 2 !. For D nC2 1
we have osc.g; W / nC1 1
>
for any W 2 .U /. Take a base fVi W i 2 !g .U / for the space U and
choose yi ; zi 2 Vi so that jf .yi / f .zi /j nC1
1
for all i 2 !. It is immediate that
A D fyi ; zi W i 2 !g and D nC2 are as promised so we settled necessity. Fact 4 is
1
proved. t
u
Fact 5. Given a set Z and S D ffj W j 2 !g RZ , assume that the sequence S is

pointwise bounded (i.e., the set S.z/ D ffj .z/ W j 2 !g is bounded in R for every
z 2 Z) and has no convergent subsequences (in RZ ). Then there exists an infinite
set A ! and p; q 2 R such that p < q and for every infinite B A, there is
z 2 Z such that the sets P .B; p; z/ D fm 2 B W fm .z/ < pg and Q.B; q; z/ D
fm 2 B W fm .z/ > qg are both infinite.
Proof. For any infinite set A ! and z 2 Z let LA .z/ D infft 2 R W the set
fm 2 A W fm .z/ > tg is finiteg and lA .z/ D supft 2 R W the set fm 2 A W fm .z/ < tg
is finiteg. Observe that LA .z/ and lA .z/ make sense because the set ffj .z/ W j 2 !g
is bounded in R. It is evident that lA .z/ LA .z/ and lA .z/ D LA .z/ if and only if
the sequence ffm .z/ W m 2 Ag is convergent.
Let f.pn ; qn / W n 2 !g be an enumeration of all pairs .p; q/ of rational numbers
such that p < q. Given an infinite A ! and p; q 2 R with p < q say that a
triad .A; p; q/ is adequate if the conclusion of our fact holds for .A; p; q/, i.e., for
any infinite B A there is z 2 Z for which the sets P .B; p; z/ and Q.B; q; z/ are
infinite.
To obtain a contradiction suppose that no triad .A; p; q/ is adequate. Then
.!; p0 ; q0 / is not adequate and therefore we can find an infinite A0 ! such
that P .A0 ; p0 ; z/ and Q.A0 ; q0 ; z/ are not both infinite for any z 2 Z. Assume
that we have constructed infinite sets A0

Ak such that P .Aj ; pi ; z/
and Q.Aj ; qj ; z/ are not both infinite for all z 2 Z and j k. Since the triad
.Ak ; pkC1 ; qkC1 / is not adequate, there exists an infinite AkC1 Ak such that the
sets P .AkC1 ; pkC1 ; z/ and Q.AkC1 ; qkC1 ; z/ are not both infinite for any z 2 Z.
After we have the sequence fAi W i 2 !g apply Fact 1 to find an infinite set
A ! with A Ai for all i 2 !. The sequence ffn W n 2 Ag is not convergent
by our hypothesis, so there exists z 2 Z such that lA .z/ < LA .z/ which implies
that there are p; q 2 Q for which lA .z/ < p < q < LA .z/. We have p D pn and
q D qn for some n 2 ! and hence one of the sets P .An ; pn ; z/ and Q.An ; qn ; z/ is
finite. It follows from A An that one of the sets P .A; pn ; z/ and Q.A; qn ; z/ is
finite which is impossible by the definition of lA .z/ and LA .z/. This contradiction
shows that there exists an infinite A ! and p; q 2 R such that p < q and the triad
.A; p; q/ is adequate. Fact 5 is proved. t
u
Fact 6. Given a Polish space M and F B1 .M / suppose that every infinite A F
has an accumulation point in B1 .M /. Then F is a compact subset of B1 .M / (the
bar denotes the closure in RM ).
Proof. It is evident that F is pseudocompact, so it is compact by Problems 401 and
407 of [TFS]). To see that F B1 .M / assume that there is f 2 F nB1 .M /. Then
there is a countable A M and > 0 such that osc.f; U / > for any U 2 .A/
(see Fact 4). Since RA is second countable, there is a sequence S D ffn W n 2 !g
F such that fn jA ! f jA. By our hypothesis, there is g 2 B1 .M / such that g is
an accumulation point of the sequence S . It is straightforward that gjA D f jA and
hence g B1 .M / by Fact 4; this contradiction proves that f 2 B1 .M /, so the
compact set F is contained in B1 .M / and Fact 6 is proved. t
u
Fact 7. If M is a Polish space and F is a compact subspace of B1 .M /, then F is

sequentially compact, i.e., every sequence in F contains a convergent subsequence.
Proof. Suppose that, on the contrary, there is a sequence S D ffi W i 2 !g F
which has no convergent subsequence. It follows from compactness of F that the
set S.x/ D ffi .x/ W i 2 !g is bounded in R for all x 2 M , so we can apply Fact 5
to find an infinite A ! and p; q 2 R such that the triad .A; p; q/ is adequate, i.e.,
for every infinite B A there is x 2 M for which the sets P .B; p; x/ D fm 2 M W
fm .x/ < pg and Q.B; q; x/ D fm 2 M W fm .x/ > qg are both infinite.
For every B A let K.B/ D fx 2 M W P .B; p; x/ and Q.B; q; x/ are infiniteg.
Then K.B/ ; for every infinite B A and it is evident that B B 0 A
implies K.B/ K.B 0 /. As a consequence,
(5) there is an infinite B A such that K.B/ D K.B 0 / for any infinite B 0 B.
Indeed, if (5) is not true, then let B0 D A and assume that for some < !1 with
> 0, we have constructed a family fB W < g of infinite subsets of A such
that
(6) <
< implies B
B and K.B
/ K.B /.
If is a limit ordinal, choose a sequence fn W n 2 !g for which n < nC1
for all n 2 ! and n ! ; by Fact 1 there exists an infinite B A such that
B Bn for every n 2 !. Now, if < , then there is n 2 ! such that
< n and hence B Bn B ; besides, K.B / K.Bn / K.B / and
K.B / K.Bn / by (6). This proves that B B and K.B / K.B / for all
< . Thus (6) holds for all .
If D
C 1, then (5) does not hold for B
and therefore there exists an infinite
B B
such that K.B / K.B
/. It is immediate that (6) holds for all in
this case as well. Consequently, our inductive procedure can be carried out to obtain
a family fB W < !1 g of infinite subsets of A such that (6) holds for all < !1 .
Then fK.B / W < !1 g is a strictly decreasing family of non-empty closed subsets
of M . Since such a family cannot exist in a second countable space, we have a
contradiction which proves (5).
Now fix an infinite B A such that (5) holds for B and let K D K.B/. If
U 2 .K/ and C is an infinite subset of B, then K.C / D K and hence there is
x 2 U \ K.C /. Since P .C; p; x/ is infinite, there is an infinite C 0 C such that
the sequence S C 0 ; x D ffm .x/ W m 2 C 0 g converges to some rx p. We also
have K.C 0 / D K, so there is y 2 U \ K.C 0 /; the set Q.C 0 ; q; y/ being infinite,
there is an infinite C 00 C 0 such that the sequence S C 00 ; y converges to some
ry q. Thus
(7) for any infinite C B and U 2 .K/, there is an infinite C 00 C such that
for some points x; y 2 U , the sequences S C 00 ; x and S C 00 ; y converge to rx
and ry respectively while rx p and ry q.
Pick a base fUi W i 2 !g .K/ of the space K. Let C0 D B; suppose that
we have infinite sets C0

Ck such that for any i k there are ri p and
si q for which S Ci ; xi ! ri and S Ci ; yi ! si for some xi ; yi 2 Ui . Now

apply (7) to C D Ck to choose an infinite set CkC1 Ck such that there are points
xkC1 ; ykC1 2 UkC1 for which S CkC1 ; xkC1 ! rkC1 and S CkC1 ; ykC1 ! skC1
where rkC1 p and skC1 q. This shows that we can inductively construct a
decreasing sequence fCi W i 2 !g of infinite subsets of B as well as sequences
fri ; si W i 2 !g R and fxi ; yi W i 2 !g M such that ri p; si q; fxi ; yi g
Ui while S Ci ; xi ! ri and S Ci ; yi ! si for every i 2 !.
Now apply Fact 1 again to find an infinite C B such that C Ci for all
i 2 !. If L D fxi ; yi W i 2 !g, then it is straightforward that S C; xi ! ri p
and S C; yi ! si q for every i 2 !. Define a function f 2 RM as follows:
f .xi / D ri and f .yi / D si for all i 2 !; if x 2 M nL, then let f .x/ D 0. Then, for
every W 2 .L/, we have Ui \ L W for some i 2 ! and hence fxi ; yi g W
which implies osc.f; W / jf .xi / f .yi /j D q p > 0. This shows that
f 2 RM nB1 .M / by Fact 4. By compactness of F the sequence ffm W m 2 C g
has an accumulation point g 2 F B1 .M /. However, gjL D f jL and hence we
can apply Fact 4 to the function g to conclude that g B1 .M /. This contradiction
shows that S must contain a convergent subsequence and hence Fact 7 is proved.
t
u
Fact 8. Let Z be a space; if m 2 N and f 2 RZ , let 'm .f /.z/ D jf .z1 /j C

C
m
jf .zm /j for any z D .z1 ; : : : ; zm / 2 Z m . Then 'm W RZ ! RZ is a continuous map
such that 'm .Cp .Z// Cp .Z / and 'm .B1 .Z// B1 .Z /.
m m
Proof. For every i 2 f1; : : : ; mg let i W Z m ! Z be the natural projection of

Z m onto its i th factor. Then 'm .f / D imD1 jf i j which shows that 'm .f /
is a continuous function on Z m if f is continuous on Z. For each z 2 Z
let pz W RZ ! R be the natural projection onto the factor determined by z.
m
Analogously, if w 2 Z m , let qw W RZ ! R be the natural projection onto the
factor determined by w. To prove continuity of 'm it suffices to show that qw 'm
is continuous for any w D .w1 P ; : : : ; wm / 2 Z m (see Problem 102 of [TFS]). Now,
m
qw .'m .f // P D 'm .f /.w/ D i D1 jf .wi /j for any f 2 R which shows that
Z
m
qw 'm D i D1 jpwi j is a continuous map because each pwi is continuous. This
proves continuity of 'm .
Finally, if f 2 B1 .Z/, then there is a sequence ffn W n 2 !g Cp .Z/ with
fn ! f . Then f'm .fn / W n 2 !g Cp .Z m / and 'm .fn / ! 'm .f / by continuity
of 'm . Thus 'm .f / 2 B1 .Z m / and therefore 'm .B1 .Z// B1 .Z m /. Fact 8 is
proved. t
u
Fact 9. Suppose that M is a Polish space and F is a compact subspace of B1 .M /.
Define u 2 RM by u.x/ D 0 for all x 2 M and assume that u 2 H where H F
and h.x/ 0 for any x 2 M and h 2 H . Then, for any " > 0, there is a countable
G H such that inffg.x/ W g 2 Gg < " for all x 2 M .
Proof. If this is false, then the set M.G/ D fx 2 M W inffg.x/ W g 2 Gg "g is
non-empty for every countable non-empty G H . Observe that G G 0 implies
M.G/ M.G 0 /. As a consequence, there exists a countable non-empty G H
such that
(8) M.G/ D M.G 0 / for any countable G 0 G.

Indeed, if such a G does not exist, then it is easy to construct by transfinite
induction a family fG W < !1 g of countable subsets of H in such a way that
< implies G G and M.G / M.G /. Therefore fM.G / W < !1 g
is a strictly decreasing !1 -sequence of non-empty closed subsets of M which is a
contradiction.
Now choose a non-empty countable G H such that (8) holds and take a
countable dense subset D of the space K D M.G/. Since u 2 H , there exists a
sequence S D ffi W i 2 !g H such that fi .d / ! 0 for every d 2 D. If g is an
accumulation point of the sequence S , then g.d / D 0 for every d 2 D. On the other
hand, if G 0 D G [ S , then M.S / M.G 0 / and hence g.x/ " for all x 2 M.G 0 /.
However, M.G 0 / is also dense in K by (8) which shows that gjK has no points of
continuity and therefore g B1 .M / by Fact 3. This contradiction shows that Fact 9
is proved. t
u
Fact 10. Let M be a Polish space. Then t.F / ! for any compact F B1 .M /.
Proof. Define u 2 RM by u.x/ D 0 for all x 2 M . Suppose that A F and f 2 A.
Define a map Lf W RM ! RM by Lf .g/ D g f for all g 2 RM . Then Lf is
a homeomorphism because RM is a linear topological space (see Problems 115 and
116 of [TFS] together with Fact 1 of S.491). Then Lf .F / is homeomorphic to F ,
and for any B A, we have u 2 Lf .B/ if and only if f 2 B. Thus we can assume,
without loss of generality, that f D u.
For any m 2 N and f 2 RM , let 'm .f /.z/ D jf .z1 /j C

C jf .zm /j for any
m
point z D .z1 ; : : : ; zm / 2 M m . Then 'm W RM ! RM is a continuous map by
Fact 8. If um is the function which is identically zero on M m , then um D 'm .u/ and
um 2 'm .A/. The space M m is Polish and 'm .A/ is contained in a compact subspace
'm .F / of B1 .M m /. Since also 'm .f / is non-negative for any f 2 A, we can apply
Fact 9 to find a countable Bm A such that inff'm .f /.z/ W f 2 Bm g < m1 for any
z 2 M m and therefore
(9) for any z1 ; : : : ; zm 2 M , there is f 2 Bm such that jf .z1 /j C

C jf .zm /j < m1
and hence jf .zi /j < m1 for all i m.
S
Consider the set B D fAm W m 2 !g; then B is a countable subset of A. To
see that u 2 B, take any U 2 .u; F /. There exist a set P D fx1 ; : : : ; xk g M
and " > 0 such that u 2 V U where V D ff 2 F W jf .xi /j < " for all i kg.
Choose m 2 N with k m and m1 < "; let z D .z1 ; : : : ; zm / 2 M m be a point with
fxi W i kg fzj W j mg. By (9), there is f 2 Bm such that jf .zj /j < m1 < "
for all j m and hence jf .xi /j < " for all i k, i.e., f 2 V \ B U \ B. This
proves that U \ B ; for any U 2 .u; F / and therefore u 2 B. Thus t.F / !
u
Fact 11. Given a Polish space M let u.x/ D 0 for all x 2 M . Assume that S D
ffn W n 2 !g Cp .M /nfug and u is an accumulation point of the sequence S .
Assume also that for every infinite A !, there is an infinite B A such that
S B D ffn W n 2 Bg is a convergent sequence (in RM ). Then there is an infinite

E ! such that S E is a convergent sequence and S E ! u.
Proof. Observe first that every infinite subset of S has an accumulation point in
B1 .M / and therefore T D S is a compact subset of B1 .M / by Fact 6 (the bar
denotes the closure in RM ). Say that a set A ! is big if u is an accumulation
point of S A . It is clear that every big set is infinite. If A !, then the expression
S A ! f says that S A is a convergent sequence and the limit of S A is f . We
claim that
(10) if fBn W n 2 !g exp.!/ where Bn is big and BnC1 Bn for all n 2 !,
then there is a big B ! such that B Bn for all n 2 !.
Let A0 D B0 and suppose that we have infinite sets A0

Ak such that
Ai Bi and Bi nAi are finite for all i k. Since C D BkC1 nAk .BkC1 nBk / [
.Bk nAk / is finite, for the set AkC1 D BkC1 nC our assumption still holds. Thus we
can construct a sequence fAi W i 2 !g ! such that Ai C1 Ai ; Ai Bi and
jBi nAi j < ! for all i 2 !. It is evident that taking away finitely many points from
a big set, we still obtain a big set, so Ai is big for every i 2 !.
Let F D ff 2 T W there is an infinite A ! such that A An for all n 2 !
and S A ! f g. Then u 2 F . Indeed, it suffices to show that for any finite P M
and " > 0, there is f 2 F such that jf .x/j < " for all x 2 P . So, fix a finite
P M and " > 0. Since every An is big, the set A0n D fm 2 An W jfm .x/j < 2" for
all x 2 P g is also big for any n 2 !, so we can choose a sequence fmi W i 2 !g !
such that mi 2 A0i and mi < mi C1 for every i 2 !. By our hypothesis, there is an
infinite A fmi W i 2 !g and f 2 T such that S A ! f . It is immediate that
A An (and hence A Bn ) for all n 2 ! so f 2 F . Besides, jf .x/j 2" < "
for all x 2 P which proves that u 2 F .
Since F is a subspace of a compact T B1 .M /, we can apply Fact 10 to find
a countable G F such that u 2 G. For every g 2 G, there is an infinite Bg !
such that Bg An (and hence Bg Bn ) for all n 2 ! and S Bg ! g. By
Fact 2, there is an infinite B ! such that Bg B for all g 2 G and B Bn
for all n 2 !. Thus g 2 S B for every g 2 G and therefore u 2 G S B which
shows that B is the promised big set and hence (10) is proved.
For any infinite set A ! and any point z 2 M , let LA .z/ D infft 2 R W the
set fm 2 A W jfm .z/j > tg is finiteg and lA .z/ D supft 2 R W the set fm 2 A W
jfm .z/j < tg is finiteg. Observe that LA .z/ and lA .z/ make sense because it follows
from compactness of T that the set ffj .z/ W j 2 !g is bounded in R.
Now, fix " > 0 and, for any infinite A !, let QA D fx 2 M W LA .x/ > "g. It
is clear that for any infinite A; B ! such that A B, we have QA QB . As a
consequence,
(11) for any big set A !, there is a minimal big set B A in the sense that
QC D QB for any big set C B.
Indeed, if (11) is not true, then let A0 A be a big set which witnesses that it
is false. Suppose that < !1 and we have a family fA W < g of big sets such
that <
< implies A
A and QA
QA . If is a limit ordinal, then
choose a sequence fn W n 2 !g such that n < nC1 for all n 2 ! and n ! . It
follows from (10) that there is a big set A ! such that A An for all n 2 !.
If < , then there is n 2 ! such that < n and hence A An A .
Furthermore, QAn QA and QAn QA which implies QA QA , so our
inductive assumption holds for all .
Now, if D
C 1, then it follows from A
A that there is a big set A A
such that QA QA
. It is clear that our inductive assumption is also fulfilled for
all and hence we can construct an !1 -sequence fA W < !1 g such that
fQA W < !1 g is a strictly decreasing !1 -sequence of closed subsets of M which
is a contradiction. Thus (11) is true. Our next step is to show that
(12) if A is a minimal big set, then Q D QA D ;, i.e., LA .x/ " for every x 2 X .
To prove the property (12), assume that Q ; and take a countable dense set
Y D fyi W i 2 !g Q (repetitions are possible in the enumeration of Y to cover the
case when Y is finite). Let us formulate explicitly an evident property, we already
used in the proof of (10).
(13) If B is a big set, then, for any finite K M and > 0, the set B 0 D fm 2 B W
jfm .x/j < for all x 2 Kg is also big.
Using (13) it is easy to choose big sets fBn W n 2 !g exp.A/ in such a way
that BnC1 Bn for all n 2 ! and jfm .yi /j < n1 for all m 2 Bn and i n.
Now apply (10) to find a big set B ! such that B Bn for all n 2 !. It is
straightforward that
. / if f is an accumulation point of S B , then f .y/ D 0 for every y 2 Y .
Choose a complete metric d0 on the space M with .d0 / D .M /. Then the
space Q with the metric d D d0 j.Q Q/ is also complete. Observe that given U 2
.Q/ and a big set C B, the set C 0 D fm 2 C W jfm .u/j > " for some u 2 U g
is also big because otherwise C0 D C nC 0 is big and jfm .u/j " for all u 2 U and
m 2 C0 ; since C0 B, we have QC0 D Q, i.e., the set fx 2 Q W LC0 .x/j > "g is
dense in Q and therefore there is m 2 C0 with jfm .x/j > " for some x 2 U which
is a contradiction. An evident induction shows that if k 2 N then
(14) for any U1 ; : : : ; Uk 2 .Q/ the set C D fm 2 B W for any i k we have
jfm .x/j > " for some x 2 Ui g is big.
Fix a base fOi W i 2 !g .Q/ of the space Q. We will construct a family
U D fU.i; j / W i 2 !; i j < !g .Q/ and a sequence fmi W i 2 !g B
(15) U.i; i / D Oi for all i 2 !;
(16) U.i; j C 1/ U.i; j / for any i 2 ! and j i ;
(17) diam.U.i; j // j1 for every i 2 ! and j i C 1;
(18) mj C1 > mi for all j 2 !;

(19) if i 2 ! and j > i , then jfmj .u/j > " for any u 2 U.i; j /.
To start the construction pick an arbitrary m0 2 B and let U.0; 0/ D O0 . Assume
that n 2 ! and we have the numbers m0 ; : : : ; mn 2 B and sets U.i; j / for all i n
and i j n such that (15)(19) are fulfilled where applicable. To satisfy (15)
let U.n C 1; n C 1/ D OnC1 ; the property (14) implies that the set C D fm 2 B W
for any i 2 f0; : : : ; ng we have jfm .x/j > " for some x 2 U.i; n/g is big and
hence there exists mnC1 > mn such that mnC1 2 C . Fix i 2 f0; : : : ; ng; the choice
of C implies that the set fx 2 U.i; n/ W jfmnC1 .x/j > "g is non-empty. Since
the function fmnC1 is continuous, there is a set U.i; n C 1/ 2 .Q/ such that
U.i; n C 1/ U.i; n/; diam.U.i; n C 1// nC1 1
and jfmnC1 .x/j > " for any
x 2 U.i; n C 1/. This completes the inductive construction of the family U. Let
D D fmi W i 2 !g.
It followsTfrom the completeness of T .Q; d / that for each i 2 !, we can choose
a point zi 2 fU.i; j / W j 2 !ni g D fU.i; j / W j 2 !ni g Oi . Since the set
Z D fzi W i 2 !g intersects Oi for every i 2 !, it is dense in Q. It follows from (19)
that fmj .zi / > " for all j > i and hence lD .zi / " for all i 2 !. Since S D T
and T is compact, we can take an accumulation point f 2 B1 .M / of the set S D .
It is immediate that jf .z/j " for all z 2 Z. Clearly f is also an accumulation
point of S B so f .y/ D 0 for all y 2 Y by . /. Thus we have found two dense
sets Y and Z in the space Q such that f jY is identically zero and jf .z/j " for all
z 2 Z. Therefore f has no point of continuity on Q which contradicts Fact 3. This
contradiction shows that (12) is true.
Finally, it follows from the properties (11) and (12) that we can construct a
sequence fAn W n 2 !g of big sets such that AnC1 An and LAn .x/ n1 for
all x 2 X . Choose a sequence J D fki W i 2 !g ! such that ki < ki C1 and
ki 2 Ai for all i 2 !. It is easy to see that S J ! u and hence Fact 11 is proved.
t
u
Fact 12. Suppose that M is a Polish space and N is a second countable space. Let
f W M ! N be a measurable map, i.e., f 1 .A/ 2 B.M / for any A 2 B.N /.
Then there is a Polish topology on the set M such that .M / and the map
f W .M; / ! N is continuous.
Proof. Let B be a countable base in N . For every B 2 B, consider the space MB
whose underlying set is M and the topology .MB / is generated by the family
B D .M / [ ff 1 .B/; M nf 1 .B/g. It is evident that MB is homeomorphic to
the space f 1 .B/ .M nf 1 .B// and therefore MB is a Borel set. Now consider
0
the space MS whose underlying set is M and the topology .M 0 / is generated by
0
the family
Q fB W B 2 Bg. Then M is homeomorphic to a closed subspace of the
space fMB W B 2 Bg by Fact 2 of T.363 and hence M 0 is a Borel set by Problems
333 and 331. There exists a Polish topology on the set M such that .M 0 / by
Problem 341. We have .M / .M 0 / ; besides, f 1 .B/ 2 .MB / .M 0 /
for every B 2 B which shows that f 1 .B/ is open in .M; / for every B 2 B and
hence the map f W .M; / ! N is continuous. Fact 12 is proved. t
u
Now it is easy to finish our solution. Suppose that M is a Polish space and X
B1 .M / is compact; denote by u the function which is identically zero on M . Take
any A X and f 2 AnA. Define a map Lf W RM ! RM by Lf .g/ D g f for
all g 2 RM . Then Lf is a homeomorphism because RM is a linear topological space
(see Problems 115 and 116 of [TFS] together with Fact 1 of S.491). Then Lf .X / is
homeomorphic to X , and for any sequence ffn W n 2 !g A, we have fn ! f
if and only if Lf .fn / ! u. Thus we can assume, without loss of generality, that
f D u.
Since t.X / ! by Fact 10, there is a sequence S D ffn W n 2 !g A
such that u is an accumulation point of S . There is a Polish topology on M
such that fn is continuous on M 0 D .M; / for all n 2 ! (see Fact 12). Thus
Cp .M / Cp .M 0 / RM and S Cp .M 0 /. It is clear that B1 .M / B1 .M 0 / and
hence S is contained in a compact space X which in turn is a subspace of B1 .M 0 /.
Thus X is sequentially compact by Fact 7 and therefore every infinite subset of
S contains a convergent subsequence. We proved that S Cp .M 0 / satisfies all
premises of Fact 11, so we can apply it to conclude that some subsequence S 0 of
the sequence S converges to u. Since S 0 A, we proved that there is a sequence in
A which converges to u D f and hence X is FrchetUrysohn, i.e., our solution is
complete.
T.385. Let X be a separable compact space. Prove that X is Rosenthal compact if
and only if, for any dense countable A X , the space Cp .AjX / is analytic.
Solution. Given a space Z, a map f W Z ! R is measurable if f 1 .P / 2 B.Z/
for any P 2 B.R/; denote by B.Z/ the set of all measurable real-valued functions
on Z. If A Z, then A is a characteristic function of A defined by A .A/ f1g
and A .ZnA/ f0g; besides, for any f 2 RZ , the function g D 1 f 2 RZ
is defined by g.x/ D 1 f .x/ for all x 2 Z. If A RZ , then let L0 .A/ D A;
if < !1 and we have fL .A/ W < g, let L .A/ D ff 2 RZ W there are
sequences fn W n 2 !g and ffn W n 2 !g RZ such that fn 2 Ln .A/
2 ! and fn ! f g. This gives us a family of sets fL .A/ W < !1 g; let
for all n S
L.A/ D fL .A/ W < !1 g.
Fact 1. If M is a metrizable space, then B.M / D L.Cp .M //.
Proof. It follows from Fact 1 of T.363 that L0 D L0 .Cp .M // D Cp .M / B.M /.
Suppose that 0 < < !1 and we proved that fL W < g B.M / where L D
L .Cp .M // for all < . If f 2 L D L .Cp .M //, then there are sequences
fn W n 2 !g and ffn W n 2 !g RM such that fn 2 Ln for all n 2 ! and
fn ! f . Given U 2 .R/ it is easy to S find a sequence fUn W n 2 !g .R/
such that U nS UnC1T for all n 2 ! and n2! Un D U . It is straightforward that
f 1 .U / D n2! kn fk1 .Un / and therefore f 1 .U / is a Borel subset of M
because so is fk1 .Un / for all k; n 2 ! by the induction hypothesis.
Thus we proved that f 1 .U / 2 B.M / for every U 2 .R/, so f is measurable
by Fact 1 of T.363. Therefore L .Cp .M // B.M / for all < !1 and hence
L.Cp .M // B.M /.
To prove that B.M / L D L.Cp .M //, let L D L .Cp .M // for all < !1 .
Our first step is to check that for any A B.M /, the function A belongs to L. We
will need the following properties of L:
(1) if f; g 2 L, then f
g 2 L and af C bg 2 L for any a; b 2 R;
(2) if ffn W n 2 !g L and fn ! f , then f 2 L;
(3) if A exp.M / and A 2 L for any A 2 A, then fM nA ; A[B ; A\B ; AnB g
L for any A; B 2 A.
The property (1) can be proved by an evident transfinite induction; we leave it as
an exercise to the reader. To see that (2) is true, choose n < !1 such that fn 2 Ln
for all n 2 ! and observe that if D supfn W n 2 !g C 1, then f 2 L .
Now, for any A; B 2 A, we have M nA D 1 A ; A\B D A
B and
AnB D .1 B /
A . Furthermore, A[B D 1 .1 A /
.1 B / so (1) implies
that (3) holds.
If A is open, then there is a sequenceS fAn W n 2 !g of closed subsets of M such
that An AnC1 for every n 2 ! and n2! An D A. By normality of M there is a
continuous fn W M ! 0; 1 such that fn .An / f1g and fn .M nA/ f0g for all
n 2 !. After a minutes reflection, one can see that fn ! A . Now, if A is closed in
M , then M nA is open in M and hence M nA 2 L; now, (3) implies A 2 L. Thus
A 2 L for every A 2 00 .M / [ 00 .M /.
Assume that 0 < < !1 and we have established that A 2 L whenever
A 2 0 .M / [ 0 .M / for some < . If A 2 0 .M /, then there is a sequence
S
fn W n 2 !g and An 2 S 0n .M / for each n 2 ! such that n2! An D A.
Let B0 D A0 and BS n D An n. i <n Ai / for all n 1. The family fBn W n 2 !g
is disjoint and A D n2! B Pn . It follows from (3) and the induction hypothesis that
Bn 2 L and hence gn D niD0 Bi 2 L for any n 2 !. Besides, gn ! A so (2)
implies that A 2 L for any A 2 0 .M /.
Finally, observe that if A 2 0 .M /, then M nA 2 0 .M / and hence M nA 2 L,
so it follows from (3) that A 2 L. Thus our inductive procedure can be continued
to show that A 2 L for every A 2 B.M /; an evident consequence is
(4) if r1 ; : : : ; rk 2 R and A1 ; : : : ; Ak 2 B.M /, then r1 A1 C : : : C rk Ak 2 L.
Now take an arbitrary f 2 B.M /. Letting fC D jf jCf 2 and f D jf jf 2 we
have f D fC f where fC .x/ 0 and f .x/ 0 for any x 2 M . The property
(1) shows that it is sufficient to prove that fC ; f 2 L, i.e., we can assume, without
loss of generality, that f .x/ 0 for any x 2 M .
For every n 2 N, let kn D n
2n and for each i kn consider the number li D 2in .
Then fl0 ; : : : ; lkn g divide 0; n into intervals of length 2n . Let S
Ai D li ; li C1 / for
all i < kn ; then the family fAi W i < kn g is disjoint and 0; n/ D fAi W i < kn g. If
Akn D n; 1/ and Bi D f 1 .Ai / for all i kn , then fBi W i kn g is a partition of
the space M ; note that this partition consists of Borel subsets of M because every
interval is a Borel subset of R. If i < kn , then f .Bi / li ; li C1 /, so for the function
gi D li
Bi , we have jf .x/ gi .x/j < 2n for any x 2 Bi . As a consequence, if
fn D g0 C

Cgkn 1 , then jf .x/ fn .x/j < 2n for any x 2 Mn D B0 [

[Bkn 1 .
S
It is clear that M D fMn W n 2 Ng and Mn MnC1 for all n 2 N, so fn ! f .
The property (4) implies that fn 2 L for every n 2 N, so f 2 L by (2) and hence
Fact 1 is proved. t
u
Fact 2. Given a Cech-complete space Z suppose that A is a family of pairs of open
subsets of Z, i.e., A .Z/ .Z/. Assume that there is a non-empty Y Z such
that A is weakly dense on Y in the sense that for any G0 ; : : : ; Gn 2 .Y /, there is
.U; V / 2 A such that U \ Gi ; V \ Gi for all i n. Then there is a family
A0 D f.Hi0 ; Hi1 / W i 2 !g A and a compact K Z such that A0 is independent
T s.i /
on K, i.e., for every s 2 D! we have fHi W i 2 !g \ K ;.
T a family fOn W n 2 !g .Z/ such that OnC1 On for all n 2 !

Proof. Choose
and Z D fOn W n 2 !g. If B D f.U; V / 2 .Z/.Z/ W .U \Z; V \Z/ 2 Ag,
then B is also weakly dense on
SY .
Let DS 0
D f;g and Dn D fDi W i ng for all n 2 N; we will also need the set
D<! D fDn W n 2 !g. If n 2 !; j 2 D and s 2 Dn , then s _ j 2 DnC1 is defined
by .s _ j /jn D s and s.n/ D j .
Our purpose is to construct inductively families f.Gi0 ; Gi1 / W i 2 !g B and
fWs W s 2 D<! g .Z/ with the following properties (the bar denotes the
closure in Z):
(5) Ws \ Y ; for any s 2 D<! ;
T s.i /
(6) if k 2 !, then W s . fGi W i kg/ \ Ok for every s 2 DkC1 ;
(7) if s; t 2 D<! and t s, then W s Wt .
To start off, let W; D Z and observe that the family B being dense on Y , there
is .G00 ; G01 / 2 B such that G00 \ Y ; G01 \ Y . We have D1 D fs0 ; s1 g where
si .0/ D i for each i 2 D. It is easy to find Wsi 2 .Z/ such that Wsi \ Y ;
and W si O0 \ G0i for every i 2 D. Now assume that n 2 N and we have families
f.Gi0 ; Gi1 / W i < ng and fWs W s 2 Dn g such that (5) is true for all s 2 Dn , the
property (6) holds for all k < n and (7) is fulfilled for all s; t 2 Dn such that t s.
It follows from the weak density of B on Y that there is .Gn0 ; Gn1 / 2 B such that
j
Gn \ Ws \ Y ; for any j 2 D and s 2 Dn . For any j 2 D we can find a set
Ws _ j 2 .Z/ such that Ws _ j \ Y ; and W s _ j Ws \ Gn \ On . Once we
j
have Ws _ j for all s 2 D and j 2 D, we obtain a family fWs W s 2 DnC1 g such that
n
(5) is true for all s 2 D nC1 , the property (6) holds for all k n and (7) is fulfilled
for all s; t 2 DnC1 such that t s.
Consequently, our inductive procedure can be continued to provide families
f.Gi0 ; Gi1 / W i 2 !g B and fWs W s 2 D<! g .Z/ forSwhich the properties
(5)(7) hold. The property (6) implies that the set Kn T D fW s W s 2 Dn g is
compact and KnC1 On for all n T 2 !. Therefore K D fKn W n 2 !g is a non-
empty compact subset of the space n2! On D Z. Let Hki D Gki \ Z for all k 2 !
and i 2 D. To see that the family A0 D f.Hi0 ; Hi1 / W i 2 !g A is independent
of K, take any s 2 D! . For every n 2 ! the set Fn D W sjn is closed in Z and
non-empty because it contains Wsjn \ Y ; [see (5)]. The property (7) implies that
T
Fn FnC1 for all n 2 ! and therefore F D n2! Fn ;. Finally observe that
T s.i / T s.i /
F . fHi W i 2 !g/ \ K by (6), so . fHi W i 2 !g/ \ K ; and Fact 2 is
proved. t
u
Fact 3. If M is a Polish space and f 2 RM nB1 .M /, then there is a closed F M
and p; q 2 R such that p < q and there are A; B F for which A D B D F while
f .x/ p for all x 2 A and f .x/ q for every x 2 B.
Proof. There exists a closed P M such that f jP is discontinuous at every x 2 P
(see Fact 3 of T.384). Let f.pn ; qn / W n 2 !g be an enumeration of all pairs .p; q/
of rational numbers such that p < q. It is an easy exercise to see that for any x 2 P
there are ax ; bx 2 R such that ax < bx and for everySU 2 .x; P / there are y; z 2 U
with f .y/ ax and f .z/ bx . Therefore P D n2! Pn where Pn D fx 2 P W
for every U 2 .x; P / there are y; z 2 U with f .y/ pn and f .z/ qn g. It
is straightforward that Pn is closed in P for every n 2 !; since P is Polish, there
is m 2 ! such that U D IntP .Pm / ;. Fix a base fWn W n 2 !g .P / in
the space U and let p D pm ; q D qm . By definition of U , there are xk ; yk 2 Wk
such that f .xk / p and f .yk / q for all k 2 !. It is immediate that the sets
A D fxk W k 2 !g; B D fyk W k 2 !g and F D A D B are as promised so Fact 3
is proved. t
u
Fact 4. Given a Polish space M suppose that E Cp .M / is a countable set and
K D E is compact (the bar denotes the closure in RM ). If the set K is not contained
in B1 .M /, then ! embeds in K.
Proof. If every infinite E 0 E has an accumulation point in the space B1 .M /, then
E B1 .M / by Fact 6 of T.384, so there is an infinite D E such that D has no
accumulation points in B1 .M /; in particular, D is a discrete subspace of Cp .M /.
On the other hand, the set D is compact, so we can fix a function f 2 DnB1 .M /.
By Fact 3, we can find a non-empty closed F M and p; q 2 R such that p < q
and there are A; B F with A D B D F for which f .x/ p for every x 2 A
and f .x/ q for all x 2 B. Take r; t 2 R with p < r < t < q and consider
the family A D f.Uh0 ; Uh1 / W h 2 Dg where Uh0 D fx 2 F W h.x/ < rg and
Uh1 D fx 2 F W h.x/ > tg are open subsets of F for any h 2 D. Assume that
W1 ; : : : ; Wn 2 .F / and take points xi 2 A \ Wi and yi 2 B \ Wi for all i n.
The set O D fg 2 RM W g.xi / < r and g.yi / > t for all i ng is open in
RM and f 2 O. Since f 2 D, there is h 2 D such that h 2 O. Consequently,
xi 2 Uh0 \ Wi and yi 2 Uh1 \ Wi for all i n which shows that the family
A is weakly dense on F (see Fact 2 for the definition of weakly dense families).
Therefore we can apply Fact 2 to find a set H D fhn W n 2 !g D such that the
family A0 D f.Uh0n ; Uh1n / W n 2 !g is independent on some compact Q F . We
claim that H is homeomorphic to !.
To prove it take any disjoint sets R; S H and choose a point s 2 D! such that
R fhn W n 2 s 1 .0/g and S fhn W n 2 s 1 .1/g. By independence of the family
T s.i /
A0 on F , there is a point z 2 fUhi W i 2 !g. If h D hn 2 R then s.n/ D 0 and
hence z 2 Uh0n , i.e., h.z/ D hn .z/ < r. This implies g.z/ r for any g 2 R.
Now, if h D hn 2 S , then s.n/ D 1 and hence z 2 Uh1n , i.e., h.z/ D hn .z/ > t,
so g.z/ t for any g 2 S . As a consequence, R \ S D ;, i.e., we proved that
H is a compact extension of a countable discrete space H for which R \ S D ;
for any disjoint R; S H . Now apply Fact 2 of S.286 to conclude that H K is
homeomorphic to !. Fact 4 is proved. t
u
Fact 5. Suppose that M is a Polish space and E Cp .M / is a countable set such
that K D E is compact (the bar denotes the closure in RM ) and K B.M /. Then
K B1 .M / and hence K is Rosenthal compact.
Proof. Given a set A let P! .A/ D fB A W jBj !g. It is evident that if
jAj c, then jP! .A/j c. It follows from nw.Cp .M // D ! that jCp .M /j c.
Now,
S if < !1 and we proved that jL .Cp .M //j c, then for the set L0 D
fL .Cp .M // W < g, we have jL0 j c. Since the set L .Cp .M // consists of
limits of sequences from L0 , we have jL .Cp .M //j jP! .L0 /j S
c. This proves
that jL .Cp .M //j c for each < !1 and hence jB.M /j D j fL .Cp .M // W
< !1 gj c
!1 D c (see Fact 1).
Now, if K is not contained in B1 .M /, then ! embeds in K by Fact 4 and hence
jKj j!j D 2c (see Problem 368 of [TFS]). On the other hand, jKj jB.M /j
c which is a contradiction showing that K B1 .M / and hence Fact 5 is proved. u t
Returning to our solution assume that X is a separable Rosenthal compact and
hence there is a Polish space N such that X B1 .N /. Fix an arbitrary dense
countable A X . By Fact 12 of T.384, there is a Polish topology on the set N
such that .N / and every f 2 A is continuous on the space M D .N; /. It
is evident that we still have X B1 .M / while A Cp .M /. Let C D Cp .X / and
CB D Cp .BjX / for any B X . We claim that
(8) a function ' W X ! R belongs to C if and only if for any " > 0 there is n 2 N
and p D .p1 ; : : : ; pn / 2 M n such that for any x; y 2 X with jx.pi / y.pi /j <
n for all i n, we have j'.x/ '.y/j < ".
1
If x 2 X is fixed, then the set U D fy 2 X W jx.pi / y.pi /j < n1 for all i ng

is an open neighborhood of x in X such that '.U / .'.x/ "; '.x/ C "/, i.e., U
witnesses continuity of ' at the point x. Therefore any ' 2 RX which satisfies (8)
is continuous on X .
Let us prove that any function ' 2 Cp .X / satisfies (8). For any n 2 N; > 0 and
p D .p1 ; : : : ; pn / 2 M n , let O.x; p; / D fy 2 X W jy.pi / x.pi /j < for all
i ng. For any x 2 X the family fO.x; p; / W p 2 M n for some n 2 N and > 0g
is a local base of X at the point x. Thus there exist px D .p1x ; : : : ; pn.x/x
/ 2 M n.x/
and x > 0 such that j'.y/ '.x/j < "=2 for Severy y 2 O.x; px ; x /. Since X
is compact, there is a finite E X such that fO.x; px ; x =3/ W x 2 Eg D X .
It is easy to find a finite set D D fp1 ; : : : ; pn g M such that n1 < minfx =3 W
x 2 Eg and pix 2 D for all x 2 E and i n.x/. Now assume that u; v 2 X and
ju.pi / v.pi /j < n1 for all i n. There is x 2 E such that u 2 O.x; px ; x =3/. Then
jv.pix / x.pix /j jv.pix / u.pix /j C ju.pix / x.pix /j n1 C 3x < x and therefore
v 2 O.x; px ; x / which shows that j'.u/ '.v/j j'.u/ '.x/j C j'.x/ '.v/j <
"=2 C "=2 D ". Thus the point p D .p1 ; : : : ; pn / witnesses that (8) holds for the
function ' and hence (8) is proved.
Now we are ready to characterize the functions from CA . It turns out that
(9) a function ' W A ! R belongs to CA if and only if for any " > 0 there is n 2 N
and p D .p1 ; : : : ; pn / 2 M n such that for any x; y 2 A with jx.pi / y.pi /j <
n for all i n, we have j'.x/ '.y/j < ".
1
If ' 2 C , then ' satisfies (8); it is immediate that 'jA satisfies (9) (note that (8)
and (9) are identical except that A is substituted everywhere in place of X ).
Now assume that ' W A ! R and ' has the property introduced in (9). For every
x 2 X let osc.'; x/ D inffdiam.'.U \ A/ W U 2 .x; X/g. If " > 0, then (9)
implies that we can choose n 2 N and p D .p1 ; : : : ; pn / 2 M n such that for any
u; v 2 A with ju.pi / v.pi /j < n1 for all i n, we have j'.u/ '.v/j < "=2. The set
U D fy 2 X W jy.pi / x.pi /j < 2n 1
for all i ng is an open neighborhood of x in
X ; if u; v 2 U \ A, then ju.pi / v.pi /j ju.pi / x.pi /j C jx.pi / v.pi /j < 2n
1
C
1
2n D 1
n which shows that j'.u/ '.v/j < "
2 and therefore diam.'.U \A// "
2 < ".
Since x 2 X and " > 0 were chosen arbitrarily, we established that osc.'; x/ D 0
for all x 2 X and hence there is a continuous '1 W X ! R such that '1 jA D ' (see
Fact 3 of T.368). Thus ' 2 CA and (9) is proved.
For all k; n 2 N consider the set R.k; n/ D f.'; p/ 2 RA M n W p D
.p1 ; : : : ; pn / and j'.u/ '.v/j k1 whenever u; v 2 A and ju.pi / v.pi /j < n1 for
all i ng. The set R.k; n/ is closed in RA M n ; to prove it take p D .p1 ; : : : ; pn / 2
M n and ' 2 RA such that .'; p/ 2 .RA M n /nR.k; n/. Then there exist u; v 2 A
such that ju.pi / v.pi /j < n1 for all i n and j'.u/ '.v/j > k1 . Observe that the set
W D f 2 RA W j.u/ .v/j > k1 g is open in RA and ' 2 W . The functions u and v
being continuous on M , the set G D fq D .q1 ; : : : ; qn / 2 M n W ju.qi / v.qi /j < n1
for all i ng is open in M n and p 2 G. It is easy to check that H D W G
is an open neighborhood of .'; p/ in RA M n such that H \ R.n; k/ D ;. This
proves that R.n; k/ is closed in the Polish space RA M n for all n; k 2 N. Let
n W RA M n ! RA be the natural projection.
It turns out that
S T
(10) if S.k/ D fn .R.n; k// W n 2 Ng for all k 2 N, then CA D fS.k/ W k 2
Ng.
Checking the property (10) is a simple application of (9). If ' 2 CA , then, for any
k 2 N, there exists n 2 N and p D .p1 ; : : : ; pn / 2 M n such that j'.u/ '.v/j < k1
whenever ju.pi / v.pi /j < n1 for all i T
n. Therefore .'; p/ 2 R.n; k/ and hence
' 2 nT.R.n; k// which shows that ' 2 fS.k/ W k 2 Ng. This proves the inclusion
CA fS.k/ WT k 2 Ng.
Now if ' 2 fS.k/ W k 2 Ng, then take " > 0; there exists k 2 N for which
1
k
< ". Since ' 2 n .R.n; k// for some n 2 N, there is p D .p1 ; : : : ; pn / 2 M n
such that .'; p/ 2 R.n; k/ and therefore j'.u/ '.v/j k1 < " for any u; v 2 A
such that ju.pi / v.p Ti /j < n for all i n. As a consequence, (9) holds for ' so
1
' 2 CA . Thus CA D fS.k/ W k 2 Ng and (10) is proved.

Finally, observe that R.n; k/ is Polish and hence n .R.n; k// is analytic for all
n; k 2 N. Therefore S.k/ is analytic by Problem 337 and hence CA is analytic by
Problem 336. We established that for any separable Rosenthal compact X and any
dense countable A X , the space CA D Cp .AjX / is analytic, i.e., we settled
necessity.
Now assume that the space CB is analytic for any countable dense B X ; fix
some countable dense set A X and a continuous onto map r W M ! CA of some
Polish space M onto CA . Let W C ! CA is the relevant restriction map. Since
is a condensation, the map D 1 W CA ! C is well-defined. If s D r,
then s W M ! C ; the map s might be discontinuous, but it is continuous if M
and C are considered with their respective discrete topologies. Therefore the map
s W RC ! RM defined by s.f / D f s for each f 2 RC is an embedding by
Problem 163 of [TFS].
Another easy observation is that if x 2 X and ex .f / D f .x/ for any f 2 C ,
then the correspondence x ! ex is an embedding of X into Cp .C / RC (see
Problem 167 of [TFS]). If X 0 D fex W x 2 X g RC , then X 0 is a homeomorphic
copy of X and hence Y D s .X 0 / RM is also a homeomorphic copy of X .
Given any a 2 A the map ea W CA ! R is continuous on CA because for any
O 2 .R/ we have .ea /1 .O/ D ff 2 CA W f .a/ 2 Og is an open set of CA .
Therefore s .ea / D ea s D ea . r/ D .ea / r is a continuous map on M
which shows that A0 D s .A/ Cp .M /.
Now if x 2 X nA, then ex is a measurable map on CA . Indeed, for any
O 2 .R/, we have U D .ex /1 .O/ D ff 2 CA W .f /.x/ 2 Og. We know
that for B D A [ fxg, the set CB is analytic. Let 0 W C ! CB be the relevant
restriction map and 0 D . 0 /1 . As before, we can establish that the map ex 0 is
a continuous map on CB because U 0 D .ea 0 /1 .O/ D ff 2 CB W f .x/ 2 Og
is an open subset of CB for any O 2 .R/. We also have the natural restriction
map 0 W CB ! CA defined by the formula 0 .f / D f jA for all f 2 CB . Let
F 0 D CB nU 0 and F D CA nU ; it is immediate that U D 0 .U 0 / and F D 0 .F 0 /.
The space CB being analytic, the sets F 0 and U 0 are also analytic and hence F and
U are analytic as well.
Since 0 is a condensation, we have F \ U D ;, so we can apply Problem 339
to the Polish space RA and disjoint analytic sets U and F . As a consequence, there
are disjoint U0 ; F0 2 B.RA / such that U U0 ; F F0 . But F [ U D CA and
therefore F0 \ CA D F and U0 \ CA D U which shows that U is a Borel subset
of CA (see Fact 1 of T.319). Thus .ex /1 .O/ is a Borel subset of CA for any
open O R which implies that ex is a measurable map (see Fact 1 of T.363).
Consequently, .ex / r is also measurable because r is continuous (in fact, it is a
trivial exercise that a composition of two measurable maps is a measurable map).
Therefore s .ex / D ex s D ex . r/ D .ex / r is a measurable map
on M which shows Y B.M /. Thus Y is a closure in RM of the set A0 Cp .M /
and we can apply Fact 5 to conclude that Y B1 .M / and hence Y is Rosenthal
compact. Since Y is homeomorphic to X , the space X is also Rosenthal compact

and hence we proved sufficiency and completed our solution.
T.386. Let X be a compact space. Assume that A and B are dense countable subsets
of X such that Cp .AjX / is analytic and Cp .BjX / is not. Prove that X contains a
subspace homeomorphic to !.
Solution. Let C D Cp .X / and CA D Cp .AjX /; fix a continuous surjective map
r W M ! CA of some Polish space M onto CA . Let W C ! CA is the relevant
restriction map. Since is a condensation, the map D 1 W CA ! C is well-
defined. If s D r, then s W M ! C ; the map s might be discontinuous, but it
is continuous if M and C are considered with their respective discrete topologies.
Therefore the map s W RC ! RM defined by s.f / D f s for each f 2 RC is an
embedding by Problem 163 of [TFS].
Another easy observation is that if x 2 X and ex .f / D f .x/ for any f 2 C ,
then the correspondence x ! ex is an embedding of X into Cp .C / RC (see
Problem 167 of [TFS]). If X 0 D fex W x 2 X g RC , then X 0 is a homeomorphic
copy of X and hence Y D s .X 0 / RM is also a homeomorphic copy of X .
Given any a 2 A the map ea W CA ! R is continuous on CA because for any
O 2 .R/ we have .ea /1 .O/ D ff 2 CA W f .a/ 2 Og is an open set of CA .
Therefore
s .ea / D ea s D ea . r/ D .ea / r
is a continuous map on M which shows that A0 D s .A/ Cp .M /.

Since A is dense in X , the set A0 is dense in Y ; if Y B1 .M /, then Y Rosenthal
compact and hence so is X which implies that Cp .BjX / is analytic (see Problem
385). This contradiction shows that Y is not contained in B1 .M / and therefore !
embeds in Y by Fact 4 of T.385. Since Y is homeomorphic to X , the space !
embeds in X as well.
T.387. Suppose that X is a compact space and A is a dense countable subset of
X such that Cp .AjX / is analytic. Prove that X is Rosenthal compact or else !
embeds in X .
Solution. If the space X is not Rosenthal compact, then there is a countable dense
B X such that Cp .BjX / is not analytic (see Problem 385). Now we can apply
Problem 386 to conclude that ! embeds in X .
T.388. Prove that a space X is K-analytic if and only if X is an image of P under
a compact-valued upper semicontinuous map, i.e.,Swhen there exists a compact-
valued upper semicontinuous ' W P ! X such that f'.p/ W p 2 Pg D X .
Solution. Necessity was proved in Fact 2 of T.346. To prove sufficiency, assume
S ' W P ! exp.X / is a compact-valued upper S
that semicontinuous map such that
f'.p/ W p 2 Pg D X . For any A P let '.A/ D f'.p/ W p 2 Ag.
The set F D f.p; x/ 2 P Z W x 2 '.p/g is closed in the space P X .
Indeed, if we are given a point .q; y/ 2 .P Z/nF , then y '.q/; since '.q/ is
compact, there is U 2 .y; X / such that U \ F D ; (the bar denotes the closure
in X ). We have F V D .X nU / \ X 2 .F; X/ and therefore there is a set
G 2 .q; P/ for which '.G/ V . The set W D G U is an open neighborhood
of the point .q; y/ in the space P X .
If .r; z/ 2 W , then '.r/ V while z 2 U X nV which shows that z '.r/.
Thus W \ F D ; and therefore every point of .P X /nF has a neighborhood
contained in .P X /nF . This proves that the set .P X /nF is open and hence
F is closed in P X .
Let W P X ! X be the natural projection. The spaces P and X are K-
analytic, so F is also K-analytic (see Problem 343). It is immediate that .F / D X ,
so X is K-analytic being a continuous image of a K-analytic space.
T.389. Prove that an arbitrary space X is K-analytic if and only if there exists
a family K D fKf W f 2 ! ! g of compact subsets of X with the following
properties:
S
(i) the family K is a cover of X , i.e., K D X ;
(ii) if a sequence fn 2 ! ! converges to f 2 ! ! and xn 2 Kfn for all n 2 !, then
the sequence fxn g has an accumulation point which belongs to Kf .
Solution. We identify P with ! ! . To prove sufficiency, assume that there exists a
family K D fKf W f 2 Pg with the properties (i) and (ii). For any f 2 P, let
'.f / D KfS . Then ' W P ! X is a compact-valued onto map; for any A P,
let '.A/ D f'.f / W f 2 Ag. To see that ' is upper semicontinuous, fix a point
f 2 P and U 2 .'.f /; X /. For every n 2 !, let O.f; n/ D fg 2 P W gjn D f jng.
Then the family fO.f; n/ W n 2 !g is a local base of P at f .
Assume that '.O.f; n// is not contained in U and fix a point xn 2 '.O.f; n//nU
for every n 2 !. For each n 2 !, there is fn 2 O.f; n/ such that xn 2 '.fn /. It
is evident that fn ! f and hence we can apply (2) to conclude that there is there
is an accumulation point x 2 '.f / of the sequence S D fxn W n 2 !g. However,
S X nU and hence all accumulation points of S lie in X nU X n'.f /; this
contradiction shows that '.O.f; n// U for some n 2 ! and hence ' is upper
semicontinuous (see Fact 1 of T.346). Thus X is K-analytic by Problem 388 and we
proved sufficiency.
Now, if X is K-analytic, then there is an upper semicontinuous compact-valued
onto map ' W P ! X (see Problem 388); let Kf D '.f / for any f 2 P. We
claim that the family K D fKf W f 2 Pg satisfies (i) and (ii). Indeed, the property
(i) is an immediate consequence of the fact that ' is onto. Suppose that a sequence
S D ffn W n 2 !g P converges to some f 2 ! and take xn 2 '.fn / for all n 2 !.
If S has no accumulation points in set '.f / D Kf , then for any y 2 Kf there is
Uy 2 .y; X/ such that the set P .Uy / D fi 2 ! W xi 2 Uy g is finite. The set Kf
being compact there are y1 ; : : : ; yk 2 Kf such that Kf U D Uy1 [

[Uyk . The
map ' is upper semicontinuous, so we can find V 2 .f; P/ for which '.V / U .
Since the sequence S converges to f , there is a number m 2 ! such that the set
S 0 D ffi W i mg is contained in V . As a consequence, '.fi / U and hence
there is s.i / 2 Nk D f1; : : : ; kg such that xi 2 Uys.i / for all i m. The set Nk
being finite, there is l 2 Nk such that s.i / D l for infinitely many i . Consequently,
P .Uyl / is infinite which is a contradiction. Thus the sequence fxn W n 2 !g has an
accumulation point in Kf and hence the family K has the property (2). This proves
necessity and makes our solution complete.
T.390. Prove that an arbitrary space X is K-analytic if and only if there exists a
space Y which maps perfectly onto P and continuously onto X .
Solution. Suppose that some Y maps perfectly onto P and continuously onto X .
Since every continuous image of a K-analytic space is K-analytic, it suffices to
show that Y is K-analytic. Take a perfect map f W Y ! P and let '.p/ D f 1 .p/
for each p 2 P. Then ' W P ! Y is a compact-valued onto map. If p 2 P and
1
U 2 .'.p/; Y /, then there is V 2 .p; P/ for S which f .V / U (see Fact 1 of
S.226). In terms of ' this says that '.V / D f'.q/ W q 2 V g U and hence '
is upper semicontinuous (see Fact 1 of T.346). Thus we can apply Problem 388 to
conclude that the space Y is K-analytic.
To prove necessity assume that X is K-analytic and fix a compact-valued upper
semicontinuous onto map ' W P ! X . The set Y D f.p; x/ 2 P Z W x 2 '.p/g
is closed in P X . Indeed, if we are given a point .q; y/ 2 .P Z/nY , then
y '.q/; since '.q/ is compact, there is U 2 .y; X / such that U \ Y D ;
(the bar denotes the closure in X ). We have Y V D .X nU / \ X 2 .Y; X/
and therefore there is G 2 .q; P/ for which '.G/ V . The set W D G U is
an open neighborhood of .q; y/ in P X . If .r; z/ 2 W , then '.r/ V while
z 2 U X nV which shows that z '.r/. Thus W \ Y D ; and therefore every
point of .P X /nY has a neighborhood contained in .P X /nY . This proves
that the set .P X /nY is open and hence Y is closed in P X .
If 1 W P X ! X be the natural projection, then .Y / D X , so X is a
continuous image of the space Y . On the other hand, if 0 W P X ! P is the
other natural projection, then 0 is a perfect map (see Fact 3 of S.288). Therefore
f D 0 jY is a perfect map as well because 0 .Y / D P and Y is closed in P X .
Thus X is a continuous image of the space Y which maps perfectly onto P.
T.391. Prove that an arbitrary space X is K-analytic if and only if X is realcompact
and has a P-directed compact cover.
Solution.
S n We identify P with the space ! ! , letting, as usual, ! 0 D f;g and ! <! D
f! W n 2 !g; if p; q 2 P, then p q stands for p.i / q.i / for all i 2 !. If
f is a function, then dom.f / is its domain; given functions f and g the expression
f g says that dom.f / dom.g/ and gjdom.f / D f . If Z is a space and we
have a sequence S D fzn W n 2 !g Z, then z 2 Z is an accumulation point for
the sequence S if for any U 2 .z; Z/ the set fn 2 ! W zn 2 U g is infinite. For
the sake of brevity, we will call a space P-dominated if it has a P-directed compact
cover. A cover fKp W p 2 Pg of a space Z is sequentially subcontinuous if for any
sequence fpn W n 2 !g P which converges to a point p 2 P and any sequence
S D fzn W n 2 !g Z such that zn 2 Kpn for all n 2 !, there is z 2 Kp which is
an accumulation point for the sequence S .
Given aSP-directed compact cover K D fKp W p 2 Pg of a space Z and s 2 ! <! ,

let Hs D fKp W s pg. Once we have the family H D fHs WT s 2 ! <! g (which
we will call the envelope of the family K), define the set Fp D fHpjn W n 2 !g
for any p 2 P and call the family K0 D fFp W p 2 Pg the regularization of the
family K.
Fact 1. Given a P-directed compact cover K D fKp W p 2 Pg of a space Z, let
H D fHs W s 2 ! <! g be it envelope and K0 D fFp W p 2 Pg its regularization.
Then, for any p 2 P and any sequence S D fzn W n 2 !g Z such that zn 2 Hpjn
for all n 2 !, there is z 2 Fp which is an accumulation point for the sequence S . In
particular, Fp is countably compact for all p 2 P.
Proof. It follows from zn 2 Hpjn that there is qn 2 P such that pjn qn and
zn 2 Fqn for all n 2 !. For any k 2 ! we have qn .k/ D p.k/ for all n k which
P the set fqn .k/ W n 2 !g is finite for any k 2 !. Therefore the number
shows that
q.k/ D fqn .k/ W n 2 !g is well-defined for every k 2 ! and hence we have a
function q 2 P such that qn q for all n 2 !. The cover K being P-directed, we
have Kqn Kq for all n 2 ! and hence fzn W n 2 !g Kq . Since the set Kq is
compact, we proved that
(1) the set L of accumulation points of the sequence S is non-empty.
Our next step is to establish that L Fp , which will, evidently, finish the proof
of this fact. So take any z 2 L and fix n 2 !. It is clear that z is also an accumulation
point of the sequence S 0 D fzk W k ng. For any y D zk 2 S 0 we have qk jn D pjn
and the set fqk .i / W k ng is finite for any i Pn. Thus we can define a point
r 2 P by r.i / D p.i / for all i < n and r.i / D fqk .i / W k ng for all i n.
Then qk r and hence Fqk Fr for all k n, i.e., the sequence fzk W k ng is
contained in the set Fr . Since pjn r, we have Fr Hpjn , so all accumulation
points of S 0 belong to Hpjn and, in particular, z 2 Hpjn . The number n 2 ! was
chosen arbitrarily, so we proved that z 2 Hpjn for all n 2 ! and therefore z 2 Fp ,
i.e., L Fp . In particular, if S Fp , then S Hpjn and hence zn 2 Hpjn for
all n 2 !. By what we proved above, S has accumulation points in Fp , so Fp is
countably compact and Fact 1 is proved. t
u
Fact 2. Given a P-directed compact cover K D fKp W p 2 Pg of a space Z its
regularization K0 D fFp W p 2 Pg is sequentially subcontinuous.
Proof. Let fHs W s 2 ! <! g be the envelope of K. Take an arbitrary sequence T D
fpn W n 2 !g P which converges to a point p 2 P and assume that zn 2 Fpn for
all n 2 !. It is evident that it suffices to prove our statement for a subsequence of
the sequence T . It follows from pn ! p that for any n 2 !, there is kn 2 ! such
that pi jn D pjn for all i kn . Thus, passing to an appropriate subsequence of the
sequence T , we can assume, without loss of generality, that
(2) pi jn D pjn for all i n.
Now it follows from zn 2 Fpn that zn 2 Hpn jn D Hpjn for all n 2 ! and hence
we can apply Fact 1 to conclude that the sequence fzn W n 2 !g has an accumulation
point z 2 Fp . Fact 2 is proved. t
u
Fact 3. If Z is P-dominated and every pseudocompact closed E Z is compact,
then Z is K-analytic.
Proof. Fix a P-directed compact cover K D fLp W p 2 Pg of the space Z. The
regularization K0 D fFp W p 2 Pg of the cover K consists of countably compact
subsets of Z by Fact 1. For any p 2 P let .p/ D Fp ; then W P ! Z is a
multivalued map. To see that it is onto take any z 2 Z; then there is p 2 P such that
z 2 Lp . It is evident that Lp Fp so z 2 Fp and hence is an onto map. It turns
out that
(3) the map is upper semicontinuous.
Indeed, take any p 2 P and U 2 ..p/; Z/. For any n 2 !, consider the set
On D fq 2 P W qjn D pjng. Then fOn W n 2 !g is a local base of P at p. If .On / is
not contained in U for any n 2 !, then pick pn 2 On and zn 2 .pn /nU D Fpn nU
for all n 2 !. Since pn ! p, there is a point z 2 Fp which is an accumulation point
of the sequence S D fzn W n 2 !g (see Fact 2). However, S X nU and hence all
accumulation points of S belong to X nU X nFp ; this contradiction shows that
.On / U for some n 2 ! and hence is upper semicontinuous (see Fact 1 of
T.346).
By our assumption about the space Z, the closure of every countably compact
subset of Z is pseudocompact and hence compact (see Fact 18 of S.351). Thus the
set Kp D F p is compact for any p 2 P and therefore fKp W p 2 Pg is a compact
cover of the space Z. Let '.p/ D Kp for every p 2 P. Then ' W P ! Z is a
compact-valued onto map. To see that ' is upper semicontinuous take any p 2 P
and U 2 .'.p/; Z/. It follows from compactness of Kp D '.p/ that there is
V 2 .Kp ; Z/ such that V U . Since .p/ D Fp Kp , we have .p/ V
and hence we can apply (3) to find O 2 .p; P/ such that .O/ V . Now, for any
q 2 O we have Fq D .q/ V and hence '.q/ D Kq D F q V U . This
shows that '.O/ U and hence the map ' is also upper semicontinuous. Therefore
Problem 388 is applicable to conclude that Z is K-analytic. Fact 3 is proved. t
u
Returning to our solution let us show first that if the space X is K-analytic,
then it is P-dominated. To do this, observe that there exists a compact-valued upper
semicontinuous onto map ' W P ! X (see Problem 388). For Q any p 2 P and i 2 !
let Ni D f0; : : : ; p.i /g; then Mp D fq 2 P W q pg D fNi W i 2 !g is a
compact subset of P. It is evident that p q implies Mp MS q so fMp W p 2 Pg
is a P-directed compact cover of P. The set Kp D '.Mp / D f'.q/ W q 2 Mp g
is compact for every p 2 P (to see this observe that 'p D 'jMp W Mp ! '.Mp /
is a compact-valued upper semicontinuous onto map and apply Problem 241). It is
immediate that p q implies Kp Kq , so fKp W p 2 Pg is a P-directed compact
cover of X . Thus, if X is K-analytic, then it is P-dominated; besides it is Lindelf
and hence realcompact (see Problems 338 and 406 of [TFS]) so we settled necessity.
Finally, if X is realcompact and P-dominated, then every pseudocompact closed

subspace of X is compact (see Problems 403 and 407 of [TFS]), so X is K-analytic
by Fact 3. This proves sufficiency and makes our solution complete.
T.392. Suppose that X can be condensed onto a metrizable space. Prove that X is
analytic if and only if it has a P-directed compact cover.
Solution. If the space X is analytic, then it is K-analytic by Problem 346 and hence
it has a P-directed compact cover by Problem 391.
Now, assume that X has a P-directed compact cover and there exists a con-
densation f W X ! M of X onto some metrizable space M . Then, for any
pseudocompact P X , the set P 0 D f .P / is compact and hence second countable
(because it is pseudocompact and metrizable, see Problem 212 of [TFS]) which
implies that the map g D f jP W P ! P 0 is a homeomorphism (see Problem
140 of [TFS]). Therefore any pseudocompact P X is Compact, so we can apply
Fact 3 of T.391 to conclude that X is K-analytic.
Thus X is Lindelf and hence so is M ; any Lindelf metrizable space is second
countable, so X condenses onto a second countable space. Even a Lindelf -space
of countable i -weight has a countable network (see Problem 266) so nw.X / D !
(because every K-analytic space is Lindelf , see Problem 261) and hence X is
analytic by Problem 346.
T.393. Let X be a compact space. Prove that Cp .X / is K-analytic if and only if it
has a P-directed compact cover.
Solution. If Cp .X / is K-analytic, then it has a P-directed compact cover by
Problem 391. On the other hand, if P Cp .X / is pseudocompact, then P is
compact by Fact 19 of S.351, so if Cp .X / has a P-directed compact cover, then
we can apply Fact 3 of T.391 to conclude that Cp .X / is K-analytic.
T.394. Give an example of a non-K-analytic space which has a P-directed compact
cover.
Solution. We identify P with ! ! ; if f; g 2 P, then f g if f .n/ g.n/ for all
n 2 !. The expression f g says that the set fn 2 ! W f .n/ > g.n/g is finite.
Denote by D the set P with the discrete topology and let Ap D fq 2 D W q pg
for any p 2 P. It is evident that p; q 2 P and p q implies Ap Aq so if
Kp D Ap for all p 2 P (the bar denotes the closure in D) then K D fKp W p 2 Pg
is a P-directed family
S of compact subsets of D.
Letting X D K, we obtain a space X D which has a P-directed compact
cover K. Since p 2 Ap for any p 2 P, we have D X and hence X is dense in
D. It follows from Fact 1 of T.097 that
(1) for any countable A D there is p 2 P such that q p for any q 2 A.
Sany countable set B X ; then there is a countable A P such that

Take
B fKq W q 2 Ag. The property (1) implies that there is p 2 P for which
q p for any q 2 A and therefore Kq Kp for any q 2 A. This shows that
S
B fKq W q 2 Ag Kp X , i.e., the closure of every countable subset of X is
compact and therefore X is countably compact.
For any p 2 P let Dp D DnAp and P consider the family D D fDp W p 2 Pg.
Given any finite B P let p.n/ D fq.n/ W q 2 Bg C 1 for any n 2 !; this
point p 2 P T
defines aS and it is immediate that p Aq for any q 2 B. Therefore
p 2 Dn. q2B Aq / D q2B Dq which shows that D is a centered family. The space
T
D being compact the set F D fD p W p 2 Pg is non-empty.
Now observe that D p \ Kp D ; for any p 2 P (see Fact 1 of S.382). As a
consequence, ; F DnX and hence X D. Therefore X is a countably
compact non-compact space (because X D D and X D). This shows that
X cannot be Lindelf and therefore X is a non-K-analytic space with a P-directed
compact cover.
T.395. Assuming MAC:CH, prove that if X is a K-analytic space such that every
compact subspace of X is metrizable, then X has a countable network. Observe that
if every compact subspace of an analytic space X is countable, then X is countable.
Solution. We identify P with ! ! ; given p; q 2 P, let p q if there is k 2 ! such
that p.n/ q.n/ for all n k. A set E P is bounded if there is p 2 P such that
q p for any q 2 E.
Fact 1. Under MA, if E P and jEj < c, then the set E is bounded.
Proof. For each q 2 E let Aq D f.i; j / 2 ! ! W j q.i /g ! ! and consider
the set Bn D f.n; j / W j 2 !g for any n 2 !. For the families
S A D fAq W q 2 Eg
and B D fBn W n 2 !g, we have jAj < c; jBj < c and Bn. A0 / is infinite for any
B 2 B and finite A0 A. Therefore we can apply Problem 051 to find an infinite set
L ! ! such that AnL is finite for any A 2 A and BnL is infinite for any B 2 B.
Observe that Problem 051 is formulated for subsets of !, but, of course, it is valid
for any countably infinite set N ; we must take a bijection ' W N ! ! and find the
respective set L0 for the families A0 D f'.A/ W A 2 Ag and B 0 D f'.B/ W B 2 Bg.
Then L D ' 1 .L0 / is as promised.
If M D .! !/nL, then M \ A is finite for any A 2 A and M \ B is infinite
for every B 2 B. Consequently, there is a function p 2 P such that .n; p.n// 2 M
for any n 2 !. It is immediate that q p for any q 2 E, so Fact 1 is proved. u t
Fact 2. If set E P is bounded, then there is a -compact Z P such that E Z.
Proof. Choose a point p 2 P for which q p for all q 2 QE. For any q 2 P the
set Hq D fr 2 P W r qg is compact because Hq D fGn W n 2 !g where
Gn D f0; : : : ; q.n/g for all n 2 !. The set Wp D fqS2 P W there exists n 2 ! such
that q.i / D p.i / for all i ng is countable so Z D fHq W q 2 Wp g is -compact.
It is easy to see that E Z, so Fact 2 is proved. t
u
Fact 3. Under MAC:CH, if E P is an uncountable set, then there is a compact
K P such that E \ K is uncountable.
Proof. We can consider that jEj D !1 < c, so Fact 1 is applicable to conclude

that E is bounded.
S By Fact 2, there exists a -compact set Z P with E Z.
Now,
S Z D n2! K n where Kn is compact for all n 2 !. Consequently, E D
n2! .E \ K n / and hence E \ Kn is uncountable for some n 2 !. Letting K D Kn ,
we obtain the promised compact K P so Fact 3 is proved. t
u
Returning to our solution assume that MA+:CH holds and all compact sub-
spaces of a K-analytic space Z are metrizable. There is an upper semicontinuous
compact-valued onto map ' W P ! Z by Problem 388. Assume first that hl.Z/ > !
and therefore there is a right-separated Y Z with jY j D !1 < c (see Problem
005). For any p 2 P the space '.p/ is metrizable and hence Y \ '.p/ is countable.
Thus there are sets Q D fq W < !1 g P and R D fz W < !1 g Z such that
z 2 '.q / for all < !1 and z z for any distinct ; < !1 .
Fact 3 guarantees that there is a compact subspace K P such that the set
U D f < !1 W q 2 Kg is uncountable.
S Therefore R0 D fz W 2 U g is also
uncountable and R '.K/ D f'.q/ W q 2 Kg. The set K 0 D '.K/ is compact
0
because the map 'jK W K ! K 0 is upper semicontinuous, compact-valued and

onto (see Problem 241). By our assumption about Z the space K 0 is metrizable and
hence second countable. However, R0 is an uncountable right-separated subspace
of K 0 which implies hl.K 0 / > ! D w.K 0 / (see Problem 005); this contradiction
shows that
(1) under MAC:CH any K-analytic space whose compact subsets are metrizable
is hereditarily Lindelf.
Finally, assume again that MA+:CH holds and all compact subspaces of a K-
analytic space X are metrizable. The space X n is K-analytic (see Problem 343) and
it is clear that all compact subspaces of X n are metrizable for every n 2 N, so we can
apply (1) to conclude that hl.X n / D ! for all n 2 N. Thus hd.Cp .X /n / D ! for
all n 2 ! by Problem 026; in particular, Cp .X / is separable and hence X condenses
onto a second countable space. The space X is Lindelf by Problem 261, so it is
stable by Problem 266 which implies that nw.X / D ! and finishes our solution.
T.396. Suppose that X is a compact space such that some outer base B of its
diagonal D f.x; x/ W x 2 X g in X X can be indexed as B D fOp W p 2 ! ! g
in such a way that Op Oq whenever p; q 2 ! ! and q p. Prove that X is
metrizable.
Solution. We identify P with ! ! . For any p 2 P and n 2 ! define sp; n 2 P by
sp; n .m/ D p.n C m/ for all m 2 !. A space Z is called P-dominated if Z has a
P-directed compact cover.
Fact 1. For any sequence fpk W k 2 !g P there exists p 2 P such that pk
sp; k for all k 2 !.
Pn
Proof. Let p.n/ D i D0 pi .n i / for every n 2 ! and observe that if k 2 !, then
X
kCn
sp; k .n/ D p.k C n/ D pi .k C n i / pk .n/
i D0
for any n 2 !; it is clear that this implies pk sp; k for all k 2 !, so Fact 1 is
proved. t
u
Returning to our solution, consider the set Kp D ff 2 RX W jf .x/j p.0/
for all x 2 X and jf .x/ f .y/j nC1 1
for any n 2 ! and x; y 2 X with
.x; y/ 2 Osp;n g for any p 2 P. We claim that
(1) Kp is a compact subset of Cp .X / for any p 2 P.
To prove the property (1) fix p 2 P and take any f 2 RX nKp . If jf .x/j > p.0/,
then V D fg 2 RX W jg.x/j > p.0/g is an open subset of RX such that f 2 V
and V \ Kp D ;. If jf .x/j p.0/ for all x 2 X , then there is n 2 ! such that
jf .x/ f .y/j > nC1 1
for some .x; y/ 2 Osp;n . Again, the set V D fg 2 RX W
jg.x/ g.y/j > nC1 g is open in RX while f 2 V and V \ Kp D ;. It turns out
1
that any f 2 RX nKp has a neighborhood contained in RX nKp , so RX nKp is open

and hence Kp is closed in RX . Furthermore, for r D p.0/ we have Kp r; r X
and hence Kp is closed in a compact space r; r X so Kp is compact.
To see that Kp Cp .X / take any f 2 Kp ; given x 2 X and " > 0 choose
n 2 ! such that nC1 1
< ". Then W D fy 2 X W .x; y/ 2 Osp;n g is an open
subset of X and x 2 W . For any y 2 W we have .x; y/ 2 Osp;n and therefore
jf .x/ f .y/j nC1 1
< " which shows that f is continuous at the point x. Since the
point x 2 X was chosen arbitrarily, we established that every f 2 Kp is continuous,
i.e., Kp Cp .X / and (1) is proved.
Now take any function f 2 Cp .X /; since X is compact, f is bounded on X ,
i.e., there is a real number N > 0 such that jf .x/j N for all x 2 X . The set
Wk D f.x; y/ 2 X X W jf .x/ f .y/j < kC1 1
g is an open neighborhood of in
X X for every k 2 !. Since B is a base of in X X , for any k 2 !, there is
pk 2 P such that Opk Wk . Apply Fact 1 to find q 2 P for which pk sq; k
and therefore Osq;k Opk Wk for all k 2 !. Now define a point p 2 P by
p.n/ D q.n/ C N for every n 2 !. It is immediate that jf .x/j N p.0/ for all
x 2 X and Osp;k Osq;k Wk for all k 2 !; this implies jf .x/ f .y/j kC1 1
whenever k 2 ! and x; y 2 Osp;k which shows that f 2 Kp . Thus the family

K D fKp W p 2 Pg is a compact cover of Cp .X /. It follows from the definition of
K that p q implies Kp Kq , i.e., K is P-directed.
Thus Cp .X / is K-analytic and hence Lindelf (see Problems 393 and 261);
since X embeds in Cp .Cp .X // (see Problem 167 of [TFS]), we can apply Baturovs
theorem (see Problem 269 of [TFS]) to conclude that
(2) ext.Y / D l.Y / for any Y X .
The set Fp D .X X /nOp is compact for all p 2 P and the family fFp W
p 2 Pg is a P-directed compact cover of X 0 D .X X /n. It is evident that
the property of being P-dominated is closed-hereditary and therefore every closed
H X 0 is P-dominated. In particular, if D is a closed discrete subspace of X 0 ,

then it is metrizable and P-dominated. Applying Problem 392 we can see that D is
analytic and hence Lindelf; any Lindelf discrete space is countable, so we proved
that every closed discrete D X 0 is countable, i.e., ext.X 0 / D !.
The proof of the following statement is standard and easy; we leave it to the
reader as an exercise.
(3) If a space Z is compact and F is closed in Z, then .F; Z/ D ! if and only if
ZnF is Lindelf.
Fix any x 2 X and observe that the subspace Yx D .fxg X /n is closed in X 0
and homeomorphic to X nfxg. Therefore ext.Yx / ext.X 0 / D ! which, together
with (2), implies that X nfxg is Lindelf and hence .x; X / D ! for any x 2 X .
As aSconsequence, ..x; x/; X X / ! for any x 2 X and hence the diagonal
D ff.x; x/g W x 2 X g is a union of G -subsets of a compact (and hence
realcompact) space X X . Thus X 0 is realcompact by Problem 408 of [TFS]; since
it is P-dominated, it is K-analytic by Problem 391. As a consequence, X 0 D .X
X /n is Lindelf and hence the diagonal of X is a G -set in X X by (3). Finally
apply Problem 091 to conclude that X is metrizable and finish our solution.
T.397. Suppose that Cp .X / is K-analytic and X is separable. Prove that Cp .X / is
analytic.
Solution. Since X is separable, the space Cp .X / condenses onto a second count-
able space (see Problem 173 of [TFS]). Now, Cp .X / is Lindelf by Problem 261;
therefore it is stable and hence nw.Cp .X // D ! (see Problem 266). Finally, apply
Problem 346 to see that Cp .X / is analytic.
T.398. Let X be a compact space such that Cp .X / is K-analytic. Prove that X is a
FrchetUrysohn space.
Solution. The space Cp .X / has the Lindelf -property and hence it is stable (see
Problems 261 and 266). Thus the space Cp .Cp .X // monolithic by Problem 154;
since X embeds in Cp .Cp .X // (see Problem 167 of [TFS]), it is also monolithic.
Furthermore, l .Cp .X // D ! (see Problem 256) and hence we have the inequality
t.X / t.Cp .Cp .X /// D ! (see Problem 149 of [TFS]). Therefore X is a
monolithic compact space of countable tightness; now Problem 120 implies that
the space X is FrchetUrysohn.
(i) Cp .Cp .X // is K-analytic;
(ii) Cp .Cp .X // is analytic;
(iii) X is finite.
Solution. If Z is a space, A Z and B exp.Z/, then the family B separates A
from ZnA if for any y; z 2 Z such that y 2 A and z 2 ZnA there is B 2 B for
which y 2 B and z B.
Fact 1. If Z is a Lindelf -space, then there exists a Lindelf -space S such

that Cp .Z/ S RZ .
Proof. Fix a family P D fPn W n 2 !g which is a network with respect to a compact
cover C of the space Z (see Problem 225). A function f W Z ! R will be called
P-bounded if for any z 2 Z there is P 2 P such that z 2 P and f .P / is bounded
in R. Let S be the set of all P-bounded functions.
To see that Cp .Z/ S take any f 2 Cp .Z/ and z 2 Z. There is C 2 C
with z 2 C . By continuity of f and compactness of C , the set C 0 D f .C / is
bounded in R, i.e., there is a > 0 such that jf .y/j < a for any y 2 C . The set
U D fy 2 Z W jf .y/j < ag is an open neighborhood of C , so there is P 2 P
such that C P U and, in particular, z 2 P . It follows from P U that
f .P / f .U / . a; a/ and hence f .P / is bounded in R. This proves that f 2 S
and therefore Cp .Z/ S .
To check that S is a Lindelf -space consider the map W R ! J D . 1; 1/
defined by .t/ D 2 arctan.t/ for all t 2 R. It is clear that is a homeomorphism
and hence the map q W Cp .Z/ ! Cp .Z; J / defined by q.f / D f for any f 2
Cp .Z/ is a homeomorphism as well (see Problem 091 of [TFS]). Let S 0 D q.S /; it
is easy to see that
(1) S 0 D ff 2 J Z W for any z 2 Z there is P 2 P such that z 2 P and f .P /
a; a for some a 2 .0; 1/g.
Observe also that we have the inclusion J Z IZ and therefore S 0 is a subspace
of a compact space IZ . Let ak D 1 2k1 2 .0; 1/ for all k 2 ! and consider the
set Bn;k D ff 2 IZ W f .Pn / ak ; ak g for all n; k 2 !. It is straightforward that
Bn;k is closed in IZ and hence compact for any n; k 2 !; we claim that the family
B D fBn;k W n; k 2 !g separates IZ nS 0 from S 0 .
Indeed, let f 2 S 0 and g 2 IZ nS 0 . If g.z/ 2 f 1; 1g for some z 2 Z, then, by
(1) there is n 2 ! such that z 2 Pn and f .Pn / a; a for some a 2 .0; 1/. There
is k 2 ! with a < ak and hence f .Pn / ak ; ak which proves that f 2 Bn;k .
On the other hand, it follows from g.z/ J that g Bn;k .
Now, if g.z/ f 1; 1g for all z 2 Z, then g 2 J Z and hence we can apply (1)
to find z 2 Z such that for any n 2 !, if z 2 Pn , then the set g.Pn / is not contained
in a; a for any a 2 .0; 1/. Again, it follows from f 2 S 0 that there is n 2 !
for which z 2 Pn and f .Pn / a; a for some a 2 .0; 1/. There is k 2 ! with
a < ak and hence f .Pn / ak ; ak which proves that f 2 Bn;k . On the other
hand, it follows from g.Pn / 6 ak ; ak that g Bn;k .
This shows that B separates S 0 from IZ nS 0 and hence S 0 is a Lindelf -space
by Problem 233. Finally, S is also Lindelf because it is homeomorphic to S 0 .
Fact 1 is proved. t
u
Returning to our solution observe that if X is finite then Cp .X / D RX is second
countable and -compact, so the space Cp .Cp .X // is analytic by Problem 367; this
proves (iii)H)(ii). The implication (ii)H)(i) follows from Problem 346 so let us
prove that (i)H)(iii).
Assume, towards a contradiction, that Cp .Cp .X // is K-analytic while X is

infinite and fix a countably infinite A X . Since X is homeomorphic to a closed
subspace of Cp .Cp .X // (see Problem 167 of [TFS]), it is also K-analytic and
hence normal. The set Y D A is closed in X and hence the restriction map
W Cp .X / ! Cp .Y / is open and onto (see Problem 152 of [TFS]). This shows
that the dual map W Cp .Cp .Y // ! Cp .Cp .X // is a homeomorphic embedding
such that .Cp .Cp .Y /// is closed in Cp .Cp .X // (see Problem 163 of [TFS]).
Thus Cp .Cp .Y // is K-analytic and hence Lindelf . Since Y is separable, we
have i w.Cp .Y // D ! and hence Cp .Y / is a realcompact space (see Problem 446
of [TFS]). For every f 2 Cp .Y /, let ef .'/ D '.f / for any ' 2 Cp .Cp .Y //. Then
the correspondence f ! ef is an embedding of Cp .Y / in Cp .Cp .Cp .Y ///
RCp .Cp .Y // ; let C be the respective image of Cp .Y /. By Fact 1, there exists a
Lindelf -space S such that Cp .Cp .Cp .Y /// S RCp .Cp .Y // .
By definition of realcompactness, the set C is closed in RCp .Cp .Y // ; since also
C S , it closed in S as well and therefore C is a Lindelf -space. The space
Cp .Y / is also a Lindelf because it is homeomorphic to C . An immediate
consequence is that Cp .Y / is stable (see Problem 266) and hence Y is monolithic
by Problem 152. Now it follows from separability of the space Y that nw.Y / D !
and hence nw.Cp .Y // D nw.Cp .Cp .Y /// D !. Thus Cp .Cp .Y // is analytic by
Problem 346; this implies that Cp .Y / is -compact by 366 and hence Y is finite by
Problem 186 of [TFS]. However, Y contains an infinite subset A; this contradiction
shows that X cannot be infinite bringing to an end the proof of (i)H)(iii) and
making our solution complete.
(i) X is hereditarily K-analytic;
(ii) X is hereditarily analytic;
(iii) X is countable.
Solution. The implications (iii)H)(ii) and (ii)H)(i) are evident, so let us prove
that (i)H)(iii). Assume that X is hereditarily K-analytic and uncountable; then it is
hereditarily Lindelf because every K-analytic space is Lindelf (see Problem
261). However, every hereditarily Lindelf -space has a countable network (see
Problem 270), so nw.X / D ! and hence X is analytic by Problem 346.
Now apply Problem 353 to conclude that the Cantor set K embeds in X . By
Fact 5 of T.250, there exists Y K which is not K-analytic. Since Y also embeds
in X , we have proved that there exists a non-K-analytic subspace in the space X ;
this contradiction shows that X is countable and settles the remaining implication
(i)H)(iii).
T.401. Observe that A.!1 / is a union of two discrete (and hence metrizable)
subspaces of countable i -weight. Deduce from this fact that metrizability, first axiom
of countability, i -weight, P -property and pseudocharacter are not finitely additive.
Solution. We have A.!1 / D !1 [ fag where all points of Y D !1 are isolated and
therefore Y is a discrete subspace of A.!1 /. Since the singleton Z D fag is also
discrete, the space A.!1 / is a union of two discrete and hence metrizable subspaces.
Since w.Y / D !1 c, the space Y condenses onto a second countable space by
Problem 102 and hence i w.Y / D !. It is clear that w.Z/ D !.
Consequently, A.!1 / is a union of two metrizable spaces; it is not first countable
at a because if fUn W Tn 2 !g .a; A.!1 //, then !1 nUn is finite for every n 2 !
and hence for G D n2! Un we have j!1 nGj ! so G fag and therefore
.a; A.!1 // D .a; A.!1 // > !.
This shows that if P 2 fmetrizability, countable character, countable pseudocha-
racterg, then P is not finitely additive. We also saw that A.!1 / D Y [ Z where
i w.Y / ! and i w.Z/ !. However, i w.A.!1 // > ! because any condensation
of the compact space is a homeomorphism (see Problem 123 of [TFS]) so A.!1 /
cannot be condensed onto a space of countable weight (not even onto a space of
countable character because .A.!1 // > !). Thus i -weight is not finitely additive.
Since Y and Z are both discrete, they are P -spaces. However, A.!1 / is not a P -
space because A.!1 /n! is a G -subset of A.!1 / which is not open in A.!1 /. This
proves that P -property is not finitely additive either.
T.402. Representing L.!1 / as a union of two metrizable subspaces, observe that
sequentiality, -character, the FrchetUrysohn property, Cech-completeness and
k-property are not finitely additive.
Solution. We have L.!1 / D !1 [ fag where all points of Y D !1 are isolated
and therefore Y is a discrete subspace of A.!1 /. Since the singleton Z D fag
is also discrete, we have L.!1 / D Y [ Z where Y and Z discrete and hence
completely metrizable subspaces. This shows that Y and Z are Cech-complete (see
Problems 204 and 269 of [TFS]) and have countable character and hence countable
-character; they are also k-spaces because they have no non-closed subsets to
disprove this. It is also immediate that Y and Z are FrchetUrysohn and hence
sequential.
Now, if B D !1 , then B is not closed in L.!1 / while there is no sequence in
B which converges outside of B because every countable subset of B is closed in
L.!1 /. Therefore L.!1 / is not sequential and hence not FrchetUrysohn. Thus
both sequentiality and FrchetUrysohn property fail to be finitely additive.
Observe also that all compact subsets of L.!1 / are finite because if K L.!1 /
is infinite, then there is a countably infinite D Knfag; it is evident that D is
a closed discrete subset of K, so K is not compact. Therefore B is a non-closed
subset of L.!1 / such that K \ B is finite and hence closed in K for any compact
K L.!1 /. This proves that L.!1 / is not a k-space. Since every Cech-complete
space is a k-space (see Fact 1 of T.210), the space L.!1 / is not Cech-complete
either. This shows that Cech-completeness and k-property are not finitely additive.
To finally see that .a; L.!1 // > ! assume that there exists a family U D
fUn W n 2 !g .L.!1 // which is a -base at a in L.!1 /. Then Un \ B ; and
hence we can choose a point zn 2 Un \ B for all n 2 !. Let Vn D fzn g for all n 2 !;
then fVn W n 2 !g is also a -base at a in L.!1 /. However, if A D fzn W n 2 !g,
then W D L.!1 /nA is a neighborhood of a such that Vn 6 W for any n 2 !. This
contradiction shows that .a; L.!1 // > ! and hence -character is not finitely
additive.
T.403. Let 2 !n! and observe that the space ! [ fg is a union of two second
countable spaces while w.! [ fg/ > !. Therefore weight is not finitely additive.
Solution. If X D ! [ fg, then X D Y [ Z where Y D ! and Z D fg.
Since Y and Z are countable discrete spaces, we have w.Y / ! and w.Z/ !.
To see that w.X / > ! observe that 2 Y and assume that we have a sequence
S D fyn W n 2 !g Y with xn ! . For any y 2 Y the set U D X nfyg is an open
neighborhood of , so there is my 2 ! such that xn 2 U for all n my . Therefore
the set Ay D fn 2 ! W xn D yg is finite for any y 2 Y which makes it possible to
choose a sequence fki W i 2 !g such that ki < ki C1 for all i 2 ! and xki xkj if
i j . If yi D xki for all i 2 ! then the sequence fyi W i 2 !g still converges to .
Now let A D fi 2 ! W i is eveng and B D fi 2 ! W i is oddg. Then P D
fyi W i 2 Ag and Q D fyi W i 2 Bg are infinite disjoint subsets of !. But they are
also sequences which both converge to and hence 2 P \ Q (the bar denotes
the closure in !). However, P \ Q D ; by Fact 1 of S.382; this contradiction
shows that there is no sequence in Y which converges to and hence .; X / > !.
Therefore w.X / .X / > !, i.e., X is not second countable. This proves that the
property of having countable weight is not finitely additive.
T.404. Give an example of a non-realcompact space which is a union of two
hereditarily realcompact subspaces.
Solution. Recall that a Mrowka space M can be represented as M D E [ D where
E is a countable dense set of isolated points of M while D is closed and discrete
in M and jM j c (see Problem 142 of [TFS]). Since w.E/ D ! and i w.D/ D !
(see Problem 102), both spaces E and D are hereditarily realcompact by Problem
446 of [TFS]. However, the Mrowka space M is not realcompact because it is
pseudocompact and non-compact (see Problems 142 and 407 of [TFS]). Thus
the Mrowka space M is not realcompact while being a union of two hereditarily
realcompact subspaces.
T.405. Prove that if ' is a cardinal function and ' 2 fnetwork weight, spread,
Lindelf number, hereditary Lindelf number, density, hereditary density, extent,
Souslin number, point-finite cellularityg, then ' is completely additive and hence
countably additive.
S
Solution. Fix an infinite cardinal and a space X D fX W < g. Assume first
that nw.X / for any < . There Sexists a network N in the space X such
that jN j for all < . If N D fN W < g, then N is a network in X
and jN j whence nw.X / . This proves that network weight is completely
additive.
Now let s.X / for all < . If D X is discrete and jDj D C , then
there is < such that jD \ X j D C ; since D \ X is a discrete subspace of
X , this contradicts s.X / . Therefore s.X / , i.e., we proved that spread is
completely additive.
In case when l.X / for any < take an open cover U of the space X .
Since fUS \ X W U 2 Ug is an open cover of X , there isSU U such that
X S U and jU j for any < . It is clear that U 0 D fU W < g U
while U 0 D X and jU 0 j . Therefore l.X / which proves that the Lindelf
number is completely additive.
To deal with hereditary Lindelf number, S assume that hl.X / for every
< and take any Y X . Then Y D fY W < g where Y D Y \ X for
all < . Since hl.X / , we have l.Y / for all < . We have already
proved that the Lindelf number is completely additive, so l.Y / ; since Y X
was taken arbitrarily, we established that hl.X / , so hereditary Lindelf number
is also completely additive.
If d.X / for everyS < , then take Y X such that jY j and Y is
dense in X . Then Y D fY W < g is dense in X and jY j . This shows that
d.X / and proves that density is completely additive.
AsS to hereditary density, let hd.X / for all < and take any Y X . Then
Y D fY W < g where Y D Y \ X for all < . Since hd.X / , we
have d.Y / for all < . We have already proved that density is completely
additive, so d.Y / ; since Y X was taken arbitrarily, we established that
hd.X / so hereditary density is also completely additive.
To settle the case of extent, suppose that ext.X / for all < . If D X is
closed, discrete and jDj D C , then there is < such that jD \ X j D C ; since
D \ X is a closed discrete subspace of X , this contradicts ext.X / . Therefore
ext.X / , i.e., we proved that extent is completely additive.
Let c.X / for all < . If c.X / > , then there is a disjoint family
U .X / such that jUj D C . Observe that for each U 2 U, there is U < such
that U \XU ;. Consequently, there is < such that the family U 0 D fU 2 U W
U D g has cardinality C . As a consequence, V D fU \ X W U 2 U 0 g .X /
is a disjoint family with jVj D C ; this contradiction with c.X / shows that
c.X / and hence the Souslin number is completely additive.
Finally, assume that p.X / for all < . If p.X / > , then there is a
point-finite family U .X / such that jUj D C . Observe that for each U 2 U,
there is U < such that U \ XU ;. Consequently, there is < such
that the family U 0 D fU 2 U W U D g has cardinality C . As a consequence,
V D fU \ X W U 2 U 0 g .X / is a point-finite family with jVj D C ; this
contradiction with p.X / shows that p.X / and hence the point-finite
cellularity is completely additive.
T.406. Prove that pseudocompleteness, -weight and the Baire property are finitely
additive but not countably additive.
S Take a countable dense subspace D of the space Cp .I/. It follows from

Solution.
D D ffd g W d 2 Dg that the space D is a countable union of compact spaces
of countable -weight. Since D has no isolated points, it is of first category in
Itself and hence a countable union can destroy both pseudocompleteness and the
Baire property, i.e., neither pseudocompleteness nor the Baire property is countably
additive.
Now, w.Cp .I// > ! by Problem 171 of [TFS] and w.D/ D w.Cp .I//
by Fact 1 of T.187. Therefore w.D/ > ! and hence -weight is not countably
additive.
Fact 1. Suppose that X is a non-empty space with X D A1 [ S

[ An . Then there
exists a disjoint family U D fU1 ; : : : ; Un g .X/ such that U is dense in X and
Ui Ai \ Ui for any i n.
Proof. Consider the family V S D fV 2 .X / W there is i n such that V
V \ Ai g. We claim that G D V is dense in X . Indeed, if G X , then H D
X nG 2 .X / and Ai \ H is nowhere dense for any i n. Let W0 D W and
construct, for any i D 0; : : : ; n, a set Wi 2 .X / such that Wi C1 Wi for all
i < n and Wi \ Ai D ; for all i D 1; : : : ; n. Then Wn 2 .X / and Wn \ Ai D ;
for all i n which is a contradiction with X D A1 [

[ An ; consequently, the
set G is dense in X .
NowSconsider a maximal disjoint subfamily W of the family V. Then the set
W D W G is dense in G. Indeed, otherwise W 0 D GnW ; and there
is x 2 W 0 such that V \ Ai is dense in V for some i n and V 2 .x; X/.
As a consequence, V 0 D V \ W 0 2 V and the family W [ fV 0 g V is disjoint
and strictly larger than W which is a contradiction. Thus W is dense in G and
hence in SX . For every W 2 W fix i D iW n such that W Ai \ W and let
Uk D fW 2 W W iW D kg for every k n. It is immediate that the family
U D fU1 ; : : : ; Un g is as promised so Fact 1 is proved. t
u
Returning to our solution fix a space X such that X D X1 [

[ Xn for some
n 2 N;Sby Fact 1 there exists a disjoint family U D fU1 ; : : : ; Un g .X/ such that
G D U is dense in X and Ui Ui \ Xi for all i n. We can assume, without
loss of generality that Ui ; for all i n.
Assume first that Xi is pseudocomplete for all i n. Then the space Ui \ Xi is
pseudocomplete being open in Xi (see Problem 466 of [TFS]); since Xi \Ui is dense
in Ui , the space Ui is pseudocomplete for every i n by Problem 467 of [TFS].
Since G D U1 [

[ Un , every point of G has a pseudocomplete neighborhood,
so G is pseudocomplete by Fact 2 of S.488. Since G is dense in X , the space X is
also pseudocomplete by Problem 467 of [TFS]. Therefore pseudocompleteness is
finitely additive.
Now suppose that each Xi has the Baire property. Then, for any i n, the space
Ui \Xi is Baire being open in a Baire space Xi (see Problem 275 of [TFS]). Thus we
can apply Problem 275 of [TFS] again to conclude that Ui is Baire for every i n. If
G is not Baire, then some H 2 .G/ is of first category; then H 0 D H \Ui ; for
some i n and hence H 0 is also of first category which is a contradiction because
H 0 is a non-empty open subset of a Baire space Ui . This contradiction shows that G
is Baire and hence X is also Baire (here we used Problem 275 of [TFS] once more).
This completes the proof of finite additivity of the Baire property.
In case when w.Xi / for every i n we have w.Xi \ Ui / for any
i n (it is an easy exercise that -weight of an open subspace of a space does not
exceed the -weight of the space). Therefore w.Ui / for all i k by Fact 1
of T.187. Take a -base Bi in the space Ui for which jBi j for every i n. If
B D B1 [

[ Bn , then B is a -base in G with jBj . Therefore w.X / D
w.G/ (we used Fact 1 of T.187 again) which proves finite additivity of -
weight and finishes our solution.
T.407. Considering any Mrowka space, prove that normality is not finitely additive.
Solution. The Mrowka space M can be represented as E [ D where E is the
set of isolated points of M and D is a closed discrete subspace of M (see Problem
142 of [TFS]). The subspaces E and D are discrete and hence normal. However, the
Mrowka space is not normal because it is pseudocompact but not countably compact
(see Problem 137 and 142 of [TFS]). Thus normality is not finitely additive.
T.408. Suppose that X X D Y [ Z, where Y and Z are first countable. Prove
that X is also first countable.
Solution. Fix an arbitrary x 2 X ; we can assume, without loss of generality, that
a D .x; x/ 2 Y . Let Vx D fxg X and assume that .x; y/ Y for some y 2 X .
Then we can find U 2 .x; X/ such that .U fyg/\Y D ; and therefore U fyg
Z which implies .U fyg/ D !. Since U fyg is homeomorphic to U , we have
.U / D ! and hence .x; X / D .x; U / D !.
Now, if Vx Y , then ! D .a; Y / D .a; Y / (see Fact 1 of S.158) and therefore
.a; Vx / .a; Y / D !. It is evident that the projection onto the second factor
maps Vx homeomorphically onto X in such a way that .a/ D x. Thus .x; X / D
.a; Vx / D !. The point x 2 X was chosen arbitrarily, so we proved that .x; X / D
! for any x 2 X and hence X is first countable.
T.409. Suppose that X X D Y [ Z, where Y and Z have countable
pseudocharacter. Prove that .X / !.
Solution. We must prove that every x 2 X is a G -set so take an arbitrary point
x 2 X . We can assume, without loss of generality, that a D .x; x/ 2 Y and hence
space X X such that H \ Y D fag. If O D fOn W
there is a G -set H in theT
n 2 !g .X X / and O D H , then T we can choose Un 2 .x; X/ such that
Un Un On for all n 2 !. Then P D fUn W n 2 !g is a G -set in X such that
a 2 P P H and therefore .P P / \ Y D fag.
If P D fxg, then .x; X / D ! and we are done; if there is y 2 P nfxg then
.P fyg/ \ Y D ; and hence P fyg Z which implies .P fyg/ D !. Since
P fyg is homeomorphic to P , we have .P / D !; this shows that .x; X /
.x; P / D ! by Fact 2 of S.358. We proved that .x; X / D ! for any x 2 X and
hence .X / !.
T.410. Suppose that X X D Y [ Z, where Y and Z have countable tightness.
Prove that t.X / !.
Solution. Take a set A X and x 2 AnA; we can assume, without loss of
generality, that a D .x; x/ 2 Y . For every y 2 X let Hy D X fyg; if there
is a point y 2 X such that .x; y/ Y \ Hy , then there is U 2 .x; X/ for which
.U fyg/\Y D ; and hence U fyg Z which shows that t.U fyg/ D !. Since
U is homeomorphic to U fyg, we have t.U / D !; it is evident that x 2 A \ U
and hence there is a countable B A \ U A such that x 2 B.
Now assume that .x; y/ 2 Y \ Hy for any y 2 X and let W X X ! X be the
natural projection onto the second factor. Since a 2 fxg A and .x; y/ 2 Y \ Hy
for any y 2 A, we have a 2 Y \ 1 .A/. It follows from t.Y / D ! that there
is a countable C Y \ 1 .A/ such that a 2 C . By continuity of , we have
x D .a/ 2 .C /. Thus B D .C / is a countable subset of A such that x 2 B.
We established that for any A X and x 2 A, there is a countable B A with
x 2 B. This proves that t.X / !.
T.411. Suppose that X X D Y [ Z, where Y and Z have countable weight. Prove
that w.X / !.
Solution. We have nw.Y / D nw.Z/ D !; since the network weight is finitely
additive by Problem 405, we have nw.X / nw.X X / D !. We will often use
the following property (whose proof is left to the reader as an exercise):
(1) If T is a space and D is dense in T , then U \ D D U for any U 2 .T /.
We have to establish first that
(2) for the set F D Y \ Z, we have w.F / D !.
Let BY and BZ be countable bases in Y and Z respectively. For any U 2 BY
take GU 2 .Y / such that GU \ Y D U ; for any V 2 BZ , take HV 2 .Z/ such
that HV \ Z D V . To see that B D fGU \ F W U 2 BY g [ fHV \ F W V 2 BZ g is
a base in F take any a 2 F and W 2 .a; F /. Since a 2 Y [ Z, we have a 2 Y or
a 2 Z. We will consider only the case when a 2 Y because the proof for a 2 Z is
identical.
Since F Y , we can pick W 0 2 .Y / with W 0 \ F D W ; there exists O 2
.a; Y / such that O W 0 (observe that the closures of subsets of Y in X X
and in Y coincide, so we use the same bar for both). Since BY is a base in Y ,
there is U 2 BY such that a 2 U O \ Y . Since G U D U by (1), we have
a 2 U G U D U O \ Y O W 0 . As a consequence, a 2 GU \ F
G U \ F W 0 \ F D W which shows that for any a 2 F and W 2 .a; F /, there
is B 2 B such that a 2 B W , i.e., B is a countable base in F so (2) is proved.
Now let Px D fxg X for any x 2 X ; if Px Y \ Z, then w.Px / w.F / !
by (2); since Px is homeomorphic to X , we have w.X / D ! and our proof is over.
Thus we can assume that
(3) for any x 2 X , there is y 2 X such that a D .x; y/ F and hence a Y or
a Z.
It follows from (3) that for every x 2 X there is y 2 X and Ux 2 .x; X/ such
that .Ux fyg/ \ Y D ; or .Ux fyg/ \ Z D ; and hence Ux fyg is contained in
one of the sets Y; Z, so w.Ux fyg/ D ! and therefore w.Ux / D ! because Ux is
homeomorphic to Ux fyg. Since l.X / nw.X / D !, there is a countable A X
S
S D fUx W x 2 Ag. Now if Cx is a countable base in Ux for all x 2 A,
such that X
then C D fCx W x 2 Ag is a countable base in X , so w.X / D ! and our solution
is complete.
T.412. Suppose that X is a separable space such that X X D Y [ Z, where Y
and Z are metrizable. Prove that X is metrizable.
S
Solution. A family A of subsets of a set T is -disjoint if A D fAn W n 2 !g
and An is disjoint for every n 2 !. If T is a space, A exp.T / and S T , then
AjS D fA \ S W A 2 Ag. We will often use the following properties (their easy
proof is left to the reader as an exercise):
(1) If T is a space and D is dense in T , then U D U \ D for any U 2 .T /.
(2) If T is a space and D is dense in T , then a family U .T / is disjoint if and
only if UjD is disjoint.
Fact 1. Suppose that T is a space such that T D T0 [

[Tn and Ti has a -disjoint
base for every i n. Then F D T 0 \

\ T n has a -disjoint base.
Proof. Let Bi be a -disjoint base of Ti for each i n. For every i n and B 2 Bi ,
fix a set OB 2 .T i / such that OB \ Ti D B; it follows fromS(2) that the family
Ci D fOB W B 2 Bi g is -disjoint. Therefore the family C D fCi jF W i ng
.F / is -disjoint; let us prove that C is a base in F .
To that end, take any x 2 F and U 2 .x; F /; there is j n such that x 2 Tj .
Since F T j , there is V 2 .T j / such that V \ F D U ; apply regularity of T j
to find W 2 .x; T j / such that W V . The family Bj being a base in Tj , there is
B 2 Bj such that x 2 B W \ Tj . We have B D O B by (1) and therefore x 2
OB O B D B W \ Tj W V which implies x 2 OB \ F V \ F D U ;
thus, for arbitrary x 2 F and U 2 .x; F /, we found a set C D OB \ F 2 C such
that x 2 C U ; this implies that C is a base in F so Fact 1 is proved. t
u
Fact 2. Suppose that T is a space and U is an open cover of T . If jUj and
w.U / for every U 2 U, then w.T / .
Proof. Choose a base BS
U of the space U such that jBU j for any U 2 U. It
is immediate that B D fBU W U 2 Ug is a base in T and jBj so Fact 2 is
proved. t
u
Returning to our solution observe that it suffices to show that the space X is
second countable; to start off, let B D .X X /nY and C D .X X /nZ. Then
B; C 2 .X X / while B Z and C Y . This shows that both B and C are
metrizable; since X X is separable, they are both separable and hence second
countable. Let W X X ! X be the natural projection onto the first factor. Then
is open so both jB and jC are open maps. It is evident that open maps do
not raise weight so w..B// D w..C // D !. The sets B 0 D .B/ and C 0 D
.C / are open in X , so we can apply Fact 2 to conclude that w.B 0 [ C 0 / !. If
B 0 [ C 0 D X , then we are done; if not, then there is x 2 X n.B 0 [ C 0 /. It is evident
that 1 .x/ D fxg X F D Y \ Z. The space F has a -disjoint base by Fact 1
and hence fxg X also has a -disjoint base. The space X being homeomorphic to
fxg X , we have a -disjoint base in X as well. But any -disjoint family of open
sets in a separable space is countable (this is an easy exercise), so X has a countable
base and our solution is complete.
T.413. Suppose that X is a compact space such that X X D Y [ Z, where Y and
Solution. It suffices to show that X is second countable. Any metrizable space has
a -discrete base by Problem 221 of [TFS]; since any -discrete base is -disjoint,
both spaces Y and Z have a -disjoint base. Thus the space F D Y \ Z also has
a -disjoint base by Fact 1 of T.412. It is evident that a -disjoint base is point-
countable;
T besides, any base B of the space F is a T1 -separating family in the sense
that fU 2 B W x 2 U g D fxg for any x 2 F . Therefore F has a point-countable
T1 -separating family of open sets, so it is metrizable and hence second countable by
Fact 1 of T.203.
Let W X X ! X be the natural projection onto the first factor. If p.F / D X ,
then X is metrizable by Fact 4 of S.368. If there is x 2 X np.F /, then, for any
y 2 X , the point .x; y/ does not belong either to Y or to Z which shows that
there is Uy 2 .y; X/ for which fxg Uy is contained either in Y or in Z and
hence fxg Uy is metrizable. Since fxg Uy is homeomorphic to Uy , the space
Uy is metrizable. The space X is compact, so there is a -compact Vy 2 .X/
such that y 2 Vy Uy . The space Vy is also metrizable; being -compact, it
is second countable. Finally, apply compactness of X to choose a finite subcover
fVy1 ; : : : ; Vyn g of the open cover fVy W y 2 X g of the space X . Now, Fact 2 of T.412
implies that w.X / ! which completes our solution.
T.414. Give an example of a non-metrizable space X such that X X is a union of
two metrizable subspaces.
Solution. Denote by L the set of limit ordinals of !1 , and choose for any 2 L, a
sequence S D fi ./ W i 2 !g .!1 nL/ \ which converges to . If 2 I D
!1 nL, then let B D ffagg; if 2 L, then B D fOn ./ [ fg W n 2 !g where
On ./ D fi ./ W i ng for every n 2 !. S
If is the topology generated by the family B D fB W < !1 g as a base, then
X D .!1 ; / is a Tychonoff zero-dimensional space for which X D L [ I where
all points of I are isolated and L is closed and discrete in X . It is also evident that
X is first countable, locally compact and the family B is a local base at for any
< !1 .
Assume that the space X is collectionwise normal. Then there is a disjoint family
fU W 2 Lg such that 2 U for all 2 L. Making every U smaller if
necessary we can assume, without loss of generality, that U D On./ ./ [ fg
for each 2 L. We have n./ ./ 2 On./ ./ for any 2 l and hence the map
' W L ! !1 defined by './ D n./ for every 2 L, is an injection.
However, L is a stationary subset of !1 by Problem 064 and './ < for any
2 L, so there exists < !1 such that the set f 2 L W './ D g is stationary
(see Problem 067). Therefore ' is not injective; this contradiction proves that X is
not collectionwise normal which implies that X is not paracompact and hence not
metrizable (see Problem 231 of [TFS]).
Now let us consider the sets Y D f.; / 2 X X W maxf; g 2 Lg and
Z D f.; / 2 X X W maxf; g 2 I g. It is immediate that Y [ Z D X X
and Y \ Z D ;. For any < !1 let M D f.;
/ 2 X X W maxf;
g D g.
It is clear that M D .. C 1/ fg/ [ .fg . C 1// for any 2 !1 . The
set C 1 D f W g is open in X for any 2 !1 ; if 2 I , then fg
in open Sin X . As a consequence, M is open in X X for any 2 I . We have
Z D fM W 2 I g and the family fM W 2 I g is disjoint, so Z is a union
of its clopen countable subspaces. Since X is first countable, every LM is second
countable and hence metrizable. The space Z is homeomorphic to fM W 2 I g
(see Problem 113 of [TFS]) so it is metrizable by Fact 1 of S.234.
It turns out that M is also open in Y for any 2 L. To see it, assume first that
< and z D .; / 2 M . There is n 2 ! such that i ./ > for all i n. As
a consequence, U D .On ./ [ fg/ . C 1/ is an open neighborhood of z; given
y D .
; 0 / 2 U , if
D , then y 2 M . If
2 On ./ then
D maxf
; 0 g 2 I
and hence y Y which shows that U \ Y M .
Analogously, if z D .; / 2 M , then V D . C 1/ .On ./ [ fg/ is
an open neighborhood of z in X X such that V \ Y M . The last case is
when z D .; / 2 M ; here the set W D .O0 ./ [ fg/ .O0 ./ [ fg/ is an
open neighborhood of z in X X and W nM I I Z which shows that
W \ Y M . We proved that every point of M has a neighborhood in Y which is
contained in M . Therefore fM W 2 Lg is a disjoint open cover of Y .
Since X is first countable, every
LM is second countable and hence metrizable.
The space Y is homeomorphic to fM W 2 Lg (see Problem 113 of [TFS]), so
it is metrizable by Fact 1 of S.234. This proves that for a non-metrizable space X ,
we have X X D Y [ Z where both Y and Z are metrizable.
S
T.415. Suppose that X ! D fXn W n 2 !g. Prove that for some n 2 !, there is
Y Xn such that there exists an open continuous map of Y onto X ! and hence
there exists an open continuous map of Y onto X . As a consequence, if X ! is a
countable union of first countable subspaces, then X is first countable.
Q
Solution. If we haveQ a product Y D fYt W t 2 T g and a set S T , then
pS W Y ! YS D fYt W t 2 S g is the natural projection defined by pS .f / D f jS
for any f 2 Y . Say that a set H Y covers all finite faces of Y if pA .H / D YA
for any finite A T .
Q
Fact 1. Suppose that we have a product Y D fYt W t 2 T g and a set H Y
which covers all finite faces of Y . Then the map p D pA jH W H ! YA is open for
any finite A T .
Q
Proof. Call a non-empty set U Y standard open if U D fUt W t 2 T g where
Ut 2 .Xt / for every t 2 T and the set supp.U / D ft 2 T W Ut Xt g is finite. The
family B of all standard open subsets of Y is a base in Y , so it suffices to prove that
p.U \ H / is open in YA for any U 2 B.
Q
We have U D t 2T Ut and henceQpA .x/.t/ D x.t/ 2 Ut for any x 2 U
and t 2 A. Therefore pA .x/ 2 V D t 2A Ut for any x 2 U which shows that
p.U \ H / pA .U / V . Now, if z 2 V , then apply the fact that H covers all
finite faces of Y to find y 2 H such that y.t/ D z.t/ for any t 2 A and y.t/ 2 Ut
for any t 2 supp.U /nA. It is evident that y 2 U \ H and p.y/ D z; the point z 2 V
was chosen arbitrarily, so V p.U \ H /, i.e., V D p.U \ H / is an open subset
of YA . Thus the map p is open and Fact 1 is proved. t
u
S
Fact 2. For an arbitrary space Z assume that Z ! D fZn W n 2 !g. Then, for
some n 2 !, there is H Zn such that H is closed in Zn and homeomorphic to
some G Z ! which covers all finite faces of Z ! .
Proof. Let pA W Z ! ! Z Q be the natural projection1for any A !. Observe
A
that for anyQy 2 ZA D fZn W n 2 Ag, the set pA .y/ is a product because
pA1 .y/ D fZn0 W n 2 !g where Zn0 D fy.n/g for all n 2 A and Zn0 D Z for
n 2 !nA. Assume first that
(1) there is a finite A ! and a point y 2 ZA such that some Zn \ pA1 .y/ covers
all finite faces of P D pA1 .y/.
Let H D Zn \ pA1 .y/; then H is closed in Zn and it is easy to see that the map
p D p!nA jpA1 .y/ W pA1 .y/ ! Z !nA is a homeomorphism such that G 0 D p.H /
covers all finite faces of Z !nA . For any bijection ' W ! ! !nA the map W
Z !nA ! Z ! defined by .f / D f ' for every f 2 Z !nA is a homeomorphism
such that G D .G 0 / covers all finite faces of Z ! . This shows that the set H Zn
is as promised, so if (1) holds, then our fact is true.
Now assume that (1) is false and hence Z0 does not cover all finite faces of Z ! ;
choose a finite A0 ! and y0 2 Z A0 such that pA10 .y0 / \ Z0 D ;. Suppose that for
some n > 0, we have finite sets A0 ; : : : ; An and points y0 ; : : : ; yn with the following
properties:
(2) Ai ! is finite and yi 2 Z Ai for all i n;
(3) Ai Ai C1 and yi C1 jAi D yi for all i < n;
(4) pA1i .yi / \ Zi D ; for all i n.
Since (1) fails, the set ZnC1 \ pA1n .yn / does not cover all finite faces of pA1n .yn /,
so we can choose a finite set B !nAn and a point z 2 Z B for which pB1 .z/ \
pA1n .yn / does not meet ZnC1 . Let AnC1 D An [ B and define ynC1 2 Z AnC1 by
ynC1 jAn D yn and ynC1 jB D z. It is immediate that (2)(4) are still fulfilled for
all i n C 1, so we can continue our inductive construction to obtain a family
fAi W i 2 !g of finite subsets of ! and a set fyi W i 2 !g for which (2)(4) hold
for all i 2 !. It follows from (3) that there is y 2 Z ! such that pAn .y/ D yn for
all nS 2 !. As a consequence, for all n 2 !, we have y Zn by (4) which implies
y n2! Zn ; this contradiction shows that (1) is true and hence Fact 2 is proved.
t
u
it is easy to finish our solution. Let Z D X ! ; then Z ! D .X ! /! D
Now S
X D fXn W n 2 !g, so by Fact 2, there is n 2 ! and Y Xn such that
!
Y is homeomorphic, so some G Z ! which covers all finite faces of Z ! . By

Fact 1, the restriction to G of the natural projection onto the first factor is open (and,
evidently, continuous and onto). Thus G maps openly onto Z D X ! and hence so
does Y .
Finally, if Xn is first countable for all n 2 !, then choose n 2 ! such that some
Y Xn maps openly onto X . It is clear that Y is first countable and it is an easy
exercise that an open continuous image of a first countable space is first countable.
Thus X is first countable and our solution is complete.
T.416. Given an arbitrary space X , suppose that X ! is a finite union of metrizable
subspaces. Prove that X is metrizable.
Solution. Given a space Z and a family A of subsets of Z, let AjY D fA \ Y W
A 2 Ag for any Y Z.
Fact 1. If Z is a space and Y is dense in Z, then for any B .Z/ and y 2 Y the
family B contains a local base at y in Z if and only if BjY contains a local base at
y in Y .
Proof. If C B is a local base at y in Z, then CjY BjY and it is evident that
CjY is a local base at y in Y . Now assume that C B and CjY is a local base at y
in Y . Given any U 2 .y; Z/, pick V 2 .y; Z/ such that V U and W 2 C for
which y 2 W \ Y V \ Y . Then y 2 W W D W \ Y V \ Y V U
and hence y 2 W U which proves that C is a local base at y in Z, so Fact 1 is
proved. t
u
For any A ! the map pA W X ! ! X A is the natural projection of X ! onto its
face X A . Call a set Y X ! strongly dense in X ! if Y covers all finite faces of X ! ,
i.e., pA .Y / D XA for any finite A !.
Suppose that X ! D Y1 [

[ Yn and Yi is metrizable for each i n. Then
(1) there is k n and metrizable subspaces Y10 : : : : ; Yk0 of the space X ! such that
X ! D Y10 [

Yk0 and Yi0 is strongly dense in X ! for every i k.
We will prove (1) by induction on n; the case of n D 1 being evident assume that
n > 1 and we proved (1) for all k < n. If there is i n such that Yi is not strongly
dense in X ! , then there is a finite A ! and z 2 X A such that pA1 .z/\Yi D ;. The
space pA1 .z/ is homeomorphic to X ! and pA1 .z/ D Z1 [

[Zi 1 [Zi C1 [

[Zn
where Zj D Yj \ pA1 .z/ for any j n; j i .
Thus X ! is represented as a union of n 1 metrizable subspaces, so the
induction hypothesis is applicable to guarantee existence of metrizable Y10 ; : : : ; Yk0
as in (1). Now, if every Yi is strongly dense in X ! , then we can take Yi0 D Yi for all
i n so (1) is proved.
It follows from (1) that we can assume, without loss of generality, that Yi is
strongly dense in X ! for every i n. Fix a family Bi .X ! / such that Bi jYi is a
-discrete base of Yi for all i n. Every Yi is dense in X ! so Bi contains a local
base in X ! at every point of Yi by Fact 1; therefore B D B1 [

[ Bn is a base
in X ! .
For any k 2 !, we identify k with the set f0; : : : ; k 1g so pk W X ! ! X k

is the respective natural projection pA for AS D f0; : : : ; k 1g. For any k 2 !, let
Ck D fpk1 .U / W U 2 .X S
k
/g; then C D fCk W k 2 !g is a base in X ! . For
any US2 .X / let Uk D fV 2 Ck W V U g for every k 2 !. It is clear that
!
U D k2! Uk and pk1 .Uk / D Uk for all k 2 !; besides, if U W , then Uk Wk

for all k 2 !. S
We have Bi D fBim W m 2 !g where Bim jYi is discrete in Yi for any m 2 !
and i n. Let Bim .k/ D fUk W U 2 Bimg for every k; m 2 ! and i n. It follows
from discreteness of Bim jYi that the family Eim .k/ D Bim .k/jYi is discrete in Yi for
all m; k 2 ! and i n.
Let Him .k/ D fpk .O/ W O 2 Bim.k/g and take any y 2 X k ; since Yi is strongly
dense in X ! , there is z 2 Yi with pk .z/ D y. The family Eim .k/ is discrete in
Yi , so there is H 2 .z; Yi / which intersects at most one element of Eim .k/. The
map pk jYi is open by Fact 1 of T.415, so G D pk .H / is an open neighborhood
of y in X k . Now, if O 2 Bim.k/ and G \ pk .O/ ;, then O \ H ; because
O D pk1 .pk .O//. Furthermore, H Yi implies that H \ .O \ Yi / ;; since
O \ Yi 2 Eim .k/, discreteness of Eim .k/ implies that there is at most one O 2 Bim.k/
such that O \ H ;. This proves that G intersects at most one element of Him .k/
and therefore Him .k/ is discrete in X k for any m; k 2 ! and i n.
It turns out that the images of the elements of Bim.k/ under pk form a discrete
S m is that Bi .k/ is discrete in X for all
m
family in X k ; an immediate consequence !
0
m; k 2 ! and i n. Thus B D fBi .k/ W m; k 2 !; i ng is a -discrete
family of open subsets of X ! . It is easy to see that every element of B is a union
of a subfamily of B 0 , so B 0 jYi is a base in Yi for all i n. Finally, apply Fact 1
to conclude that B 0 is a -discrete base of X ! which shows that X ! is metrizable
(see Problem 221 of [TFS]). This implies that X is metrizable so our solution is
complete.
T.417. Given a countably compact space X , suppose that X ! is a countable union
of metrizable subspaces. Prove that X is metrizable.
Solution. Given a space Z and a family A of subsets of Z, let AjY D fA \ Y W
A 2 Ag for any Y Z. If A !, then pA W X ! ! X A is the natural projection of
X ! onto its face X A . Call a set Y X ! strongly dense in X ! if Y covers all finite
.Y / D X A for any finite A !.
faces of X ! , i.e., pA S
We have X D fXn W n 2 !g where Xn is metrizable for all n 2 !. This
!
implies .X / D ! (see Problem 415) and hence it follows from Fact 1 of S.322 that
X n is countably compact for every n 2 N. Now apply Fact 2 of T.415 to find n 2 !
and Xn0 Xn which is homeomorphic to a strongly dense M X ! .
S the space M is metrizable, we can choose a family B .X / such that
!
Since
B D n2! Bn where Bn jM is discrete in M and BjM is a base of M .
For any k 2 ! we identify k with the set f0; : : : ; k 1g, so pk W X ! ! X k
is the respective natural projection pA for AS D f0; : : : ; k 1g. For any k 2 !, let
Ck D fpk1 .U / W U 2 .XSk /g; then C D fCk W k 2 !g is a base in X ! . For
any U 2 .X ! /, let Uk D fV 2 Ck W V U g for every k 2 !. It is clear that
S
U D k2! Uk and pk1 .Uk / D Uk for all k 2 !; besides, if U W , then Uk Wk
for all k 2 !.
Let Bn .k/ D fUk W U 2 Bn g and Cn .k/ D fpk .U / W U 2 Bn .k/g for every
n; k 2 !. It follows from discreteness of Bn jM that the family Bn .k/jM is discrete
in M for all n; k 2 !.
The set M being strongly dense in X ! , for any z 2 X k , there is y 2 M for
which pk .y/ D z. Since the family Bn .k/jM is discrete, there is H 2 .y; M /
which intersects at most one element of Bn .k/jM ; the set G D pk .H / is open in
X k because the map pk jM W M ! X k is open by Fact 1 of T.415. If G\pk .O/ ;
for some O 2 Bn .k/, then H \ O ; because pk1 .pk .O// D O. Since H M ,
we have H \ .O \ M / ;; furthermore, O \ M 2 Bn .k/jM , so O \ M is the
unique element of Bn .k/jM which meets H . The family Bn .k/ is disjoint (because
so is Bn .k/jM ), so pk .O/ is the unique element of Cn .k/ which meets G. This
proves that Cn .k/ is a discrete family of open sets of X k ; since X k is countably
compact, Cn .k/ has to be finite
S and hence so is Bn .k/ for all n; k 2 !.
As a consequence, B 0 D fBn .k/ W n; k 2 !g is countable; observe that every
element of B is a union of a subfamily of B 0 , so B 0 jM is a countable base in M . The
space X is an open continuous image of a second countable space M (see Fact 1 of
T.415) so w.X / D ! and hence X is metrizable.
T.418. Give an example of a non-metrizable space X such that X ! is a countable
union of its metrizable subspaces.
Solution. If A is a family of subsets of a space Z and we have a subspace Y Z,
then AjY D fA \ Y W A 2 Ag. Any n 2 ! is identified with the set f0; : : : ; n 1g.
For the sake of brevity we call a space -metrizable if it is a countable union of
its metrizable subspaces. Denote by L the set of all limit ordinals of !1 . For any
2 I D !1 nL there exist unique n 2 ! and 2 L [ f0g such that D C n.
To see it, let 0 D ; if 0 D 0 then we are done. If not, then there is 1 2 !1
with 0 D 1 C 1. Suppose that we have a sequence 0 ; : : : ; n !1 such that
kC1 C 1 D k for any k < n. If n is a limit ordinal or zero, then D 0 D n C n
and we are done. If not, then there is nC1 2 !1 for which nC1 C1 D n , so we can
continue to construct our sequence inductively. However, this construction cannot
have infinitely many steps because !1 has no infinite decreasing sequences. Thus n
has to be limit or zero for some n and hence D n C n, so we can take D n .
The uniqueness of and n is an easy exercise.
Let I0 D f0g and In D f 2 I W D C n forSsome 2 L [ f0gg for every
n 2 N. It is clear that In \ Im D ; if n m and fIn W n 2 !g D I . Observe
that
(1) for every 2 L, there is a sequence S D fn ./ W n 2 !g I \ such that
n ./ ! and n ./ 2 In for every n 2 !.
Indeed, if is an isolated point of L (with the interval topology induced from
!1 ), then for the ordinal D supf
2 L [ f0g W
< g we have 2 L [ f0g and
C n ! , so we can take 0 ./ D 0 and n ./ D C n for all n 2 N.
If supf
2 L W
< g D , then choose a sequence fn W n 2 !g L such that
n ! and n < nC1 for every n 2 !. It is immediate that n C n 2 In \ and
n C n ! , so letting 0 ./ D 0 and n ./ D n C n for all n 2 N we conclude
the proof of (1).
For any 2 I , let B D ffagg; if 2 L, then B D fOn ./ [ fg W n 2 !g
where On ./ D fi ./ W i ng for every n 2 !. S
If is the topology generated by the family B D fB W < !1 g as a base, then
X D .!1 ; / is a Tychonoff zero-dimensional space for which X D L [ I where
all points of I are isolated and L is closed and discrete in X . It is also evident that
X is first countable, locally compact and the family B is a local base at for any
< !1 .
Assume that the space X is collectionwise normal. Then there is a disjoint family
fU W 2 Lg such that 2 U for all 2 L. Making every U smaller if
necessary we can assume, without loss of generality, that U D On./ ./ [ fg
for each 2 L. We have n./ ./ 2 On./ ./ for any 2 L and hence the map
' W L ! !1 defined by './ D n./ ./ for every 2 L is an injection.
However, L is a stationary subset of !1 by Problem 064 and './ < for
any 2 L, so there exists < !1 such that the set f 2 L W './ D g is
stationary (see Problem 067). Thus ' is not injective; this contradiction proves that
X is not collectionwise normal which implies that X is not paracompact and hence
not metrizable (see Problem 231 of [TFS]).
We will prove that X ! is -metrizable. As a first step note that X D L [ I is a
union of two discrete and hence metrizable subspaces. Therefore X n is a union of
finitely many metrizable subspaces which shows that S X n L! is also a finite union
of metrizable spaces for any n 2 !. Thus the set Y D fX n L!nn W n 2 Ng X !
is -metrizable.
For any f 2 X ! and n 2 N let Mn .f / D maxff .i / W i < ng and consider
the set Z D ff 2 X ! W for any k 2 !, there is n > k such that f .n/ 2 I and
Mn .f / 2 I g. Our purpose is to show thatQ Z is metrizable.
Call a set U X ! standard if U D i 2! Ui and there exists n D l.U / 2 N
such that Ui D X for all i n and Ui 2 B for all i < n. Thus Ui is a basic
neighborhood of its top point max.Ui / for all i < n. Denote by E the family of
all standard subsets of X ! ; it is evident that E is a base in X ! .
We will need the map n W X ! ! X which is the natural projection of the space
Q ordinal 2 I and m; n 2 ! such that m < n the
!
X onto its nth factor. For fixed
family O.; n; m/ D fU D i 2! Ui 2 E W l.U / D n; Um D fg and max.Uj /
for all j < ng is countable, so we can enumerate it as fU.; n; m; i / W i 2 !g.
Now, consider the family C.n; m; i; j; k/ D fU.; n; m; i / \ n1 ./ W 2 Ij ;
2 Ik g. We claim that
S
(2) C D fC.n; m; i; j; k/ W n; m; i; j; k 2 ! and m < ng contains a local base in
!
X at every point of Z.
Q
To prove (2) take any f 2 Z and W D i 2! Wi 2 E such that f 2 W .
It is clear that we can consider that Wi 2 Bf .i / for every i < l.W /. It follows
from the definition of Z that there is n 2 ! such that n > l.W /; f .n/ 2 I and
Mn .f / D 2 I . There are j; k 2 ! such that 2 Ij and D f .n/ 2 Ik .
Furthermore, f .m/ D for some m < n; let Vi D Wi for all i < l.W /. For every
i 2 fl.W /; : : : ; ng choose
Q Vi 2 Bf .i / arbitrarily and let Vi D X for all i > n. It
is immediate that V D i 2! Vi 2 O.; n; m/ and hence there is i 2 ! for which
V D U.; n; m; i / so G D V \ n1 ./ 2 C.n; m; i; j; k/ and f 2 G W so (2)
is proved.
To show that C.n; m; i; j; k/ is discrete in XS !
for any n; m; i; j; k 2 ! with
m < n take any f 2 X and assume that f 2 C.n; m; i; j; k/. Let D f .m/
!
and D f .n/; if S Ij , then m1 .!1 nIj / is an open neighborhood of f in X !

which does not meet C.n; m; i; j; k/. Analogously, if S Ik , then n1 .!1 nIk / is
an open neighborhood of f in X ! which does not meet C.n; m; i; j; k/.
Thus 2 Ij and 2 Ik and it is easy to see that H D m1 ./ \ n1 ./ is an
open neighborhood of f in X ! which meets at most one element of C.n; m; i; j; k/
because no H 2 C.n; m; i; j; k/ distinct from U.; n; m; i / \ n1 ./ can meet H .
This proves that CjZ is a -discrete base of Z, so the space Z is metrizable.
Now consider the set T D ff 2 X ! nY W there is k 2 ! such that for any n > k
if f .n/ 2 I , then Mn .f / 2 Lg. It is easy to see that X ! D Y [ Z [ T , so it remains
to show that T is -metrizable. Let Tk D ff 2 X ! nY W for any S n > k if f .n/ 2 I ,
then Mn .f / 2 Lg for any k 2 !. It is immediate that T D k2! Tk , so it suffices
to show that every Tk is metrizable.
Given a number nQ2 !, an ordinal 2 L and a non-empty set F n let
G.; F; n/ D fU D i 2! Ui 2 E W l.U / D n; max.Uj / D for any j 2 F and
max.Ui / < min.Uj / for any j 2 F and i 2 nnF g; since this family is countable,
we can enumerate its elements as fV .; F; n; i / W i 2 !g. Now let D.F; n; i; m/ D
fV .; F; n; i / \ n1 ./ W 2 L and 2 Im g for any n; m; i 2 ! and a finite
non-empty F n. Next we show that
S
(3) the family D D fD.F; n; i; m/ W n; i; m 2 !; n > k and ; F ng
contains a local base in X ! at any point f 2 Tk .
Q
To prove (3) take any f 2 Tk and W D i 2! Wi 2 E such that f 2 W ; as
before, we can assume that Wi 2 Bf .i / for any i < l.W /. Since f Y , there is
n > k C l.W / such that D f .n/ 2 I . By definition of Tk the ordinal D Mn .f /
belongs to L; clearly, F D fi < n W f .i / D g is a non-empty subset of n. Observe
that f .i / D is a limit ordinal for any i 2 F while f .j / < for any j 2 nnF .
Thus we can find a family fV0 ; : : : ; Vn1 g B such that f .i / 2 Vi Wi for all
i < n and max.Vi / <Qmin.Vj / whenever j 2 F and i 2 nnF . Now, if Vi D X for
all i n, then V D i 2! Vi 2 G.; F; n/. Therefore V D V .; F; n; i / for some
i 2 !; besides, there is m 2 ! for which D f .n/ 2 Im , so W 0 D V \ n1 ./ 2
D.F; n; i; m/ and f 2 W 0 W so (3) is proved.
To finish the proof of metrizability of Tk it suffices to show that the family
D.F; n; i; m/jTk is discrete in Tk for any n; m; i 2 !; n > k and finite non-empty
F n. So take f 2 Tk such that every neighborhood of f in Tk meets an element
of D.F; n; i; m/jTk . If D f .n/ Im , then n1 .!1 nIm / is an open neighborhood
of f in X ! which does not meet any element of D.F; n; i; m/. Thus 2 Im ; by
definition of Tk , we have D Mn .f / 2 L; let F 0 D fj 2 n W f .j / D g.

For any j < n, choose Uj 2 Bf .j / so that max.Ui / < min.Uj / whenever Q
j 2 F 0 and i 2 nnF 0 and let Uj D X for all j n. If U D i 2! Ui
and W D U \ n1 ./, then W is a neighborhood of f in X ! . Assume that
g 2 Tk \ W \ V . 0 ; F; n; i / \ n1 .
/ for some 0 2 L and
2 Im . Then
D g.n/ D . Furthermore, Mn .g/ 2 L because g 2 Tk . Now g 2 U implies

g.j / 2 Uj for every j < n and hence Mn .g/ 2 Uj for some j 2 F 0 because if
j 2 nnF 0 , then g.j / max.Uj / < min Ui g.i / for any i 2 F 0 .
It follows from the definition of V . 0 ; F; n; i / that g 2 V . 0 ; F; n; i / implies
Mn .g/ D 0 ; since 0 2 Uj for some j 2 F 0 , we have 0 D because the unique
limit ordinal of Uj is . Therefore the only element of D.F; n; i; m/ which meets
W \ Tk is V .; F; n; i / \ n1 ./. This shows that D.F; n; i; m/jTk is discrete in
Tk and hence Tk is metrizable. Thus X ! D Y [ Z [ T is -metrizable and our
T.419. For an arbitrary space X and any f; g 2 C .X /, let
d.f; g/ D supfjf .x/ g.x/j W x 2 X g:
Prove that d is a complete metric on C .X / and the topology, generated by d ,

coincides with the topology of Cu .X /.
Solution. It follows from what was proved in Problem 248 of [TFS] that d is a
complete metric on C .X /. The coincidence of u and the topology .d / generated
by the metric d was proved in Fact 1 of T.357. It is worth mentioning that the
u
correspondence A ! A satisfies all conditions of Problem 004 of [TFS], so it
is indeed a closure operator which generates the topology u on C .X /; this was
proved in Problem 084 of [TFS].
T.420. Let P be an F -hereditary property, i.e., X ` P implies Y ` P whenever Y
is an F -subspace of X . Suppose that Cp .X / is a finite union of subspaces which
have the property P. Prove that Cp .X / is a finite union of dense subspaces each
one of which has the property P.
Solution. It is evident that it suffices to prove by induction on n 2 N the following
statement:
S.n/: if Cp .X / D Y1 [

[ Yn and every Yi has P, then there is m n such
that Cp .X / D Y10 [

[ Ym0 where every Yi0 is dense in Cp .X / and has P.
Since S.1/ is evident, assume that we have proved S.n/ for all n < k and
Cp .X / D Y1 [

[ Yk where Yi ` P for all i k. If every Yi is dense in
Cp .X / there is nothing to prove, so assume that Y j Cp .X / for some j k.
There exist x1 ; : : : ; xp 2 X and rational open intervals O1 ; : : : ; Op R such that
the set W D ff 2 Cp .X / W f .xi / 2 Oi for all i pg does not intersect Yj and
hence W Y1 [

[ Yj 1 [ Yj C1 [

[ Yk .
Let K D fx1 ; : : : ; xp g and C0 D ff 2 Cp .X / W f jK 0g. It was Q proved in
Fact 1 of S.409 that Cp .X / ' C0 RK . On the other hand, W ' C0 . i k Oi /
(see Fact 1 of S.494). Since Oi ' R for every i k, we have W ' Cp .X /.

Furthermore, every Oi is an F -set in R, so W is an F -subset of Cp .X / (this is a
consequence of the following easy fact which we leave to the reader as an exercise:
QA1 ; : : : ; Am are spaces and
if QBi Ai is an F -subset of Ai for all i m, then
i m B i is an F -subset of i m Ai ).
As a consequence, the set Zi D Yi \ W is an F -subset of Yi and hence Zi ` P
for every i k. Thus W D Z1 [

[ Zj 1 [ Zj C1 [

[ Zk ; since W is
homeomorphic to Cp .X /, we proved that Cp .X / can be represented as a union
k 1 spaces with the property P. Applying the induction hypothesis, we can find
m k 1 and dense sets Y10 ; : : : ; Ym0 Cp .X / such that Cp .X / D Y10 [

[ Ym0
and Yi0 ` P for every i m. This proves S.k/ and shows that S.n/ is true for every
n 2 N.
T.421. Let P be a hereditary property. Suppose that Cp .X / is a finite union of
subspaces which have the property P. Prove that there is n 2 N and " > 0 such that
u
Cp .X; . "; "// D Y1 [

[ Yn , where Yi ` P and Yi C.X; . "; "// for each
i 2 f1; : : : ; ng.
Solution. Given a space Z let dZ .f; g/ D supfjf .z/ g.z/j W z 2 Zg for
any functions f; g 2 C .Z/; then dZ is a metric on C .Z/ which generates the
topology uZ of the space Cu .Z/ (see Problem 419). For any function f 2 C .Z/
and r > 0, let BZ .f; r/ D fg 2 C .Z/ W dZ .f; g/ < rg and denote by 0Z the
function which is identically zero on Z. If P Q C .Z/, we will say that P is
uZ -dense in Q if the closure of P in Cu .Z/ contains Q.
Fact 1. If Z is a space, then for any " > 0 we have BZ .0Z ; "/ C.Z; . "; "// and
the set BZ .0Z ; "/ is uZ -dense in C.Z; "; " / for any " > 0.
Proof. If f 2 BZ .0Z ; "/, then jf .z/j D jf .z/ 0Z .z/j dZ .f; 0Z / < " for any
z 2 Z which proves that f 2 C.Z; . "; "// and hence BZ .0Z ; "/ C.Z; . "; "//.
Choose a sequence f"n W n 2 !g such that 0 < "n < " for all n 2 ! and
"n ! ". For any f 2 C.Z; "; " / define fn W Z ! R as follows: fn .z/ D f .z/ if
jf .z/j < "n ; if f .z/ "n , then fn .z/ D "n , and if f .z/ "n , then fn .z/ D "n .
It is an easy consequence of Problem 028 of [TFS] that fn 2 C.Z/ and hence
fn 2 C .Z/ for any n 2 !; besides, d.fn ; 0Z / "n < " for all n 2 ! and hence
ffn W n 2 !g BZ .0Z ; "/. We omit an easy verification of the fact that fn f
which shows that every f 2 C.Z; "; " / is a uniform limit of some sequence from
BZ .0Z ; "/ and therefore BZ .0Z ; "/ is uZ -dense in C.Z; "; " /. Fact 1 is proved.
t
u
Fact 2. Given a space Z and f 2 C .Z/ let L.g/ D g C f for any g 2 C .Z/.
Then dZ .g; h/ D dZ .L.g/; L.h// for any g; h 2 C .Z/ and hence L W Cu .Z/ !
Cu .Z/ is a homeomorphism.
Proof. We have jL.g/.z/ L.h/.z/j D jg.z/Cf .z/ .h.z/Cf .z//j D jg.z/ h.z/j
for any z 2 Z and g; h 2 C .Z/; taking the respective suprema, we obtain the
equality dZ .g; h/ D dZ .L.g/; L.h//. An immediate consequence is that gn g
if and only if L.gn / L.g/ for any sequence fgn W n 2 !g C .Z/ and
g 2 C .Z/. This shows that L W Cu .Z/ ! Cu .Z/ is a homeomorphism so Fact 2
is proved. t
u
Fact 3. If Z is a non-empty space
S and A is a finite cover of Z, then there is A0 A
and U 2 .Z/ such that U A and A \ U is dense in U for any A 2 A0 .
0
Proof. If this is not true, then

S
. / for any A0 A and U 2 .Z/ with U A0 there is A 2 A0 and
V .Z/ such that V U and V \ A D ;.
We have A D fA1 ; : : : ; An g for some n 2 N; by . /, there exists V1 2 .Z/
with V1 \ A1 D ;. Suppose that k < n and we have V1 ; : : : ; Vk 2 .Z/ such
that V1

Vk and Vi \ Ai D ; for all i k. Then Vk AkC1 [

[ An ,
so we can apply . / to the family A0 D fAkC1 ; : : : ; An g and the set Vk to find
VkC1 2 .Vk / such that VkC1 \ AkC1 D ;. This shows that we can inductively
construct V1 ; : : : ; Vn 2 .Z/ such S
that V1

Vn and S
Vi \ Ai D ; for all
i n. It is evident that Vn Zn. A/ which contradicts A D Z. Fact 3 is
proved. t
u
Returning to our solution suppose that Cp .X / D B1 [

[ Bm and Bi ` P
for every i m. Letting S Ai D Bi \ C .X / for every i m, we obtain the

equality Cu .X / D i m Ai ; since P is hereditary, every Ai has the property P
(as a subspace of Cp .X /). Applying Fact 3 to the space Cu .X / and the family
A DS fA1 ; : : : ; Am g we can find a non-empty W 2 uX and A0 A such that
W A0 and A \ W is uX -dense in W for any A 2 A0 .
Thus there is f 2S W and > 0 for which V D BX .f; / W . Since V is open
in W , we have V A0 and A \ V is uX -dense in V for any A 2 A0 . Let L.g/ D
g f for any g 2 C .X /. Then L W C .X / ! C .X / is a homeomorphism both
with respect to uX and .Cp .X //. For .Cp .X // this follows from Problem 079 of
[TFS] and for uX this was proved in Fact 2.
Observe that G D L.V / C.X; . ; //; it follows S 0 from Fact 2 that G D
BX .0X ; /. If B 0 D fL.A/ W A 2 A0 g, then G B and B \ G is uX -dense
in G for any B 2 B 0 . Therefore, for any O 2 uX with O G the set B \ O is
0
u -dense in O for any B 2 B . For " D 2 , we have C.X; . "; "// B.0X ; /
X
S
B . Besides, B \ BX .0X ; "/ is u -dense in BX .0X ; "/ for any B 2 B 0 ; applying
0 X
Fact 2 we conclude that B \ BX .0X ; "/ is uX -dense in C.X; . "; "// and hence
B \ C.X; . "; "// has to be uX -dense in C.X; . "; "// for any B 2 B 0 .
If B 0 D fB1 ; : : : ; Bn g and Yi D BSi \ C.X; . "; "// for all i n, then it follows
from the inclusion C.X; . "; "// B 0 that C.X; . "; "// D Y1 [

[ Yn . Every
Bi , considered as a subspace of Cp .X /, is homeomorphic to L1 .Bi / 2 A. Thus
Bi ` P for all i n. Since P is a hereditary property, the set Yi has P for all i n.
We already saw that Yi is uX -dense in C.X; . "; "//, so our solution is complete.
T.422. Suppose that Cp .X / is a finite union of its paracompact (not necessarily
closed) subspaces. Prove that Cp .X / is Lindelf and hence paracompact.
Solution. If Z is a space, A exp.Z/ and Y Z, then AjY D fA \ Y W A 2 Ag.

If A; B exp.Z/, then A is inscribed in B if for any A 2 A there is B 2 B such
that A B.
Fact 1. If Z is a paracompact space and Y Z is an F -set in Z, then Y is also
paracompact.
Proof. Let U be an open cover S of Y ; for any U 2 U, choose OU 2 .Z/ with
OU \ Y D U . We have Y D n2! Yn where Yn is closed in Z for any n 2 !. The
family Vn D fOU W U 2 Ug [ fZnYn g is an open cover of Z for any n 2 !, so we
can find an open locally finite refinement Wn of the cover Vn .
If Wn0 D
SfW 0 2 Wn W W \ Yn ;g, S then Wn0 is inscribed in V D fOU W U 2 Ug
and Yn Wn for every n 2 !. Thus fWn0 jY W n 2 !g is a -locally finite open
refinement of U and hence Y is paracompact by Problem 230 of [TFS]. Fact 1 is
proved. t
u
Returning to our solution assume that Cp .X / is a finite union of it paracompact
subspaces. It followsSfrom Fact 1 and Problem 420 that there is a finite family A
such that Cp .X / D A while every A 2 A is paracompact and dense in Cp .X /.
Therefore c.A/ D c.Cp .X // D ! and hence A is Lindelf for any A 2 A (see
Fact 2 of S.219). Consequently, Cp .X / is Lindelf being a finite union of Lindelf
spaces (see Problem 405). Since any Lindelf space is paracompact (see Problem
230 of [TFS]), this proves that Cp .X / is paracompact.
T.423. Suppose that Cp .X / D Y1 [

[Yn , where Yi is realcompact for each i n.
Prove that Cp .X / is realcompact.
Solution. Any F -subspace of a realcompact space is realcompact by Problem 408
of [TFS]. Therefore we can apply Problem 420 to convince ourselves that we can
assume, without loss of generality, that every Yi is dense in Cp .X / and hence in
RX . Since RX is a Moscow space by Problem 424 of [TFS] (applied to X with the
discrete topology), every Yi is !-placed in RX , i.e., for any f 2 RX nYi , there is a
G -set H in RX such that f 2 H RX nYi (see Problem 425 of [TFS]).
Therefore we can find for any f 2 RX nCp .X / and i n a G -subset Hi of the
space RX such that f 2 Hi RX nYi . If Hf D H1 \

\ Hn , then Hf is a G -
subset of RX such that f 2 Hf S RX nCp .X /. This shows that in a realcompact
space RX , the set RX nCp .X / D fHf W f 2 RX nCp .X /g is a union of G -subsets
of RX . Therefore Cp .X / is realcompact by Problem 408 of [TFS].

[ Yn , where Yi is homeomorphic to Ri for
each i n. Prove that X is discrete.
Solution. It follows from Problem 401, 465 and 470 of [TFS] that R is pseudo-
complete and realcompact for any cardinal . Applying Problem 423 we can see
that Cp .X / is realcompact. It is an immediate consequence of Problem 406 that
Cp .X / is pseudocomplete. Finally apply Problem 486 of [TFS] to conclude that X
is discrete.

[ Yn , where Yi is hereditarily realcompact
for each i n. Prove that i w.Cp .X // D .Cp .X // D ! and hence Cp .X / is
Solution. It follows from Problem 420 that we can assume, without loss of
generality, that every Yi is dense in Cp .X /. Fix any f 2 Cp .X /; it is clear that
Yi0 D Yi nff g is realcompact and dense in Cp .X / for any i n.
Since Cp .X / is a Moscow space by Problem 424 of [TFS], we can apply
Problem 425 of [TFS] to see that Yi0 is !-placed in Cp .X / and hence there is
a G -subset Hi of the space Cp .X / such that f 2 Hi Cp .X /nYi0 for every
i n. Consequently,SH D H1 \

\ Hn is a G -subset of Cp .X / such that
f 2 H Cp .X /n. i n Yi0 / D ff g. Thus ff g is a G -set in Cp .X / for any
f 2 Cp .X / and hence i w.Cp .X // D .Cp .X // D !. Any space of countable i -
weight is hereditarily realcompact by Problem 446 of [TFS] so Cp .X / is hereditarily
realcompact.
T.426. Given an infinite cardinal suppose that Cp .X / is a finite union of its -
monolithic (not necessarily closed) subspaces. Prove that Cp .X / is -monolithic.
Solution. Since -monolithity is a hereditary property, there exist n 2 N; " > 0
and -monolithic spaces Y1 ; : : : ; Yn such that Cp .X; . "; "// D Y1 [

[ Yn and
u
C D Cp .X; . "; "// Y i for every i n (see Problem 421). If A C and
jAj , then let Ai D A \ Yi for any i n (the bar denotes the closure in C ).
f
Since Yi is uniformly dense in C , for every f 2 C , there is a sequence Si Yi
f
which converges uniformly to f . However, it is sufficient for us to know that Si is
f
countable and f 2 clC .Si / for any f 2 C and i n.
S f
Now, if Bi D fSi W f 2 Ag, then Bi Yi and jBi j ; besides, we
have Ai A B i and therefore Ai clYi .Bi / which shows that nw.Ai /
because Yi is -monolithic for all i n. Since network weight is finitely additive
(see Problem 405), we have nw.A/ D nw.A1 [

[ An / which proves that C
is -monolithic. Thus the space Cp .X / is also -monolithic being homeomorphic
to C .
T.427. Given an infinite cardinal suppose that Cp .X / is a finite union of its
spread()-monolithic (not necessarily closed) subspaces. Prove that Cp .X / itself
is spread()-monolithic.
Solution. Since spread()-monolithity is hereditary, there exist n 2 N; " > 0 and
spread()-monolithic spaces Y1 ; : : : ; Yn such that Cp .X; . "; "// D Y1 [

[ Yn
u
and C D Cp .X; . "; "// Y i for every i n (see Problem 421). If A C and
f
f
f
S f
Now, if Bi D fSi W f 2 Ag, then Bi Yi and jBi j ; besides, we
have Ai A B i and therefore Ai clYi .Bi / which shows that s.Ai /
because Yi is spread()-monolithic for all i n. Since spread is finitely additive
(see Problem 405), we have s.A/ D s.A1 [

[ An / which proves that C
is spread()-monolithic. Thus the space Cp .X / is also spread()-monolithic being
homeomorphic to C .
T.428. Given an infinite cardinal suppose that Cp .X / is a finite union of its
hd./-monolithic (not necessarily closed) subspaces. Prove that the space Cp .X /
is hd./-monolithic.
Solution. Since hd./-monolithity is hereditary, there exist n 2 N; " > 0 and
hd./-monolithic spaces Y1 ; : : : ; Yn such that Cp .X; . "; "// D Y1 [

[ Yn and
u
f
f
f
S f
Now, if Bi D fSi W f 2 Ag, then Bi Yi and jBi j ; besides, we have
Ai A B i and therefore Ai clYi .Bi / which shows that hd.Ai / because
Yi is hd./-monolithic for all i n. Since hereditary density is finitely additive
(see Problem 405), we have hd.A/ D hd.A1 [

[ An / which proves that
the space C is hd./-monolithic. Thus the space Cp .X / is also hd./-monolithic
being homeomorphic to C .
T.429. Given an infinite cardinal suppose that Cp .X / is a finite union of its hl./-
monolithic (not necessarily closed) subspaces. Prove that the space Cp .X / is hl./-
monolithic.
Solution. Since hl./-monolithity is hereditary, there exist n 2 N; " > 0 and
hl./-monolithic spaces Y1 ; : : : ; Yn such that Cp .X; . "; "// D Y1 [

[ Yn and
u
f
Since Yi is uniformly dense in C , for every f 2 C there is a sequence Si Yi
f
f
S f
Now, if Bi D fSi W f 2 Ag, then Bi Yi and jBi j ; besides, we have
Ai A B i and therefore Ai clYi .Bi / which shows that hl.Ai / because
Yi is hl./-monolithic for all i n. Since hereditary Lindelf number is finitely
additive (see Problem 405), we have hl.A/ D hl.A1 [

[ An / which proves
that the space C is hl./-monolithic. Thus the space Cp .X / is also hl./-monolithic
being homeomorphic to C .
T.430. Suppose that Cp .X / is a finite union of its Dieudonn complete subspaces.
Prove that Cp .X / is realcompact and hence Dieudonn complete.
Solution. If Z is a Dieudonn complete space and Y Z is an F -subspace of Z,

then Y is also Dieudonn complete (see Problem 460 of [TFS]). Thus we can apply
Problem 420 to conclude that there exist n 2 N; " > 0 and Dieudonn complete
spaces Y1 ; : : : ; Yn such that Cp .X / D Y1 [

[ Yn and Yi is dense in Cp .X / for
each i n.
As a consequence, c.Yi / D c.Cp .X // D ! and therefore Yi is realcompact for
all i k (see Problem 458 of [TFS]). Thus we can apply Problem 423 to conclude
that Cp .X / is realcompact and hence Dieudonn complete (see Problem 454 of
[TFS]).
S
T.431. Let X be an arbitrary space. Suppose that Cp .X / D fZn W n 2 !g. Prove
that there exists f 2 Cp .X / and " > 0 such that for some n 2 !, the set .Zn C
f / \ C.X; . "; "// is dense in Cu .X; . "; "// and hence also in Cp .X; . "; "//.
Solution. Given g; h 2 C .X / let d.g; h/ D supfjg.x/ h.x/j W x 2 X g. Then
d is a complete metric on C .X / which generates the topology u of uniform
convergence on X (see Problem 419). We denote by 0 the function which is
identically zero on X . Let B.g; r/ D fh 2 C .X / W d.g; h/ < rg for any
f 2 C .X / and r > 0. The set B.g; r/ is a ball with respect to d , so it is open in
Cu .X / for any f 2 C .X / and r > 0. If P Q C .X /, say that P is u -dense
u
in Q if Q P . S
Let Yn D Zn \ C .X / for every n 2 !; then Cu .X / D n2! Yn . Since Cu .X /
is Cech-complete, it has the Baire property, so it is impossible that all elements of
the sequence fYn W n 2 !g be nowhere dense in Cu .X /.
u
As a consequence, there is a non-empty set V 2 u with V Yn \ V . Since
the metric d generates the topology u , we can find v 2 V and " > 0 such that
u
W D B.v; "/ V and hence W Yn \ W . The map L W C .X / ! C .X /

defined by L.g/ D g v for all g 2 C .X / is an isometry with respect to the
metric d , i.e., d.g; h/ D d.L.g/; L.h// for any functions g; h 2 C .X / (see Fact 2
of T.421). An immediate consequence is that L.W / D B.0; "/.
Furthermore, the map L is a homeomorphism, so the set L.Yn / \ B.0; "/ D
L.Yn \ W / is u -dense in B.0; "/. It is evident that B.0; "/ C.X; . "; "//; since
it was proved in Fact 1 of T.421 that B.0; "/ is u -dense in C D C.X; . "; "//, the
set L.Yn / \ B.0; "/ is u -dense in C as well and therefore a larger set L.Yn / \ C
is u -dense in C too. Finally, if f D v, then L.Yn / \ C D .Yn C f / \ C D
.Zn C f / \ C which shows that .Zn C f / \ C is u -dense in C . Since the identity
map i W Cu .X; . "; "// ! Cp .X; . "; "// is continuous (see Problem 086 of [TFS]),
the set .Zn C f / \ C is dense in Cp .X; . "; "//, so we proved all the promised
properties of Zn and f .
S
T.432. Let Cp .X / D fZn W n 2 !g, where each Zn is closed in Cp .X /. Prove
that some Zn contains a homeomorphic copy of Cp .X /.
Solution. Apply Problem 431 to find n 2 !; f 2 Cp .X / and " > 0 such that
the set .Zn C f / \ C.X; . "; "// is dense in C D C.X; . "; "//. It was proved in
Problem 079 of [TFS] that the map L W Cp .X / ! Cp .X / defined by L.g/ D f Cg
for all g 2 Cp .X / is a homeomorphism, so Zn C f D L.Zn / is closed in Cp .X /.

Thus C L.Zn / and hence C 0 D L1 .C / Zn . Since C ' Cp .X / and C 0 ' C ,
the space C 0 is a homeomorphic copy of Cp .X / which is contained in Zn .
S
T.433. Let P be a hereditary property. Suppose that Cp .X / D fZn W n 2 !g,
where each Zn is closed in Cp .X / and has P. Prove that Cp .X / also has P.
Solution. By Problem 432, there is n 2 ! such that Cp .X / embeds in Zn . Since
Zn has P and P is hereditary, the space Cp .X / has P as well.
S
T.434. Suppose that Cp .X / D fZn W n 2 !g, where each Zn is locally compact.
Prove that X is finite.
Solution. If we are given a function f 2 RX , a set K X is finite and > 0,
then O.f; K; / D fg 2 RX W jg.x/ f .x/j < for all x 2 Kg. It is clear that the
family fO.f; K; / W K is a finite subset of X and > 0g is a local base in RX at
the point f for any f 2 RX . If A Cp .X /, then A is its closure in Cp .X / while
cl.A/ denotes the closure of A in RX .
Fact 1. The space R! cannot be represented as a countable union of its locally
compact subspaces.
Proof. Every locally compact second countable spaces is -compact, so if R! is a
countable union of its locally compact subspaces, then it has to be -compact which
it is not (see Fact 2 of S.186). Fact 1 is proved. t
u
By the result of Problem 431, there exist n 2 !; " > 0 and v 2 Cp .X / such that
the set .Zn Cv/\C.X; . "; "// is dense in C D C.X; . "; "//. Since Y D .Zn Cv/
is locally compact, we can choose a point f 2 Y \ C and H 2 .f; Y / such that
clY .H / is compact. Consequently, H D clY .H / is compact as well.
The set Z D Y \ C is dense in C and G D H \ Z D H \ C 2 .Z/, so
we can choose O 2 .C / such that O \ Z D G. It follows from density of Z in C
that clC .G/ D clC .O/. Therefore O D clC .O/ D clC .G/ D G H is a compact
subset of Cp .X /. Thus A O is compact for any A O.
Since the set O is open in the space C and f 2 O, we can choose > 0 and
a finite set K X such that .f .x/ ; f .x/ C / . "; "/ for all x 2 K while
U D O.f; K; / \ C O. It turns out that
Q
(1) E D . "; "/X nK x2K .f .x/ ; f .x/ C // cl.U /.
Indeed, for any g 2 E; r > 0 and a finite L X , there exists h 2 C such
that hjL D gjL; then h 2 U \ O.g; L; r/, i.e., U \ O.g; L; r/ ; for any basic
neighborhood O.g; L; r/ of the point g in RX . Therefore g 2 cl.U /; since g 2 E
was chosen arbitrarily, the property (1) is proved.
Observe also that U is a compact dense subset of cl.U / so U D cl.U / which
shows that E U Cp .X /. Now let g.x/ D 0 for any x 2 X nK and g.x/ D
f .x/ for every x 2 K. Then g 2 E and hence g 2 C . It is easy to see that for the
set I D fh 2 Cp .X / W jg.x/ h.x/j 2 for all x 2 X g, we have I C ; besides,
I is closed in Cp .X / and I E U , so I is closed in the compact space U which
implies that I is compact.
We have I ' Cp .X; I/ by Fact 3 of S.398, so Cp .X; I/ is compact and hence X

is discrete (see Problem 396 of [TFS]). If X is infinite, then R! embeds in Cp .X / D
RX as a closed subspace so R! is -locally compact which contradicts Fact 1 and
S
T.435. Suppose that Cp .X / D fZn W n 2 !g, where each Zn is locally
pseudocompact. Prove that Cp .X / is -pseudocompact.
Solution. If we are given a function f 2 Cp .X /, a set K X is finite and > 0,
then O.f; K; / D fg 2 Cp .X / W jg.x/ f .x/j < for all x 2 Kg. It is clear
that the family fO.f; K; / W K is a finite subset of X and > 0g is a local base in
Cp .X / at the point f for any f 2 Cp .X /.
By the result of Problem 431, there exist n 2 !; " > 0 and v 2 Cp .X / such that
the set .Zn Cv/\C.X; . "; "// is dense in C D C.X; . "; "//. Since Y D .Zn Cv/
is locally pseudocompact, we can choose a point f 2 Y \ C and H 2 .f; Y / such
that clY .H / is pseudocompact. Since clY .H / is a dense subspace of H (the bar
denotes the closure in Cp .X /), the set H is also pseudocompact (see Fact 18 of
S.351).
The set Z D Y \ C is dense in C and G D H \ Z D H \ C 2 .Z/, so we
can choose O 2 .C / such that O \ Z D G. It follows from density of Z in C that
clC .G/ D clC .O/. Therefore O D clC .O/ D clC .G/ D G H .
Since the set O is open in the space C and f 2 O, we can choose > 0
and a finite set K X such that f .x/ 2; f .x/ C 2 . "; "/ for all points
x 2 K while U D O.f; K; 2/ \ C O. If r D maxfjf .x/j W x 2 Kg C then
r ; r C . "; "/ and f .x/ 2 r; r for all x 2 K. Therefore there exists a
function g 2 C.X; r; r / such that gjK D f jK.
It is easy to see that for the set I D fh 2 Cp .X / W jg.x/ h.x/j for all
x 2 X g, we have I C .
Furthermore, I is a retract of the space Cp .X / by Fact 3 of S.398; an easy
consequence of I O H is that I is also a retract (and hence a continuous
image) of a pseudocompact space H . This implies that I is pseudocompact. We
have I ' Cp .X; I/ by Fact 3 of S.398, so Cp .X; I/ is also pseudocompact.
If X is not pseudocompact, then Cp .X / ' Cp .X / R! by Fact 6 of T.132.
Therefore there is R Cp .X / S such that R is a retract of Cp .X / and R ' R! . If
Ti D Zi \ R for all i 2 !, then i 2! Ti D R. Since R has the Baire property, there
is W 2 .R/ such that Tk \ W is dense in W for some k 2 !.
Choose a point w 2 Tk \ W and V 2 .w; Tk / such that clTk .V / is pseudocom-
pact. Since clTk .V / is a dense subset of V , the set V is also pseudocompact (see
Fact 18 of S.351). The set N D V \ W D V \ .W \ Tk / is open in Tk \ W R,
so there is M 2 .R/ such that M W and M \ .Tk \ W / D M \ Tk D N . Since
Tk \ W is dense in W , the set N D Tk \ M is dense in M , so clR .M / D clR .N /.
Let r W Cp .X / ! R be a retraction. Then r.V / V \ R V \ W D N . The
set r.V / is pseudocompact in a second countable space R, so it is compact; since
N r.V /, the set clR .N / D clR .M / is compact. It turns out that the closure of
some non-empty open subset of R is compact and hence the same is true for some
E 2 .R! / (because R is homeomorphic to R! ). Therefore every h 2 E is a point
of local compactness of R! ; since R! is homogeneous, it is locally compact and
hence -compact which is a contradiction (see Fact 2 of S.186).
Thus X is pseudocompact, so applying Problems 398 and 399 of [TFS], we can
conclude that Cp .X / is -pseudocompact and finish our solution.
S
T.436. Suppose that Cp .X / D fZn W n 2 !g, where each Zn is realcompact and
closed in Cp .X /. Prove that Cp .X / is realcompact.
Solution. Our solution can be easily derived from the following fact.
Fact 1. For any space Z, if Cp .Z; I/ is realcompact, then Cp .Z/ is also realcom-
pact.
Proof. Let On D 1; 1 C 2n1 / [ .1 2n1 ; 1 ; it is clear that On 2 .I/, so
the set Wnz D ff 2 Cp .Z/ W f .z/ 2 On g is an open subset of Cp .Z/ T for every
n 2 ! and z 2 Z. The equality Fz D ff 2 Cp .Z/ W f .z/ 2 f 1; 1gg D n2! Wnz
implies that Fz is a G -subset of Cp .Z/ for any z 2 Z.S
Now observe that Cp .Z; . 1; 1// D Cp .Z; I/n. fFz W z 2 Zg/ and hence
the space Cp .Z; . 1; 1// is realcompact by Problem 408 of [TFS]. Since Cp .Z/ is
homeomorphic to Cp .Z; . 1; 1//, the space Cp .Z/ is also realcompact so Fact 1 is
proved. t
u
By the result of Problem 431, there exist n 2 !; " > 0 and v 2 Cp .X / such
that the set .Zn C v/ \ C.X; . "; "// is dense in C D C.X; . "; "//. Since Y D
.Zn C v/ is closed in Cp .X / we have Cp .X; "; " / D C Y ; the space E D
Cp .X; "; " / is closed in Cp .X / and hence in Y , so E is realcompact by Problem
403 of [TFS].
Furthermore E is homeomorphic to Cp .X; I/ by Fact 3 of S.398, so Cp .X; I/ is
realcompact which shows that Cp .X / is also realcompact (see Fact 1).
T.437. Prove that any metacompact collectionwise normal space is paracompact.
Solution. If Z is a space, A exp.Z/ and z 2 Z, then ord.z; A/ is the cardinality
of the family A.z/ D fA 2 A W z 2 Ag. A family B exp.Z/ is inscribed in A if
for any B 2 B, there is A 2 A such that B A.
Suppose that X is a collectionwise normal metacompact space. Given an open
cover U of the space X , it has a point-finite open refinement
S V. For any n 2 N, let
Xn D fx 2 X W ord.x; V/ D ng. It is clear that X D n2! Xn . Observe first that
(1) the set Yn D X1 [

[ Xn is closed in X for any n 2 N.
Indeed, Tif x 2 X nYn , then there are distinct sets V1 ; : : : ; VnC1 2 V such that
x 2 V D i nC1 Vi which shows that V X nYn and hence X nYn is open in X
so (1) is proved.
For Tany point x 2 X there is a unique number n 2 N such that x 2 Xn ; let
Wx D fV 2 V W x 2 V g and Vx D Wx \ Xn . We claim that
(2) if n 2 N and x; y 2 Xn , then either Vx D Vy or Vx \ Vy D ;.

T T
To see that (2) holds, assume that z 2 Vx \ Vy ; then z 2 . V.x// \ . V.y//
However, jV.x/j D n and jV.y/j D n, so if V.x/nV.y/ ;, then z belongs to
more than n elements of V which contradicts z 2 Xn . Analogously, it is impossible
that V.y/nV.x/ ;, so V.x/ D V.y/ and hence Vx D Vy so (2) is proved.
It follows from (1) and (2) that En D fVx W x 2 Xn g is a disjoint open cover of
Xn . Consequently, Vx is closed in Xn for every x 2 Xn . For every n 2 N, choose
an indexation fVsn W s 2 Sn g of the family En such that Vsn Vtn for any distinct
s; t 2 Sn . Given n 2 N and s 2 Sn , there is x 2 Xn with Vsn D Vx ; let Wsn D Wx .
Then Vsn Wsn for any s 2 Sn and the family Wn D fWsn W s 2 Sn g .X/ is
inscribed in V and hence in U.
Observe that all elements of E1 are closed in X1 and hence in X by (1). Any
disjoint open cover of a space is a discrete family in that space, so E1 is discrete in
X1 and hence in X . The space X being collectionwise normal, we can find a discrete
family H1 D fUs1 W s 2 S1 g .X/ such that Vs1 Us1 Ws1 for any s 2 S1 .
Assume that n 2 N and for any i < n we have constructed a discrete S Sfamily
Hi D fUsi W s 2 Sn g .X/ which is inscribed in U and Yi Hi D j i . Hj /
for all i < n. The set H D H1 [

[ Hn1 is open in X , so P D Xn nH D Yn nH
is closed in X by (1). Consequently, G D fVsn \ P W s 2 Sn g is a disjoint open cover
of P , so all elements of G are closed in P and hence in X .
Any disjoint open cover of a space is a discrete family in that space, so G is
discrete in P and hence in X . The space X being collectionwise normal, we can
find a discrete family Hn D fUsn W s 2 Sn g .X/ such that Vsn \ P Usn Wsn
for any s 2 Sn .
ThusS Hn covers P and therefore Yn D H [ P is covered by H [ Qn where
Qn D Hn ; since Wn is inscribed in U, the family Hn is also inscribed in U.
This shows that we can construct inductively a sequence fHn W n 2 !g of discrete
families S subsets of X such that Hn is inscribed in U for all n 2 ! and
S of open
Yn i n . Hi / S for all n 2 N.
Therefore H D n2! Hn is a -discrete open cover of X inscribed in U, i.e., H
is a -discrete open refinement of U. The open cover U of the space X was chosen
arbitrarily, so every open cover of X has a -discrete open refinement; applying
Problem 230 of [TFS] we conclude that X is paracompact.
T.438. Prove that if Cp .X / is normal and metacompact, then it is Lindelf.
Solution. Normality of Cp .X / implies that Cp .X / is collectionwise normal (see
Problem 295 of [TFS]). Therefore Cp .X / is paracompact (see Problem 437) and
hence Lindelf by Problem 219 of [TFS].
T.439. Prove that Cp .!/ is not metacompact.
Solution. An immediate consequence of Problem 382 of [TFS] is that
p.Cp .!// D !, i.e., every point-finite family of non-empty open subsets of
Cp .!/ is countable. If Cp .!/ is metacompact, then every open cover of Cp .!/
has a point-finite open refinement which has to be countable by our observation.

Thus Cp .!/ is Lindelf.
If T is the two arrows space (see Problem 384 of [TFS]), then T is separable,
compact and ext.Cp .T // D c. However, Cp .T / embeds in Cp .!/ as a closed
subspace by Problem 375 of [TFS]. Consequently, ext.Cp .!// ext.Cp .T // D c
which contradicts the Lindelf property of Cp .!/. This contradiction shows that
Cp .!/ is not metacompact.
T.440. Prove that Cp .L.// is not metacompact for any uncountable .
Solution. Recall that L./ D [ fag where all point of the set are isolated and
U 2 .a; L.// if and only if a 2 U L./ and L./nU is countable. For any
A L./, let A W Cp .L.// ! Cp .A/ be the restriction map defined by A .f / D
f jA for any f 2 Cp .L.//. Since L./ is a P -space, the space Cp .L./; I/ is
countably compact (see Problem 397 of [TFS]), so it follows from Problem 479 of
[TFS] that Cp .L.// is pseudocomplete and hence Baire by Problem 464 of [TFS].
Thus p.Cp .L./// D c.Cp .L./// D ! by Problem 282 of [TFS]. Now, if
Cp .L.// is metacompact, then every open cover of Cp .L.// has a point-finite
open refinement which has to be countable by p.Cp .L./// D !. Therefore
Cp .L.// is Lindelf and hence t.L.// D ! by Problem 189 of [TFS] which
is a contradiction because a 2 cl./ while no countable subset of contains a in its
closure.
T.441. Prove that neither the Baire property nor pseudocompleteness is countably
S pseudocompact space X such that

Solution. Recall that there exists an infinite
Cp .X / is -pseudocompact, i.e., Cp .X / D n2! Zn where Zn is pseudocompact
for all n 2 ! (see Problem 400 of [TFS]). The space Cp .X / does not have the
Baire property by Problem 284 of [TFS]. Since every pseudocompact space is
pseudocomplete and hence Baire (see Problem 472 and 464 of [TFS]), every Zn
is pseudocomplete, so Cp .X / witnesses that neither pseudocompleteness nor the
Baire property is countably additive in Cp .X /.
T.442. Prove that -weight and -character are not countably additive in spaces
Cp .X /.
Solution. Let X be the Cantor set K and denote by C the family of all non-empty
clopen subsets of X . Since X is second countable and zero-dimensional, we can
choose a countable base B C of the S space X . Call a non-empty family B 0 B
0 0
adequate if B is finite, disjoint and B X . It is easy to see that the set Q D
f.B1 ; : : : ; Bn ; q1 ; : : : ; qn / W n 2 N; fB1 ; : : : ; Bn g is an adequate family and qi 2 Q\
I for all i ng is countable, so we can enumerate it as Q D fwi W i 2 !g. For each
i 2 ! define a function fi W X ! R as follows: if wi D .B1 ; : : : ; Bn ; q1 ; : : : ; qn /,
then let S fi .x/ D qk if x 2 Bk for some k n and fi .x/ D i C 2 for any
x 2 X n. i n Bi /.
Observe that the set D D ffi W i 2 !g is a discrete subspace S of Cp .X /. To see it

take any i 2 !; if wi D .B1 ; : : : ; Bn ; q1 ; : : : ; qn / and B D i n Bi , then pick any
x 2 X nB. The set Oi D ff 2 Cp .X / W jf .x/ fi .x/j < 12 g is open in Cp .X /
and Oi \ D D ffi g for all i 2 !, so D is discrete and hence Z D D has countable
-weight because fffi g W i 2 !g is a -base in Z.
Given any function f 2 Cp .X; I/ consider any U 2 .f; Cp .X //. There are
x1 ; : : : ; xn 2 X and " > 0 such that V D fg 2 Cp .X / W jf .xi / g.xi /j < " for
all i ng U . It is easy to find an adequate family fB1 ; : : : ; Bn g B such that
xk 2 Bk for all k n. Choose q1 ; : : : ; qn 2 Q \ I such that jf .xk / qk j < "
for all k n. There is i 2 ! such that .B1 ; : : : ; Bn ; q1 ; : : : ; qn / D wi and hence
jfi .xk / f .xk /j < " for all k n which proves that fi 2 V \ D U \ D. Thus
U \ D ; for every U 2 .f; Cp .X // and therefore f 2 D. Since the function
f 2 Cp .X; I/ was chosen arbitrarily, we proved that Cp .X; I/ Z.
For any n 2 N, the map 'n W Cp .X / ! Cp .X / defined by 'n .f / D n
f for
every f 2 Cp .X / is a homeomorphism, so Zn D 'n .Z/ has a countable -weight
for all n 2 N. Since Cp .X; n; n / D 'n .Cp .X;
S S I// Zn for all n 2 N, we have
fZn W n 2 Ng D Cp .X / because Cp .X / D fCp .X; n; n / W n 2 Ng; this last
equality is due to the fact that SX is compact.
It turns out that Cp .X / D fZn W n 2 Ng and .Zn / w.Zn / D ! for all
n 2 N. However, .Cp .X // D w.Cp .X // D jX j D c > ! (see Fact 1 of T.158)
and hence Cp .X / witnesses that both -weight and -character are not countably
additive in Cp .X /.
T.443. Suppose that Cp .X / is a countable union of its Cech-complete (not neces-
sarily closed) subspaces. Prove that X is countable and discrete (and hence Cp .X /
is Cech-complete).
S
Solution. Assume that Cp .X / D fPn W n 2 !g and every Pn is Cech-complete.
It follows from Problem 431 that there is n 2 !; " > 0 and f 2 Cp .X / such that
the set .Pn Cf /\C.X; . "; "// is dense in C D C.X; . "; "//. Since Y D Pn Cf
is Cech-complete and C is dense in E D C.X; "; " /, the set Y \ E is dense in E.
Furthermore, E is closed in Cp .X / and therefore Z D E \ Y is a Cech-complete
dense subspace of E.
The space E is homeomorphic to Cp .X; I/ (see Problem 287 of [TFS]), so
Cp .X; I/ also has a dense Cech-complete subspace and hence X is discrete by
Problem 287 of [TFS]. This implies that Cp .X / D RX has the Baire property and
therefore all sets Pm cannot be nowhere dense; fix U 2 .Cp .X // such that Pm \U
is dense in U for some m 2 !. There exists a set K D fx1 ; : : : ; xk g X and non-
empty rational intervals O1 ; : : : ; Ok such that V D ff 2 Cp .X / W f .xi / 2 Oi
for all i kg U . It is immediate that V D RX nK O1

Ok ; since
Oi ' R for all i k, the space V is homeomorphic to RX . Besides, the set Pm \ V
is Cech-complete and dense in V which shows that Cp .X / D RX has a dense
Cech-complete subspace. Finally apply Problem 265 of [TFS] to conclude that X is
discrete and countable, so Cp .X / ' R! is Cech-complete.
T.444. Given an infinite cardinal suppose that Cp .X / is a union of countably many

(not necessarily closed) subspaces of character . Prove that .Cp .X // and
hence jX j .
S
Solution. Assume that Cp .X / D n2! Zn and .Zn / for all n 2 !. By
Problem 431 there exist n 2 !; " > 0 and f 2 Cp .X / such that .Zn C f / \
C.X; . "; "// is dense in C D C.X; . "; "//. If Y D .Zn Cf /\C , then .Y /
and Y is dense in C . Consequently, .g; C / D .g; Y / for any g 2 Y (see
Fact 2 of S.265).
Since C is homeomorphic to the space Cp .X / (see Fact 1 of S.295), there is some
h 2 Cp .X / such that .h; Cp .X // and hence .Cp .X // because Cp .X /
is homogeneous (see Problem 079 of [TFS]). Therefore jX j D .Cp .X // by
Problem 169 of [TFS].
T.445. Prove that weight is countably additive in spaces Cp .X /.
S
Solution. Given an infinite cardinal assume that Cp .X / D n2! Zn and
w.Zn / for all n 2 !. Since .Zn / w.Zn / for all n 2 !, we can apply
Problem 444 to conclude that w.Cp .X // D jX j D .Cp .X // (see Problem
169 of [TFS]) and therefore w.Cp .X // . This proves that weight is countably
T.446. Prove that metrizability is countably additive in spaces Cp .X /.
S
Solution. Assume that Cp .X / D n2! Zn and Zn is metrizable for all n 2 !.
Then .Zn / ! for all n 2 !, so we can apply Problem 444 to conclude that
w.Cp .X // D jX j D .Cp .X // ! (see Problem 169 of [TFS]) and hence the
space Cp .X / is metrizable.
T.447. Prove that tightness is countably additive in spaces Cp .X /.
Solution. Given a function f 2 Cp .X /, a finite set K X and a number > 0,
let O.f; K; / D fg 2 Cp .X / W jg.x/ f .x/j < for all x 2 Kg. It is clear that
the family fO.f; K; / W K X is finite and > 0g is a local base of Cp .X / at the
point f . Given any sets A B Cp .X / we will say that A is uniformly dense in
u
B if B A . S
Let be an infinite cardinal; assume that Cp .X / D n2! Zn and t.Zn /
for all n 2 !. By Problem 431 there exist n 2 !; " > 0 and v 2 Cp .X / such that
Y D .Zn C v/ \ C.X; . "; "// is uniformly dense in C D C.X; . "; "//.
Recall that a family U exp.X / is an !-cover of X if for any finite K X
there is U 2 U such that K U . If all elements of U are open in X , then U
is called an open !-cover of X . Fix an open !-cover U of the space X and let
D D ff 2 Y W f 1 .. 9" ; "// U for some U 2 Ug. Let w.x/ D 89 " for every
x 2 X ; then w 2 C and hence there is f 2 Y such that jf .x/ w.x/j < 19 " for all
x 2 X . It is evident that f .x/ 79 " for all x 2 X .
To show that f 2 D, take any W 2 .f; Cp .X //; there is a finite K X
and > 0 such that O.f; K; / W . The family U being an !-cover of X , there
is U 2 U such that K U . It is easy to construct a function g 2 C such that
gj.X nU / 0 and g.x/ D f .x/ for every x 2 K. Let D minf; 19 "g; since
Y is uniformly dense in C , there is h 2 Y such that jh.x/ g.x/j < for all
x 2 X . Thus jh.x/ f .x/j < for all x 2 K and therefore h 2 O.f; K; /.
Besides, h.x/ < 19 " for all x 2 X nU which shows that h1 .. 19 "; "// U , i.e.,
h 2 D\O.f; K; / D\W . Consequently, D\W ; for any W 2 .f; Cp .X //
so f 2 D.
Since t.Y / , there is E D such that jEj and f 2 E. For every h 2 E
there is Uh 2 U for which Sh D h1 .. 19 "; "// Uh . Then U 0 D fUh W h 2 Eg U
and jU 0 j . If K X is finite, then G D fg 2 Cp .X / W g.x/ > 19 " for any
x 2 Kg is an open neighborhood of f in Cp .X /, so there is e 2 E for which
e 2 G. Since e.x/ > 19 " for all x 2 K, we have K Se Ue 2 U 0 and therefore
U 0 is an !-cover of X . We proved that for any open !-cover U of the space X , there
is U 0 U such that jU 0 j and U 0 is an !-cover of X . Now apply Problems 148
and 149 of [TFS] to conclude that t.Cp .X // and finish our solution.
T.448. Prove that pseudocharacter is countably additive in spaces Cp .X /.
Solution. If Z is a space, f 2 C .Z/ and " > 0, let IZ .f; "/ D fg 2 C .Z/ W
jf .z/ g.z/j " for all z 2 Zg. For any f; g 2 C .Z/ we write f g if
f .z/ g.z/ for all z 2 Z.
Fact 1. Suppose thatTZ is a space, f0 ; : : : ; fn 2 C .Z/ and "0 ; : : : ; "n are positive
numbers. Let Yi D ki IZ .fk ; "k / for every i n and assume additionally that
fi C1 2 Yi for all i < n. Then .f; Yn / D .Cp .Z// for any f 2 Yn .
Proof. The inequality .f; Yn / .Cp .Z// is true for any f 2 Yn because Yn
is a subspace of Cp .Z/. Now, assume that .f; Yn / for some f 2 Yn and an
infinite cardinal . It follows from Fact 1 of S.426 that there exists A Z such that
jAj and G.A; f / \ Yn D ff g where G.A; f / D fg 2 C .Z/ W gjA D f jAg.
If A is not dense in Z, then fix any z 2 ZnA and a function p 2 C.Z; 0; 1 /
such that p.z/ D 1 and p.A/ D f0g. Let m.y/ D maxffi .y/ "i W i ng and
M.y/ D minffi .y/ C "i W i ng for any point y 2 Z. Then m; M 2 C .Z/
and
(1) m.y/ < M.y/ for any y 2 Z.
Indeed, it follows from fn 2 Yn that fi .y/ "i fn .y/ fi .y/ C "i for all
i n and therefore m.y/ M.y/ for all y 2 Z. Now if m.y/ D M.y/ for some
y 2 Z, then there exist distinct l; k n such that fl .y/ "l D fk .y/C"k and hence
jfl .z/ fk .z/j D "l C "k > maxf"l ; "k g. If l < k, then we obtain a contradiction
with fk 2 IZ .fl ; "l /; if k < l, then our inequality contradicts fl 2 IZ .fk ; "k / so
(1) is proved.
Observe that for any g 2 C .Z/ we have g 2 Yn if and only if m g M . In
particular, m.z/ f .z/ M.z/, so either m.z/ < f .z/ or f .z/ < M.z/.
In the first case choose a positive number with < f .z/ m.z/ and consider
the functions g D f p and h D max.g; m/. It is immediate that m h f , so
h 2 Yn . If y 2 A, then p.y/ D 0 and hence g.y/ D f .y/ D h.y/, i.e., hjA D f jA.
However, h.z/ D f .z/ f .z/ which shows that h f while h 2 G.A; f /\Yn
Finally, if f .z/ < M.z/, then choose a positive number with < M.z/ f .z/
and consider the functions g D f C p and h D min.g; M /. It is immediate that
f h M , so h 2 Yn . If y 2 A, then p.y/ D 0 and hence g.y/ D f .y/ D h.y/,
i.e., hjA D f jA. However, h.z/ D f .z/ C f .z/ which shows that h f
while h 2 G.A; f / \ Yn which is again a contradiction. Thus A is dense in Z and
therefore .Cp .Z// D d.Z/ so Fact 1 is proved. t
u
S assume that X is a space and is an infinite cardinal
such that Cp .X / D n2! Cn0 and .CS 0
n / for every n 2 !. Let Cn D Cn \
0

C .X / for all n 2 !; then Cp .X / D n2! Cn . Since pseudocharacter is a
hereditary property, we can assume that Cn \ Cm D ; whenever m n.
If f; g 2 C .X /, let d.f; g/ D supfjf .x/ g.x/j W x 2 X g. Then d is
a complete metric on C .X / and the topology .d / generated by d coincides
with the uniform convergence topology u on C .X / (see Problem 419). For any
A C .X / we denote by diam.A/ the diameter of A with respect to the metric
d , i.e., diam.A/ D supfd.f; g/ W f; g 2 Ag. Assume, towards a contradiction, that
.Cp .X // > and denote by 0X the function which is identically zero on X . An
immediate consequence of Fact 1 is that .Cp .X; n; n // D .IX .0X ; n// D
.Cp .X // for all n 2 N and hence .Cp .X // D .Cp .X // > . Besides,
the space Cp .X / is homogeneous by Problem 079 of [TFS], so .f; Cp .X // D
.Cp .X // > for any f 2 Cp .X /.
Denote by C the family of all closed non-empty G -subsets of Cp .X /. We will
often implicitly use the fact that C is a base in the family of all G -subsets of Cp .X /
in the sense that if A is a G -subset of Cp .X / and f 2 A, then f 2 H A for
some H 2 C (see Fact 2 of S.328).
Suppose that Y Cp .X / and .f; Y / > for any f 2 Y . If A is a G -set in

Cp .X / and .f; A \ Y / for some f 2 A \ Y , then .f; Y / which is a
contradiction. Therefore
(2) if Y Cp .X / and .f; Y / > for any f 2 Y , then for any G -set A
Cp .X / we have .f; Y \ A/ > for any f 2 Y \ A.
For the sake of brevity, call a set Y Cp .X / deep if .f; Y / > for any
f 2 Y . Choose any g 2 C0 and let P be a closed G -set in the space Cp .X /
such that P \ C0 D fgg. The set P is deep by (2) so choose f0 2 P0 nfgg and
H 2 C such that f0 2 H Cp .X /nfgg. It is clear that f0 2 H0 D H \ P 2 C
and H0 \ C0 D ;. Suppose that n 2 N and we have f0 ; : : : ; fn1 2 Cp .X / and
H0 ; : : : ; Hn1 2 C with the following properties:
(3) H0

Hn1 ; T
(4) fi 2 Hi \ Yi where Yi D ki IX .fk ; 2k / for all i n 1;
(5) Hi \ Yi \ Ci D ; for all i < n.
If Hn1 \ Yn1 \ Cn D ;, then for the sets Hn D Hn1 and fn D fn1 the
conditions (3)(5) still hold so the inductive step is fulfilled.
Otherwise, take any function g 2 Cn \ Hn1 \ Yn1 and pick P 2 C such

that P \ Cn D fgg. The set Q D Hn1 \ P \ Yn1 is deep by Fact 1 and (2),
so we can choose fn 2 Qnfgg and H 2 C such that fn 2 H and g H . Then
Hn D Hn1 \P \H 2 C and it is immediate that (3)(5) still hold for the functions
f0 ; : : : ; fn and the sets H0 ; : : : ; Hn . Therefore our inductive procedure provides
sequences ffi W i 2 !g Cp .X / and fHi W i 2 !g C such that (3)(5) hold for
all i 2 !. Let Fi D Hi \ Yi for all i 2 !. Then Fi is closed in Cp .X / and hence in
Cu .X /; besides, Fi ; because fi 2 Fi by (4). Furthermore, diam.IX .f; "// 2"
for any " > 0 and f 2 Cp .X / which shows that diam.Fi / 2i C1 for all i 2 !
by (4). Thus fFi W i 2 !g is a decreasing sequence of non-empty closedTsubsets of
a complete metric space Cu .X /. Since diam.Fi / ! 0, we have S F D i 2! Fi
; (see Problem 236 of [TFS]). Now, if f 2 F , then f i 2! Ci by (5); this
contradiction shows that .Cp .X // and makes our solution complete.
T.449. Prove that i -weight and diagonal number are countably additive in spaces
Cp .X /.
S
Solution. Given an infinite cardinal , suppose that Cp .X / D fCn W n 2 !g and
.Cn / (or i w.Cn / ) for all n 2 !. It is easy to see that pseudocharacter does
not exceed diagonal number (or i -weight respectively) for any space, so .Cn /
.Cn / (or .Cn / i w.Cn / respectively) for all n 2 !. Therefore we can
apply Problem 448 to conclude that .Cp .X // . Consequently, .Cp .X // D
.Cp .X // (or i w.Cp .X // D .Cp .X // respectively) by Problem 173 of
[TFS].
T.450. Prove that the FrchetUrysohn property is countably additive in spaces
Cp .X /.
S
Solution. Suppose that Cp .X / D n2! Cn where Cn is a FrchetUrysohn space
for each n 2 !. By Problem 431, we can find " > 0; n 2 ! and v 2 Cp .X / such
that D D .Cn C v/ \ C.X; . "; "// is uniformly dense in C D C.X; . "; "// (recall
that this means that every function from C is a uniform limit of a sequence from the
set D). If we are given a function f 2 Cp .X /, a finite set K X and > 0, then
O.f; K; / D fg 2 Cp .X / W jg.x/ f .x/j < for all x 2 Kg. It is evident that the
family fO.f; K; / W K is a finite subset of X and > 0g is a local base of Cp .X /
at the point f .
Since D is uniformly dense in C , there is w 2 D such that w.x/ 89 " for every
x 2 X . Recall that a family A is an !-cover of X if for any finite K X there is
A 2 A such that K A. Suppose that U .X/ is an !-cover of X and consider
the set P D ff 2 D W f 1 .. 19 "; "// U for some U 2 Ug.
To see that w 2 P , take any W 2 .w; Cp .X //. There exists a finite K X and
2 .0; 19 "/ such that O.w; K; / W . Since U is an !-cover of X , we can take
U 2 U with K U . It is easy to construct a function f 2 C such that f jK D wjK
and f .X nU / D f0g. Since D is uniformly dense in C , there is g 2 D for which
jg.x/ f .x/j < for all x 2 X . It is immediate that jg.x/j < 19 " for all x 2 X nU
and hence f 1 .. 19 "; "// U which shows that g 2 P . Besides, jg.x/ w.x/j D
jg.x/ f .x/j < for all x 2 K and hence g 2 O.w; K; / \ P W \ P . Thus

any W 2 .w; Cp .X // intersects the set P , i.e., w 2 P .
The space D being FrchetUrysohn there is a sequence ffk W k 2 !g P such
that fk ! w. Pick Uk 2 U such that fk1 .. 19 "; "// Uk for all k 2 !. Given any
x 2 X the set G D fh 2 Cp .X / W h.x/ > 19 "g is an open neighborhood of the point
w, so there is m 2 ! such that fk 2 G for all k m. Thus x 2 fk1 .. 91 "; "// Uk
for all k m. Therefore every point of X belongs to all sets Uk except finitely
many (this is usually denoted by limfUk W k 2 !g D X ). It turns out that for every
open !-cover U of the space X , there is a sequence fUk W k 2 !g U such that
limfUk W k 2 !g D X . Therefore we can apply Problem 144 of [TFS] to conclude
that Cp .X / is a FrchetUrysohn space.
S
T.451. Suppose that X is a metrizable space and Cp .X / D fYi W i 2 !g, where
Yi is hereditarily realcompact (not necessarily closed) for every i 2 !. Prove that
nw.Cp .X // D i w.Cp .X // D ! and hence Cp .X / is hereditarily realcompact.
Solution. If ext.XS / > !, then R!1 embeds in Cp .X / by Fact 1 of S.215 and
therefore R!1 D fZn W n 2 !g where every Zn is hereditarily realcompact.
Observe first that
S
(1) if R!1 D n2! Rn and every Rn is dense in R , then all Rn cannot be
!1
Indeed, every Rn is a Moscow space (see Problems 423 and 424 of [TFS]). If Rn
is hereditarily realcompact for every n 2 !, then given a point x 2 Rn , the space
Rn nfxg is realcompact and dense in a Moscow space Rn , so Rn nfxg must be !-
placed in Rn (see Problem 425 of [TFS]) which implies that fxg is a G -set in Rn .
This proves that .Rn / ! for all n 2 !. Since R!1 D Cp .D.!1 // (where D.!1 /
is the set !1 with the discrete topology), we can apply Problem 448 to conclude that
.R!1 / D !; this contradiction shows that (1) is true.
For every A !1 the map A W R!1 ! RA is the natural projection. By (1),
there is n0 2 ! such that Zn0 is not dense in R!1 , so there is a finite set A0 !1
and x0 2 RA0 for which A10 .x/ \ Zn0 D ;. Suppose that < !1 , and for every
< , we have n 2 !, a finite set A !1 and x 2 RA with the following
properties:
(2) <
< implies n nS
and A \ A
D ;; S
(3) for any < , let B D fA
W
g and y D fx
W
g 2 RB ;
then B1 .y / \ Zn D ;.
S S
Let y D fx W < g; the set B D fA W < g is countable and y 2
1
RB , so j!1 nBj
T D1!1 and therefore B .y/ is homeomorphic to R!1 . Furthermore,
1 1
B .y/ fB .y / W < g which shows that B .y/ \ Zn D ; for any
< .
If N D fn 2 ! W Zn \B1 .y/ ;g, then N \fn W < g D ; and (1) implies
that it is impossible that Zn \ B1 .y/ be dense in B1 .y/ for all n 2 N . Therefore
there exists a finite set A !1 nB and x 2 RA such that for y D y [ x and
B D B [ A , we have B1 .y / \ Zn D ; for some n 2 N . It is evident that (2)

and (3) still hold for all , so our inductive construction can be continued to
obtain, among other things, a set fn W < !1 g ! such that n n
for distinct
;
2 !1 [see (2)].
This contradiction shows that R!1 cannot be represented as a countable union
of its hereditarily realcompact subspaces and hence ext.X / D !. Since X is
metrizable, we have nw.Cp .X // D nw.X / D w.X / D ! (see Problem 172 and
214 of [TFS]) and hence i w.Cp .X // D d.X / D ! which, together with Problem
446 of [TFS], implies that Cp .X / is hereditarily realcompact.
S
T.452. Let X be a pseudocompact space. Suppose that Cp .X / D fZn W n 2 !g,
where each Zn is paracompact and closed in Cp .X /. Prove that Cp .X / is Lindelf.
Solution. By Problem 431, there exists n 2 !; " > 0 and v 2 Cp .X / such that
the set .Zn C v/ \ C.X; . "; "// is dense in C D C.X; . "; "//. Since Z D Zn C
v is closed in Cp .X /, we have Cp .X; "; " / D C Z (see Fact 1 of T.421)
and hence Cp .X; "; " / is paracompact being a closed subspace of a paracompact
space Z. The space Cp .X; I/ is homeomorphic to Cp .X; "; " / (see Fact 3 of
S.398), so Cp .X; I/ is paracompact and hence Lindelf because c.Cp .X; I// D !
(see Problems 111 and 092 of [TFS] together with Fact 2 of S.219).
SLindelf for any n 2 N (here we use Fact 3 of
Therefore Cp .X; n; n / is also
S.398 again) and hence Cp .X / D fCp .X; n; n W n 2 Ng is a Lindelf space as
well (the last equality holds because X is pseudocompact).
T.453. Give an example of a non-normal space which is a countable union of its
closed normal subspaces.
Solution. The Mrowka space M can be represented as M D D [ E where E is
countable and the set D is discrete and closed in M (see Problem 142 of [TFS]).
Thus the family fDg [ ffxg W x 2 Eg is countable and consists of discrete (and
hence normal) closed subspaces of M . Therefore M is a countable union of its
closed normal subspaces. However, M is not normal because it is pseudocompact
but not countably compact (see Problems 142 and 137 of [TFS]).
S
T.454. Let X be a compact space. Suppose that Cp .X / D fZn W n 2 !g, where
each Zn is normal and closed in Cp .X /. Prove that Cp .X / is Lindelf.
Solution. By Problem 431, there exists n 2 !; " > 0 and v 2 Cp .X / such that the
set .Zn C v/ \ C.X; . "; "// is dense in C D C.X; . "; "//. Since Z D Zn C v is
closed in Cp .X /, we have Cp .X; "; " / D C Z (see Fact 1 of T.421), and hence
Cp .X; "; " / is normal being a closed subspace of a normal space Z. The space
Cp .X; I/ is homeomorphic to Cp .X; "; " / (see Fact 3 of S.398), so Cp .X; I/ is
also normal and hence ext.Cp .X; I// D ! by Problem 296 of [TFS].
Since X is compact, we can apply Baturovs theorem (see 269) to conclude that
Cp .X; I/ is Lindelf. Thus Cp .X; n; n / is also Lindelf
S for any n 2 N (here we
used Fact 3 of S.398 again) and therefore Cp .X / D fCp .X; n; n / W n 2 Ng is
Lindelf as well.
T.455. Let X be a metrizable space. Suppose that Cp .X / is a countable union of its

(not necessarily closed) normal subspaces. Prove that X is second countable and
hence Cp .X / is normal.
Q
Solution. Let Mt beQ a space for any t 2 T ; in the product M D t 2T Mt the map
pA W M ! MA D t 2A Mt is the natural projection for any A T . A set E M
covers all countable faces of M if pA .E/ D MA for any countable A T .
Q
Fact 1. If Mt is a second countable space for any t 2 T and E M D t 2T Mt
covers all countable faces of M , then E is C -embedded in M .
Proof. It is easy to see that any set which covers all countable faces of a product
is dense in that product, so E is dense in M . Given a continuous f W E ! R,
we can apply Problem 299 of [TFS] to find a countable A T and a continuous
map g W pA .E/ D MA ! R such that f D g .pA jE/. It is immediate that
h D g pA 2 C.M / and hjE D f , so Fact 1 is proved. t
u
Fact 2. Suppose that a property P is closed-hereditary and R!1 is a countable union
of its subspaces with the property P. Then
S there exists a sequence S D fYn W n 2 !g
of subspaces of R!1 such that R!1 D S and the set Yi covers all countable faces
of R!1 and has P for any i 2 !. The sequence S will be called P-representation of
the space R!1 .
Proof. For every A !1 the map A W R!1 ! RA isSthe natural projection. Assume
that no P-representation of R!1 exists while R!1 D fZn W n 2 !g and Zn ` P (
Zn has P) for all n 2 !. For any countable A !1 and x 2 RA , the space A1 .x/ is
closed in R!1 and homeomorphic to R!1 nA ; it is easy to construct a homeomorphism
h W R!1 nA ! R!1 such that h.P / covers all countable faces of R!1 whenever it
covers all countable faces of R!1 nA . Thus our assumption implies that
(1) for any countable set A !1 and x 2 RA , there exists m 2 ! such that
Zm \ A1 .x/ ; while A[B
1
.y/ \ Zm D ; for some countable B !1 nA
and y 2 R A[B
such that yjA D x.
It follows from (1) that there is n0 2 ! such that Zn0 does not cover all countable
faces of R!1 , so there is a countable set A0 !1 and x0 2 RA0 for which A10 .x0 / \
Zn0 D ;. Suppose that < !1 , and for every < we have n 2 !, a countable
set A !1 and x 2 RA with the following properties:
(2) <
< implies n nS
and A \ A
D ;; S
(3) for any < , let B D fA
W
g and y D fx
W
g 2 RB .
then B1 .y / \ Zn D ;.
S S
Let y D fx W < T g; the set B D fA W < g is countable and y 2 RB .
Furthermore, B1 .y/ fB1 .y / W < g which shows that B1 .y/ \ Zn D ;
for any < .
If N D fn 2 ! W Zn \B1 .y/ ;g, then N \fn W < g D ; and (1) implies
that it is impossible for Zn \ B1 .y/ to cover all countable faces of B1 .y/ for all
n 2 N . Therefore there exists a countable set A !1 nB and x 2 RA such that,
for y D y [ x and B D B [ A we have B1 .y / \ Zn D ; for some n 2 N .

It is evident that (2) and (3) still hold for all so our inductive construction
can be continued to obtain, among other things, a set fn W < !1 g ! such that
n n
for distinct ;
2 !1 [see (2)].
This contradiction shows that R!1 has a P-representation and hence Fact 2 is
proved. t
u
Fact 3. There exists a closed discrete D R!1 such that jDj D !1 while A .D/ is
countable for any countable A !1 . Here B W R!1 ! RB is the natural projection
for any B !1 .
Proof. Since ! is a closed subset of R, the space ! !1 is a closed subset of R!1 , so
it suffices to construct the promised set in ! !1 . It was proved in Problem 068 that
there exists an !1 -sequence fs W < !1 g such that s W ! ! is an injection for
each < !1 and s j s whenever < < !1 (recall that if we have sets T; S
and maps p; q W R ! S , then p q says that the set fr 2 R W p.r/ q.r/g is
finite). For every countable ordinal ! fix a bijection b W ! ! and define an
element f 2 ! !1 as follows: f ./ D sC1 ./ for all ; if > , then let
f ./ D b1 ./.
We claim that the set D D ff W ! < !1 g is as promised. First assume that
< < !1 and take any
> . Then f .
/ D b
1 ./ b
1 ./ D f .
/ (the
inequality in the middle is true because b
1 is a bijection). This proves that f f
if so jDj D !1 .
To prove that D is closed and discrete in ! !1 , it suffices to show that every
f 2 ! !1 has a neighborhood which intersects at most one element of D. So, take
any f 2 ! !1 and let ./ D b .f .// for any !. By our definition of b , we
have ./ < for all ! and hence there is a stationary S !1 and < !1
such that ./ D for all 2 S .
Therefore there exist ; 2 S such that < < and f ./ D f . / D n. The
set W D fg 2 ! !1 W g./ D g. / D ng is an open neighborhood of f in ! !1 . If
, then f W because f j. C 1/ is an injection so f ./ f . /.
If < then f . / D b 1 ./ while n D b 1 ./; since b 1 is a bijection
and < , we have f . / n, i.e., again f W .
Finally, if < , then f ./ D b1 ./, so if f 2 W , then b1 ./ D b1 ./
and hence D because b is a bijection. Thus the only element of D which can
belong to W is f , so the set D is, indeed, closed and discrete in ! !1 .
To see that B .D/ is countable for any countable B !1 , it suffices to show
that the set Q D ff j W ! < !1 g is countable for any < !1 . Now,
Q ff j W g [ ff j W < < !1 g. The first set in this union is evidently,
countable. The second one is equal to the set P D fsC1 j W < < !1 g.
However, p s for any p 2 P and therefore P P 0 D fs 2 ! W s s g. Since
it is evident that P 0 is countable, the set D has all required properties, so Fact 3 is
proved. t
u
Returning to our solution, suppose first that ext.X / > !; then R!1 embeds in
Cp .X / as a closed subspace (see Fact 1 of S.215). An immediate consequence is
that R!1 is a countable union of its normal

S subspaces. This makes it possible to
apply Fact 2 to conclude that R!1 D fZn W n 2 !g where Zn is normal and
covers all countable faces of R!1 for all n 2 !. By Fact 1 every Zn is C -embedded
in R!1 .
Now apply Fact 3 to find a closed discrete D R!1 such that jDj D !1 , but
B .D/ is countable for any countable B !1 . There is n 2 ! such that D \ Zn
is uncountable, so we can assume, without loss of generality, that D Zn . Take
any injection u W D ! R; since u is continuous on D, by normality of Zn there
is a continuous v W Zn ! R such that vjD D u. Since Zn is C -embedded in
R!1 , there is a continuous w W R!1 ! R such that wjZn D v and hence wjD D
u. By Problem 299 of [TFS] there is a countable B !1 and a continuous map
s W RB ! R such that s B D w. Thus w.D/ D u.D/ D s.B .D// is a
countable subset of R because B .D/ is countable. However u.D/ is uncountable
because u is an injection. This contradiction shows that our assumption is false, i.e.,
ext.X / D ! and hence w.X / D ! (see Problem 214 of [TFS]). As a consequence,
nw.Cp .X // D nw.X / D !; since any space with a countable network is Lindelf,
Cp .X / is Lindelf and hence normal so our solution is complete.
T.456. Let X be an arbitrary space. Given a uniformly dense Y Cp .X /, prove
that t.X / l.Y /.
Solution. Take any A X and x 2 A. Let D l.Y /; it is evident that the set
F D ff 2 Y W f .x/ 12 g is closed in Y so l.F / . For every a 2 A the
S
set Va D ff 2 Y W f .a/ > 13 g is open in Y . Besides, F fVa W a 2 Ag;
indeed, if g 2 F , then g.x/ 12 and hence the set W D g 1 .. 13 ; C1// is an open
neighborhood of x. Since x 2 A, there is a 2 A \ W and hence g 2 Va .
S Now, l.F / implies that there is a set B A such that jBj and F
fVa W a 2 Bg. If x B, then there is f 2 Cp .X / such that f .x/ D 1 and
f .a/ D 0 for all a 2 B. The set Y is uniformly dense in Cp .X /, so there is
g 2 Y for which jg.y/ f .y/j < 13 for all y 2 X . Consequently, g.x/ > 23 12
S
and g.a/ < 13 for any a 2 B which shows that g 2 F n. fVa W a 2 Bg/; this
contradiction proves that x 2 B. Thus, for any A X and x 2 A, we found a set
B A such that jBj and x 2 B. Therefore t.X / D l.Y /.
T.457. For an arbitrary space X and a uniformly dense Y Cp .X /, prove that
nw.Y / D nw.Cp .X // and d.Y / D d.Cp .X //.
Solution. Our solution will be derived from the following statement.
Fact 1. If Z is a space and A is uniformly dense in Cp .Z/, then Z is homeomorphic
to a closed subspace of Cp .A/.
Proof. It is easy to see that A C f is uniformly dense in Cp .Z/ for any f 2 Cp .Z/;
since ACf is homeomorphic to A, we can assume that u 2 A where u is identically
zero on Z, i.e., u.z/ D 0 for all z 2 Z.
If z 2 Z and F is a closed subspace of Z with z F , then there is f 2 C.Z/
such that f .z/ D 1 and f .F / f0g. Since A is uniformly dense in Z, we can find
g 2 A such that jg.y/ f .y/j < 13 for any y 2 Z. It is evident that g.z/ 2 23 ; 43
while g.F / 2 13 ; 13 and therefore g.z/ g.F /. Thus A separates the points and
closed sets in Z and hence the map ' W Z ! Cp .A/ defined by '.z/.f / D f .z/ for
any z 2 Z and f 2 A is a homeomorphic embedding (see Problem 166 of [TFS]).
To see that '.Z/ is closed in Cp .A/ take any 2 Cp .A/n'.Z/. The function
is continuous at u, so there is U 2 .u; A/ such that .U / . 12 ; 12 /; recalling
that A Cp .Z/, we can see that there exists a finite K Z and " > 0 such that
O D ff 2 A W f .K/ . "; "/g U . Since '.z/ for any z 2 K, there is
V 2 .K; Z/ for which '.V /. It is easy to construct a function f 2 Cp .Z/
such that f .K/ f0g and f .ZnV / f1g. For D minf 14 ; "g take g 2 A such
that jg.z/ f .z/j < for any z 2 Z. Then g.K/ . "; "/ and g.ZnV / 34 ; 54 .
Consequently, g 2 O U which implies .g/ 2 12 ; 12 while '.z/.g/ 2 34 ; 54
for each z 2 ZnV which shows that '.ZnV /.
Consequently, '.Z/ D '.V / [ '.ZnV /, i.e., we established that '.Z/
for any point 2 Cp .A/n'.Z/, i.e., '.Z/ is closed in Cp .A/ and Fact 1 is proved.
t
u
Returning to our solution observe that nw.Y / nw.Cp .X // because network
weight is hereditary (see Problem 159 of [TFS]). On the other hand, the space X
embeds in Cp .Y / by Fact 1 and therefore nw.Cp .X // D nw.X / nw.Cp .Y // D
nw.Y / which proves that nw.Y / D nw.Cp .X // (it is worth mentioning that we
did not need a closed embedding in Cp .Y /: the existence of any embedding is
sufficient).
As to densities, the set Y is dense in Cp .X / (see Problem 344 of [TFS]) which,
evidently, implies d.Cp .X // d.Y /. Now assume that d.Cp .X // D and take a
dense A Cp .X / with jAj . It follows from uniform density of Y in the space
Cp .X / that for any f 2 A there exists a sequence Sf Y such that Sf ! f and
S
hence f 2 S f . If B D fSf W f 2 Ag, then B Y and jBj .
Thus Y Cp .X / D A B which shows that the set B is dense in Y ; since
jBj D d.Cp .X //, we proved that d.Y / d.Cp .X //, i.e., d.Y / D d.Cp .X //
and hence our solution is complete.
hd.Y / D hd.Cp .X //; hl.Y / D hl.Cp .X // and s.Y / D s.Cp .X //.
Solution. It is evident that '.Y / '.Cp .X // for any ' 2 fhd; hl; sg. Now assume
that s.Y / D and there is a discrete D Cp .X / with jDj D C . For any d 2 D
there is a finite set Kd X and nd 2 N such that O.d; Kd ; n1d / \ D D fd g where
O.d; Kd ; n1d / D ff 2 Cp .X / W jf .x/ d.x/j < n1d for any x 2 Kd g for every
d 2 D. There are E D and n 2 N such that jEj D C and nd D n for any
d 2 E.
It follows from uniform density of Y in Cp .X / that for any d 2 E there is
wd 2 Y such that jwd .x/ d.x/j < 3n 1
for any x 2 X . Since s.Y / , the
set fwd W d 2 Eg Y cannot be discrete, so there are distinct d; e 2 E for
which we 2 O.wd ; Kd ; 3n 1
/. Consequently, je.x/ d.x/j D je.x/ we .x/j C
jwe .x/ wd .x/j C jwd .x/ d.x/j < 3n 1

C 3n
1
C 3n
1
D n1 for any x 2 Kd , i.e.,
e 2 O.d; Kd ; n1 / which is a contradiction with the choice of the set O.d; Kd ; n1d /.
This contradiction shows that there are no discrete subsets of Cp .X / of cardinality
C , i.e., s.Cp .X // D s.Y /. Therefore s.Y / D s.Cp .X //.
Now assume that hd.Y / D while hd.Cp .X // > and hence there exists a
left-separated L Cp .X / such that jLj D C (see Problem 004). Let < be a well-
order on L which witnesses that L is left-separated. For any h 2 L there is a finite
Ph X and mh 2 N such that O.h; Ph ; m1h / \ L Lh D fg 2 L W h gg.
There are M L and m 2 N for which jM j D C and mh D m for any h 2 M .
It follows from uniform density of Y in Cp .X / that for any h 2 M there is uh 2 Y
such that juh .x/ h.x/j < 3m 1
for any x 2 X . Given distinct g; h 2 M let ug < uh
if and only if g < h. It is clear that < is a well-order on the set N D fuh W h 2 M g.
Since hd.Y / , the set N cannot be left-separated, so there are g; h 2 M such
that g < h and ug 2 O.uh ; Ph ; 3m 1
/.
Consequently, jg.x/ h.x/j jg.x/ ug .x/jCjug .x/ uh .x/jCjuh .x/ h.x/j <
1
3m
C 3m 1
C 3m
1
D m1 for any x 2 Ph and therefore g 2 O.h; Ph ; m1 / which contradicts
the choice of O.h; Ph ; m1 /. This contradiction shows that there are no left-separated
subsets of Cp .X / of cardinality C , so hd.Cp .X // D hd.Y / (see Problem
004) and therefore hd.Y / D hd.Cp .X //.
Finally, suppose that hl.Y / D while hl.Cp .X // > and hence there is a
right-separated R Cp .X / such that jRj D C (see Problem 005). Let < be a
well-order on R which witnesses that R is right-separated. For any h 2 R there is a
finite Fh X and kh 2 N such that O.h; Fh ; k1h / \ R Rh D fg 2 L W g hg.
There are S R and k 2 N for which jS j D C and kh D k for any h 2 S . It
follows from uniform density of Y in Cp .X / that for any h 2 S there is vh 2 Y
such that jvh .x/ h.x/j < 3m 1
for any x 2 X . Given distinct g; h 2 M let vg < vh
if and only if g < h. It is clear that < is a well-order on the set T D fvh W h 2 S g.
Since hl.Y / , the set T cannot be right-separated, so there are g; h 2 M such
that h < g and vg 2 O.vh ; Fh ; 3k 1
/.
Consequently, jg.x/ h.x/j jg.x/ vg .x/jCjvg .x/ vh .x/jCjvh .x/ h.x/j <
1
3k
C 3k1
C 3k
1
D k1 for any x 2 Fh , and therefore g 2 O.h; Fh ; k1 / which contradicts
the choice of O.h; Fh ; k1 /. This contradiction shows that there are no right-separated
subsets of Cp .X / of cardinality C , so hl.Cp .X // D hl.Y / (see Problem 005)
and therefore hl.Y / D hl.Cp .X //.
T.459. Suppose that X is a space and Y Cp .X / is uniformly dense in Cp .X /.
Prove that if Y is a Lindelf -space, then Cp .X / is also Lindelf .
Solution. The statement of this problem is an easy consequence of the following
fact.
Fact 1. Suppose
T that Z is a space and A is a uniformly dense subset of Cp .Z/. Then
Cp .Z/ D fCn W n 2 Ng where Cn RZ is a continuous image of A n1 ; n1 Z
for every n 2 N.
Proof. Define a map 'n W A n1 ; n1 Z ! RZ by 'n .f; g/ D f C g for each

.f; g/ 2 A n1 ; n1 Z . It is clear that 'n is a continuous map for all n 2 N; let

Cn D 'n A n1 ; n1 Z . Now, if n 2 N and h 2 Cp .Z/, then there exists f 2 A
such that jh.z/ f .z/j < n1 for all z 2 Z. Consequently, g D h f 2 n1 ; T
n
1 Z
and
h D f C g which shows that Cp .Z/ Cn for all n 2 N, i.e., Cp .Z/ fCn W
n 2 Ng. T
Now, if h 2 fCn W n 2 Ng, then there is a sequence S D ffn W n 2 Ng
Cp .Z/ such that jfn .z/ h.z/j n1 for all z 2 Z and n 2 N. Thus the sequence S
converges uniformly to h and hence h 2 Cp .Z/ (see Problem 029 of [TFS]). Fact 1
is proved. t
u
Returning to our solution applyT Fact 1 to find a sequence fCn W n 2 Ng of
subsets of RX such that Cp .X / D fCn W n 2 Ng and Cn is a continuous image
of Zn D Y n1 ; n1 X for every n 2 N. Since Y is a Lindelf -space, the space
Zn is also Lindelf (see Problem 256) for any n 2 N. Therefore Cn is a Lindelf
-space for all n 2 N by Problem 243 and hence Cp .X / is Lindelf by Problem
258.
T.460. Suppose that X is a space and Y Cp .X / is uniformly dense in Cp .X /.
Prove that
(i) if Y is K-analytic, then Cp .X / is K-analytic;
(ii) if Y is analytic, then Cp .X / is analytic.
Solution. Apply FactT1 of T.459 to find a sequence fCn W n 2 Ng of subsets of RX
such that Cp .X / D fCn W n 2 Ng and Cn is a continuous image of the space
Zn D Y n1 ; n1 X for every n 2 N.
If Y is a K-analytic space, then the space Zn is also K-analytic (see Problem
343) for any n 2 N. Therefore Cn is a K-analytic space for every n 2 N (because a
continuous image of a K-analytic space is a K-analytic space) and hence Cp .X / is
K-analytic by Problem 344. This proves (i).
As to (ii), if Y is analytic, then nw.Cp .X // D ! by Problem 457; besides, Cp .X /
is K-analytic by (i), so it is analytic by Problem 346.
t.Y / D t.Cp .X //.
Solution. It follows from Problem 159 of [TFS] that t.Y / t.Cp .X //; to prove
the inverse inequality, assume that t.Y / D . Recall that a family A exp.X / is
called an !-cover of X if for every finite K X , there is A 2 A such that K A.
An !-cover A of a space is open if all its elements are open.
Now take an arbitrary open !-cover U of the space X ; for any f 2 Y , let S.f / D
fx 2 X W f .x/ 13 g. Consider the set P D ff 2 Y W S.f / U for some U 2 Ug.
Since Y is uniformly dense in Cp .X /, there is w 2 Y such that w.x/ 23 for
all x 2 X . We claim that w 2 P . Indeed, if O 2 .w; Cp .X //, then there is a
finite K X and " > 0 such that W D ff 2 Cp .X / W jf .x/ w.x/j < "
for all x 2 Kg O. Since U is an !-cover of X , there is U 2 U such that
K U ; it is easy to construct a function f 2 Cp .X / for which f jK D wjK and

f .X nU / f0g. Let D minf 31 ; "g; there is g 2 Y such that jg.x/ f .x/j <
for all x 2 X . Therefore jg.x/ w.x/j < " for all x 2 K and hence g 2 W .
Furthermore, jg.x/ f .x/j < 13 for all x 2 X nU and hence g.x/ < 13 for all
x 2 X nU which shows that S.g/ U . Thus g 2 W \ P O \ P , so O \ P ;
for any O 2 .w; Cp .X // and hence w 2 P .
We have t.Y / , so there exists Q P such that jQj and w 2 Q. For
every f 2 Q there is Uf 2 U for which S.f / Uf ; let U 0 D fUf W f 2 Qg. Then
U 0 U and jU 0 j . Given a finite K X , the set V D ff 2 Y W jf .x/ w.x/j <
1
3
g is an open neighborhood of w in Cp .X /. Since w 2 Q, there is f 2 Q such that
f 2 V and therefore jf .x/ w.x/j < 13 for all x 2 K which implies f .x/ > 13 for
all x 2 K and hence K S.f / Uf . This proves that U 0 is also an !-cover of X .
Thus every open !-cover of X has an !-subcover of cardinality and therefore
t.Cp .X // D t.Y / (see Problems 148 and 149 of [TFS]). As a consequence,
t.Y / D t.Cp .X //.
T.462. Let X be an arbitrary space with ext .X / . Prove that t.Y / for any
compact Y Cp .X /.
Solution. For any x 2 X let ex .f / D f .x/ for any f 2 Y . Then ex 2 Cp .Y /
for any x 2 X and the map e W X ! Cp .Y / defined by e.x/ D ex for all x 2 X
is continuous (see Problem 166 of [TFS]). If Z D e.X /, then ext .Z/ (it
is an easy exercise that if a space P is a continuous image of a space Q, then
ext .P / ext .Q/).
Since Z Cp .Y /, we have Z n .Cp .Y //n for any n 2 N; if Yi is a
homeomorphic copy of Y for each i D 1; : : : ; n, then .Cp .Y //n is homeomorphic to
the space Cp .Y1

Yn / (see Problem 114 of [TFS]). Since Kn D Y1

Yn is
a compact space, we proved that Z n embeds in Cp .Kn / for some compact space Kn
for any n 2 N. Thus we can apply Baturovs theorem (see Problem 269) to conclude
that l.Z n / D ext.Z n / for all n 2 N. As a consequence l .Z/ and hence
t.Cp .Z// (see Problem 149 of [TFS]).
Given f; g 2 Y with f g there is a point x 2 X such that f .x/ g.x/.
Therefore ex 2 Z and ex .f / ex .g/ which proves that Z separates the points of
Y and hence Y embeds in Cp .Z/ (see Fact 2 of S.351). As a consequence, t.Y /
t.Cp .Z// D .
T.463. Suppose that X has the Gerlits property '. Prove that all continuous images
and all closed subspaces of X have '.
Solution. Let f W X ! Y be a continuous S onto map; suppose that a family U
.Y / is an !-cover of Y such that U D n2! Un and S Un UnC1 for all n 2 !. If
Vn D ff 1 .U / W U 2 Un g for all n 2 !, then V D n2! Vn is an open !-cover of
X such that Vn VnC1 for any n 2 !. The space X has the Gerlits property ', so
there is a sequence fXn W n 2 !g exp.X / such that Xn ! X and Xn is !-covered
by Vn for every n 2 !. Let Yn D f .Xn /; it is immediate that Yn ! Y and Yn is
!-covered by Un for each n 2 !. Therefore Y ` '.
Now assume thatS F is a closed subset of X and a family U .F / is an !-cover

of F such that U D n2! Un and Un UnC1 for all n 2 !. Let O.U / D X n.F nU /
for any U 2 .F /; it is evident that O.U / 2 .X/ and O.U / \ F D U . The
family SVn D fO.U / W U 2 Un g consists of open subsets of X for any n 2 !; if
V D n2! Vn , then V is an open !-cover of X such that Vn VnC1 for all n 2 !.
It follows from X ` ' that there is a sequence fXn W n 2 !g exp.X / such that
Xn ! X and Xn is !-covered by Vn for any n 2 !. Let Yn D Xn \ F for all n 2 !;
it is evident that Yn ! F and Yn is !-covered by Un for every n 2 !. This proves
that F ` '.
T.464. Prove that Cp .X / is a FrchetUrysohn space if and only if X has the Gerlits
property ' and t.Cp .X // D !.
Solution. We will first establish that ' is equivalent to some formally weaker
property ' 0 .
Fact 1. If Z is an arbitrary space, then Z ` ' if and only if Z ` ' 0 whereS
' 0 is the
following property: for any open !-cover U of the space Z such that U D n2! Un
and Un UnC1 for any n 2 !, there are sequences fZn W n 2 !g exp.Z/ and
fkn W n 2 !g ! such that Zn ! Z; kn < knC1 and Zn is !-covered by Ukn for
any n 2 !.
Proof. It is evident that ' implies ' 0 , soSassume that Z ` ' 0 and take any open
!-cover U of the space Z such that U D n2! Un and Un UnC1 for every n 2 !.
It follows from the property ' 0 that there are sequences fZn0 W n 2 !g exp.Z/
and fkn W n 2 !g ! such that Zn0 ! Z; kn < knC1 and Zn0 is !-covered by Ukn
for all n 2 !.
Let Zi D ; for all i < k0 ; assume that m 2 ! and we have defined Zi for all i <
km in such a way that Zi is !-covered by Ui for all i < km . Letting Zi D Zk0 m for
all i D km ; : : : ; kmC1 1 we define the set Zi for all i < kmC1 and it follows from
Ukm Ui that Zi is !-covered by Ui for all i D km ; : : : ; kmC1 1. Consequently,
Zi is !-covered by Ui for all i < kmC1 , and hence our inductive construction can
be continued giving us a sequence fZi W i 2 !g exp.Z/ such that Zi ! Z and
Zi is !-covered by Ui for all i 2 !. Therefore Z ` ' and Fact 1 is proved. t
u
Returning to our solution assume that Cp .X / is a FrchetUrysohn space. Then
t.Cp .X // ! because every FrchetUrysohn space has countable tightness.
S Now
suppose that U is an open !-cover of the space X such that U D n2! Un and
Un UnC1 for any n 2 !. By Problem 144 of [TFS] there is a sequence fUn W
n 2 !g U such that Un ! X . It is easy to choose a strictly increasing sequence
fkn W n 2 !g ! such that Un 2 Ukn for all n 2 !. It is clear that Un is !-covered
by Ukn for each n 2 !, so X ` ' by Fact 1. This proves necessity.
Now assume that t.Cp .X // D ! and X ` '. Given an open !-cover U of the
space X , there is a countable V U such that V is also an !-cover of X (see
Problem 148 and 149 S of [TFS]); let V D fVn W n 2 !g and Vn D fVi W i ng for
every n 2 !. Then n2! Vn D V and Vn VnC1 for every n 2 !. Since X has ',
there is a sequence fXn W n 2 !g exp.X / such that Xn ! X and Xn is !-covered
by Vn . However, a set is !-covered by a finite family if and only if it is contained in

an element of that family. Thus there is Un 2 Vn for which Xn Un for all n 2 !.
An immediate consequence is that Un ! X , so for any open !-cover U of the space
X , we have a sequence fUn W n 2 !g U such that Un ! X . Therefore Cp .X / is
a FrchetUrysohn space (see Problem 144 of [TFS]) so sufficiency is settled and
T.465 (GerlitsPytkeev theorem). Prove that the following conditions are equiva-
lent for any space X :
(ii) Cp .X / is a sequential space;
(iii) Cp .X / is a k-space.
Solution. The implications (i)H)(ii)H)(iii) are clear.
Fact 1. Let Z be any space; then, for any pseudocompact P Cp .Z/ and z 2 Z
the set P .z/ D ff .z/ W f 2 P g is compact and hence bounded in R.
Proof. The map ez W Cp .Z/ ! R defined by ez .f / D f .z/ for all f 2 Cp .Z/
is continuous by Problem 166 of [TFS], so P .z/ D ez .P / is a continuous image
of P . Thus P .z/ is pseudocompact; being second countable it is compact and hence
Fact 1 is proved. t
u
Recall that a family U .X/ is called an open !-cover of X if for any finite
K X , there is U 2 U such that K U . We will first show that
(1) if Cp .X / is a k-space, then X has the Gerlits property '.
Assume towards a contradiction that Cp .X / is a k-space and X does not have '.
0
It was proved in Fact 1 of T.464 that ' is equivalent to Sthe following property ' :
for any open !-cover U of the space X such that U D n2! Un and Un UnC1 for
all n 2 !, there are sequences fXn W n 2 !g exp.X / and fkn W n 2 !g !
such that Xn ! X; kn < knC1 and Xn is !-covered by Ukn for all n 2 !. Since
0
' does not holds
S for X , the property ' does not hold either, so there is an open
!-cover U D n2! Un of the space X which witnesses this. In particular, X is not
!-covered by Un for all n 2 !.
For any f 2 Cp .X / and n 2 N let Y .f; n/ D fx 2 X W f .x/ < ng and
consider the set An D ff 2 Cp .X / W Y .f; n/ is !-covered by Un g for any n 2 N.
If g 2 Cp .X /nAn , then there is a finite set K Y .g; n/ such that K 6 U for any
U 2 Un . The set O D ff 2 Cp .X / W f .x/ < n for all x 2 Kg is open in Cp .X /
and g 2 O S Cp .X /nAn which proves that An is closed in Cp .X / for all n 2 N.
Let A D fAn W n 2 Ng.
If u 2 Cp .X / is the function which is identically zero on X , then the set
Y .u; n/ D X is not !-covered by Un for each n 2 N and hence u 2 Cp .X /nA.
However, u 2 A; to see this take any W 2 .u; Cp .X //. There exists a finite K X
and " > 0 such that V D ff 2 Cp .X / W jf .x/j < " for all x 2 Kg W . There
is n 2 N and U 2 Un such that K U . It is easy to find a function f 2 Cp .X /
for which f jK D ujK and f .x/ D n C 1 for all x 2 X nU . It is evident that

Y .f; n/ U and therefore f 2 A \ V A \ W . Thus W \ A ; for any
W 2 .u; Cp .X // which proves that u 2 A.
Since u 2 AnA, the set A is not closed in Cp .X /, so we can apply k-property of
Cp .X / to find a compact set C Cp .X / such that C \ A is not closed. For any
x 2 X the set C.x/ D ff .x/ W f 2 C g is compact by Fact 1, so there is n.x/ 2 N
such that C.x/ n.x/; n.x/ . Let Xn D fx 2 XSW n.x/ ng for every n 2 N; it
is immediate that Xn XnC1 for any n 2 N and fXn W n 2 Ng D X . Therefore
Xn ! X ; since U witnesses that ' 0 does not hold, there is m 2 N such that Xm is
not !-covered by Un for any n 2 !.
We claim that C \ An D ; for any n > m. Indeed, if f 2 An for some n >
m, then Y .f; n/ is !-covered by Un ; since Un does not !-cover Xm , there is x 2
Xm nY .f; n/. We have f .x/ n > m and hence f C because g.x/ m for any
g 2 C and x 2 Xm . S
As a consequence, C \ A D fC \ Ak W k mg is a closed subset of Cp .X /.
This contradiction shows that X has ' 0 and hence ' proving the property (1). Our
next step is to establish that
(2) if Cp .X / is a k-space, then ext .X / D !.
Assume, towards a contradiction, that there is an uncountable closed discrete
set D X n for some n 2 N. For every m 2 N, consider the set Pm D ff 2
Cp .X / W j.Y .f; m//n \ Dj mg for every m 2 N. The set Pm is closed in the space
Cp .X / for any m 2 N; indeed, if f 2 Cp .X /nPm , then there are y1 ; : : : ; ymC1 2
.Y .f; m//n \ D. We have yi D .y1i ; : : : ; yni / for all i m C 1 and f .yji / < m for
every i m C 1 and j n. The set K D fyji W i m C 1; j ng is finite, so
W D fh 2 Cp .X / W h.y/ < m for all y 2 Kg is open in Cp .X /. Given any h 2 W ,
it is immediate that yi 2 .Y .h; m//n \ D for each i m C 1 which shows that
h Pm . Thus W \ Pm D ;S and hence Pm is closed in Cp .X /.
Observe that u P D fPm W m 2 Ng because Y .u; m/ D X and therefore
the set .Y .u; m//n \ D D D is uncountable for any m 2 N. However, u 2 P ; to
see it take any O 2 .u; Cp .X //. There is a finite set L X and " > 0 such that
G D ff 2 Cp .X / W jf .x/j < " for all x 2 Lg O. The set Ln X n is finite,
so it has a neighborhood in X n which contains finitely many points of D. Therefore
there is k 2 N and V 2 .L; X/ such that k > 1 and jV n \ Dj k. Choose a
function f 2 Cp .X / such that f jK D ujK and f .x/ D k C 1 for all x 2 X nV .
Then Y .f; k/ V which implies j.Y .f; k//n \Dj k, i.e., f 2 Pk \G Pk \O.
Since P \ O ; for any O 2 .u; Cp .X //, we proved that u 2 P nP and hence
P is not closed in Cp .X /.
Now, Cp .X / is a k-space, so there exists a compact C Cp .X / such that C \P
is not closed in C and hence in Cp .X /. This implies that C is not contained in
P1 [

[ Pm for any m 2 N; as a consequence, there is an increasing sequence
fmi W i 2 Ng for which Pmi \ C ; Sfor all i 2 N. Pick a function fi 2 Pmi \ C
for all i 2 N and observe that X n n. f.Y .fi ; mi //n W i 2 Ng/ ; because each
.Y .fi ; mi //n covers only finitely many points of D and D is uncountable.
S
Take any z D .z1 ; : : : ; zn / 2 X n n. f.Y .fi ; mi //n W i 2 Ng/ and observe that for
any i 2 N, there is k 2 f1; : : : ; ng such that fi .zk / mi and therefore supffi .zj / W
j D 1; : : : ; ng mi . Thus the set S D fsupffi .zj / W j D 1; : : : ; ng W i 2 Ng is not
bounded in R; on the other hand, S C.z1 / [

[ C.zn / and hence S is bounded
in R by Fact 1. This contradiction shows that ext .X / !, i.e., (2) is proved.
Finally assume that Cp .X / is a k-space. Then ext .X / ! by (2) and hence
t.F / ! for any compact F Cp .X / (see Problem 462). Now, if A is not closed
in Cp .X /, then choose a compact F Cp .X / such that F \ A is not closed; since
t.F / D !, there is a countable B F \ A such that Bn.F \ A/ ;. The set
F being compact, we have B F and therefore BnA ;. We proved that for
any non-closed A Cp .X /, there is a countable B A such that BnA ;, so
t.Cp .X // ! (see Lemma of S.162).
We finally have t.Cp .X // D !; since also X ` ' by (1), we can apply Problem
464 to conclude that Cp .X / is a FrchetUrysohn space. This settles (iii)H)(i) and
S
T.466. Suppose that X is a -compact space such that Cp .X / D n2! Yn and Yn
is a k-space for every n 2 !. Prove that Cp .X / is a FrchetUrysohn space. In
particular, if X is -compact and Cp .X / is a countable union of sequential spaces,
then Cp .X / is a FrchetUrysohn space.
Solution. The space Cp .X / is Whyburn by Problem 216; it is evident that Whyburn
property is hereditary, so every Yn is Whyburn and hence FrchetUrysohn by
Problem 210. Now apply Problem 450 to conclude that Cp .X / is a FrchetUrysohn
space.
T.467. Suppose that we have arbitrary spaces X and Y and a continuous map ' W
X ! Y . Let ' .f / D f ' for any function f 2 Cp .Y /; this gives us a map
' W Cp .Y / ! R D ' .Cp .Y // Cp .X /. Define r' W Cp .Cp .X // ! Cp .Cp .Y //
by r' ./ D .jR/ ' for any 2 Cp .Cp .X //. Prove that r' is a continuous
ring homomorphism such that r' jX D ' (here we identify the spaces X and Y
with their canonical copies in Cp .Cp .X // and Cp .Cp .Y // respectively). Prove that
the map ' is the unique continuous ring homomorphism with this property, i.e.,
if s W Cp .Cp .X // ! Cp .Cp .Y // is a continuous ring homomorphism such that
sjX D ', then s D r' .
Solution. It is straightforward that ' is a ring homomorphism; besides, the map
' is continuous by Problem 163 of [TFS]. Applying Problem 163 of [TFS] once
more, we can convince ourselves that the map W Cp .R/ ! Cp .Cp .Y // defined by
./ D ' for any 2 Cp .R/ is a ring homomorphism. Since any restriction map
is also a continuous ring homomorphism, this proves that the map r' is a continuous
ring homomorphism.
Recall that the canonical identification of X with a subspace of Cp .Cp .X // is
determined by the map e W X ! Cp .Cp .X // defined by e.x/.f / D f .x/ for any
x 2 X and f 2 Cp .X /. The mapping e W X ! e.X / is a homeomorphism and
e.X / is a closed subspace of Cp .Cp .X // (see Problem 167 of [TFS]).
Analogously, let q.y/.f / D f .y/ for any y 2 Y and f 2 Cp .Y /. Then the map
q W Y ! q.Y / Cp .Cp .Y // is a homeomorphism. To check that r' extends ', we
must show that r' .e.X // q.Y / and q 1 r' e D '; since q is a homeomorphism,
it is equivalent to proving that r' e D q ', so take any x 2 X and f 2 Cp .Y /.
Then r' .e.x//.f / D e.x/.' .f // D ' .f /.x/ D f .'.x// D q.'.x//.f / which
shows that r' .e.x//.f / D q.'.x//.f / for any f 2 Cp .Y / and therefore r' e D
q '. Thus, if we identify e.X / with X and q.Y / with Y , then r' jX D '.
Now assume that s W Cp .Cp .X // ! Cp .Cp .Y // is a ring homomorphism such
that sje.X / D r' je.X /. For any set A Cp .Cp .X //, consider the sets P .A/ D
ff1
: : :
fn W n 2 N; fi 2 A for all i ng and R.A/ D f0 C1
g1 C

Cm
gm W
m 2 N; i 2 R and gi 2 P .A/ for all i mg.
It was proved in Fact 1 of S.312 that R.A/ is an algebra in Cp .Cp .X // with
A R.A/. Now observe that e.X / separates the points of Cp .X /, so R.e.X // is
an algebra which is dense in Cp .Cp .X // by Problem 192 of [TFS]. Since s and r'
are ring homomorphisms, we have r' ./ D s./ for any 2 R.e.X // (to see it
recall that sje.X / D r' je.X / and every 2 R.e.X // is obtained from elements
of e.X / applying finitely many products and multiplications; both maps s and r'
commute with products and multiplications so s./ D r' ./).
As a consequence r' and s coincide on a dense subspace R.e.X // of the space
Cp .Cp .X // and hence r' D s by Fact 0 of S.351. This proves uniqueness and makes
T.468. Suppose that X is an !-monolithic compact space. Prove that for every
surjective continuous map ' W X ! Y , the map r' W Cp .Cp .X // ! Cp .Cp .Y // is
surjective.
Solution. Given spaces Z; T and a continuous map W Z ! T , let .f / D f
for any f 2 Cp .T /. This defines a continuous map W Cp .T / ! Cp .Z/ which is
an embedding if is onto (see Problem 163 of [TFS]). As usual, for any subspace
A Z, let Cp .AjZ/ D ff jA W f 2 Cp .Z/g Cp .A/.
Fact 1. Let K be an !-monolithic compact space. If W K ! L is a continuous
onto map, then E D .Cp .L// is C -embedded in Cp .K/.
Proof. For any C D L let CD W Cp .DjL/ ! Cp .C jL/ be the restriction map
and let C D CL . Analogously, if A K, then pA W Cp .K/ ! Cp .AjK/ is the
relevant restriction map.
Since is closed and onto, the map is a homeomorphic embedding and E is
closed in Cp .K/ (see Problem 163 of [TFS]). If 2 Cp .E/, then W Cp .L/ !
R is continuous and hence there is a countable B L and a continuous map ' 0 W
Cp .BjL/ ! R such that D ' 0 B (see Problem 300 of [TFS]). Let G D B;
it is evident that for ' D ' 0 BG , we have D ' G . Choose a countable
A K such that .A/ D B; if F D A then .F / D G. We claim that
(1) pF .E/ D .jF / .Cp .G//.
Indeed, if f 2 Cp .G/, then by normality of L there is g 2 Cp .L/ such that
gjG D f . Then h D g 2 E and pF .h/ D f D .jF / .f / which shows
that .jF / .Cp .G// pF .E/. If, on the other hand, h 2 E, then h D g for
some g 2 Cp .L/. It is evident that for f D gjG we have f .jF / D pF .h/ and
hence (1) is proved.
The spaces F and G being compact, the map jF W F ! G is closed and hence
pF .E/ D .jF / .Cp .G// is closed in Cp .F /. Since X is !-monolithic, we have
w.F / D ! and hence nw.Cp .F // D ! which proves that Cp .F / is normal. The
map .jF / W Cp .G/ ! pF .E/ is a homeomorphism, so its inverse
W pF .E/ !
Cp .G/ is continuous. Since Cp .GjL/ D Cp .G/ by normality of L, the map ' is
defined on the whole Cp .G/, so '
W PF .E/ ! R can be continuously extended
over Cp .F /, i.e., there is 2 Cp .Cp .F // such that jpF .E/ D '
. If D pF ,
then 2 Cp .Cp .K//; we claim that jE D .
Indeed, take any function h 2 E; there is g 2 Cp .L/ with g D h. We
have .h/ D . .g// D '.G .g//. Furthermore, if f D G .g/, then pF .h/ D
f .jF / D .jF / .f /. This implies that
.h/ D .pF .h// D ..jF / .f // D '.

..jF / .f /// D '.f / D '.G .g// D .h/:
Thus .h/ D .h/ for any h 2 E and hence is a continuous extension of the
function . Since a continuous W E ! R was chosen arbitrarily, the set E is C -
embedded in Cp .K/ and Fact 1 is proved. t
u
Returning to our solution let R D ' .Cp .Y // Cp .X /. Then r' ./ D
.jR/ ' for any 2 Cp .Cp .X // (see Problem 467). The map ' W Cp .Y / ! R
is a homeomorphism by Problem 163 of [TFS] so its inverse
W R ! Cp .Y /
is continuous. The set R is C -embedded in Cp .X / by Fact 1, so for any 2
Cp .Cp .Y // the continuous map
W R ! R can be continuously extended
over Cp .X /, i.e., there is 2 Cp .Cp .X // such that jR D
. We have
r' ./ D .jR/' D .
/' D .
' / D , i.e., for any 2 Cp .Cp .Y //,
there is 2 Cp .Cp .X // such that r' ./ D which shows that r' is surjective and
finishes our solution.
T.469. Given spaces X and Y , let ' W X ! Y be a continuous onto map. Prove that
the mapping r' W Cp .Cp .X // ! r' .Cp .Cp .X /// Cp .Cp .Y // is open if and only
if ' is R-quotient.
Solution. For any f 2 Cp .Y / let ' .f / D f '. Then ' W Cp .Y / ! Cp .X / is an
embedding by Problem 163 of [TFS]. If R D ' .Cp .Y // Cp .X /, then r' ./ D
.jR/ ' for any 2 Cp .Cp .X // (see Problem 467). Consider the restriction map
W Cp .Cp .X // ! Cp .R/ and let ./ D ' for any 2 Cp .R/. Then W
Cp .R/ ! Cp .Cp .Y // is a homeomorphism by Problem 163 of [TFS]. Therefore
r' D is open if and only if the map is open which happens if and only
if R is closed in Cp .X / (see Problem 152 of [TFS]). Finally apply Problem 163 of
[TFS] to conclude that R is closed in Cp .X / if and only if ' is R-quotient.
T.470. Suppose that there exists a continuous map of Cp .X / onto Cp .Y /. Prove that
nw.Y / nw.X /.
Solution. Network weight is not raised by continuous maps (see Problem 157
of [TFS]), so if there is a continuous onto map ' W Cp .X / ! Cp .Y /, then
nw.Cp .Y // nw.Cp .X // and hence nw.Y / D nw.Cp .Y // nw.Cp .X // D
nw.X /.
i w.Y / i w.X /.
Solution. The density is not raised by continuous maps (see Problem 157 of [TFS]),
so if there is a continuous onto map ' W Cp .X / ! Cp .Y /, then d.Cp .Y //
d.Cp .X // and hence i w.Y / D d.Cp .Y // d.Cp .X // D i w.X / (see Problem
174 of [TFS]).
s .Y / s .X /; hl .Y / hl .X / and hd .Y / hd .X /.
Solution. We will apply three times the following evident fact.
(1) If a space T is a continuous image of a space Z, then T n is a continuous image
of the space Z n for every n 2 N.
It follows from (1) and Problem 157 of [TFS] that s is not raised by continuous
maps, so if there is a continuous onto map ' W Cp .X / ! Cp .Y /, then s .Cp .Y //
s .Cp .X // and hence s .Y / D s .Cp .Y // s .Cp .X // D s .X / (see Problem
025).
Now apply the property (1) and Problem 157 of [TFS] again to see that hd
is not raised by continuous maps, so if ' W Cp .X / ! Cp .Y / is continuous and
onto, then we have the inequality hd .Cp .Y // hd .Cp .X // and consequently,
hl .Y / D hd .Cp .Y // hd .Cp .X // D hl .X / (see Problem 026).
Finally, apply (1) and Problem 157 of [TFS] to see that hl is not raised
by continuous maps, so if ' W Cp .X / ! Cp .Y / is continuous and onto, then
hl .Cp .Y // hl .Cp .X // and hence hd .Y / D hl .Cp .Y // hl .Cp .X // D
hd .X / (see Problem 027).
if X is -monolithic, then Y is also -monolithic.
Solution. If X is -monolithic, then Cp .X / is -stable by Problem 152; since there
exists a continuous onto map ' W Cp .X / ! Cp .Y /, the space Cp .Y / is also -
stable by Problem 123. Finally, apply Problem 152 again to conclude that Y is
-monolithic.
T.474. Suppose that there exists a quotient map of Cp .X / onto Cp .Y /. Prove that
l .Y / l .X / and q.Y / q.X /.
Solution. We have the inequality t.Cp .Y // t.Cp .X // by Problem 162 of [TFS]

which shows that l .Y / D t.Cp .Y // t.Cp .X // D l .X / (see Problem 149
of [TFS]). Analogously, we have tm .Cp .Y // tm .Cp .X // by Problem 420 of
[TFS] and hence it follows from Problem 434 of [TFS] that q.Y / D tm .Cp .Y //
tm .Cp .X // D q.X /.
T.475. Suppose that there exists a quotient map of Cp .X / onto Cp .Y /. Prove that if
X is l ./-monolithic, then Y is also l ./-monolithic.
Solution. Since X is l ./-monolithic, the space Cp .X / is t./-quotient-stable by
Problem 183. It is evident that the property of being t./-quotient-stable is preserved
by quotient maps, so if there exists a quotient map ' W Cp .X / ! Cp .Y /, then
Cp .Y / is t./-quotient-stable as well, and hence we can apply Problem 183 again
to conclude that Y is l ./-monolithic.
T.476. Suppose that there exists a continuous open map of Cp .X / onto Cp .Y /.
Prove that jY j jX j.
Solution. We have w.Cp .Y // w.Cp .X // by Problem 161 of [TFS] and hence it
follows from Problem 169 of [TFS] that jY j D w.Cp .Y // w.Cp .X // D jX j.
T.477. Suppose that there exists a continuous open map of Cp .X / onto Cp .Y /.
Prove that if X is -scattered, then Y is also -scattered.
Solution. It follows from the fact that X is -scattered that the space Cp .X / is
w./-open-stable (see Problem 187). It is evident that being w./-open-stable is
preserved by open maps, so if there exists an open map ' W Cp .X / ! Cp .Y /,
then the space Cp .Y / is also w./-open-stable. Finally, apply Problem 187 again to
conclude that Y is -scattered.
T.478. Suppose that there exists a continuous closed map of Cp .X / onto Cp .Y /.
Prove that if X is -stable, then Y is also -stable.
Solution. Since X is -stable, the space Cp .X / is -monolithic by Problem 154.
There exists a closed continuous onto map ' W Cp .X / ! Cp .Y /, so the space
Cp .Y / is also -monolithic by Problem 121. Finally, apply Problem 154 again to
conclude that Y is -stable.
T.479. Give an example of spaces X and Y for which there is a continuous map of
Cp .X / onto Cp .Y / while jY j > jX j.
Solution. Let X be the set ! with the discrete topology. If Y D I, then Cp .Y / is
analytic by Problem 367 and hence R! maps continuously onto Cp .Y / by Problem
360. Since Cp .X / is homeomorphic to R! , we have jX j D ! < c D jY j while
Cp .X / maps continuously onto Cp .Y /.
Cp .X / onto Cp .Y / while l .Y / > l .X /.
Solution. Let X D !1 C 1 and Y D !1 where both ordinals are endowed with their
interval topology. Then Y X is countably compact and X D Y (see Problem
314 of [TFS]). Therefore the restriction map W Cp .X / ! Cp .Y / is surjective
and hence Cp .X / maps continuously onto Cp .Y /. Since Y is a countably compact
non-compact space, we have l.Y / > ! and therefore l .X / D ! < l.Y / l .Y /.
Cp .X / onto Cp .Y / while q.Y / > q.X /.
314 of [TFS]). Therefore the restriction map W Cp .X / ! Cp .Y / is surjective and
hence Cp .X / maps continuously onto Cp .Y /. The space X is compact so q.X / D
!; since Y is a countably compact non-compact space, we have q.Y / > ! (see
Problems 401 and 407 of [TFS]) and therefore q.X / < q.Y /.
Cp .X / onto Cp .Y / while X is compact and Y is not -compact.
314 of [TFS]). Therefore the restriction map W Cp .X / ! Cp .Y / is surjective and
hence Cp .X / maps continuously onto Cp .Y /. The space X is compact; since Y is a
countably compact non-compact space, it is not Lindelf and hence not -compact,
T.483. Give an example of spaces X and Y for which there is an open continuous
map of Cp .X / onto Cp .Y / while d.Y / > d.X /.
Solution. Let X D ! and Y D !n! X ; then X is compact and Y is closed
in X (see Fact 1 of S.370). Therefore the restriction map W Cp .X / ! Cp .Y / is
open and surjective (see Problem 152 of [TFS]). This proves that there is an open
continuous map of Cp .X / onto Cp .Y / while d.X / D ! and d.Y / c.Y / D c >
d.X / (see Problem 371 of [TFS]).
map of Cp .X / onto Cp .Y / while tm .Y / > tm .X /.
in X (see Fact 1 of S.370). Therefore the restriction map W Cp .X / ! Cp .Y /
is open and surjective (see Problem 152 of [TFS]). This proves that there is an
open continuous map of Cp .X / onto Cp .Y /; we have tm .X / d.X / D ! (see
Problem 418 of [TFS]) while tm .Y / > ! by Problem 430 of [TFS] and hence
tm .X / < tm .Y /.
map of Cp .X / onto Cp .Y / while c.Y / > c.X / and p.Y / > p.X /.
in X (see Fact 1 of S.370). Therefore the restriction map W Cp .X / ! Cp .Y /
is open and surjective (see Problem 152 of [TFS]). This proves that there is an
open continuous map of Cp .X / onto Cp .Y /; we have c.X / d.X / D ! while
c.Y / D c by Problem 371 of [TFS] and hence c.X / < c.Y /. Furthermore, we
have c.K/ D p.K/ for any compact space K by Problem 282 of [TFS] so p.X / D
c.X / < c.Y / D p.Y /.
Cp .X / onto Cp .Y / while X is discrete and Y is not discrete.
Solution. Let X be the set ! with the discrete topology. If Y D I, then Cp .Y / is
analytic by Problem 367 and hence R! maps continuously onto Cp .Y / by Problem
360. Since Cp .X / is homeomorphic to R! , the space Cp .X / maps continuously
onto Cp .Y / while X is discrete and Y isnt.
T.487. Suppose that there exists a perfect map of Cp .X / onto Cp .Y /. Prove that
d.X / D d.Y /.
Solution. The following easy facts will be useful for our solution.
Fact 1. If Z is a space with w.Z/ , then .K; Z/ for any compact K Z.
Proof. Take a base B in ZS
with jBj . The family U D fU 2 .K; Z/ W there is
a finite B 0 B with U D B 0 g has cardinality . If O 2 .K; Z/, then choose
for any x 2 K a set Ux 2 B such thatS x 2 Ux O. Since K is compact, there is
a finite A K for which K U D fUx W x 2 Ag. It is clear that U 2 U and
K U O, so U is a base of K in Z and Fact 1 is proved. t
u
Fact 2. For any space Z and compact K Cp .Z/, we have .K; Cp .Z// D
.Cp .Z//.
Proof. Assume first that .K; Cp .Z// for some compact K Cp .Z/.
Since the space Cp .Z/ is homogeneous (i.e., for any f; g 2 Cp .Z/, there is a
homeomorphism ' W Cp .Z/ ! Cp .Z/ such that '.f / D g (see Problem 079
of [TFS])), we can assume that u 2 K where u 0. Given a standard set U D
x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .Z/ W f .xi / 2 Oi for all i ng 2 .Cp .Z//,
let supp.U / T D fx1 ; : : : ; xn g. Let V be a family of neighborhoods of K such that
jVj and V D K. Fix any V 2 V. For every f 2 K pick a standard open set
Uf such that f 2 Uf U . Taking any finite subcover fUf1 ; : : : ; Ufm g of the open
cover fUf W f 2 Kg of the compact set K, we obtain a set WV D Uf1 [

[ Ufm
with K WV S V and the set AV D supp.Uf1 / [

[ supp.Ufm /. It is evident
that the set A D fAV W V 2 Vg has cardinality , so it suffices to prove that
A D Z because then .Cp .Z// D d.Z/ (see Problem 173 of [TFS]).
Suppose that x 2 ZnA. The map ex W Cp .Z/ ! R defined by ex .f / D f .x/
is continuous and therefore the set ex .K/ is bounded in R. Choose any r > 0 such
that jf .x/j < r for all f 2 K and find some g 2 Cp .Z/ such that g.x/ D r and
g.A/T f0g. It follows from g.x/ D r that T g K. However gjA D ujA implies
g 2 V which contradicts the fact that V D K. Thus .Cp .Z// D d.Z/
which shows that .Cp .Z// .K; Cp .Z//.
Now, if .Cp .Z// and K Cp .Z/ is compact, then there is a condensation
' W Cp .Z/ ! M such that w.M / (see Problem 173 of [TFS]). The set L D
'.K/ is compact, so .L; M / T .L; M / by Fact 1. Fix a family V
.L; M / such that jVj and V D L. Then,Tfor the family U D f' 1 .V / W
V 2 Vg, we have U .K; Cp .Z//; jUj and U D K so .K; Cp .Z// .
This shows that .K; Cp .Z// D .Cp .Z// for any compact K Cp .Z/ so Fact 2
is proved. t
u
Returning to our solution, assume that ' W Cp .X / ! Cp .Y / is a perfect map.

If .Cp .X // , then take any f 2 Cp .Y /; the set K D ' 1 .f / is compact, so
.K;TCp .Z// by Fact 2. Fix a family U .K; Cp .X // such that jUj
and U D K. For every U 2 U, the set OU D Cp .Y /n'.Cp .X /nU / is open in
Cp .Y / and contains f and ' 1 .OU / U (see Fact 1 of S.226). An immediate
consequence is that for the family V D fOU W U 2 Ug, we have jVj and
T
V D ff g, so .f; Cp .Y // D .Cp .Y // which proves that .Cp .Y //
.Cp .X //.
Now, if .Cp .Y // , then take any f 2 Cp .Y /. The set K D ' 1 .f / is
compact and .K; Cp .X // , so we can apply Fact 2 to see that .Cp .X // .
Therefore .Cp .X // .Cp .Y //, i.e., d.X / D .Cp .X // D .Cp .Y // D d.Y /
(see Problem 173 of [TFS]) and hence our solution is complete.
nw.X / D nw.Y /.
Solution. It follows from Problem 470 that nw.Y / nw.X /, so let ' W Cp .X / !
Cp .Y / be a perfect map and assume that nw.Y / . Then d.X / D d.Y /
nw.Y / by Problem 487, and hence there is a condensation r W Cp .X / ! M
such that w.M / (see Problem 173 of [TFS]). For the map D r' W
Cp .X / ! M Cp .Y /, let C D .Cp .X //. Then W Cp .X / ! C is a
condensation because so is r; furthermore, is perfect because ' is perfect (see
Fact 1 of T.266). Consequently, is a homeomorphism (see Problem 155 of [TFS])
and hence nw.X / D nw.Cp .X // D nw.C / nw.M Cp .Y // . This proves
that nw.X / nw.Y / and therefore nw.X / D nw.Y /.
jX j D jY j.
Solution. We will need the following general fact about perfect maps.
Fact 1. Given spaces Z and T , if f W Z ! T is a perfect map, then w.T / w.Z/.
In other words, perfect maps do not increase weight (recall that all perfect maps in
this book are surjective).
Proof. Fix a base B in the space Z such that jBj D w.Z/. For any U 2 .Z/
let f # .U / D T nf .ZnU /. It is evident that f # .U / is open in T for any U 2 S
.Z/.
Therefore the family C D ff # .U / W there is a finite B 0 B such that U D B 0 g
consists of open subsets of T ; it is immediate that jCj .
To see that C is a base in T , take a point t 2 T and W 2 .t; T /. For any
z 2 K D f 1 .t/, choose Uz 2 B for which z 2 S Uz f 1 .W /. Since K is
compact,S there is a finite A K such that K fUz W z 2 Ag; for the set
U D fUz W z 2 Ag, we have V D f # .U / 2 C and t 2 V W (see Fact 1
of S.226). Therefore C is a base in T which shows that w.T / jCj . Thus
w.T / w.Z/, so Fact 1 is proved. t
u
Fact 2. For any space Z and compact K Cp .Z/ we have .Cp .Z// D
.K; Cp .Z//.
Proof. If .Cp .Z// , then w.Cp .Z// by Problem 169 of [TFS] and hence
.K; Cp .Z// by Fact 1 of T.487. Now assume that .K; Cp .Z// for some
compact K Cp .Z/.
Given a standard set U D x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .Z/ W f .zi / 2 Oi
for all i ng 2 .Cp .Z//, let supp.U / D fx1 ; : : : ; xn g. Fix a base B of
neighborhoods of K in Cp .Z/ with jBj . If V 2 B, then pick for every f 2 K a
standard set Uf such that f 2 Uf V . Taking any finite subcover fUf1 ; : : : ; Ufm g
of the open cover fUf W f 2 Kg of the compact set K, we obtain a set GV D
Uf1 [

[ Ufm with K GV S V and the set AV D supp.Uf1 / [

[ supp.Ufm /.
It is evident that the set A D fAV W V 2 Bg has cardinality , so it suffices to
prove that A D Z because then .Cp .Z// D w.Cp .Z// D jZj (see Problem
169 of [TFS]).
Suppose that x 2 ZnA. The map ex W Cp .Z/ ! R defined by ex .f / D f .x/
is continuous and therefore the set ex .K/ is bounded in R. Choose any r > 0
such that jf .x/j < r for all f 2 K and observe that W D xI . r; r/ is an
open neighborhood of K. There exists V 2 B such that K V W and hence
GV D Uf1 [

[ Ufm W . This implies Uf1 D x1 ; : : : ; xn I O1 ; : : : ; On W
while x fx1 ; : : : ; xn g. Apply Problem 034 of [TFS] to find g 2 Cp .Z/ such that
g.xi / 2 Oi for all i n and g.x/ D r. It is immediate that g 2 V nW which is a
contradiction. Thus A D Z and hence jZj which implies .Cp .Z// D jZj
, i.e., .Cp .Z// .K; Cp .Z//. Thus .K; Cp .Z// D .Cp .Z// and Fact 2 is
proved. t
u
Returning to our solution observe that w.Cp .Y // w.Cp .X // by Fact 1 and
therefore jY j D w.Cp .Y // w.Cp .X // D jX j (see Problem 169 of [TFS]). To
prove that jX j jY j assume that ' W Cp .X / ! Cp .Y / is a perfect map and
.Cp .Y // . Take any f 2 Cp .Y / and let K D ' 1 .f /. There is a base B at
the point f in the space Cp .Y / with jBj . The family C D f' 1 .B/ W B 2 Bg
has cardinality . To see that C is a base of K in the space Cp .X / take any
U 2 .K; Cp .X //.
The set W D Cp .Y /nf .Cp .X /nU / is open in the space Cp .Y / while f 2 W
and ' 1 .W / U (see Fact 1 of S.266). Pick a set B 2 B with B W ; then V D
' 1 .B/ 2 C and K V U . Thus C is a base of K in Cp .X / such that jCj .
This proves that .K; Cp .X // and therefore w.Cp .X // D .Cp .X //
by Fact 2. Finally, jX j D w.Cp .X // (see Problem 169 of [TFS]) and hence
jX j jY j, i.e., jX j D jY j, so our solution is complete.
hd .X / D hd .Y /.
S a space P and an infinite cardinal , a set A P is an F -subset

Solution. Given
of P if A D F where every F 2 F is closed in P and jF j .
Fact 1. Given spaces Z; T and an infinite cardinal if hl.Z/ and w.T / ,

then every G 2 .Z T / is an F -subset of Z T .
Proof. SLet B be a base in T such that jBj ; for any B 2 B, consider the set
OB D fU 2 .Z/ W U B Gg. The set OB is open in Z, so OB is an F -set
in Z (see Problem 001). Since hl.T / w.T / , we can also apply Problem 001
S in T . Thus OB B is an F -set in Z T for any
to conclude that B is an F -set
B 2 B. Consequently, G D fOB B W B 2 Bg is an F -subset of Z T and
Fact 1 is proved. t
u
Fact 2. Given spaces Z; T and an infinite cardinal , if there is a perfect map f W
Z ! T and l.T / , then l.Z/ .
Proof. Let U be an open cover S of the space Z. For any t 2 T there is a finite
Ut U such thatSf 1 .t/ Ut ; by Fact 1 of S.226, there is Wt 2 .t; T / such
that f 1 .Wt / Ut .
S fWt W t 2 T g is an open cover
Since
1
of T , there is A T such that jAj
and fWt W t 2 Ag D T . If S tV D f .W t / for any t 2 A, then
SfVt W t 2 Ag is a
cover of Z such that Vt Ut for all t 2 A. Therefore U 0 D fUt W t 2 Ag is a
subcover of U with jU 0 j . This shows that l.Z/ so Fact 2 is proved. t
u
Returning to our solution observe that it follows from Problem 472 that we have
hd .Y / hd .X /. To prove the inverse inequality, we will need the following
property:
(1) For any spaces Z and T if there exists a perfect map ' W Cp .Z/ ! Cp .T /, then
hl.Cp .Z// hl.Cp .T //.
To show that (1) holds let D hl.Cp .T //; then .Cp .Z// D d.Z/ D d.T / D
.Cp .T // D by Problem 487. Therefore there is a condensation r W Cp .Z/ ! M
such that w.M / (see Problem 173 of [TFS]). For the map D r' W Cp .Z/ !
M Cp .T /, let Q D .Cp .Z//. Then W Cp .Z/ ! Q is a condensation because so
is r; besides, is perfect because ' is perfect (see Fact 1 of Problem 266 of [TFS]).
Consequently, is a homeomorphism (see Problem 155 of [TFS]) and hence Cp .Z/
embeds in M Cp .T /. Therefore Fact 1 is applicable to conclude that every open
subset of Cp .Z/ is an F -subset of Cp .Z/ (Fact 1 actually says that this holds for
the whole product, but it is evident that the relevant property is hereditary). Since
l.Cp .Z// by Fact 2, we can apply Problem 001 to see that hl.Cp .Z// ;
this proves the property (1).
Finally, assume that hd .Y / ; then hl .Cp .Y // (see Problem 027). The
space .Cp .X //n can be also perfectly mapped onto .Cp .Y //n for any n 2 N (see
Fact 4 of S.271). Given n 2 N, let Dn be a discrete space of cardinality n; since
.Cp .Z//n is homeomorphic to Cp .Z Dn / for any space Z (see Problem 114 of
[TFS]), the space Cp .X Dn / maps perfectly onto the space Cp .Y Dn / for any n 2
N. This makes it possible to apply (1) to conclude that hl.Cp .X Dn // for all
n 2 N, i.e., hl .Cp .X // . Therefore hd .X / D hl .Cp .X // (see Problem
027) which implies hd .X / hd .Y /. Consequently, hd .X / D hd .Y / so our
T.491. Can Cp .!n!/ be mapped continuously onto Cp .!1 /?

Solution. No, the space Cp .!n!/ cannot be mapped continuously onto Cp .!1 /.
We will deduce it from the following fact.
Fact 1. For any cardinal with cf./ > ! we have p.Cp .// D . Here is
endowed with its interval topology.
Proof. For every < let e .f / D f ./ for any f 2 Cp ./. Then e W Cp ./ !
R is continuous by Problem 167 of [TFS] and hence ' D e eC1 is also
continuous on Cp ./, so the set U D ff 2 Cp ./ W ' .f / > 1g is open in
Cp ./. It is easy to see that for every < there a function f 2 Cp ./ such that
f ./ D 2 and f . C 1/ D 0; as a consequence, f 2 U which proves that U ;
for any < . To see that the family U D fU W < g is point-finite suppose not
and fix a function f 2 Cp ./ such that A D f 2 W f 2 U g is infinite. Choose a
countably infinite B A; since is not !-cofinal, there is < such that B .
The space C1 is compact by Problem 306 of [TFS], so there is a faithfully indexed
sequence S D fn W n 2 !g B which has an accumulation point .
The function f being continuous, there is
< such that jf ./ f ./j < 13
for all such that
< < . Since is an accumulation point of the sequence S ,
there is n 2 ! for which
< n < . Then j'n .f /j D jf .n / f .n C 1/j
jf .n / f ./j C jf .n C 1/ f ./j < 13 C 13 < 1; however f 2 Un and
hence 'n .f / > 1 which is a contradiction. Thus U is a point-finite family of non-
empty open subsets of Cp ./; since jUj D , we have p.Cp .// . The inverse
inequality follows from p.Cp .// w.Cp .// D (see Problem 169 of [TFS]) so
Fact 1 is proved. t
u
Now observe that the restriction map W Cp .!/ ! Cp .!n!/ is continuous
and onto (see Problem 152 of [TFS] and Fact 1 of S.370). We have p.Cp .!// D !
by Problem 382 of [TFS], so p.Cp .!n!// D ! because point-finite cellularity
is not raised by continuous maps. For the same reason any continuous image of
Cp .!n!/ must have countable point-finite cellularity. However, p.Cp .!1 // D !1
by Fact 1 so Cp .!1 / is not a continuous image of Cp .!n!/.
T.492. Suppose that there exists a perfect irreducible map ' W Cp .X / ! R for
some cardinal . Prove that X is discrete.
Solution. The space R is pseudocomplete by Problems 465 and 470 of [TFS], so
we can take a pseudocomplete sequence S D fBn W n 2 !g of -bases in R . The
family Cn D f' 1 .U / W U 2 Bn g is a -base in Cp .X / for every n 2 ! (see Fact 2
of S.373). To see that the sequence fCn W n 2 !g is pseudocomplete assume that
Cn 2 Cn and C nC1 Cn for each n 2 !. There is Bn 2 Bn for which Cn D ' 1 .Bn /
for all n 2 !. The map 'T being closed, we have B nC1 D '.C nC1 / '.Cn / D Bn
for every n 2 !, so A D n2! BT n ; because the sequence
T S is pseudocomplete.
If f 2 A, then ; ' 1 .f / n2! Cn and therefore n2! Cn ;. This proves
that Cp .X / is pseudocomplete.
Fact 1. Any perfect preimage of a realcompact space is realcompact.
Proof. Let f W S ! T be a perfect map such that T is realcompact. There exists

a continuous map g W S ! T such that gjS D f (see Problem 257 of [TFS]).
Besides, g.S nS / T nT (see Fact 3 of S.261) and therefore g 1 .T / D S . This
makes it possible to apply Fact 1 of S.408 (for R D S; Z D T and B D T ) to
see that S is realcompact and finish the proof of Fact 1. t
u
To end our solution observe that Cp .X / is realcompact by Fact 1 because so is
R (see Problem 401 of [TFS]). Thus Cp .X / is pseudocomplete and realcompact
so we can apply Problem 486 of [TFS] to conclude that X is discrete.
Q
T.493. Let M D t 2T Mt where Mt is a metrizable space for all t 2 T ; assume
that ' W M ! Cp .X / is a closed continuous onto map. Prove that forQ every t 2 T
we can choose a closed separable Nt Mt in such a way that '. t 2T Nt / D
Cp .X /. In particular, a space Cp .X / is a closed continuous image of a product of
(completely) metrizable spaces if and only if it is a closed continuous image of a
product of separable (completely) metrizable spaces.
Solution. Let t W M ! Mt be the natural projection for all t 2 T .
Fact 1. Let Y be a space such that every closed bounded subspace of Y is compact.
Suppose that Z is a space in which any point is a limit of a nontrivial convergent
sequence. Then any closed map h W Y ! Z is irreducible on some closed subset of
Y , i.e., there is a closed F Y such that h.F / D Z and hF D hjF is irreducible.
Proof. For every y 2 Z fix a sequence Sy D fyn W n 2 !g Znfyg converging to
S
y. We will prove first that the set Py D h1 .y/ \ fh1 .yn / W n 2 !g is compact
for every y 2 Z. Indeed, if for some y 2 Z the set Py is not compact, then it
is not bounded being closed in Y . Therefore there is f 2 Cp .Y / such that f is
not bounded on Py and hence there is a sequence fpn W n 2 !g Py such that
jf .pn /j > n for any n 2 !. It is easy to choose a strictly increasing sequence
fkn W n 2 !g ! such that f .pkn / f .pkm / whenever n m and ff .pkn / W
n 2 !g is a closed discrete subset of R. Let rn D f .pkn / for all n 2 !; since R is
collectionwise normal, we can choose a discrete family fOn W n 2 !g .R/ such
that rn 2 On for all n 2 !. If Un D f 1 .On / for all n 2 ! then U D fUn W n 2
!g .Y / is discrete in Y ; if xn D pkn , then xn 2 Un \ Py for all n 2 !.
If ASis an arbitrary finite subset of !, then for each natural number n, we have
1
Un \ . fhS .yk / W k 2 !nAg/ ;. This makes it possible to choose a point
zn 2 Un \ . fh1 .yk / W k 2 !g/ in such a way that h.zm / h.zn / if n m.
The family U being discrete the set D D fzn W n 2 !g is closed and discrete in Y .
The set h.D/ is also closed because h is a closed map. Note that h.D/ has also to
be discrete because h.C / is closed for any C D. However h.D/ is a nontrivial
sequence converging to y, a contradiction with the fact that h.D/ is closed and
discrete. This proves that Py is compact for all y 2 Z.
Claim. Suppose that H is a closed subset of Y such that h.H / D Z. Then H \
Py ; for all y 2 Z.
Proof of the claim. Fix any y 2 Z. It follows from h.H / D Z that it is possible to
choose a point tn 2 H \ h1 .yn / for all n 2 !. The map h is closed and therefore
ftn W n 2 !g \ h1 .y/ ;. But H ftn W n 2 !g and ftn W n 2 !g \ h1 .y/ Py .
Thus H \ Py ; and the claim is proved. t
u
Suppose that we have a family F of closed subsets of Y such T that F is totally
ordered by inclusion and h.H / D Z for every H 2 F . Then h. F / D Z. Indeed,
HT\ Py ; for anyTy 2 Z and H 2 F . We proved that the set P Ty is compact, so
. F /\h1 .y/ . F /\P T y ; for all y 2 Z; consequently,
T . F /\h 1
.y/
; which implies y 2 h. F / for every y 2 Z, i.e., h. F / D Z. Finally, use
Zorns lemma to find a closed F Y which is maximal (with respect to the inverse
inclusion) in the family of all closed sets H Y such that h.H / D Z. It is evident
that hF is irreducible so Fact 1 is proved. t
u
Returning to our solution observe that
(1) every closed bounded subspace B of the space M is compact.
Indeed, for every t 2 T the set Bt D t .B/ is bounded in Mt by Fact 1 of S.399.
Therefore CtQ D B t is compact (see Fact 1 of S.415 and Problem 406 of [TFS]), so
the set C D t 2T Ct is also compact. The set B is closed in M and B C which
shows that B is closed in C and hence compact so (1) is proved.
For every f 2 Cp .X /, let fn D f C n1 for any n 2 N; then ffn W n 2 Ng is
a nontrivial sequence which converges to f . This, together with (1), shows that we
can apply Fact 1 to the map ' to find a closed F M such that '.F / D Cp .X /
and 'jF is irreducible. By Fact 1 of S.228 we have c.F / D c.Cp .X // D !; let
Ft D t .F / and Nt D F t for all t 2 T . Then c.Nt / D c.Ft / D ! and hence Nt is
a closed separable
Q subspace of Mt for any t 2 T (see Problem 214 of [TFS]). Since
F N D t 2T Nt , we have '.N / '.F / D Cp .X / and hence '.N / D Cp .X /.
The map ' 0 D 'jN maps N continuously onto Cp .X /; it is evident that a restriction
of a closed map to a closed subspace is a closed map onto the image of that subspace
so Cp .X / is a closed continuous image of N and our solution is complete.
T.494. Let M be a product of completely metrizable spaces. Suppose that there
exists a continuous closed onto map ' W M ! Cp .X /. Prove that X is discrete.
In particular, X is discrete if Cp .X / is a closed continuous image of R for some
cardinal .
Solution. It follows from Problem 493 that we can assume, without loss of
space M is a product of second countable completely metrizable
generality, that theQ
spaces, i.e., M D fMt W t 2 T g where w.Mt / D ! and Mt is Cech-complete
Q for
every t 2 T . For any non-empty S T the map pS W M ! MS D fMt W t 2 S g
is the natural projection; we write pt instead of pft g for any t 2 T . As usual,
A W Cp .X / ! Cp .AjX / Cp .A/ is the restriction map for any A X .
Fact 1. If h W Y ! Z is a closed irreducible map and Z is Pseudocompact, then Y
is also pseudocompact.
Proof. If the space Y is not pseudocompact, then there exists a discrete family fOn W
n 2 !g .Y /. The set Wn D Znh.Y nOn / is open, non-empty and h1 .Wn /
On for any n 2 !. Choose Vn 2 .Z/ such that V n Wn for all n 2 !. If
Fn D h1 .V n /, then Fn On for all n 2S!, so fFn W n 2 !g is a discrete family of
closed subsets of Y . Consequently, F D n2! Fn is closed in Y which implies that
S
n2! V n D h.F / is closed in Z.
Now take any z S 2 Z. If z 2 Wn for some n 2 !, then Wn \ Vi D ; for all
i n. If z 2 Zn. n2! Wn /, then O D Znh.F / is an open neighborhood of z
which does not intersect any element of V D fVn W n 2 !g. This proves that every
z 2 Z has a neighborhood which intersects at most one element of V. Therefore
V is an infinite discrete family of non-empty open subsets of Z which contradicts
pseudocompactness of Z. Thus Y is also pseudocompact and Fact 1 is proved. u t
Returning to our solution, observe first that Cp .X / is stable because so is M (see
Problem 268). Thus X is !-monolithic, i.e., nw.A/ D ! for any countable A X .
We will establish that, in fact, A is countable for any countable A X . The first
step is to show that
(1) every closed bounded subspace B of the space M is compact.
Indeed, for every t 2 T , the set Bt D t .B/ is bounded in Mt by Fact 1 of
S.399.QTherefore Ct D B t is compact (see Fact 1 of S.415 and V1.406), so the set
C D t 2T Ct is also compact. The set B is closed in M and B C which shows
that B is closed in C and hence compact so (1) is proved.
For any f 2 Cp .X / let fn D f C n1 for all n 2 N; then ffn W n 2 Ng is a
nontrivial sequence which converges to f . Thus
(2) every f 2 Cp .X / is a limit of a nontrivial convergent sequence.
The properties (1) and (2) show that we can apply Fact 1 of T.493 to find a closed
Y M such that '.Y / D Cp .X / and 'jY is an irreducible map. Now assume that
A X is a countable closed subset of X . The map A W Cp .X / ! Cp .AjX / is
open by Problem 152 of [TFS]; apply Problem 299 of [TFS] to find a countable
S T and a continuous map W MS ! Cp .AjX / such that pS D A '. If
YS D pS .Y / and Z D Y S , then it follows from '.Y / D Cp .X / that .YS / D C D
Cp .AjX / and hence .Z/ D C . Let p D pS jY W Y ! YS . We claim that
(3) IntC ..U // ; for any U 2 .Z/.
To see that the property (3) is satisfied, observe that U 0 D U \ YS 2 .YS / and
hence V D p 1 .U 0 / is a non-empty open subset of Y . The map ' 0 D 'jY being
irreducible, there is W 2 .Cp .X // such that W ' 0 .V / D '.V / (see Fact 1 of
S.383). Since the map A is open, A .W / is a non-empty open subset of C such
that A .W / .U / so (3) is proved.
An immediate consequence of (3) is that 0 D jZ W Z ! C is weakly open, i.e.,
IntC . 0 .U // is dense in 0 .U / for any U 2 .Z/. The space Z is closed in a Cech-
complete space MS , so it is Cech-complete and hence pseudocomplete. Applying
Fact 1 of S.471 we can conclude that the space C is also pseudocomplete, and hence
it has a dense Cech-complete subspace D by Problem 468 of [TFS]. Since D is also

dense in Cp .A/, the space A is discrete by Problem 265 of [TFS].
Furthermore, C D Cp .AjX / is an algebra and, in particular, f g 2 Cp .AjX /
for any f; g 2 Cp .AjX /. Take any h 2 RA nCp .AjX /; the map Th W RA ! RA
defined by Th .g/ D h C g is a homeomorphism (Problem 079 of [TFS]) and
Th .Cp .AjX //\Cp .AjX / D ;. Indeed, if g 2 Cp .AjX / and f D hCg 2 Cp .AjX /,
then h D f g 2 Cp .AjX / which is a contradiction. Therefore RA has two disjoint
dense Cech-complete subspaces D and Th .D/ which contradicts Problem 264 of
[TFS]. Thus Cp .AjX / D RA and hence we proved that
(4) every countable closed A X is discrete and C -embedded in X .
Now it is easy to show that A is countable for any countable A X . Indeed,
we have nw.A/ D !, so if jAj > !, then there is a nontrivial convergent sequence
S A (see Fact 1 of S.497); since S is countable, closed and non-discrete, this
contradicts (4) and proves that jAj D ! for any countable A X . Applying (4)
once more we can see that A is discrete and C -embedded in X for any countable
A X . Consequently, any countable A X is also closed, discrete and C -
embedded in X and hence Cp .X / is pseudocomplete by Problem 485 of [TFS].
It follows from pseudocompleteness of Cp .X / that Cp .X; I/ is pseudocompact
(see Problem 476 of [TFS]). If H D ' 1 .Cp .X; I//, then
D 'jH W H !
Cp .X; I/ is a closed map. We leave it to the reader as an easy exercise to prove that
every point of Cp .X; I/ is a limit of a nontrivial convergent sequence. Furthermore,
if F is closed and bounded in H , then it is compact being closed and bounded in M
[see (1)].
Thus we can apply Fact 1 of T.493 to conclude that there is a closed G H such
that
.H / D Cp .X; I/ and
jG is irreducible. Since H is closed in M , the set G is
also closed in M . By Fact 1 the space G is pseudocompact and hence bounded in M .
Therefore G is compact [see (1)], so Cp .X; I/ is also compact being a continuous
image of a compact space G. Finally apply Problem 396 of [TFS] to conclude that
X is discrete and finish our solution.
T.495. Let X be a pseudocompact space. Suppose that Cp .X / contains a dense
subspace which is a continuous image of a product of separable spaces. Prove that
X is compact and metrizable.
Q exists a family fZt W t 2 T g of separable spaces such that for the

Solution. There
space Z D t 2T Zt , there is a continuous map ' W Z ! Cp .X / for which the set
Y D '.Z/ is dense in Cp .X /. Let e.x/.f / D f .x/ for any x 2 X and f 2 Y .
Then e.x/ 2 Cp .Y / for any x 2 X and the map e W X ! Cp .Y / is an injection (see
Problem 166 of [TFS] and Fact 2 of S.351); let X 0 D e.X /. Since ' W Z ! Y is
continuous and onto, the dual map ' W Cp .Y / ! Cp .Z/ embeds Cp .Y / in Cp .Z/
(see Problem 163 of [TFS]) and therefore X 0 is a pseudocompact space which
embeds in Cp .Z/. This implies that X 0 is metrizable (and hence second countable
and compact) by Problem 307 of [TFS]; any condensation of a pseudocompact space
onto a second countable space is a homeomorphism (Problem 140 of [TFS]), so

e W X ! X 0 is a homeomorphism. Therefore X is compact and metrizable being
homeomorphic to a compact metrizable space X 0 .
T.496. Let X be a pseudocompact space. Suppose that Cp .X / is a closed continuous
image of a product of metrizable spaces. Prove that X is countable.
Solution. By Problem 493 the space Cp .X / is a closed continuous image of a
product of separable metric spaces. This implies that X is compact and metrizable
(see Problem 495). Fix a family fMt W t 2 T g of separable
Q metrizable spaces such
that there is a closed continuous
Q onto map ' W M D t 2T M t ! Cp .X /. For any
S T let pS W M ! MS D t 2S Mt be the projection of M onto its face MS .
We have nw.Cp .X // D !, so there is a condensation W Cp .X / ! Z of Cp .X /
onto a second countable space Z. By Problem 299 of [TFS], there is a countable
S T and a continuous map W MS ! Z such that pS D '. For the
map
D 1 , we have
W MS ! Cp .X / and
pS D '; therefore
is
surjective. If U 2 .Cp .X //, then
1 .U / D pS .' 1 .U // is an open subset of MS
because ' is continuous and pS is open. As a consequence,
is a continuous map.
Besides, if F is closed in MS , then
.F / D '.pS1 .F // is closed in Cp .X / because
pS is continuous and ' is closed. Consequently, Cp .X / is a closed image of the
metrizable space MS and hence X is countable by Problem 228 of [TFS].
T.497. Let X be a pseudocompact space. Suppose that Cp .X / is an open continuous
image of a product of separable metrizable spaces. Prove that X is countable.
Solution. The space X is compact and metrizable by Problem 495. By our
hypothesis there is a family fMt W t 2 T g of separable Q metrizable spaces such
that there is an open continuous
Q onto map ' W M D t 2T Mt ! Cp .X /. For any
S T let pS W M ! MS D t 2S Mt be the projection of M onto its face MS .
We have nw.Cp .X // D !, so there is a condensation W Cp .X / ! Z of Cp .X /
onto a second countable space Z. By Problem 299 of [TFS] there is a countable
S T and a continuous map W MS ! Z such that pS D '. For the map
D 1 , we have
W MS ! Cp .X / and
pS D '; therefore
is surjective. If
U 2 .Cp .X //, then
1 .U / D pS .' 1 .U // is an open subset of MS because ' is
continuous and pS is open. As a consequence,
is a continuous map. Besides, if G
is open in MS , then
.G/ D '.pS1 .G// is open in Cp .X / because pS is continuous
and ' is open. Consequently, Cp .X / is an open image of the metrizable space MS
and hence X is countable by Problem 229 of [TFS].
T.498. Let M be a product of separable completely metrizable spaces. Assuming
that there is a finite-to-one open map ' W M ! Cp .X /, prove that X is discrete.
Solution. Suppose that we are given spaces Y; Z and a map f W Y ! Z. We will
say that y 2 Y is a point of local homeomorphism of f if there is U 2 .y; Y /
such that f jU W U !Sf .U / is a homeomorphism. As usual, D D f0; 1g; we let
D0 D f;g and D< D fDn W n < g for any !. If n 2 !; i 2 D and s 2 Dn ,
then s _ i 2 DnC1 is defined by .s _ i /jn D s and s.n/ D i .
Fact 1. Given spaces Y and Z let h W Y ! Z be an open finite-to-one map. If Z

is pseudocomplete, then any non-empty open subset of Y contains a point of local
homeomorphism of h.
Proof. Let fBn W n 2 !g be a pseudocomplete sequence in Z. The map h being
open, it suffices to prove that for every U 2 .Y / there is an open non-empty O
U such that hjO is an injection. To obtain a contradiction assume that this is not
true for some U 2 .Y /. Then for every O 2 .U / there are distinct y0 ; y1 2 O
with h.y0 / D h.y1 /. Let us construct inductively families fUs W s 2 D<! g .Y /
and fVn W n 2 !g .Z/ with the following properties:
(1) Vn 2 Bn and V nC1 Vn for all n 2 !;
(2) the family fUs W s 2 Dn g is disjoint for all n 2 !;
(3) if n 2 ! and s 2 Dn , then h.Us / Vn and Us _ i Us for any i 2 D;
(4) If n 2 ! and s 2 DnC1 , then h.Us / Vn .
If we let U; D U and choose any V0 2 B0 such that V0 h.U; /, then (1)(4)
are fulfilled for n D 0 (the second parts of (1) and (3) as well as the property (4) are
satisfied vacuously). Now assume that m 2 N and we have families fUs W s 2 D<m g
and fVn W n < mg such that (1)(4) are fulfilled for all n < m (to save space let us
agree that the last statement actually says that (2) and the first part of (1) are satisfied
for all n < m while the second parts of (1) and (3) as well as the property (4) are
true for all n < m 1).
Let fsi W 0 i < 2m1 g be an enumeration of the set Dm1 . By our assumption
and (3), there are distinct y0 ; y1 2 Us0 \ h1 .Vm1 / such that h.y0 / D h.y1 /.
Choose O0i 2 .yi ; Us0 \ h1 .Vm1 // for every i 2 D such that O00 \ O01 D ;; then
W0 D h.O00 / \ h.O01 / 2 .Vm1 /.
Now assume that k < 2m1 and we have open non-empty sets Oji and Wj for all
j < k and i 2 D with the following properties:
(5) Oj0 \ Oj1 D ; and Oj0 [ Oj1 Usj for any j < k;
T
(6) Wj D fh.Opi / W p j; i 2 Dg 2 .Vm1 / for all j < k.
It follows from (6) and (3) that G D Usk \ h1 .Wk1 / ;, so we can
choose distinct y0 ; y1 2 G such that h.y0 / D h.y1 /. Take disjoint Ok0 2
.y0 ; Usk \ h1 .Vm1 // and Ok1 2 .y1 ; Usk \ h1 .Vm1 //; it is evident that (5)
and (6) are fulfilled for all j k. Thus we can construct inductively a family
fOji W j < 2m1 ; i 2 Dg with the properties (5) and (6) fulfilled for all j < 2m1 .
Consequently, W D W2m1 1 ;, so we can choose Vm 2 Bm such that V m W .
If s 2 Dm1 , then s D sj for some j < 2m1 ; let Us _ i D Oji for all i 2 D. This
defines the set Us for every s 2 Dm and it is immediate from the choice of Oji that
h.Us / Vm1 for each s 2 Dm . It also follows from (6) and the choice of Vm that
h.Us / Vm for all s 2 Dm . Therefore (3) is satisfied for all n m. Since (1) and
(2) are clear, we proved that for the families fUs W s 2 D<mC1 g and fVi W i < m C 1g
the properties (1)(4) are fulfilled for all n m.
Thus our inductive procedure gives us families fUs W s 2 D<! g .Y /

and fVn W n 2 !g .Z/ T with (1)(4). The sequence
T fBn W n 2 !g being
pseudocomplete, we have n2! Vn ;. If x 2 n2! Vn , then for any n 2 N
we have h1 .x/ \ Us ; for any s 2 Dn by (3). The property (2) shows that
jh1 .x/j 2n for any n 2 N and hence the set h1 .x/ is infinite. This contradiction
shows that U must have points of local homeomorphism and finishes the proof of
Fact 1. t
u
Returning to our solution observe that Cp .X / is pseudocomplete by Problem
475 of [TFS]. Therefore Fact 1 can be applied to the map ' to conclude that there
is a non-empty open U M such that 'jU W U ! '.U / is a homeomorphism.
Observe that the space M is realcompact (see Problems 406 and 402 of [TFS]); we
can choose V 2 .U / such that V is an F -set in M (see Fact 1 of T.252). The
space V is realcompact by Problem 408 of [TFS] and therefore '.V / is a non-empty
open realcompact subspace of Cp .X /. This implies that Cp .X / is realcompact (see
Problem 428 of [TFS]), so we can apply Problem 486 of [TFS] to conclude that X
is discrete.
T.499. Let M be a product of separable completely metrizable spaces. Assuming
that there is a finite-to-one open map ' W Cp .X / ! M , prove that X is discrete.
Solution. It follows from Problem 465 of [TFS] that every completely metrizable
space is pseudocomplete. Therefore any product of completely metrizable spaces is
pseudocomplete by Problem 470 of [TFS]. This shows that M is pseudocomplete
and hence there is a non-empty open U Cp .X / such that 'jU W U ! '.U /
is a homeomorphism (see Fact 1 of T.498). The set '.U / is pseudocomplete being
open in M (see Problem 466 of [TFS]), so U is a non-empty open pseudocomplete
subspace of Cp .X /. By homogeneity of Cp .X /, every f 2 Cp .X / has a
pseudocomplete open neighborhood, so Cp .X / is pseudocomplete by Fact 2 of
S.488.
Observe that the space M is realcompact (see Problems 406 and 402 of [TFS]);
we can choose W 2 .'.U // such that W is an F -set in M (see Fact 1 of T.252).
The space W is realcompact by Problem 408 of [TFS]; the set W 0 D ' 1 .W /\U
U is open in Cp .X / and '.W 0 / D W which shows that W 0 is homeomorphic to W .
Consequently, W 0 is a non-empty open realcompact subspace of Cp .X / which
implies that Cp .X / is realcompact (see Problem 428 of [TFS]), so we can apply
Problem 486 of [TFS] to conclude that X is discrete.
T.500. Let H be a G -subspace of R for some . Prove that if Cp .X / is
homeomorphic to a retract of H , then X is discrete. In particular, if Cp .X / is
homeomorphic to a retract of R for some cardinal , then X is discrete.
Solution. Fix a retraction r W H ! C such that C is homeomorphic to Cp .X /;
from now on we will identify C with the space Cp .X /. For any set A X the map
A W C ! Cp .A/ is the restriction; given any E C denote the space A .E/ by
0
EA . Furthermore, if A A0 X , then AA W C A0 ! C A is also the restriction
map. Call a set E C factorizable in C if for any continuous map p W E ! M
of the space E onto a second countable space M there is a countable A X and a

continuous map u W EA ! M such that u .A jE/ D p.
Given a set B the map pB W R ! RB is the natural projection of R
0
onto its face RB ; if G R , then GB D pB .G/. If B B 0 , then pBB W
0 0
RB ! RB is the natural projection of RB onto its face RB . We will say that a set
G R is factorizable in R if for any continuous map p W G ! M of the space

G onto a second countable space M , there is a countable B and continuous

map v W GB ! M for which v .pB jG/ D p. If B and f 2 RB , then
let G.B; f / D pB1 .f /; a set G R is called a standard G -subset of R if
G D G.B; f / for some countable B and f 2 RB . We let m D f1; : : : ; mg
for any m 2 N.
Fact 1. Suppose that Z is a space and A Z is countable. Let A W Cp .Z/ !
Cp .A/ be the restriction map and assume that A .Cp .Z; I// has a dense Cech-
complete subspace. Then A is a discrete subspace of Z.
Proof. Let P A .Cp .Z; I// be a dense Cech-complete subspace of the space
C 0 D A .Cp .Z; I//. It is easy to see that C 0 is dense in Cp .A; I/ and hence P is a
dense Cech-complete subspace of Cp .A; I/. Therefore A is discrete by Problem 287
of [TFS] and Fact 1 is proved. t
u
S
Fact 2. Given a space Zt for all t 2 T assume that QT D fTn W n 2 !g and
Tn TnC1 for all n 2 !. For any S T let ZS D fZt W t 2 S g; if n 2 ! then
n W ZTnC1 ! ZTn is the natural projection. Assume additionally that Y is a space
and fn W Y ! ZTn is a continuous map such that n fnC1 D fn forSall n 2 !.
Given any y 2 Y we have fn .y/ fnC1 .y/ for all n 2 !, so f .y/ D n2! fn .y/
is well-defined and f .y/ 2 ZT . Then the map f W Y ! ZT is continuous.
Proof. For any t 2 T and n 2 ! let pt W ZT ! Zt and ptn W ZTn ! Zt be the
respective natural projections. It suffices to show that pt f is continuous for all
t 2 T (see Problem 102 of [TFS]), so take an arbitrary t 2 T . There is n 2 ! such
that t 2 Tn ; it is immediate that pt f D ptn fn ; it follows from continuity of fn
that ptn fn and hence pt f is a continuous map, so Fact 2 is proved. t
u
Fact 3. Given a space Z assume that An Z and An AnC1 for every number
n 2 !. Let n W Cp .AnC1 jZ/ ! Cp .An jZ/ be the restriction map for all n 2 !.
S map fn W Y ! Cp .An jZ/
Suppose also that Y is a space and there is a continuous
such that n fnC1 D fn for all n 2 !. Let A D n2! An ;Sfor any y 2 Y we
have fn .y/ fnC1 .y/ for all n 2 !, so the function f .y/ D n2! fn .y/ is well-
defined. Then the map f W Y ! RA is continuous; in particular, if f .y/ 2 Cp .AjZ/
for all y 2 Y , then the map f W Y ! Cp .AjZ/ is continuous.
Proof. This is an immediate consequence of Fact 2 if we note that fn W Y ! RAn for
any n 2 ! and every n is the restriction to Cp .AnC1 jZ/ of the natural projection
of RAnC1 onto its face RAn . Fact 3 is proved. t
u
Fact 4. Let E be a non-empty subset of R . For each f 2 E, take a countable set

0
Bf and let Gf D G.Bf ; f jBS
S f / Then there is a countable E E such that
0
fGf W f 2 E g is dense in G D fGf W f 2 Eg.
S
Proof. For any D E let GhDi D fGf W f 2 Dg. Take an arbitrary f0 2 E and
let T0 D ff0 g; B0 D Bf0 . Assume that we have countable sets T0

Tn E
and B0

Bn such that
. / pBi .GhTi C1 i/ is dense in pBi .G/ for any i < n.
The space pBn .G/ is second countable, so there is a countable set TnC1 E
such that Tn TnC1 and the set pBn .GhTnC1 i/ is dense in the space pBn .G/. If we
let BnC1 D Bn [ fBf W f 2 TnC1 g, then . / will hold for all i n, so our inductive
procedure gives us sequences fTi W i 2 !g and fBi W i 2 !g such that TS i E and
Bi are countable and . / holds for all i 2 !. We claim that E 0 D n2! Tn is
as promised.
Indeed, take any g 2 G, and O 2 .g; R /; there is a finite set L and
O 2 .g./; R/ S for every 2 L such that O 0 D ff 2 R W f ./ 2 O for all 2
Lg O. If B D n2! Bn , then there is m 2 ! such that L0 D B \ L D Bm \ L.
Since pBm .GhTmC1 i/ is dense in pBm .G/ by . /, there is f 2 TmC1 for which
there is v 2 Gf with v./ 2 O for all 2 L0 . Observe that if L0 D .LnL0 /, then
L0 \B D ; and define vO 2 R by vO ./ D v./ for all 2 nL0 and vO ./ D g./ for
every 2 L0 . Then pB .Ov/ D pB .v/ which implies pBf .Ov/ D pBf .v/ and therefore
2 Gf \ O 0 Gf \ O, we proved that every neighborhood of g
vO 2 Gf . Since vO S
intersects the set fGf W f 2 E 0 g, so Fact 4 is proved. t
u
Fact 5. If G is a non-empty G -subset of R and F is a retract of G, then both G
and F are factorizable in R .
Proof. Fix a retraction s W G ! F ; it follows from Fact 1 of S.426 that for any
f 2 G there is a countable set Bf such that f 2 Gf D G.B Sf ; f jBf / G.
Fact 4 guarantees that there is a countable G 0 G such that Q D fGf W f 2 G 0 g
is dense in G.S
Let N D fBf W f 2 G 0 g; then N is a countable subset of ; besides, we have
Q D pN1 .pN .Q//. Since Q is dense in G, the set pN .Q/ is dense in pN .G/, so
if R D pN .Q/, then G pN1 .R/. The map pN is open, so the set G is dense
in pN1 .R/ D R RnN . It turns out that G is a dense subspace of a product of
second countable spaces, so G is factorizable in the relevant product by Problem
299 of [TFS]. Since all factors of R RnN except countably many coincide with
the factors of R , the set G is also factorizable in R .
To see that F is factorizable in R take a continuous onto map ' W F ! M
such that w.M / D !. Since G is factorizable, there is a countable S and a
continuous map v W GS ! M for which v .pS jG/ D ' s. We have FS GS ; if
v0 D vjFS , then v0 .pS jF / D ', so F is factorizable and Fact 5 is proved. t
u
Fact 6. Suppose that G is a non-empty G -subset of R . Assume that a set F C
is factorizable in C and there is a retraction W G ! F . Then, for any countable
A0 X and B0 , there exist countable sets A X and B such that
A0 A; B0 B and there exists a homeomorphism s W FB ! F A and a

retraction W GB ! FB such that s .pB jF / D A jF and .pB jF / D .pB jG/,
i.e., the diagram
. /

G ! F
pB # pB # & A
s
GB ! FB ! F A
is commutative.
Proof. To save space, we will use the relevant projections without mentioning that,
in fact, their restrictions to some sets are considered. Those sets will be always
clear from the context (we start using this convention in the diagram . /). In the
expression continuous map we will often omit continuous, so the word map
actually says continuous map.
The set F is factorizable in R by Fact 5, so there exists a countable B1
such that B0 B1 and there is a continuous map s1 W FB1 ! F A0 such that
s1 pB1 D A0 . By factorizability of G, there is a countable D1 B1 and a
continuous map 1 W GD1 ! FB1 for which 1 pD1 D pB1 . The set F is
factorizable in C , so there is a countable A1 A0 and a map q1 W F A1 ! FB1 for
which q1 A1 D pB1 .
To proceed inductively, assume that m 2 N and we have constructed families
fAi W i 2 m g; fBi W i 2 m g and fDi W i 2 m g of countable sets with the
(2) A0 A1

Am X ;
(3) B0 B1 D1 B2

Dm1 Bm Dm ;
(4) for each i 2 m , there is a map si W FBi ! F Ai 1 such that si pBi D Ai 1 ;
(5) for each i 2 m , there is a map i W GDi ! FBi such that i pDi D pBi ;
(6) for each i 2 m , there is a map qi W F Ai ! FBi such that qi Ai D pBi .
The set F is factorizable in R by Fact 5, so there exists a countable BmC1
such that Dm BmC1 and there is a map smC1 W FBmC1 ! F Am such that
smC1 pBmC1 D Am . By factorizability of the set G, there is a countable set DmC1
BmC1 and a map mC1 W GDmC1 ! FBmC1 with mC1 pDmC1 D pBmC1 . The
set F is factorizable in C , so there is a countable AmC1 Am and a map qmC1 W
F AmC1 ! FBmC1 for which qmC1 AmC1 D pBmC1 . It is evident that the families
fAi W i 2 mC1 g; fBi W i 2 mC1 g and fDi W i 2 mC1 g have the properties
(2)(6), so our inductive procedure gives us families fAiS W i 2 Ng; fBi W i 2SNg and
fDi W i 2 Ng for which (2)(6) hold. We claim that A D i 2! Ai and B D i 2! Bi
are as promised.
For any i 2 N let i D i .pD B
i
jGB /. Then i W GB ! FBi ; if f 2 GB , then
f D pB .g/ for some g 2 G and we have
(7)
B B
pBii C1 .i C1 .f // DpBii C1 .i C1 .pD
B
i C1
.pB .g////
B B
DpBii C1 .i C1 .pDi C1 .g/// D pBii C1 .pBi C1 ..g/// D pBi ..g//
Di .pDi .g// D i .pD
B
i
.pB .g/// D i .pD
B
i
.f // D i .f /
B S
which shows that pBii C1 i C1 .f / D i .f / for any i 2 N, so .f / D fi .f / W
i 2 Ng is well-defined. Besides, it follows from (7) that i .f / D pBi ..g// for
every i 2 N and hence
S
(8) .f / D fpBi ..g// W i 2 Ng D pB ..g// 2 FB .
The elements f 2 GB and g 2 G with pB .g/ D f were chosen arbitrarily,
B
so we proved that pBii C1 i C1 D i for all i 2 N, and hence W GB ! FB is
a continuous map (see Fact 2); besides, it follows from (8) that pB D pB .
Now, if f 2 FB , then f D pB .g/ for some g 2 F , so we can apply (8) to see that
.f / D pB ..g// D pB .g/ D f , i.e., W GB ! FB is a retraction.
For any i 2 N, let i D si .pBBi jFB /. Then i W FB ! F Ai 1 ; if f 2 FB ,
then f D pB .g/ for some g 2 F and we have
(9)
AAii1 .i C1 .f // DAAii1 .si C1 .pBBi C1 .pB .g////
DAAii1 .si C1 .pBi C1 .g/// D AAii1 .Ai .g// D Ai 1 .g/

Dsi .pBi .g// D si .pBBi .pB .g/// D si .pBBi .f // D i .f /
S
which shows that AAii1 i C1 .f / D i .f / for any i 2 N, so s.f / D fi .f / W
i 2 Ng is well-defined. Besides, it follows from (9) that i .f / D Ai 1 .g/ for every
i 2 N and hence
S
(10) s.f / D fAi 1 .g/ W i 2 Ng D A .g/ 2 F A .
The elements f 2 FB and g 2 F with pB .g/ D f were chosen arbitrarily, so
we proved that AAii1 i C1 D i for all i 2 N and hence s W FB ! F A is a
continuous map (see Fact 3) with s pB D A . This shows that the diagram . / is
commutative.
To finally prove that s is a homeomorphism, for any i 2 N, consider the map
i D qi .AAi jF A /. Then i W F A ! FBi ; if f 2 F A , then f D A .g/ for
some g 2 F and we have
(11)
B B
pBii C1 .i C1 .f // D Bii C1 .qi C1 .AAi C1 .A .g////
B B
D pBii C1 .qi C1 .Ai C1 .g/// D pBii C1 .pBi C1 .g// D pBi .g/
D qi .Ai .g// D qi .AAi .A .g/// D qi .AAi .f // D i .f /
B S
which shows that pBii C1 i C1 .f / D i .f / for any i 2 N, so .f / D fi .f / W
i 2 Ng is well-defined. Besides, it follows from (11) that i .f / D pBi .g/ for every
i 2 N and hence
S
(12) .A .g// D .f / D fpBi .g/ W i 2 Ng D pB .g/ 2 FB .
The elements f 2 F A and g 2 F with A .g/ D f were chosen arbitrarily,
B
so we proved that pBii C1 i C1 D i for all i 2 N and hence W F A ! FB is a
continuous map (see Fact 2). Now, if f 2 F A and A .g/ D f , then s..f // D
s.pB .g// D A .g/ D f (we used (12) and commutativity of the diagram . /);
finally, if f 2 FB and pB .g/ D f , then .s.f // D .A .g// D pB .g/ D f [we
used (10) and (12)] which shows that s is a homeomorphism and is the inverse
of s. Fact 6 is proved. t
u
S
Returning to our solution let u.x/ D 0 for all x 2 X ; we have H D n2! Hn
where Hn is a G -subset of R for all n 2 !. There is m 2 ! for which u 2 Hm ;
we claim that there exist countable sets A X and B such that
(13) there is a function h 2 RB for which Q D G.B; h/ Hm and we have
r.Q/ D Q \ C D ff 2 C W f jA D ujAg.
Let W .L/ D ff 2 C W f jL D ujLg for any L X . Since u 2 Hm and Hm \ C
is a G -subset of C , there is a countable A0 X such that W .A0 / Hm (see Fact 1
of S.426). Observe that Z0 D r 1 .W .A0 // \ Hm is a G -subset of Hm and hence of
R , so there is B0 and h0 2 RB0 for which u 2 G.B0 ; h0 / Z0 . Assume that
k 2 ! and we have families fAi W i kg and fBi W i kg of countable sets with
(14) A0

Ak X and B0

Bk ;
(15) for each i k, there is hi 2 RBi such that hi hi C1 for all i < k and
u 2 G.Bi ; hi / r 1 .W .Ai // \ Hm for all i k;
(16) W .Ai C1 / G.Bi ; hi / for all i < k.
Since E D G.Bk ; hk / \ C is a G -subset of the space C with u 2 E, there
is a countable set AkC1 Ak such that W .AkC1 / G.Bk ; hk /. It is easy to see
that r 1 .W .AkC1 // \ G.Bk ; hk // is a G -subset of G.Bk ; hk / and hence of R , so
there is a countable set BkC1 Bk and hkC1 2 RBkC1 such that hk hkC1 and
u 2 G.BkC1 ; hkC1 / r 1 .W .AkC1 // \ G.Bk ; hk /. It is immediate that (14)(16)
are still fulfilled for fAi W i k C 1g and fBi W i k C 1g, so we can construct
inductively families fAi W i 2 !g and fBi W i 2 !g such that (14)(16) hold for
S S S
all i 2 !. To see that A D i 2! Ai ; B D i 2! Bi and h D i 2! hi are as
promised,Tobserve that Q D G.B; h/ Hm by (15). It follows from (16) that
W .A/ i 2! G.Bi ; hi / D Q and hence W .A/ Q \ C r.Q/. On the other
hand, r.Q/ r.G.Bi ; hi // W .Ai / for all i 2 ! by (15), so r.Q/ W .A/ D
T
i 2! W .Ai /. This shows that r.Q/ D Q\C D W .A/ and therefore (13) is proved.
Let Y D A; then Y is a closed separable subspace of X . We claim that
(17) for any countable non-empty L X nY , we have L .W .A// D RL .
It follows from the property (13) that r1 D rjQ is a retraction of the space Q
onto W .A/, so we can apply Fact 6 to find countable sets D and N X such
that L N; B D, and there is a retraction W pD .Q/ ! pD .W .A// and a
homeomorphism s W pD .W .A// ! N .W .A//. Now, pD .Q/ D fhg RDnB is a
Cech-complete space and hence so is pD .W .A// being a closed subspace of pD .Q/.
Thus N .W .A// is Cech-complete. All functions from W .A/ are identically zero on
Y , so it is easy to see that NNnY W N .W .A// ! N nY .W .A// is a homeomorphism.
Therefore Z D N nY .W .A// is Cech-complete and dense in Cp .N nY jX / and
hence in RN nY . Since Z is an algebra in RN nY , if p 2 RN nY nZ, then p C Z is
a dense Cech-complete subspace of RN nY which does not intersect another dense
Cech-complete subspace Z of RN nY . This contradiction with Problem 264 of [TFS]
shows that N nY .W .A// D RN nY ; since L N nY , we have L .W .A// D
LN nY .Z/ D LN nY .RN nY / D RL and hence (17) is proved.
The space W .A/ is realcompact being closed in R ; since X nY jW .A/ is a
homeomorphism and W D X nY .W .A// is dense in a Moscow space RX nY , it
must be !-placed in RX nY by Problem 425 of [TFS]. Thus, if g 2 RX nY nW , then
there is a G -subset R of the space RX nY such that g 2 R RX nY nW . This implies
that there is a countable L X nY for which L .g/ L .W / D RL [see (17)]
which is a contradiction. As a consequence, Cp .X nY / X nY .W .A// D RX nY
which shows that Cp .X nY / D RX nY , i.e., X nY is discrete. Another consequence
of X nY .W .A// D RX nY is that the function f which is equal to 1 on X nY and
zero on Y is continuous on X , i.e., Y nX is clopen in X . This shows that
(18) X D Y .X nY / and X nY is discrete.
Consequently, the space Cp .Y / is a retract of Cp .X / and hence of H . This
shows that we are in the same situation with Y as we were with X , so we can
apply Facts 16 as if we were assuming that Y D X .
Let W H ! C0 D Cp .Y / be a retraction. The set Hn \ C0 is a G -subspace
of C0 , so HQ n D 1S.Hn \ C0 / \ Hn is a G -subset of Hn and hence of R for all
n 2 !. Then HQ D n2! HQ n is G in R and C0 is still a retract of HQ . This shows
that we can assume, without loss of generality, that Hn0 D .Hn / D Hn \ C0 for
every n 2S!.
Since n2! Hn0 D Cp .Y /, there exist m 2 !; " > 0 and h0 2 Cp .Y / such that
.Hm0 C h0 / \ Cp .Y; . "; "// is dense in Cp .Y; . "; "// and hence in Cp .Y; "; " /
(see Problem 431). It is easy to see that there exists a homeomorphism ' W C0 ! C0
such that '.Cp .Y; "; " // D Cp .Y; I/; since Hm0 Ch0 is homeomorphic to a retract
of Hm , we can conclude that
(19) there is a set H 0 C0 homeomorphic to a retract of Hm , such that the set

C 0 D H 0 \ Cp .Y; I/ is dense in C1 D Cp .Y; I/,
so we will consider that .Hm / D H 0 . The space Y being separable, we have
the inequality i w.H 0 / i w.Cp .Y // D !. For any f 2 C 0 , there is a countable
Bf such that Gf D G.Bf ; f jBf / Hm (see Fact 1 of S.426) S
It followsSfrom Fact 4 that for some countable C 00 C 0 the set fGf W f 2 C 00 g
is dense in fGf W f 2 C 0 g. It is clear that Gf ' R , so Gf is stable for any
S
f 2 C 00 by Problem 268. Therefore QQ D fGf W f 2 C 00 g is also stable by
Problem 124. The set RQ D .Q/ Q is stable; since C 00 R,Q the set RQ \ C 0 is dense in
0 Q 0
C which together with i w.R/ i w.H / D ! implies that nw.R/ Q D ! and hence
0
C is separable.
Let F0 C 0 be a countableTdense subset of C 0 (it is clear that F0 is also dense
in Cp .Y; I/). We have Hm D n2! On where the set On is open in the space R
for all n 2 !. For each f 2 F0 and n 2 !, there exists a finite set B.n; f /
and W .n; f / 2 .RB.n;f / / such that f 2 WQ .n; f / D pB.n;f 1
/ .W .n; f // On .
S
If N0 D S fB.n; f / W f 2 F 0 ; n 2 !g, then N 0 is a countable subset of . Let
Wn D fWQ .n; f / W f 2 F0 g for every n 2 !; it is evident that pN10 .pN0 .Wn // D
T
Wn for all n 2 !. Thus K D fpN0 .Wn / W n 2 !g is a Cech-complete subspace
of RN0 such that K 0 D K RnN0 Hm and F 0 D K 0 \ C 0 is dense in C 0
because F0 K 0 \ C 0 . Consequently, F 0 is dense in C1 D Cp .Y; I/ because C 0 is
dense in C1 . Furthermore, K 0 is stable being a product of Lindelf -spaces (see
Problem 268); therefore nw..K 0 // i w..K 0 // i w.C0 / D ! which shows
that nw..K 0 // D !. The set F D Cp .Y; I/ \ .K 0 / is closed in .K 0 /, so it is a
G -set in .K 0 / (it is an easy exercise that every closed subspace of a space with a
countable network is G ), so G D 1 .F / \ K 0 is a G -set in K 0 and hence in R .
We have .G/ D F D G \ C1 ; besides, F K 0 \ C1 K 0 \ C 0 F0 and hence
F is dense in C1 . This proves that
(20) there exists a G -set G in the space R for which we have G Hm and
F D .G/ D G \ C1 while F0 F and hence F is dense in C1 .
T subset of the set G in the same way we found the set

We still have to find a nice
K 0 Hm . We have G D n2! En where the set En is open in the space R for
all n 2 !. For each f 2 F0 and n 2 ! there exists a finite set D.n; f / and
Q.n; f / 2 .RD.n;f / / such that f 2 Q.n; Q 1
f / D pD.n;f
S / .Q.n; f // En . Then
N D fD.n; S f / W f 2 F0 ; n 2 !g is a countable subset of . For every n 2 N
let QnTD fQ.n; Q f / W f 2 F0 g; it is evident that pN1 .pN .Qn // D Qn . Thus
R D fpN .Qn / W n 2 !g is a Cech-complete subspace of the space RN such that
K1 D R RnN G and K1 \ C1 is dense in C1 because F0 K1 \ C1 .
We are going to prove next that
(21) every countable L Y is discrete.
To do it observe that F is factorizable in R by Fact 5; since F is dense in
Cp .Y; I/, it is factorizable in C0 being dense in IY (see Problem 299 of [TFS]).
Therefore we can apply Fact 6 to the sets L; N to obtain countable sets L1 Y and
N1 such that L L1 ; N N1 and there exists a retraction 1 W GN1 ! FN1
and a homeomorphism W FN1 ! F L1 .
The space K2 D pN1 .K1 / D R RN1 nN is Cech-complete; since K1 \ F is
dense in F , the set K2 \ FN1 pN1 .K1 \ F / is dense in FN1 . The set FN1 is
closed in pN1 .G/, so K2 \ FN1 is closed in the Cech-complete space K2 . Therefore
K2 \ FN1 is a dense Cech-complete subspace of FN1 . The spaces FN1 and F L1
are homeomorphic, so F L1 has a dense Cech-complete subspace; now, F L1 is
dense in L1 .Cp .Y; I//, so L1 .Cp .Y; I// also has a dense Cech-complete subspace
and therefore we can apply Fact 1 to conclude that L1 is discrete. Thus L L1 is
also discrete and (21) is settled.
Finally observe that a separable space is discrete if and only if all its countable
subspaces are discrete, so Y is discrete by (21) and hence X D Y .X nY / is
discrete by (18) which shows that our solution is complete.
Chapter 3
Bonus Results: Some Hidden Statements
The reader has, evidently, noticed that an essential percentage of the problems of
the main text is formed by purely topological statements, some of which are quite
famous and difficult theorems. A common saying among Cp -theorists is that any
result on Cp -theory contains only 20 % of Cp -theory; the rest is general topology.
It is evident that the author could not foresee all topology which would be needed
for the development of Cp -theory, so a lot of material had to be dealt with in the
form of auxiliary assertions. After accumulating more than seven hundred such
assertions, the author decided that some deserve to be formulated together to give a
big picture of the additional material that can be found in solutions of problems.
This section presents 100 topological statements which were proved in the solu-
tions of problems without being formulated in the main text. In these formulations
the main principle is to make them clear for an average topologist. A student could
lack the knowledge of some concepts of the formulation so the index of this book
can be used to find the definitions of the necessary notions.
After every statement we indicate the exact place (in this book) where it
was proved. We did not include any facts from Cp -theory because more general
statements are proved sooner or later in the main text.
The author considers that most of the results that follow are very useful and have
many applications in topology. Some of them are folkloric statements and quite a
few are published theorems, sometimes famous ones. For example, Fact 1 of T.015
is a HajnalJuhsz theorem [see Theorem 4.9 in Hodel (1984)]; Fact 2 of T.092 is
a result proved in Tkachenko (1978). Fact 5 of T.298 is a partial (but still a very
nontrivial) case of Shapirovskys theorem [see Theorem 3.18 in Juhsz (1980)].
Besides, the result of Fact 9 of T.298 is published in Juhsz and Szentmiklssy
(1992).
To help the reader find a result he/she might need, we have classified the material
of this section according to the following topics: standard spaces, metrizable
spaces, compact spaces and their generalizations, properties of continuous maps,
completeness and convergence properties, product spaces, cardinal invariants and
528 3 Bonus Results: Some Hidden Statements
set theory. The last subsection is entitled Raznoie (which in Russian means
miscellaneous) and contains unclassified results. The author hopes that once we
understand in which subsection a result should be, then it will be easy to find it.
3.1 Standard Spaces
By standard spaces we mean the real line, its subspaces and its powers, Tychonoff
and Cantor cubes as well as ordinals together with the Alexandroff and StoneCech
compactifications of discrete spaces.
T.040. Fact 2. Under the Continuum Hypothesis there exists an HFD space
Z D fz W < !1 g D!1 .
T.040. Fact 4. Any HFD space Z D fz W < !1 g D!1 is hereditarily separable.
T.042. Fact 1. No non-empty open subset of ! is a union of !1 -many of nowhere
dense subsets of ! .
T.063. Fact 3. Let I be the space 0; 1 with the natural topology. No Luzin space
can be mapped continuously onto I . No additional axioms are needed for the proof
of this fact.
T.063. Fact 4. Let Z be an arbitrary space. If Z is not zero-dimensional, then it
maps continuously onto I .
T.069. Fact 2. Given any f W !1 ! !1 , the set Cf D f < !1 W f g is a
club. Here f D ff ./ W < g.
T.097. Fact 1. Let A be a countable subset of ! ! . Then there exists f 2 P such that
g < f for any g 2 A. Here P D ff 2 ! ! W i < j implies f .i / < f .j /g.
T.126. Fact 1. Let D be a discrete space of cardinality !1 . Then there exist countably
compact subspaces X and Y of the space D such that D D X \Y and hence X Y
contains an uncountable clopen discrete subspace.
T.131. Fact 2. Every infinite closed subspace of ! has cardinality 2c and hence
! has no nontrivial convergent sequences.
T.203. Fact 2. If is an uncountable cardinal and Z D A./, then Z 2 n.Z/ is not
normal.
T.322. Fact 4. In the family f! [ fg W 2 !n!g, there exists a subfamily of
2c -many pairwise non-homeomorphic spaces. Here ! [ fg is considered with the
subspace topology inherited from ! for all 2 !n!.
T.349. Fact 1. If P is a countable dense subspace of the real line R, then there
exists a homeomorphism f W Q ! P such that x; y 2 Q and x < y implies
f .x/ < f .y/. In particular, if P and Q are countable dense subspaces of R, then
P ' Q.
3.2 Metrizable Spaces 529
T.371. Fact 2. Let D be an infinite discrete space. Then

(a) for any A D the set A is clopen in D (the bar denotes the closure in the
space D);
(b) if z 2 DnD and Dz D fU \ D W U 2 .z; Z/g, then a set A D belongs to
Dz if and only if z 2 A;
(c) the family Dz is an ultrafilter on D for any z 2 DnD.
T.455. Fact 2. Suppose that a property P is closed-hereditary and R!1 is a countable
union of its subspaces with the property P. Then there S exists a sequence S D
fYn W n 2 !g of subspaces of R!1 such that R!1 D S, the set Yi covers all
countable faces of R!1 and has P for any i 2 !. The sequence S will be called
P-representation of the space R!1 .
T.455. Fact 3. There exists a closed discrete D R!1 such that jDj D !1 , while
A .D/ is countable for any countable A !1 . Here B W R!1 ! RB is the natural
projection for any B !1 .
3.2 Metrizable Spaces
The results of this section deal with metrics, pseudometrics or metrizable spaces in
some way. We almost always assume the Tychonoff separation axiom so our second
countable spaces are metrizable and hence present here too.
T.055. Fact 1. For any second countable space Z there exists a metric d on Z and a
base B D fUn W n 2 !g in the space Z such that .d / D .Z/ and diamd .Un / ! 0
when n ! 1.
T.131. Fact 4. Let M be a non-compact second countable space. Then there exists
a continuous onto map f W M ! Z such that .Z/ > !. Consequently, M is not
.!)-stable.
T.220. Fact 1. If M is an uncountable second countable space, then there is an
uncountable M 0 M such that every U 2 .M 0 / is uncountable.
T.220. Fact 2. If w.M / ! < jM j, then the space Z D L.!1 / M is not weakly
Whyburn.
T.285. Fact 1. Let Z be a metrizable space. If is not a countably cofinal cardinal
and w.Z/ , then there is a closed discrete D Z such that jDj D .
T.333. Fact 1. Suppose that M is a complete metric space. Given any space Z and
a continuous map f W A ! M for some dense A Z, there is a G -set G of
the space Z such that A G and there is a continuous map g W G ! M with
gjA D f .
T.333. Fact 2. Let M and N be complete metric spaces. If A M; B N and

there is a homeomorphism f W A ! B, then there exist A0 M; B 0 N such
that A A0 ; B B 0 , the set A0 is G in M , the set B 0 is G in N and there is a
homeomorphism h W A0 ! B 0 such that hjA D f .
T.348. Fact 1. Let .M; / be a complete metric space. Suppose that we have a family
U D fUs W s 2 D<! g .M / with the following properties:
(a) the family U.n/ D fUs W s 2 Dn g is disjoint for all n 2 !;
(b) if 0 k < n < !; s 2 Dk ; t 2 Dn and s t, then U t Us ;
(c) diam.U.n// ! 0 when n ! 1.
T S
Then the space L.U/ D f U.n/ W n 2 !g is homeomorphic to K.
T.368. Fact 3. Given any space Z suppose that .M; d / is a complete metric space
and f W A ! M is a continuous map for some dense A Z. For a point z 2 Z
say that osc.f; z/ D 0 if for any " > 0 there is W 2 .z; Z/ such that diam.f .W \
A// < ". If G D fz 2 Z W osc.f; z/ D 0g, then A G and there is a continuous
map g W G ! M with gjA D f .
3.3 Compact Spaces and Their Generalizations
This section contains some statements on compact, countably compact and pseudo-
compact spaces.
T.041. Fact 4. Let X be a compact space of countable tightness. Then there exists a
countable set A X and a non-empty G -set H X such that H A.
T.082. Fact 2. A space Z is called a CS -space if all closed subspaces of Z are
separable. If K is a compact CS -space, then K is hereditarily separable.
T.090. Fact 2. Any infinite compact space has a non-closed countable subspace.
T.203. Fact 1. A countably compact space Z is metrizable if and only if there exists
a point-countable T1 -separating family U .Z/.
T.211. Fact 1. Let Z be a compact space. If z is a non-isolated point of Z, then there
exists a regular cardinal and a -sequence S D fz W < g Znfzg such that
S ! z.
T.235. Fact 2. Every countably compact space Z with .Z/ D ! is compact and
hence metrizable.
T.298. Fact 4. Given an infinite cardinal the following conditions are equivalent
for any compact space X :
(a) the space X can be mapped continuously onto I ;
(b) there exists a continuous onto map f W F ! D for some closed F X ;
(c) there exists a -dyadic family of closed subsets of X .
3.4 Properties of Continuous Maps 531
T.298. Fact 5. If X is a compact space such that .x; X / > ! for each x 2 X ,
then there is a closed subset P X for which there is a continuous onto map
f W P ! D!1 .
T.298. Fact 7. Suppose that X is a compact space such that there is a continuous
onto map f W X ! D!1 . Then X contains a convergent !1 -sequence.
T.298. Fact 8. If is an infinite cardinal and X is a compact space with t.X / >
then there is a closed Y X which maps continuously onto C C 1.
T.298. Fact 9. If X is a compact space with t.X / > !, then X contains a convergent
!1 -sequence.
T.357. Fact 2. If K is a compact space, then w.Cu .K// D w.K/. In particular, if K
is metrizable, then Cu .K/ is a Polish space.
T.357. Fact 3. Every locally compact space condenses onto a compact space.
T.494. Fact 1. If h W Y ! Z is a closed irreducible map and Z is pseudocompact,
then Y is also pseudocompact.
3.4 Properties of Continuous Maps
We consider the most common classes of continuous maps: open, closed, perfect
and quotient. The respective results basically deal with preservation of topological
properties by direct and inverse images.
T.132. Fact 7. If Z is an infinite space, then there is f 2 C.Z/ such that f .Z/ is
infinite.
T.139. Fact 2. Assume that Z and T are Tychonoff spaces and f W Z ! T is a
continuous onto map. Then there exists a Tychonoff space T 0 such that for some
R-quotient continuous onto map g W Z ! T 0 and a condensation h W T 0 ! T , we
have f D h g.
T.201. Fact 1. A closed continuous image of a normal space is a normal space.
T.316. Fact 1. If Y is a sequential space, then a continuous onto map f W Z ! Y
is closed if and only if f .D/ is not a nontrivial convergent sequence for any closed
discrete D Z.
T.354. Fact 2. Given spaces Z and Y assume that F D fFt W t 2 T g is a closed
locally finite cover of Z and we have a family fft W t 2 T g of functions such that
ft W Ft ! Y is continuous for allSt 2 T and ft j.Ft \ Fs / D fs j.Ft \ Fs / for any
s; t 2 T . Then the function f D t 2T ft W Z ! Y is also continuous.
T.363. Fact 3. Let Y and Z be second countable spaces. If f W Y ! Z is a
measurable onto map and Y is analytic, then the space Z is also analytic.
T.363. Fact 4. If Z is a cosmic space, then there exist second countable spaces M
and N and condensations f W M ! Z and g W Z ! N such that the mappings
f 1 and g 1 are measurable.
T.245. Fact 1. Let Z be a paracompact p-space; if f W Z ! T is a perfect
map (recall that our definition of a perfect map implies that f is continuous and
onto), then T is also a paracompact p-space. In other words, a perfect image of a
paracompact p-space is a paracompact p-space.
T.245. Fact 2. Let Z be a normal space and assume that F is a non-empty closed
subset of Z; for any A Z, let A D .AnF / [ fF g. Given z 2 Z, let pF .z/ D z if
z 2 ZnF and pF .z/ D F if z 2 F . It is clear that pF W Z ! ZF D fF g [ .ZnF /.
Then
(i) the family F D fU 2 .Z/ W U ZnF g [ fU W U 2 .F; Z/g is a topology
on the set ZF ;
(ii) the space Z=F D .ZF ; F / is T1 and normal (and hence Tychonoff) and the
map pF W Z ! Z=F is continuous, closed and onto.
The operation of obtaining the space Z=F from a space Z is called collapsing the
set F to a point.
T.246. Fact 1. Let Y be a paracompact space. Suppose that Z is a space in which
any point is a limit of a nontrivial convergent sequence. Then any closed map h W
Y ! Z is irreducible on some closed subset of Y , i.e., there is a closed F Y
such that h.F / D Z and hF D hjF is irreducible.
T.266. Fact 1. Let Z be a space and suppose that fa W Z ! Ya isQa continuous map
for all a 2 A. For the map f D ffa W a 2 Ag W Z ! Y D fYa W a 2 Ag let
Z 0 D f .Z/. If the map fb is perfect for some b 2 A, then f W Z ! Z 0 is also
perfect.
T.268. Fact 1. Given spaces Y and Z, if q W Y ! Z is an R-quotient map, then, for
any space M , a map p W Z ! M is continuous if and only if p q is continuous.
T.489. Fact 1. Given spaces Z and T , if f W Z ! T is a perfect map then
w.T / w.Z/. In other words, perfect maps do not increase weight (recall that all
perfect maps in this book are surjective).
T.492. Fact 1. Any perfect preimage of a realcompact space is realcompact.
3.5 Completeness and Convergence Properties
This section deals mainly with Cech-complete spaces. Some results on convergence
properties are presented as well.
T.041. Fact 2. If X is a sequential space and jAj c for some A X , then jAj c.
T.041. Fact 3. Any sequential space has countable tightness.
3.6 Product Spaces 533
T.045. Fact 1. If Z is a FrchetUrysohn space without isolated points, then there

exists a closed separable dense-in-itself subspace Y Z.
T.046. Fact 1. Let X be a dense-in-itself space with the Baire property such that
c.X / ! and w.X / !1 . Then, under CH, there is a dense Luzin subspace in the
space X .
T.203. Fact 3. If Z is a locally compact paracompact space, then there isSa disjoint
family U of clopen subsets of Z such that each U 2 U is -compact and U D Z.
T.210. Fact 1. Every Cech-complete space is a k-space.
T.223. Fact 3. Every locally compact space is p-space.
T.272. Fact 1. For any non-scattered Cech-complete space Z, there exists a dense-
in-itself compact K Z. In other words, if Z is Cech-complete and every compact
subspace of Z is scattered, then Z is itself scattered.
T.298. Fact 1. If a space X has a small diagonal, then no convergent !1 -sequence
is embeddable in X .
T.371. Fact 3. Let Z be a homogeneous space, i.e., for every x; y 2 Z, there is a
homeomorphism h W Z ! Z such that h.x/ D y. If Z is of second category in
itself, then it is a Baire space.
T.385. Fact 4. Given a Polish space M , suppose that E Cp .M / is a countable
set and K D E is compact (the bar denotes the closure in RM ). If the set K is not
contained in B1 .M /, then ! embeds in K.
T.385. Fact 5. Suppose that M is a Polish space and E Cp .M / is a countable set
such that K D E is compact (the bar denotes the closure in RM ) and K B.M /.
Then K B1 .M / and hence K is Rosenthal compact.
3.6 Product Spaces
The space Cp .X / being dense in RX , the results on topological products form

a fundamental part of Cp -theory. The main line here is to classify spaces which
could be embedded in (or expressed as a continuous image of) a nice subspace of a
product.
T.109. Fact 1. Suppose that XtQis a space such that nw.Xt / D ! for every t 2 T
and consider theQ space X D fXt W t 2 T g; given any A T , the map pA W
X ! XA D fXt W t 2 Ag is the natural projection of X onto the face XA
defined by pA .x/ D xjA for every x 2 X . Suppose that Y is a dense subspace of
the space X and f W Y ! M is a continuous map of Y onto a space M such that
w.M / D !. Then there is a set S T and a continuous map g W pS .Y / ! M
such that jS j and f D g .pS jY /.
Q
T.110. Fact 1. Let Xt be a space for each t 2 T ; given a point x 2 X D fXt W t 2
T g, let .X; x/ DQfy 2 X W jft 2 T W x.t/ y.t/gj < !g. The natural projection
pS W X ! XS D t 2S Xt is defined by pS .x/ D xjS for any x 2 X . Suppose that
Y X and .X; x/ Y for any x 2 Y . Then the map pS jY W Y ! pS .Y / is
open for any S T .
T.268. Fact 2. Assume
Q that we have a family fXt W t 2 T g of spaces such that the
product
Q X A D t 2A t is Lindelf for every finite A T . Then for any a 2 X D
X
t 2T X t , the space D .X; a/ is Lindelf.
Q
T.268. Fact 3. Given a space Xt for every t 2 T , let X D fXt W t 2 T g. Suppose
that Y is a Lindelf subspace of the space X and f W Y ! M is a continuous map
of Y onto a space M such that w.M / D !. Then there is a set S T and a
Q 4. Assume that we have a family fXt W t 2 T g of spaces

T.268. Fact Q such that the
product t 2A Xt is Lindelf for every finite A T . Let X D fXt W t 2 T g
and suppose that Y 2 f.X; a/; .X; a/g. Then, for any infinite cardinal and
any continuous map f W Y ! M of Y to a space M with w.M / , there is
a set S T and a continuous map g W pS .Y / ! M such that jS j and
f D g .pS jY /.
T.268. Fact 5. QAssume that we have a family fXt W t 2 T g of spaces
Qsuch that the
product XA D t 2A Xt is Lindelf for every finite A T . Let X D fXt W t 2 T g
and suppose that f W X ! M is a continuous map of X to a space M such that
w.M / . Then there is a set S T and a continuous map g W XS ! M such
that jS j and f D g pS .
T.298. Fact 6. Let Mt be a second Q
Q countable space for every t 2 T . If M D
t 2T M t , then let p S W M ! M S D t 2S Mt be the natural projection of M onto
its face MS (recall that pS is defined by pS .x/ D xjS for any x 2 M ). Then, for
any open U M , the set U depends on countably many coordinates, i.e., there
exists a countable S T such that pS1 .pS .U // D U .
Q
T.415. Fact 1. Suppose that we have a product Y D fYt W t 2 T g and a set
H Y which covers all finite faces of Y . Then the map p D pA jH W H ! YA is
open for any finite A T .
S
T.415. Fact 2. Assume that Z ! D fZn W n 2 !g. Then, for some n 2 !, there is
H Zn homeomorphic to some G Z ! which covers all finite faces of Z ! .
Q
T.455. Fact 1. If w.Mt / ! for any t 2 T and E M D t 2T Mt covers all
countable faces of M , then E is C -embedded in M .
3.7 Cardinal Invariants and Set Theory 535
3.7 Cardinal Invariants and Set Theory
To classify function spaces using cardinal invariants often gives crucial information.
This section includes both basic, simple results on the topic as well as very difficult
classical theorems.
T.015. Fact 1. For any space Z, we have jZj 2 .Z/s.Z/
.
T.036. Fact 1. Suppose that Y and Z are spaces such that s.Y Z/ for some
infinite cardinal . Then either hd.Y / or hl.Z/ .
T.039. Fact 1. Let Z be a space such that c.Z/ ! and w.Z/ !1 . If CH holds,
then there exists a family N of nowhere dense closed subspaces of Z such that
jN j !1 , and, for any nowhere dense F Z, there is N 2 N such that F N .
We will say that the family N is cofinal in the family of all nowhere dense subsets
of Z.
T.040. Fact 1. Assume that U D fUn W n 2 !g is a family of infinite sets. Then there
exists a disjoint family V D fVn W n 2 !g such that Vn is infinite and Vn Un for
all n 2 !. We will say that the family V is a (disjoint) -net for U.
T.050. Fact 1. Let Z be a space with c.Z/ D !. If MAC:CH holds, then !1 is a
precaliber of the space Z.
T.050. Fact 2. Given an infinite cardinal
Q , suppose that c.Zs / for all
2 S . Assume additionally that c. s2A Zs / for each finite A S . Then
s Q
c. s2S Zs / .
T.082. Fact 1. A space Z is called a CS -space if all closed subspaces of Z are
separable. If Z is a CS -space, then s.Z/ D !.
T.092. Fact 2. Let Z be any space. Assume that w.Y / ! for every Y Z with
jY j !1 . Then w.Z/ D !.
T.102. Fact 1. Let Z be an infinite space with w.Z/ D . Then, for any base B for
the space Z, there is B 0 B such that jB 0 j and B 0 is a base in Z. In other
words, any base of a space contains a base of minimal cardinality.
T.187. Fact 1. If Z is a space and Y is dense in Z, then w.Y / D w.Z/.
T.235. Fact 1. A space Z has a G -diagonal T if and only if there is a sequence
fDn W n 2 !g of open covers of Z such that fSt.z; Dn / W n 2 !g D fzg for any
z 2 Z. The sequence fDn W n 2 !g is called a G -diagonal sequence for Z.
T.412. Fact 2. Suppose that T is a space and U is an open cover of T . If jUj
and w.U / for every U 2 U, then w.T / .
3.8 Raznoie (Unclassified Results)
Last, but not least, we place here some interesting results which do not fit in any of
the previous subsections.
T.098. Fact 2. Let Z be any space (no axioms of separation are assumed). If, for
any z 2 Z, there exists U 2 .z; Z/ such that U is compact and Hausdorff, then Z
is a Tychonoff space.
T.139. Fact 1. The least upper bound of any family of completely regular (not
necessarily Tychonoff) topologies on a set Z is a completely regular topology on Z.
SFact 1. Let Z be any space; if A is a family of resolvable subspaces of Z,

T.219.
then A is also resolvable.
T.219. Fact 2. If Z is an irresolvable space, then there is a non-empty open
hereditarily irresolvable U Z.
T.219. Fact 3. A space Z is ultradisconnected if and only if for any crowded
set A Z, if ZnA is also crowded, then A is clopen. In particular, every
ultradisconnected T1 -space is irresolvable.
T.219. Fact 5. The conditions below are equivalent for any crowded T1 -space Z:
(1) Z is maximal;
(2) Z is perfectly disconnected;
(3) every crowded subspace of Z is open;
(4) Z is ultradisconnected and nodec;
(5) Z is ultradisconnected and every discrete subspace of Z is closed;
(6) Z is submaximal and extremally disconnected;
(7) Z is extremally disconnected, hereditarily irresolvable and nodec.
T.219. Fact 6. For any Tychonoff crowded space Z, there exists a maximal Tychonoff
topology on the set Z such that .Z/ .
T.219. Fact 7. A space Z is maximal Tychonoff if and only if Z is Tychonoff and
ultradisconnected.
T.219. Fact 8. Suppose that Z is a countable Tychonoff hereditarily irresolvable
space. Then the set T D fz 2 Z W z 2 DnD for some discrete subspace D Zg is
nowhere dense in Z.
T.219. Fact 9. There exists a countable Tychonoff maximal space.
T.219. Fact 10. If Z is a Tychonoff submaximal countable space, then Z A.!/ is
not weakly Whyburn.
T.222. Fact 1. If Z is a p-space, then it is of pointwise countable type, i.e., for every
z 2 Z, there is a compact K Z such that z 2 K and .K; Z/ !.
3.8 Raznoie (Unclassified Results) 537
T.300. Fact 1. Suppose that Z is a space in which there exists a countable closed
network F with respect to a compact cover C of the space Z such that every C 2 C
is metrizable. If, additionally, the space Z has a small diagonal, then nw.Z/ D !.
T.412. Fact 1. Suppose that T is a space such that T D T0 [ : : : [ Tn and Ti has a
-disjoint base for every i n. Then F D T 0 \ : : : \ T n has a -disjoint base.
Chapter 4
Open Problems
The unsolved problems form an incentive for the development of any area of
mathematics. Since this book has an ambitious purpose to embrace all or almost
all modern Cp -theory, it was impossible to avoid dealing with open questions.
In this book, we have a wide selection of unsolved problems of Cp -theory.
Of course, unsolved means unsolved to the best of the knowledge of the author.
I give a classification by topics, but there is no mention whatsoever of whether the
given problem is difficult or not. One good parameter is the year of publication, but
sometimes the problem is not solved for many years because of lack of interest or
effort and not because it is too difficult.
I believe that almost all unsolved problems of importance in Cp -theory are
present in this chapter. The reader understands, of course, that there is a big
difference between the textbook material of the first four chapters and open
questions to which an author must be assigned. I decided that it was my obligation
to make this assignment and did my best to be frowned at (or hated!) by the least
possible number of potential authors of open problems.
This volume contains 100 unsolved problems which are classified by topics
presented in eight sections the names of which outline what the given group of
problems is about. At the beginning of each subsection we define the notions which
are not defined in the main text. Each published problem has a reference to the
respective paper or book. If it is unpublished, then my opinion on who is the author is
expressed. The last part of each problem is a very brief explanation of its motivation
and/or comments referring to the problems of the main text or some papers for
additional information. If the paper is published and the background material is
presented in the main text, we mention the respective exercises. If the main text
contains no background, we refer the reader to the original paper. If no paper is
mentioned in the motivation part, then the reader must consult the paper/book in
which the unsolved problem was published.
540 4 Open Problems
To do my best to assign the right author to every problem, I implemented the

following simple principles:
1. If the unsolved problem is published, then I cite the publication and consider
myself not to be involved in the decision about who is the author. Some problems
are published many times and I have generally preferred to cite the articles in
journals/books which are more available for the Western reader. Thus it may
happen that I do not cite the earliest paper where the problem was formulated.
Of course, I mention it explicitly, if the author of the publication attributes the
problem to someone else.
2. If, to the best of my knowledge, the problem is unpublished, then I mention the
author according to my personal records. The information I have is based upon
my personal acquaintance and communication with practically all specialists in
Cp -theory. I am aware that it is a weak point and it might happen that the problem
I attributed to someone was published (or invented) by another person. However,
I did an extensive work ploughing through the literature to make sure that this
does not happen.
4.1 Analyticity and Similar Properties
Quite a few methods here come from descriptive set theory and functional analysis.
Some problems are likely to be very difficult because they stem from the old
question whether every cosmic space embeds in an analytic one.
4.1.1. Suppose that Cp .X / embeds in an analytic space. Must the space X be
-compact?
Published in Arhangelskii and Calbrix (1999)
Related to Problems 2.3652.368
4.1.2. Let P be the space of irrationals. Is it true that Cp .P/ embeds in an analytic
space?
4.1.3. Is it true that Cp .Cp .P// embeds in an analytic space?

4.1.4. Given a space X suppose that there exists a K-analytic space Y such that
Cp .X / Y RX . Must X be -bounded? Is the same conclusion true if Y
is a K -space?
4.2 Whyburn Property in Function Spaces 541
4.1.5. Let X be a Lindelf space such that Cp .X / is K-analytic. Must X be

-compact?
4.1.6. Let X be a Lindelf space for which there exists a K-analytic space Y such
that Cp .X / Y RX . Must X be -compact?
4.1.7. Suppose that Cp .X / is a K -space. Must X be -compact?

Published in Arhangelskii (1988a)
4.2 Whyburn Property in Function Spaces
The nice algebraic structure of Cp .X / implies very strong dependencies between

local and convergence properties of Cp .X /. The most famous result is the coinci-
dence of the k-property and FrchetUrysohn property in the spaces Cp .X /. Many
problems below reflect the intention to generalize this result.
4.2.1. Is it true that the Whyburn property of Cp .X / implies countable tightness of
Cp .X /?
Published in Bella and Yaschenko (1999)
4.2.2. Suppose that Cp .Cp .X // is a Whyburn space. Must X be finite?

Published in Tkachuk and Yaschenko (2001)
Motivated by the fact that this is true when X is cosmic (Problem 2.218)
!
4.2.3. Is Cp .R / a weakly Whyburn space?
Motivated by the fact that Cp .R! / is not a Whyburn space (Problem 2.217)
4.2.4. Call a space Z discretely generated if, for any A Z and any z 2 A, there is
a discrete D A such that z 2 D. Suppose that Cp .X / is a Whyburn space.
Must Cp .X / be discretely generated?
Motivated by the fact that this is true if Cp .X/ has countable tightness
542 4 Open Problems
4.2.5. Suppose that Cp .X / is a Whyburn space for some second countable space X .
Is it true that all finite powers of X are Hurewicz spaces?
Motivated by the fact that X has to be a Hurewicz space (Problem 2.217)
4.2.6. Suppose that Cp .X / is a Whyburn space. Is it true that .Cp .X //! is also a
Whyburn space?
Motivated by the fact that the FrchetUrysohn property in Cp .X/ is countably

multiplicative (Problem 1.145)
4.3 Uniformly Dense Subspaces
It is standard in general topology to try to prove some nice properties of a space Z

if it has a nice dense subspace. The same approach is valid for the spaces Cp .X /.
For example, Cp .X / is second countable if it has a dense metrizable subspace. The
most interesting line of research in this direction is to find properties P such that
every Cp .X / has a dense subspace with the property P.
4.3.1. Suppose that Cp .X / has a uniformly dense Lindelf subspace. Is it true in
ZFC that Cp .X / is Lindelf? Recall that A Cp .X / is uniformly dense
in Cp .X / if, for any f 2 Cp .X /, there is a sequence ffn gn2! A which
converges uniformly to f .
Published in Tkachuk (2003)
Motivated by Problems 2.458 and 2.459 (this is true if the uniformly dense subspace
in question is Lindelf or hereditarily Lindelf)
4.3.2. Suppose that Cp .X / has a uniformly dense normal subspace. Is it true in

ZFC that Cp .X / is normal?
4.3.3. Suppose that Cp .X / has a uniformly dense perfectly normal subspace. Must
Cp .X / be perfectly normal?
in question is hereditarily Lindelf)
4.3 Uniformly Dense Subspaces 543
4.3.4. Suppose that Cp .X / has a uniformly dense hereditarily normal subspace.

Must Cp .X / be hereditarily normal?
in question is hereditarily Lindelf)
4.3.5. Suppose that a space Cp .X / has a uniformly dense subspace Y such that
i w.Y / !. Must Cp .X / have countable i -weight?
Motivated by Problem 2.457 (this is true if the uniformly dense subspace in question
is cosmic)
Comment It is not sufficient to require that .Y / !
4.3.6. Suppose that Cp .X / has a uniformly dense realcompact subspace. Must

Cp .X / be realcompact?
Motivated by Problem 2.459 (this is true if the uniformly dense subspace in question
is Lindelf )
4.3.7. Suppose that Cp .X / has a uniformly dense !-monolithic subspace. Must

Cp .X / be !-monolithic?
Motivated by Problem 2.457 (this is true if the respective uniformly dense subspace
is cosmic)
4.3.8. Suppose that Cp .X / has a uniformly dense set Y such that ext.Y / !. Is it
true in ZFC that ext.Cp .X // D !?
4.3.9. Suppose that Cp .X / has a uniformly dense pseudoradial subspace. Is it true

that Cp .X / is pseudoradial?
Motivated by the fact that this is true if the uniformly dense subspace in question is
sequential
544 4 Open Problems
4.4 Countable Spread and Similar Properties
Hereditary density being finitely multiplicative in spaces Cp .X /, a considerable

effort has been made to prove the same for spread and hereditary Lindelf number.
However, in spite of a wide variety of positive results, the general question still
remains open.
4.4.1. Let Cp .X / be perfectly normal. Is it true in ZFC that Cp .X / Cp .X / is
perfectly normal?
4.4.2. Suppose that every closed subspace of Cp .X / is separable. Is it true that

every closed subspace of Cp .X / Cp .X / is separable?
4.4.3. Suppose that every closed subspace of Cp .X / is separable and X is zero-

dimensional. Is it true that every closed subspace of Cp .X / Cp .X / is
separable?
4.4.4. Suppose that every closed subspace of Cp .X / Cp .X / is separable. Is it true

that every closed subspace of .Cp .X //! is separable?
4.4.5. Suppose that Lp .X / is perfectly normal. Is it true that Cp .Cp .X // is

perfectly normal?
4.4.6. Suppose that every closed subspace of Cp .X; I/ is separable. Is it true that
every closed subspace of Cp .X / is separable?
4.4.7. Suppose that every closed subspace of Cp .X / is separable and Y is a (closed)

subspace of X . Is it true in ZFC that every closed subspace of Cp .Y / is
separable?
4.4 Countable Spread and Similar Properties 545
4.4.8. Suppose that Cp Cp .X / is perfectly normal. Is it true that every closed

subspace of Cp Cp Cp .X / is separable?
4.4.9. Suppose that every closed subspace of Cp .X / is separable. Is then Lp .X /

perfectly normal?
4.4.10. Suppose that Lp .X / is perfectly normal. Is it true that every closed

subspace of Cp .X / is separable?
4.4.11. Suppose that all closed subspaces of Cp .X / are separable. Is it true in ZFC
that Cp .X / is hereditarily separable?
4.4.12. Suppose that all closed subspaces of X ! are separable. Is it true in ZFC that
all closed subspaces of Cp Cp .X / are separable?
4.4.13. Is it true that hl.Cp .X / Cp .X // D hl.Cp .X // for any space X ?

Published in Arhangelskii (1989)
4.4.14. Is it true that s.Cp .X / Cp .X // D s.Cp .X // for any space X ?

4.4.15. Suppose that Cp .X / is hereditarily Lindelf. Is it true in ZFC that Cp .X /

Cp .X / is hereditarily Lindelf?

546 4 Open Problems
4.4.16. Let X be a hereditarily Lindelf space. Is it true that every compact sub-
space of Cp .X / is (hereditarily) separable? Must every compact subspace
of Cp .X / have countable spread?
Published in Arhangelskii (1998b)
Motivated by the fact that Cp .X/ is hereditarily separable in case all finite powers
of X are hereditarily Lindelf (Problem 2.026)
4.4.17. Let X be a space of countable spread. Is it true that every compact subspace
of Cp .X / is (hereditarily) separable? Must every compact subspace of
Cp .X / have countable spread?
Motivated by the fact that Cp .X/ has countable spread in case all finite powers of
X have countable spread (Problem 2.025)
4.4.18. Let X be a hereditarily Lindelf space. Is it true that every separable

compact subspace of Cp .X / is hereditarily separable?
4.4.19. Suppose that X n is hereditarily Lindelf for all n 2 N. Is it true in ZFC that
there is a Lindelf Z such that Cp .X / Z RX ?
Published in Yaschenko (1992a) (attributed to Arhangelskii)
Motivated by the fact that under MAC:CH the space Z D Cp .X/ is Lindelf
(Problems 2.027 and 2.059)
4.4.20. Let X be a hereditarily Lindelf space. Is it true that Cp .X / condenses onto

a hereditarily separable space?
4.4.21. If X is a space and Y X , let Cp .Y jX / D ff jY W f 2 Cp .X /g. Is it true

that s.Cp .Y // D s.Cp .Y jX // for any Y X ?
Motivated by the fact that it is consistent with ZFC that the spread of Cp .Y / is
countable if and only if the spread of Cp .Y jX/ is countable
4.4.22. Assume MAC:CH. Suppose that spread of Cp .X / is countable. Is it true

that Cp .X /! is hereditarily Lindelf and hereditarily separable?
Motivated by the fact that this is true under the axiom SA (Problem 2.036)
4.5 Metacompactness and Its Derivatives 547

that t.X / D !?
Motivated by the fact that a stronger statement is true under SA (Problem 2.036)

that X is separable?
Motivated by the fact that a stronger statement is true under SA (Problem 2.036)
4.5 Metacompactness and Its Derivatives
The results on metacompactness in Cp .X / are scarce. So far it is not known whether

metacompactness of Cp .X / implies its Lindelf property. Even compactness of
X does not help. This subsection contains the most evident questions about
metacompactness and similar properties in Cp .X /.
4.5.1. Suppose that Cp .X / is metacompact. Must it be Lindelf? Is it true if X is
compact?
Author V.V. Tkachuk
Motivated by Problem 1.219 and the fact that any normal metacompact Cp .X/ is
Lindelf (Problem 2.438)
4.5.2. Suppose that Cp .X / is metacompact. Must it be realcompact? Is it true if X

is compact?
Author V.V. Tkachuk
Motivated by the fact that nothing is known about metacompactness in Cp .X/
4.5.3. Suppose that Cp .X / is metacompact. Must Cp .X / Cp .X / be metacom-

pact? Is this true if X is compact?
Author V.V. Tkachuk
Motivated by the analogous problem about Lindelf Cp .X/
4.5.4. Suppose that Cp .X / is Lindelf. Must Cp .X / Cp .X / be metacompact? Is

this true if X is compact?
Author V.V. Tkachuk
Motivated by the fact that almost nothing is known about Cp .X/ Cp .X/ in case
when Cp .X/ is Lindelf
548 4 Open Problems
4.5.5. Let X be a compact space such that Cp .X / is metacompact. Is it true that

Cp .Y / is metacompact for any closed Y X ?
Author V.V. Tkachuk
4.5.6. Let X be a compact space such that Cp .X / is metacompact. Is it true that

t.X / D !?
Author V.V. Tkachuk
Motivated by Problem 1.189
4.5.7. Suppose that Cp;n .X / is metacompact for all n 2 N. Must Cp .X / be

Lindelf? What happens if X is compact?
Author V.V. Tkachuk
4.5.8. Suppose that Cp .X / is hereditarily metacompact. Must Cp .X / be (heredi-

tarily) Lindelf? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that a hereditarily paracompact space Cp .X/ is hereditarily
Lindelf (Problems 1.219, 1.292, 2.001)
4.5.9. Suppose that the space Cp .X / is hereditarily metacompact. Must Cp .X /

Cp .X / be hereditarily metacompact? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the analogous problem about hereditarily Lindelf Cp .X/
4.5.10. Suppose that Cp;n .X / is hereditarily metacompact for all n 2 N. Must

Cp .X / be (hereditarily) Lindelf? What happens if X is compact?
Author V.V. Tkachuk
4.5.11. Is Cp .!1 C 1/ (hereditarily) metacompact?

Author V.V. Tkachuk
4.5.12. An arbitrary space Z is called -metacompact if any open cover of Z has

an open -point-finite refinement. Suppose that Cp .X / is -metacompact.
Does this imply that Cp .X / is metacompact? Must Cp .X / be Lindelf?
What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that if Cp .X/ is normal and -metacompact then it is Lindelf
(see Burke 1984)
4.5 Metacompactness and Its Derivatives 549
4.5.13. Suppose that the space Cp .X / is -metacompact. Must Cp .X / Cp .X / be

-metacompact? Is this true if X is compact?
Author V.V. Tkachuk
Motivated by the analogous problem about Lindelf Cp .X/
4.5.14. Suppose that X is a space such that Cp .X / is Lindelf. Must the space
Cp .X / Cp .X / be -metacompact? Is this true if X is compact?
Author V.V. Tkachuk
Motivated by the fact that almost nothing is known about Cp .X/ Cp .X/ in case
when Cp .X/ is Lindelf
4.5.15. Let X be a Corson (or Eberlein) compact space such that Cp .X / is

hereditarily -metacompact. Is it true that X is metrizable?
Author V.V. Tkachuk
Motivated by the fact that if Cp .X/ is hereditarily Lindelf then X is metrizable
4.5.16. Suppose that Cp;n .X / is -metacompact for all n 2 N. Must Cp .X / be

metacompact (or Lindelf)? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that nothing is known about -metacompactness in Cp .X/
4.5.17. Suppose that Cp .X / is hereditarily -metacompact. Must Cp .X / be (hered-

itarily) Lindelf? What happens if X is compact?
Author V.V. Tkachuk
4.5.18. Suppose that Cp .X / is hereditarily -metacompact. Must the space

Cp .X / Cp .X / be hereditarily -metacompact? What happens if X is
compact?
Author V.V. Tkachuk
4.5.19. Suppose that Cp;n .X / is hereditarily -metacompact for all n 2 N. Must

Cp .X / be (hereditarily) Lindelf? What happens if X is compact?
Author V.V. Tkachuk
4.5.20. Is Cp .!1 C 1/ (hereditarily) -metacompact?

Author V.V. Tkachuk
550 4 Open Problems
4.5.21. A space Z is called submetacompact

S if every open cover U of the space Z
has an open refinement V D n2! Vn such that every Vn covers Z and, for
any z 2 Z, there exists an n 2 ! for which the set fV 2 Vn W z 2 V g is finite.
Suppose that Cp .X / is submetacompact. Must Cp .X / be -metacompact
(or metacompact or Lindelf)? What happens when X is compact?
Author V.V. Tkachuk
Motivated by the fact that if Cp .X/ is normal and submetacompact then it is

Lindelf (see Burke 1984)
4.5.22. Suppose that Cp .X / is submetacompact. Must Cp .X /Cp .X / be submeta-

compact? Is this true if X is compact?
Author V.V. Tkachuk
Motivated by the analogous problem about Lindelf space Cp .X/
4.5.23. Suppose that Cp .X / is Lindelf. Must the space Cp .X / Cp .X / be

submetacompact? Is this true if X is compact?
Author V.V. Tkachuk
Motivated by the fact that nothing is known on Cp .X/Cp .X/ in case when Cp .X/
is Lindelf
4.5.24. Suppose that Cp;n .X / is submetacompact for all n 2 N. Must Cp .X / be

-metacompact (or Lindelf)? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that nothing is known about submetacompactness in Cp .X/
4.5.25. Suppose that Cp .X / is hereditarily submetacompact. Must Cp .X / be

(hereditarily) Lindelf? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that a hereditarily paracompact Cp .X/ is hereditarily Lindelf

(Problems 1.219, 1.292, 2.001)
4.5.26. Suppose that Cp .X / is hereditarily submetacompact. Must the space

Cp .X / Cp .X / be hereditarily submetacompact? What happens if X is
compact?
Author V.V. Tkachuk
4.5.27. Suppose that Cp;n .X / is hereditarily submetacompact for all n 2 N. Must

Cp .X / be (hereditarily) metacompact? What happens if X is compact?
Author V.V. Tkachuk
Motivated by the fact that a hereditarily paracompact Cp .X/ is hereditarily Lindelf

(Problems 1.219, 1.292, 2.001)
4.6 Mappings Which Involve Cp -Spaces 551
4.5.28. Is it true that Cp .X / is submetacompact for any compact X ?

Motivated by the fact that nothing is known about submetacompactness in Cp .X/
4.6 Mappings Which Involve Cp -Spaces
The existence of an algebraic structure compatible with the topology of Cp .X /

radically improves its topological properties. In particular, if Cp .X / is an open
continuous image of some nice space Z, then Cp .X / might have even better
properties than Z. For example, if Cp .X / is an open image of a metrizable space,
then it is second countable. Also, if a space Z is a continuous image of Cp .X /,
then we can expect very strong restrictions on Z if, say, Z is compact. A lot of
research has been done in this area and this subsection contains a compilation of the
respective open questions.
4.6.1. Let K be a linearly ordered compact space. Is it true that every compact
continuous image of Cp .K/ is metrizable?
Published in Tkachenko and Tkachuk (2005)
4.6.2. Let X be a Lindelf P -space. Is it true that every compact continuous image
of Cp .X / is metrizable?
4.6.3. Let X be a hereditarily Lindelf space. Is it true in ZFC that every compact
continuous image of Cp .X / is metrizable?
4.6.4. Suppose that K is a compact space with t.K/ ! and ' W Cp .X / ! K is a

continuous map. Is it true that '.Cp .X // is cosmic or even metrizable?
4.6.5. Suppose that Cp .X / is Lindelf, K is a compact space of countable tightness

and ' W Cp .X / ! K is a continuous map. Is it true that '.Cp .X // is cosmic
or even metrizable?
4.6.6. Suppose that X is compact, K is a compact space of countable tightness and

' W Cp .X / ! K is a continuous map. Is it true that '.Cp .X // is cosmic or
even metrizable?
552 4 Open Problems
4.6.7. Is it true that, for any cardinal and any compact space K of countable
tightness, if ' W R ! K is a continuous map, then '.R / is cosmic or even
metrizable?
4.6.8. Suppose that Cp .X / is Lindelf and there exists a condensation of Cp .X /

onto a -compact space Y . Must Y be cosmic?
Motivated by the fact that this is true if Cp .X/ is a Lindelf -space
4.6.9. Suppose that Cp .X / is a Lindelf -space and ' W Cp .X / ! Y is a

continuous surjective map of Cp .X / onto a -compact space Y . Must Y
be cosmic?
Motivated by the fact that this is true if ' is a condensation
4.6.10. Suppose that K is a perfectly normal compact space. Is it true that every
-compact continuous image of Cp .X / has a countable network?
4.6.11. Suppose that Cp .X / condenses onto a space of countable -weight. Must

X be separable?
Motivated by the fact that it is true if i w.Cp .X// D !
4.7 Additivity of Topological Properties
The existence of an algebraic structure compatible with the topology of Cp .X /

implies that many non-additive topological
S properties turn out to be additive in
Cp .X /. For example, if Cp .X / D n2! Mn and every Mn is metrizable, then
Cp .X / is metrizable, and hence, X is countable. On the other hand, countable
-weight is not countably additive in Cp .X / even when X is compact. Therefore it
is an interesting task to find out what topological properties are finitely or countably
4.7.1. Suppose that Cp .X / D A [ B, where A and B are Whyburn spaces. Must
Cp .X / be a Whyburn space?
Motivated by this is true if A and B are FrchetUrysohn spaces (Problem 2.450)

4.7 Additivity of Topological Properties 553
4.7.2. Suppose that Cp .X / D A [ B, where A and B are metacompact. Must

Cp .X / be metacompact? What happens if X is compact?
Published in CasarrubiasSegura (1999)
Motivated by the fact that paracompactness is finitely additive in spaces Cp .X/

(Problem 2.422)
4.7.3. Suppose that Cp .X / D A [ B, where A and B are normal. Must Cp .X / be

normal? What happens if X is compact?
Motivated by the fact that some non-additive properties are finitely additive in spaces
Cp .X/ (Problems 2.4222.430)
4.7.4. Suppose that Cp .X / D A [ B, where A and B are perfectly normal. Must

Cp .X / be perfectly normal? What happens if X is compact?
Cp .X/ (Problems 2.4222.430)
4.7.5. Suppose that Cp .X / D A [ B, where A and B are hereditarily normal. Must

Cp .X / be perfectly normal? What happens if X is compact?
Cp .X/ (Problems 1.292 and 2.4222.430)
4.7.6. Suppose that Cp .X / D A [ B, where A and B are collectionwise normal.

Must Cp .X / be normal? What happens if X is compact?
Cp .X/ (Problems 2.4222.430)
S
4.7.7. Suppose that Cp .X / D n2! Yn and every Yn is realcompact. Must Cp .X /
be realcompact?
Motivated by the fact that realcompactness is finitely additive in in spaces Cp .X/

(Problem 2.423)
S
4.7.8. Suppose that Cp .X / D n2! Yn and every Yn is paracompact. Must Cp .X /
be Lindelf?
Motivated by the fact that paracompactness is finitely additive in in spaces Cp .X/

(Problem 2.422)
554 4 Open Problems
S
4.7.9. Suppose that Cp .X / D n2! Yn and every Yn is metacompact. Must Cp .X /
be metacompact?
Motivated by the fact that paracompactness is finitely additive in in spaces Cp .X/

(Problem 2.422)
S
4.7.10. Suppose that Cp .X / D n2! Yn and every Yn is !-monolithic. Must
Cp .X / be !-monolithic?
Motivated by the fact that !-monolithity is finitely additive in in spaces Cp .X/

(Problem 2.426)
4.8 Raznoie (Unclassified Questions)
It is usually impossible to completely classify a complex data set such as the open
problems in Cp -theory. This last group of problems contains the open questions
which do not fit into any of the twenty-four previous subsections.
4.8.1. Is it consistent with ZFC that w.X / D s.Cp .X // for any compact space X ?
Motivated by the fact that under MAC:CH if the spread of Cp .X/ is countable then
X is metrizable (Problems 2.016 and 2.062)
4.8.2. Let X be a Lindelf !-stable space. Is then X n Lindelf for all n 2 N? Must
X X be Lindelf?
Motivated by the fact that if X is Lindelf -space then X ! is Lindelf
4.8.3. Let X be a Lindelf !-stable space. Suppose that F is a compact subspace

of Cp .X /. Must the tightness of F be countable?
Motivated by the fact that this is true if X is a Lindelf -space
4.8.4. Is it true in ZFC that there exists a compact separable space X such that some
Y Cp .X / is Lindelf and has uncountable network weight?
Comment Such a space exists consistently (Problems 2.0982.099)

4.8 Raznoie (Unclassified Questions) 555
4.8.5. Suppose that X is a zero-dimensional space such that Cp .X; D/ is a k-space

(sequential space). Is it true that Cp .X; D/ is FrchetUrysohn?
Author V.V. Tkachuk
Related to Problem 2.465

Bibliography
The bibliography of this book is intended to reflect the state of the art of modern Cp -theory;
besides, it is obligatory to mention the work of all authors whose results, in one form or another,
are cited here. The bibliographic selection for this volume has 300 items to solve the proportional
part of the task.
ALAS, O.T., COMFORT, W.W., GARCIA -FERREIRA , S., HENRIKSEN , M., W ILSON , R.G.,
WOODS, R.
[2000] When is jC.X Y /j D jC.X/jjC.Y /j? Houston J. Math. 26:1(2000), 83115.
ALAS, O.T., GARCIA -FERREIRA, S., TOMITA , A.H.
[1999] The extraresolvability of some function spaces, Glas. Mat. Ser. III
34(54):1(1999), 2335.
ALAS, O.T., TAMARIZ-M ASCARA, A.
[2006] On the Cech number of Cp .X/ II, Questions Answers Gen. Topology 24:1(2006),
3149.
ARGYROS, S., M ERCOURAKIS, S., NEGREPONTIS, S.
[1983] Analytic properties of Corson compact spaces, General Topology and Its Rela-
tions to Modern Analysis and Algebra, 5. Berlin, 1983, 1224.
ARGYROS, S., NEGREPONTIS, S.
[1983] On weakly K-countably determined spaces of continuous functions, Proc. Amer.
Math. Soc., 87:4(1983), 731736.
ARHANGELSKII , A.V.
[1976] On some topological spaces occurring in functional analysis (in Russian), Uspehi
Mat. Nauk, 31:5(1976), 1732.
[1978] The structure and classification of topological spaces and cardinal invariants (in
Russian), Uspehi Mat. Nauk, 33:6(1978), 2984.
[1981] Classes of topological groups (in Russian), Uspehi Mat. Nauk, 36:3(1981),
127146.
[1984] Continuous mappings, factorization theorems and function spaces (in Russian),
Trudy Mosk. Mat. Obsch., 47(1984), 321.
[1986] Hurewicz spaces, analytic sets and fan tightness of function spaces (in Russian),
Doklady AN SSSR, 287:3(1986), 525528.
[1987] A survey of Cp -theory, Questions and Answers in General Topology, 5(1987),
1109.
Problem Books in Mathematics, DOI 10.1007/978-3-319-04747-8,
558 Bibliography
[1989] Topological Function Spaces (in Russian), Moscow University P.H., Moscow,
1989.
[1989b] Hereditarily Lindelof spaces of continuous functions, Moscow University Math.
Bull., 44:3(1989b), 6769.
[1990a] Problems in Cp -theory, in: Open Problems in Topology, North Holland, Amster-
dam, 1990a, 603615.
[1990b] On the Lindelf degree of topological spaces of functions, and on embeddings
into Cp .X/, Moscow University Math. Bull., 45:5(1990b), 4345.
[1992a] Topological Function Spaces (translated from Russian), Kluwer Academic Pub-
lishers, Dordrecht, 1992a.
[1992b] Cp -theory, in: Recent Progress in General Topology, North Holland, Amsterdam,
1992, 156.
[1996a] On Lindelf property and spread in Cp -theory, Topol. and Its Appl., 74:
(13)(1996a), 8390.
[1996b] On spread and condensations, Proc. Amer. Math. Soc., 124:11(1996b), 3519
3527.
[1997] On a theorem of Grothendieck in Cp -theory, Topology Appl. 80(1997), 2141.
[1998a] Some observations on Cp -theory and bibliography, Topology Appl., 89(1998a),
203221.
[1998b] Embeddings in Cp -spaces, Topology Appl., 85(1998b), 933.
[2000a] On condensations of Cp -spaces onto compacta, Proc. Amer. Math. Soc.,
128:6(2000a), 18811883.
[2000b] Projective -compactness, !1 -caliber and Cp -spaces, Topology Appl.,
104(2000b), 1326.
[1999] Arhangelskii, A.V., Calbrix, J. A characterization of -compactness of a cosmic
space X by means of subspaces of RX , Proc. Amer. Math. Soc., 127:8(1999),
24972504.
ARHANGELSKII , A.V., CHOBAN M.M.
[1990] On the position of a subspace in the whole space, Comptes Rendus Acad. Bulg.
Sci., 43:4(1990), 1315.
[1996] On continuous mappings of Cp -spaces and extenders, Proc. Steklov Institute
Math., 212(1996), 2831.
ARHANGELSKII , A.V., OKUNEV, O.G.
[1985] Characterization of properties of spaces by properties of their continuous images
(in Russian) Vestnik Mosk. Univ., Math., Mech., 40:5(1985), 2830.
ARHANGELSKII , A.V., PAVLOV, O.I.
[2002] A note on condensations of Cp .X/ onto compacta, Comment. Math. Univ.
Carolinae, 43:3(2002), 485492.
ARHANGELSKII , A.V., PONOMAREV, V.I.
[1974] Basics of General Topology in Problems and Exercises (in Russian), Nauka,
Moscow, 1974.
ARHANGELSKII , A.V., TKACHUK , V.V.
[1985] Function Spaces and Topological Invariants (in Russian), Moscow University
P.H., Moscow, 1985.
[1986] Calibers and point-finite cellularity of the spaces Cp .X/ and some questions of
S. Gulko and M. Huek, Topology Appl., 23:1(1986), 6574.
ARHANGELSKII , A.V., USPENSKIJ, V.V.
[1986] On the cardinality of Lindelf subspaces of function spaces, Comment. Math.
Univ. Carolinae, 27:4(1986), 673676.
Bibliography 559
ASANOV, M.O., VELICHKO , M.V.

[1981] Compact sets in Cp .X/ (in Russian), Comment. Math. Univ. Carolinae,
22:2(1981), 255266.
BATUROV D.P.
[1987] On subspaces of function spaces (in Russian), Vestnik Moskovsk. Univ., Math.,
Mech., 42:4(1987), 6669.
[1988] Normality of dense subsets of function spaces, Vestnik Moskovsk. Univ., Math.,
Mech., 43:4(1988), 6365.
[1990] Normality in dense subspaces of products, Topology Appl., 36(1990), 111116.
[1990] Some properties of the weak topology of Banach spaces, Vestnik Mosk. Univ.,
Math., Mech., 45:6(1990), 6870.
BELLA , A., YASCHENKO , I.V.
[1999] On AP and WAP spaces, Comment. Math. Univ. Carolinae, 40:3(1999), 531536.
BESSAGA , C., PELCZINSKI , A.
[1960] Spaces of continuous functions, 4, Studia Math., 19(1960), 5362.
[1975] Selected Topics in Infinite-Dimensional Topology, PWN, Warszawa, 1975.
BOUZIAD , A., CALBRIX , J.
[1995] Images usco-compactes des espaces Cech-complets de Lindelf, C. R. Acad. Sci.
Paris Sr. I Math. 320:7(1995), 839842.
BURKE, D.K.
[1984] Covering properties, Handbook of SetTheoretic Topology, edited by K. Kunen
and J.E. Vaughan, Elsevier Science Publishers B.V., 1984, 347422.
[2007] Weak-bases and D-spaces, Comment. Math. Univ. Carolin. 48:2(2007),
281289.
BURKE, D.K., LUTZER , D.J.
[1976] Recent advances in the theory of generalized metric spaces, in: Topology: Proc.
Memphis State University Conference, Lecture Notes in Pure and Applied Math.,
Marcel Dekker, New York, 1976, 170.
BUZYAKOVA , R.Z.
[2006a] Spaces of continuous step functions over LOTS, Fund. Math., 192(2006a), 2535.
[2006b] Spaces of continuous characteristic functions, Comment. Math. Universitatis
Carolinae, 47:4(2006b), 599608.
[2007] Function spaces over GO spaces, Topology Appl., 154:4(2007), 917924.
CALBRIX , J.
[1996] k-spaces and Borel filters on the set of integers (in French), Trans. Amer. Math.
Soc., 348(1996), 20852090.
CASARRUBIASSEGURA, F.
[1999] Realcompactness and monolithity are finitely additive in Cp .X/, Topology Proc.,
24(1999), 89102.
[2001] On compact weaker topologies in function spaces, Topology and Its Applications,
115(2001), 291298.
CASCALES, B.
[1987] On K-analytic locally convex spaces, Arch. Math., 49(1987), 232244.
CASCALES, B., KAKOL, J., SAXON , S.A.
[2002] Weight of precompact subsets and tightness, J. Math. Anal. Appl., 269(2002),
500518.
560 Bibliography
CASCALES, B., ORIHUELA , J.

[1991] Countably determined locally convex spaces, Portugal. Math. 48:1(1991), 7589.
CHOBAN , M.M.
[2005] On some problems of descriptive set theory in topological spaces, Russian Math.
Surveys 60:4(2005), 699719.
CHRISTENSEN , J.P.R.
[1974] Topology and Borel Structure, North Holland P.C., Amsterdam, 1974.
[1981] Joint continuity of separably continuous functions, Proceedings of Amer. Math.
Soc., 82:3(1981), 455461.
CIESIELSKI , K.
[1993] Linear subspace of R without dense totally disconnected subsets, Fund. Math.
142 (1993), 8588.
COMFORT, W.W., FENG , L.
[1993] The union of resolvable spaces is resolvable, Math. Japon. 38:3(1993), 413414.
COMFORT, W.W., HAGER , A.W.
[1970a] Estimates for the number of real-valued continuous functions, Trans. Amer. Math.
Soc., 150(1970a), 619631.
[1970b] Dense subspaces of some spaces of continuous functions, Math. Z. 114(1970b),
373389.
[1970c] Estimates for the number of real-valued continuous functions, Trans. Amer. Math.
Soc. 150(1970c), 619631.
COMFORT, W.W., NEGREPONTIS, S.A.
[1974] The theory of ultrafilters. Die Grundl. Math. Wiss., 211, Springer, New York,
1974.
[1982] Chain Conditions in Topology, Cambridge Tracts in Mathematics, 79, New York,
1982.
CONTRERAS-CARRETO , A., TAMARIZ-M ASCARA , A.
[2003] On some generalizations of compactness in spaces Cp .X; 2/ and Cp .X; Z/, Bol.
Soc. Mat. Mexicana (3)9:2(2003), 291308.
DEBS, G.
[1985] Espaces K-analytiques et espaces de Baire de fonctions continues, Mathematika,
32(1985), 218228.
DIJKSTRA , J., GRILLOT, T., LUTZER , D., VAN M ILL, J.
[1985] Function spaces of low Borel complexity, Proc. Amer. Math. Soc., 94:4(1985),
703710.
DIJKSTRA , J., M OGILSKI , J.
[1991] Cp .X/-representation of certain Borel absorbers, Topology Proc., 16(1991),
2939.
DI M AIO , G., HOL , L., HOL, D., M C COY, R.A.
[1998] Topologies on the space of continuous functions, Topology Appl. 86:2(1998),
105122.
DIMOV, G., TIRONI , G.
[1987] Some remarks on almost radiality in function spaces, Acta Univ. Carolin. Math.
Phys. 28:2(1987), 4958.
Bibliography 561
DOUWEN , E.K. VAN

[1984] The integers and topology, Handbook of SetTheoretic Topology, K. Kunen and
J.E. Vaughan, editors, Elsevier Science Publishers B.V., 1984, 111167.
DOUWEN , E.K. POL, R.
[1977] Countable spaces without extension properties, Bull. Polon. Acad. Sci., Math.,
25(1977), 987991.
DOW, A.
[2005a] Closures of discrete sets in compact spaces, Studia Sci. Math. Hungar.
42:2(2005a), 227234.
[2005b] Property D and pseudonormality in first countable spaces, Comment. Math.
Univ. Carolin. 46:2(2005b), 369372.
DOW, A., PAVLOV, O.
[2006] More about spaces with a small diagonal, Fund. Math. 191:1(2006), 6780.
[2007] Perfect preimages and small diagonal, Topology Proc. 31:1(2007), 8995.
EFIMOV, B.A.
[1977] Mappings and imbeddings of dyadic spaces, I, Math. USSR Sbornik, 32:1(1977),
4557.
ENGELKING , R.
[1977] General Topology, PWN, Warszawa, 1977.
FORT, M.K.
[1951] A note on pointwise convergence, Proc. Amer. Math. Soc., 2(1951), 3435.
FOX , R.H.
[1945] On topologies for function spaces, Bull. Amer. Math. Soc., 51(1945), 429432.
FREMLIN , D.H.
[1977] K-analytic spaces with metrizable compacta, Mathematika, 24(1977), 257261.
[1994] Sequential convergence in Cp .X/, Comment. Math. Univ. Carolin., 35:2(1994),
371382.
GARTSIDE, P.
[1997] Cardinal invariants of monotonically normal spaces, Topology Appl.
77:3(1997), 303314.
[1998] Nonstratifiability of topological vector spaces, Topology Appl. 86:2(1998),
133140.
GARTSIDE, P., FENG , Z.
[2007] More stratifiable function spaces, Topology Appl. 154:12(2007), 24572461.
GERLITS, J.
[1983] Some properties of C.X/, II, Topology Appl., 15:3(1983), 255262.
GERLITS, J., NAGY, ZS.
[1982] Some properties of C.X/, I, Topology Appl., 14:2(1982), 151161.
GERLITS, J., JUHSZ, I., SZENTMIKLSSY, Z.
[2005] Two improvements on Tkacenkos addition theorem, Comment. Math. Univ.
Carolin. 46:4(2005), 705710.
GERLITS, J., NAGY, ZS., SZENTMIKLOSSY, Z.
[1988] Some convergence properties in function spaces, in: General Topology and Its
Relation to Modern Analysis and Algebra, Heldermann, Berlin, 1988, 211222.
562 Bibliography
GILLMAN , L., JERISON , M.

[1960] Rings of Continuous Functions, D. van Nostrand Company Inc., Princeton, 1960.
GORDIENKO , I.YU .
[1990] Two theorems on relative cardinal invariants in Cp -theory, Zb. Rad. Filozofskogo
Fak. Niu, Ser. Mat., 4(1990), 57.
GRAEV, M.I.
[1950] Theory of topological groups, I (in Russian), Uspehi Mat. Nauk, 5:2(1950), 356.
GRUENHAGE, G.
[1976] Infinite games and generalizations of first-countable spaces, General Topology
and Appl. 6:3(1976), 339352.
[1997] A non-metrizable space whose countable power is -metrizable, Proc. Amer.
Math. Soc. 125:6(1997), 18811883.
[1998] Dugundji extenders and retracts of generalized ordered spaces, Fundam. Math.,
158(1998), 147164.
[2002] Spaces having a small diagonal, Topology Appl., 122(2002), 183200.
[2006] The story of a topological game, Rocky Mountain J. Math. 36:6(2006), 1885
1914.
GRUENHAGE, G., M A , D.K.
[1997] Bairness of Ck .X/ for locally compact X, Topology Appl., 80(1997), 131139.
GULKO , S.P.
[1979] On the structure of spaces of continuous functions and their hereditary paracom-
pactness (in Russian), Uspehi Matem. Nauk, 34:6(1979), 3340.
GULKO , S.P., SOKOLOV, G.A.
[1998] P -points in N and the spaces of continuous functions, Topology Appl.,
85(1998), 137142.
[2000] Compact spaces of separately continuous functions in two variables, Topology
and Its Appl. 107:1-2(2000), 8996.
HAGER , A.W.
[1969] Approximation of real continuous functions on Lindelf spaces, Proc. Amer.
Math. Soc., 22(1969), 156163.
HAGLER , J.
[1975] On the structure of S and C.S/ for S dyadic, Trans. Amer. Math. Soc.,
214(1975), 415428.
HAO -XUAN , Z.
[1982] On the small diagonals, Topology Appl., 13:3(1982), 283293.
HAYDON , R.G.
[1990] A counterexample to several questions about scattered compact spaces, Bull.
London Math. Soc., 22(1990), 261268.
HEATH , R.W., LUTZER , D.J.
[1974a] The Dugundji extension theorem and collectionwise normality, Bull. Acad. Polon.
Sci., Ser. Math., 22(1974a), 827830.
[1974b] Dugundji extension theorem for linearly ordered spaces, Pacific J. Math.,
55(1974b), 419425.
HEATH , R.W., LUTZER , D.J., ZENOR , P.L.
[1975] On continuous extenders, Studies in Topology, Academic Press, New York, 1975,
203213.
Bibliography 563
HEWITT, E.
[1948] Rings of real-valued continuous functions, I, Trans. Amer. Math. Soc.,
64:1(1948), 4599.
HODEL, R.
[1984] Cardinal Functions I, in: Handbook of Set-Theoretic Topology, Ed. by K. Kunen
and J.E. Vaughan, Elsevier Science Publishers B.V., 1984, 161.
HUEK , M.
[1972] Realcompactness of function spaces and .P Q/, General Topology and Appl.
2(1972), 165179.
[1977] Topological spaces without -accessible diagonal, Comment. Math. Univ. Car-
olinae, 18:4(1977), 777788.
[1979] Mappings from products. Topological structures, II, Math. Centre Tracts, Ams-
terdam, 115(1979), 131145.
[1997a] Productivity of some classes of topological linear spaces, Topology Appl.
80:1-2(1997), 141154.
[2005] Cp .X/ in coreflective classes of locally convex spaces, Topology and Its Appli-
cations, 146/147(2005), 267278.
IVANOV, A.V.
[1978] On bicompacta all finite powers of which are hereditarily separable, Soviet
Math., Doklady, 19:6(1978), 14701473.
JUHSZ, I.
[1971] Cardinal functions in topology, Mathematical Centre Tracts, 34, Amsterdam,
1971.
[1980] Cardinal Functions in TopologyTen Years Later, Mathematical Centre Tracts,
North Holland P.C., Amsterdam, 1980.
[1992] Cardinal functions, Recent Progress in General Topology, North-Holland, Ams-
terdam, 1992, 417441.
JUHSZ, I., M ILL, J. VAN
[1981] Countably compact spaces all countable subsets of which are scattered, Com-
ment. Math. Univ. Carolin. 22:4(1981), 851855.
JUHSZ, I., SOUKUP, L., SZENTMIKLSSY, Z.
[2007] First countable spaces without point-countable -bases, Fund. Math.
196:2(2007), 139149.
JUHSZ, I., SZENTMIKLSSY, Z.
[1992] Convergent free sequences in compact spaces, Proc. Amer. Math. Soc.,
116:4(1992), 11531160.
[1995] Spaces with no smaller normal or compact topologies, 1993), Bolyai Soc. Math.
Stud., 4(1995), 267274.
[2002] Calibers, free sequences and density, Topology Appl. 119:3(2002), 315324.
JUST, W., SIPACHEVA , O.V., SZEPTYCKI , P.J.

[1996] Non-normal spaces Cp .X/ with countable extent, Proceedings Amer. Math. Soc.,
124:4(1996), 12271235.
KALAMIDAS, N.D.
[1985] Functional properties of C.X/ and chain conditions on X, Bull. Soc. Math.
Grce 26(1985), 5364.
[1992] Chain condition and continuous mappings on Cp .X/, Rendiconti Sem. Mat.
Univ. Padova, 87(1992), 1927.
564 Bibliography
KALAMIDAS, N.D., SPILIOPOULOS, G.D.

[1992] Compact sets in Cp .X/ and calibers, Canadian Math. Bull., 35:4(1992),
497502.
KUNEN , K.
[1980] Set Theory. An Introduction to Independence Proofs, Studies Logic Found.
Mathematics, 102(1980), North Holland P.C., Amsterdam, 1980
[1981] A compact L-space under CH, Topology Appl., 12(1981), 283287.
KUNEN , K., DE LA VEGA , R.
[2004] A compact homogeneous S-space, Topology Appl., 136(2004), 123127.
KURATOWSKI , C.
[1966] Topology, vol. 1, Academic Press Inc., London, 1966.
LELEK , A.
[1969] Some cover properties of spaces, Fund. Math., 64:2(1969), 209218.
LINDENSTRAUSS, J., TZAFRIRI , L.
[1977] Classical Banach Spaces I, Springer, Berlin, 1977.
LUTZER , D.J., M C COY, R.A.
[1980] Category in function spaces I, Pacific J. Math., 90:1(1980), 145168.
LUTZER , D.J., M ILL, J. VAN , POL, R.
[1985] Descriptive complexity of function spaces, Transactions of the Amer. Math. Soc.,
291(1985), 121128.
LUTZER , D. J., M ILL, J. VAN , TKACHUK , V.V.
[2008] Amsterdam properties of Cp .X/ imply discreteness of X, Canadian Math. Bull.
51:4(2008), 570578.
M ALYKHIN , V.I.
[1987] Spaces of continuous functions in simplest generic extensions, Math. Notes,
41(1987), 301304.
[1994] A non-hereditarily separable space with separable closed subspaces, Q & A in
General Topology, 12(1994), 209214.
[1998] On subspaces of sequential spaces (in Russian), Matem. Zametki, 64:3(1998),
407413.
[1999] Cp .I / is not subsequential, Comment. Math. Univ. Carolinae, 40:4(1999),
785788.
M ALYKHIN , V.I., SHAKHMATOV, D.B.
[1992] Cartesian products of Frchet topological groups and function spaces, Acta
Math. Hungarica, 60(1992), 207215.
M ARCISZEWSKI , W.
[1983] A pre-Hilbert space without any continuous map onto its own square, Bull. Acad.
Polon. Sci., 31:(912)(1983), 393397.
[1988] A remark on the space of functions of first Baire class, Bull. Polish Acad. Sci.,
Math., 36:(12)(1988), 6567.
[1993] On analytic and coanalytic function spaces Cp .X/, Topology and Its Appl.,
50(1993), 241248.
[1995a] On universal Borel and projective filters, Bull. Acad. Polon. Sci., Math.,
43:1(1995a), 4145.
[1995b] A countable X having a closed subspace A with Cp .A/ not a factor of Cp .X/,
Topology Appl., 64(1995b), 141147.
Bibliography 565
[1997a] A function space Cp .X/ not linearly homeomorphic to Cp .X/ R, Fundamenta

Math., 153:2(1997a), 125140.
[1997b] On hereditary Baire products, Bull. Polish Acad. Sci., Math., 45:3(1997b),
247250.
[1998a] P -filters and hereditary Baire function spaces, Topology Appl., 89(1998a),
241247.
[1998b] Some recent results on function spaces Cp .X/, Recent Progress in Function
Spaces., Quad. Mat. 3(1998b), Aracne, Rome, 221239.
[2002] Function Spaces, in: Recent Progress in General Topology II, Ed. by M. Huek
and J. van Mill, Elsevier Sci. B.V., Amsterdam, 2002, 345369.
[2003] A function space Cp .X/ without a condensation onto a -compact space, Proc.
Amer. Math. Society, 131:6(2003), 19651969.
M ARCISZEWSKI , W., PELANT, J.
[1997] Absolute Borel sets and function spaces, Trans. Amer. Math. Soc., 349:9(1997),
35853596.
M C COY, R.A.
[1975] First category function spaces under the topology of pointwise convergence, Proc.
Amer. Math. Soc., 50(1975), 431434.
[1978a] Characterization of pseudocompactness by the topology of uniform convergence
on function spaces, J. Austral. Math. Soc., 26(1978a), 251256.
[1978b] Submetrizable spaces and almost -compact function spaces, Proc. Amer. Math.
Soc., 71(1978b), 138142.
[1978c] Second countable and separable function spaces, Amer. Math. Monthly,
85:6(1978c), 487489.
[1980a] Countability properties of function spaces, Rocky Mountain J. Math., 10(1980a),
717730.
[1980b] A K-space function space, Int. J. Math. Sci., 3(1980b), 701711.
[1986] Fine topology on function spaces, Internat. J. Math. Math. Sci. 9:3(1986),
417424.
M C COY, R.A., NTANTU , I.
[1986] Completeness properties of function spaces, Topology Appl., 22:2(1986),
191206.
[1988] Topological Properties of Spaces of Continuous Functions, Lecture Notes in
Math., 1315, Springer, Berlin, 1988.
M ICHAEL, E.
[1966] @0 -spaces, J. Math. and Mech., 15:6(1966), 9831002.
[1973] On k-spaces, kR -spaces and k.X/, Pacific J. Math., 47:2(1973), 487498.
[1977] @0 -spaces and a function space theorem of R. Pol, Indiana Univ. Math. J.,
26(1977), 299306.
M ILL, J. VAN
[1984] An introduction to !, Handbook of Set-Theoretic Topology, North-Holland,
Amsterdam, 1984, 503567.
[1989] Infinite-Dimensional Topology. Prerequisites and Introduction, North Holland,
Amsterdam, 1989.
[1999] Cp .X/ is not G : a simple proof, Bull. Polon. Acad. Sci., Ser. Math., 47(1999),
319323.
[2002] The Infinite-Dimensional Topology of Function Spaces, North Holland Math.
Library 64, Elsevier, Amsterdam, 2002.
M ORISHITA , K.
[2014] The kR -property of function spaces, Preprint.
566 Bibliography
[1992a] The minimal support for a continuous functional on a function space, Proc. Amer.
Math. Soc. 114:2(1992a), 585587.
[1992b] The minimal support for a continuous functional on a function space. II, Tsukuba
J. Math. 16:2(1992b), 495501.
NAKHMANSON , L. B.
[1982] On continuous images of -products (in Russian), Topology and Set Theory,
Udmurtia Universty P.H., Izhevsk, 1982, 1115.
[1984] The Souslin number and calibers of the ring of continuous functions (in Russian),
Izv. Vuzov, Matematika, 1984, N 3, 4955.
[1985] On Lindelf property of function spaces (in Russian), Mappings and Extensions
of Topological Spaces, Udmurtia University P.H., Ustinov, 1985, 812.
NOBLE, N.
[1969a] Products with closed projections, Trans Amer. Math. Soc., 140(1969a), 381391.
[1969b] Ascoli theorems and the exponential map, Trans Amer. Math. Soc., 143(1969b),
393411.
[1971] Products with closed projections II, Trans Amer. Math. Soc., 160(1971),
169183.
[1974] The density character of function spaces, Proc. Amer. Math. Soc., 42:1(1974),
228233.
NOBLE, N., ULMER , M.
[1972] Factoring functions on Cartesian products, Trans Amer. Math. Soc., 163(1972),
329339.
NYIKOS, P.
[1981] Metrizability and the FrchetUrysohn property in topological groups, Proc.
Amer. Math. Soc., 83:4(1981), 793801.
OKUNEV, O.G.
[1984] Hewitt extensions and function spaces (in Russian), Cardinal Invariants and
Mappings of Topological Spaces (in Russian), Izhvsk, 1984, 7778.
[1985] Spaces of functions in the topology of pointwise convergence: Hewitt extension
and -continuous functions, Moscow Univ. Math. Bull., 40:4(1985), 8487.
[1993a] On Lindelf -spaces of functions in the pointwise topology, Topology and Its
Appl., 49(1993a), 149166.
[1993b] On analyticity in cosmic spaces, Comment. Math. Univ. Carolinae, 34(1993b),
185190.
[1995] On Lindelf sets of continuous functions, Topology Appl., 63(1995), 9196.
[1996] A remark on the tightness of products, Comment. Math. Univ. Carolinae,
37:2(1996), 397399.
[1997] On the Lindelf property and tightness of products, Topology Proc., 22(1997),
363371.
[2002] Tightness of compact spaces is preserved by the t -equivalence relation, Com-
ment. Math. Univ. Carolin. 43:2(2002), 335342.
[2005a] Frchet property in compact spaces is not preserved by M -equivalence, Com-
ment. Math. Univ. Carolin. 46:4(2005a), 747749.
[2005b] A -compact space without uncountable free sequences can have arbitrary
tightness, Questions Answers Gen. Topology 23:2(2005b), 107108.
OKUNEV, O.G., SHAKHMATOV, D.B.
[1987] The Baire property and linear isomorphisms of continuous function spaces (in
Russian), Topological Structures and Their Maps, Latvian State University P.H.,
Riga, 1987, 8992.
Bibliography 567
OKUNEV, O.G., TAMANO , K.

[1996] Lindelf powers and products of function spaces, Proceedings of Amer. Math.
Soc., 124:9(1996), 29052916.
OKUNEV, O.G., TAMARIZ-M ASCARA, A.
[2004] On the Cech number of Cp .X/, Topology Appl., 137(2004), 237249.
OKUNEV, O.G., TKACHUK , V.V.
[2001] Lindelf -property in Cp .X/ and p.Cp .X// D ! do not imply countable
network weight in X, Acta Mathematica Hungarica, 90:(12)(2001), 119132.
[2002] Density properties and points of uncountable order for families of open sets in
function spaces, Topology Appl., 122(2002), 397406.
PASYNKOV, B.A.
[1967] On open mappings, Soviet Math. Dokl., 8(1967), 853856.
PELANT, J.
[1988] A remark on spaces of bounded continuous functions, Indag. Math., 91(1988),
335338.
POL, R.
[1977] Concerning function spaces on separable compact spaces, Bull. Acad. Polon.
Sci., Sr. Math., Astron. et Phys., 25:10(1977), 993997.
[1978] The Lindelf property and its analogue in function spaces with weak topology,
Topology. 4-th Colloq. Budapest, 2(1978), Amsterdam, 1980, 965969.
[1989] Note on pointwise convergence of sequences of analytic sets, Mathem., 36(1989),
290300.
PONTRIAGIN , L.S.
[1984] Continuous Groups (in Russian), Nauka, Moscow, 1984.
PYTKEEV, E.G.
[1976] Upper bounds of topologies, Math. Notes, 20:4(1976), 831837.
[1982a] On the tightness of spaces of continuous functions (in Russian), Uspehi Mat.
Nauk, 37:1(1982a), 157158.
[1982b] Sequentiality of spaces of continuous functions (in Russian), Uspehi Matematich-
eskih Nauk, 37:5(1982b), 197198.
[1985] The Baire property of spaces of continuous functions, (in Russian), Matem.
Zametki, 38:5(1985), 726740.
[1990] A note on Baire isomorphism, Comment. Math. Univ. Carolin. 31:1(1990), 109
112.
[1992a] On FrchetUrysohn property of spaces of continuous functions, (in Russian),
Trudy Math. Inst. RAN, 193(1992a), 156161.
[1992b] Spaces of functions of the first Baire class over K-analytic spaces (in Russian),
Mat. Zametki 52:3(1992b), 108116.
[2003] Baire functions and spaces of Baire functions, Journal Math. Sci. (N.Y.)
136:5(2006), 41314155
PYTKEEV, E.G., YAKOVLEV, N.N.
[1980] On bicompacta which are unions of spaces defined by means of coverings,
Comment. Math. Univ. Carolinae, 21:2(1980), 247261.
RAJAGOPALAN , M., W HEELER , R.F.
[1976] Sequential compactness of X implies a completeness property for C.X/, Cana-
dian J. Math., 28(1976), 207210.
568 Bibliography
REZNICHENKO , E.A.
[1987] Functional and weak functional tightness (in Russian), Topological Structures
and Their Maps, Latvian State University P.H., Riga, 1987, 105110.
[1989] A pseudocompact space in which only the subsets of not full cardinality are not
closed and not discrete, Moscow Univ. Math. Bull. 44:6(1989), 7071.
[1990] Normality and collectionwise normality in function spaces, Moscow Univ. Math.
Bull., 45:6(1990), 2526.
[2008] Stratifiability of Ck .X/ for a class of separable metrizable X, Topology Appl.
155:17-18(2008), 20602062.
RUDIN , M.E.
[1956] A note on certain function spaces, Arch. Math., 7(1956), 469470.
RUDIN , W.
[1973] Functional Analysis, McGraw-Hill Book Company, New York, 1973.
SAKAI , M.
[1988a] On supertightness and function spaces, Commentationes Math. Univ. Carolinae,
29:2(1988a), 249251.
[1988b] Property C 00 and function spaces, Proc. Amer. Math. Soc., 104:3(1988b),
917919.
[1992] On embeddings into Cp .X/ where X is Lindelf, Comment. Math. Univ.
Carolinae, 33:1(1992), 165171.
[1995] Embeddings of -metrizable spaces into function spaces, Topol. Appl., 65(1995),
155165.
[2000] Variations on tightness in function spaces, Topology Appl., 101(2000), 273280.
[2003] The Pytkeev property and the Reznichenko property in function spaces, Note di
Matem. 22:2(2003/04), 4352
[2006] Special subsets of reals characterizing local properties of function spaces,
Selection Principles and Covering Properties in Topology, Dept. Math., Seconda
Univ. Napoli, Caserta, Quad. Mat., 18(2006), 195225,
[2007] The sequence selection properties of Cp .X/, Topology Appl., 154(2007),
552560.
[2008] Function spaces with a countable cs -network at a point, Topology and Its Appl.,
156:1(2008), 117123.
[2014] On calibers of function spaces, Preprint.
SHAKHMATOV, D.B.
[1986] A pseudocompact Tychonoff space all countable subsets of which are closed and
C -embedded, Topology Appl., 22:2(1986), 139144.
SHAPIROVSKY, B.E.
[1974] Canonical sets and character. Density and weight in compact spaces, Soviet
Math. Dokl., 15(1974), 12821287.
[1978] Special types of embeddings in Tychonoff cubes, subspaces of -products and
cardinal invariants, Colloquia Mathematica Soc. Janos Bolyai, 23(1978), 1055
1086.
[1981] Cardinal invariants in bicompacta (in Russian), Seminar on General Topology,
Moscow University P.H., Moscow, 1981, 162187.
SIPACHEVA , O.V.
[1988] Lindelf -spaces of functions and the Souslin number, Moscow Univ. Math.
Bull., 43:3(1988), 2124.
[1989] On Lindelf subspaces of function spaces over linearly ordered separable
compacta, General Topology, Spaces and Mappings, Mosk. Univ., Moscow,
1989, 143148.
Bibliography 569
[1992] On surlindelf compacta (in Russian), General Topology. Spaces, Mappings and
Functors, Moscow University P.H., Moscow, 1992, 132140.
SOKOLOV, G.A.
[1986] On Lindelf spaces of continuous functions (in Russian), Mat. Zametki,
39:6(1986), 887894.
[1993] Lindelf property and the iterated continuous function spaces, Fundam. Math.,
143(1993), 8795.
STONE, A.H.
[1963] A note on paracompactness and normality of mapping spaces, Proc. Amer. Math.
Soc., 14(1963), 8183.
TALAGRAND , M.
[1979] Sur la K-analyticit des certains espaces doperateurs, Israel J. Math., 32(1979),
124130.
[1984] A new countably determined Banach space, Israel J. Math., 47:1(1984), 7580.
[1985] Espaces de Baire et espaces de Namioka, Math. Ann., 270(1985), 159164.
TAMANO , K., TODORCEVIC , S.
[2005] Cosmic spaces which are not -spaces among function spaces with the topology
of pointwise convergence, Topology Appl. 146/147(2005), 611616.
TAMARIZM ASCARA , A.
[1996] Countable product of function spaces having p-Frechet-Urysohn like properties,
Tsukuba J. Math. 20:2(1996), 291319.
[1998] -pseudocompactness in Cp -spaces, Topology Proc., 23(1998), 349362.
[2006] Continuous selections on spaces of continuous functions, Comment. Math. Univ.
Carolin. 47:4(2006), 641660.
TANI , T.
[1986] On the tightness of Cp .X/, Memoirs Numazu College Technology, 21(1986),
217220.
TKACHENKO , M.G.
[1978] On the behaviour of cardinal invariants under the union of chains of spaces (in
Russian), Vestnik Mosk. Univ., Math., Mech., 33:4(1978), 5058.
[1979] On continuous images of dense subspaces of topological products (in Russian),
Uspehi Mat. Nauk, 34:6(1979), 199202.
[1982] On continuous images of dense subspaces of -products of compacta (in
Russian), Sibirsk. Math. J., 23:3(1982), 198207.
[1985] On continuous images of spaces of functions, Sibirsk. Mat. Zhurnal, 26:5(1985),
159167.
[2005] Tkachenko, M.G., Tkachuk, V.V. Dyadicity index and metrizability of compact
continuous images of function spaces, Topol. Appl., 149:(13)(2005), 243257.
TKACHUK , V.V.
[1984b] Characterization of the Baire property in Cp .X/ in terms of the properties of the
space X (in Russian), Cardinal Invariants and Mappings of Topological Spaces
(in Russian), Izhevsk, 1984b, 7677.
[1986a] The spaces Cp .X/: decomposition into a countable union of bounded subspaces
and completeness properties, Topology Appl., 22:3(1986a), 241254.
[1986b] Approximation of RX with countable subsets of Cp .X/ and calibers of the space
Cp .X/, Comment. Math. Univ. Carolinae, 27:2(1986b), 267276.
570 Bibliography
[1987a] The smallest subring of the ring Cp .Cp .X// which contains X [ f1g is dense
in Cp .Cp .X// (in Russian), Vestnik Mosk. Univ., Math., Mech., 42:1(1987a),
2022.
[1987b] Spaces that are projective with respect to classes of mappings, Trans. Moscow
Math. Soc., 50(1987b), 139156.
[1988] Calibers of spaces of functions and metrization problem for compact subsets of
Cp .X/ (in Russian), Vestnik Mosk. Univ., Matem., Mech., 43:3(1988), 2124.
[1991] Methods of the theory of cardinal invariants and the theory of mappings applied
to the spaces of functions (in Russian), Sibirsk. Mat. Zhurnal, 32:1(1991),
116130.
[1994] Decomposition of Cp .X/ into a countable union of subspaces with good
properties implies good properties of Cp .X/, Trans. Moscow Math. Soc.,
55(1994), 239248.
[1995] What if Cp .X/ is perfectly normal? Topology Appl., 65(1995), 5767.
[1998] Mapping metric spaces and their products onto Cp .X/, New Zealand J. Math.,
27:1(1998), 113122.
[2000] Behaviour of the Lindelf -property in iterated function spaces, Topology
Appl., 107:3-4(2000), 297305.
[2001] Lindelf -property in Cp .X/ together with countable spread of X implies X is
cosmic, New Zealand J. Math., 30(2001), 93101.
[2003] Properties of function spaces reflected by uniformly dense subspaces, Topology
Appl., 132(2003), 183193.
[2005a] A space Cp .X/ is dominated by irrationals if and only if it is K-analytic, Acta
Math. Hungarica, 107:4(2005a), 253265.
[2005b] A nice class extracted from Cp -theory, Comment. Math. Univ. Carolinae,
46:3(2005b), 503513.
[2007a] Condensing function spaces into -products of real lines, Houston Journal of
Math., 33:1(2007a), 209228.
[2007b] Twenty questions on metacompactness in function spaces, Open Problems in
Topology II, ed. by E. Pearl, Elsevier B.V., Amsterdam, 2007a, 595598.
[2007c] A selection of recent results and problems in Cp -theory, Topology and Its Appl.,
154:12(2007c), 24652493.
[2009] Condensations of Cp .X/ onto -compact spaces, Appl. Gen. Topology,
10:1(2009), 3948.
TKACHUK , V.V., SHAKHMATOV, D.B.
[1986] When the space Cp .X/ is -countably compact? (in Russian), Vestnik Mosk.
Univ., Math., Mech., 41:1(1986), 7072.
TKACHUK , V.V., YASCHENKO , I.V.
[2001] Almost closed sets and topologies they determine, Comment. Math. Univ.
Carolinae, 42:2(2001), 395405.
TODORCEVIC , S.
[1989] Partition Problems in Topology, Contemporary Mathematics, American Mathe-
matical Society, 84(1989). Providence, Rhode Island, 1989.
[1993] Some applications of S and L combinatorics, Ann. New York Acad. Sci.,
705(1993), 130167.
[2000] Chain-condition methods in topology, Topology Appl. 101:1(2000), 4582.
TYCHONOFF A.N.
[1935] ber einer Funktionenraum, Math. Ann., 111(1935), 762766.
Bibliography 571
USPENSKIJ, V.V.
[1978] On embeddings in functional spaces (in Russian), Doklady Acad. Nauk SSSR,
242:3(1978), 545546.
[1982a] On frequency spectrum of functional spaces (in Russian), Vestnik Mosk. Univ.,
Math., Mech., 37:1(1982a), 3135.
[1982b] A characterization of compactness in terms of the uniform structure in function
space (in Russian), Uspehi Mat. Nauk, 37:4(1982b), 183184.
[1983b] A characterization of realcompactness in terms of the topology of pointwise con-
vergence on the function space, Comment. Math. Univ. Carolinae, 24:1(1983),
121126.
VALOV, V.M.
[1986] Some properties of Cp .X/, Comment. Math. Univ. Carolinae, 27:4(1986),
665672.
[1997a] Function spaces, Topology Appl. 81:1(1997), 122.
[1999] Spaces of bounded functions, Houston J. Math. 25:3(1999), 501521.
VALOV, V., VUMA , D,
[1996] Lindelf degree and function spaces, Papers in honour of Bernhard
Banaschewski, Kluwer Acad. Publ., Dordrecht, 2000, 475483.
[1998] Function spaces and Dieudonn completeness, Quaest. Math. 21:3-4(1998),
303309.
DE LA VEGA , R., KUNEN , K.
[2004] A compact homogeneous S-space, Topology Appl., 136(2004), 123127.
VELICHKO , N.V.
[1981] On weak topology of spaces of continuous functions (in Russian), Matematich.
Zametki, 30:5(1981), 703712.
[1982] Regarding the theory of spaces of continuous functions (in Russian), Uspehi
Matem. Nauk, 37:4(1982), 149150.
[1985] Networks in spaces of mappings (in Russian), Mappings and Extensions of
Topological Spaces, Udmurtia University P.H., Ustinov, 1985, 36.
[1995b] On normality in function spaces, Math. Notes, 56:5-6(1995), 11161124.
[2001] On subspaces of functional spaces, Proc. Steklov Inst. Math. 2(2001), 234240.
[2002] Remarks on Cp -theory, Proc. Steklov Inst. Math., 2(2002), 190192.
VIDOSSICH , G.
[1969a] On topological spaces whose function space is of second category, Invent. Math.,
8:2(1969a), 111113.
[1969b] A remark on the density character of function spaces, Proc. Amer. Math. Soc.,
22(1969b), 618619.
[1970] Characterizing separability of function spaces, Invent. Math., 10:3(1970),
205208.
[1972] On compactness in function spaces, Proc. Amer. Math. Soc., 33(1972), 594598.
W HITE, H.E., JR .
[1978] First countable spaces that have special pseudo-bases, Canadian Mathematical
Bull., 21:1(1978), 103112.
W ILDE, M. DE
[1974] Pointwise compactness in space of functions and R.C. James theorem, Math.
Ann., 208(1974), 3347.
572 Bibliography
YASCHENKO , I.V.
[1992c] On fixed points of mappings of topological spaces, Vestnik Mosk. Univ., Math.,
Mech., 47:5(1992a), 93.
[1992d] Cardinality of discrete families of open sets and one-to-one continuous mappings,
Questions and Answers in General Topology, 2(1992b), 2426.
[1994] On the monotone normality of functional spaces, Moscow University Math. Bull.,
49:3(1994), 6263.
ZENOR , PH .
[1980] Hereditary m-separability and hereditary m-Lindelf property in product spaces
and function spaces, Fund. Math., 106(1980), 175180.
List of special symbols
For every symbol of this list we refer the reader to a place where it was defined.
There could be many such places, but we only mention one here. Note that a symbol
is often defined in the first volume of this book entitled Topology and Function
Spaces; we denote it by TFS. We never use page numbers; instead, we have the
following types of references:
(a) To an introductory part of a section
For example,
exp X

1.1
says that exp X is defined in the Introductory Part of Section 1.1.
Of course,
Cp .X /

TFS-1.1
says that Cp .X / is defined in the Introductory Part of Section 1.1 of the book
TFS.
(b) To a problem
For example,
Cu .X /

TFS-084
says that the expression Cu .X / is defined in Problem 084 of the book TFS.
(c) To a solution
For example,
O.f; K; "/

S.321
says that the definition of O.f; K; "/ can be found in the Solution of Problem
321 of the book TFS.
The expression,
HFD

T.040
says that the definition of HFD can be found in the Solution of Problem 040 of
this volume.
574 List of special symbols
Every problem is short, so it wont be difficult to find a reference in it. An

introductory part is never longer than two pages so, hopefully, it is not hard to find a
reference in it either. Please keep in mind that a solution of a problem can be pretty
long, but its definitions are always given in the beginning.
The symbols are arranged in alphabetical order; this makes it easy to find the
expressions
L B.x; r/ and X , but it is not immediate what to do if we are looking
for t 2T Xt . I hope that the placement of the expressions which start with Greek
letters or mathematical symbols is intuitive enough to be of help to the reader. Even
if it is not, then there are only three pages to plough through. The alphabetic order is
by line and not by column. For example, the first three lines contain symbols which
start with A or something similar and lines 35 are for the expressions beginning
with B, or B.
A./

TFS-1.2 a.X /

TFS-1.5
V
AD.X /

TFS-1.4 A

T.300
AjY

T.092 Bd .x; r/

TFS-1.3
B.x; r/

TFS-1.3 (B1)(B2)

TFS-006
X

TFS-1.3 B.X /

1.4
clX .A/

TFS-1.1 cl .A/

TFS-1.1
C.X /

TFS-1.1 C .X /

TFS-1.1
C.X; Y /

TFS-1.1 Cp .X; Y /

TFS-1.1
Cu .X /

TFS-084 Cp .Y jX /

TFS-1.5
Cp .X /

TFS-1.1 Cp .X /

TFS-1.1
c.X /

TFS-1.2 conv.A/

1.2
CH

1.1 .X /

TFS-1.2
.A; X /

1.2 .x; X /

1.2
D./

TFS-1.2 d.X /

TFS-1.2
dom.f /

1.4 diam.A/

TFS-1.3
D

1.4 F

TFS-1.5
X

TFS-1.2 .X /

TFS-1.2
ij
n .Z/

T.019 n .Z/

T.019
}

1.1 }C

1.1
t 2T ft

TFS-1.5 ext.X /

TFS-1.2
exp X

TFS-1.1 Fin.A/

S.326
f < g

1.1 f g

1.4
List of special symbols 575
fn !
!f

TFS-1.1 f jY

TFS-022
F

TFS-1.3 F

TFS-1.3
' .X / for a card. inv. '

1.1 F

1.4
G

TFS-1.3 G

TFS-1.3
G

TFS-1.5 GCH

1.1
h .U /

S.228
#
HFD

T.040
h'.X / for a card. inv. '

1.1 I

TFS-1.1
Int.A/

TFS-1.1 IntX .A/

TFS-1.1
i w.X /

TFS-1.2 J./

TFS-1.5
K

1.4 K

1.4
l.X /

TFS-1.2 L./

TFS-1.2
lim S for S D fAn W n 2 !g

1.5 (LB1)(LB3)

TFS-007
max.f; g/

TFS-1.1 min.f; g/

TFS-1.1
(MS1)(MS3)

TFS-1.3 m.X /

TFS-1.5
MA./

1.1 MA

1.1
nw.X /

TFS-1.2 N

TFS-1.1
O.f; K; "/

S.321 O.f; x1 ; : : : ; xn ; "/

TFS-1.2
L L
t 2T Xt

TFS-1.4 fXt W t 2 T g

TFS-1.4
p.X /

TFS-1.2 (PO1)(PO3)

TFS-1.4
Q Q Q
pS W t 2T Xt ! t 2S Xt TFS-1.4 pt W s2T Xs ! Xt

TFS-1.2
Q Q
t 2T Xt

TFS-1.2 fXt W t 2 T g

TFS-1.2
Q
t 2T gt

TFS-1.5 0 .X /

TFS-2.4
.A; X /

TFS-1.2 .x; X /

TFS-1.2
.X /

TFS-1.2 .x; X /

TFS-1.4
w.X /

TFS-1.4 .X /

TFS-1.4
Y W Cp .X / ! Cp .Y /

TFS-1.5 P

1.1
P

1.4 Q

TFS-1.1
q.X /

TFS-1.5 R

TFS-1.1
r for r W X ! Y

TFS-1.5 hS i

S.489
s.X /

TFS-1.2 SA

TFS-2.1
576 List of special symbols
St.x; U/

TFS-1.3 St.A; U/

TFS-1.3
./

TFS-1.5 0 .X /

1.4
./

TFS-1.5 ./

TFS-1.5
t0 .X /

TFS-1.5 tm .X /

TFS-1.5
t.X /

TFS-1.2 .x; X/

TFS-1.1
.X/

TFS-1.1 .A; X/

TFS-1.1
.X /

TFS-1.2 .d /

TFS-1.3
S
X

TFS-1.5 ffi W i 2 I g

1.4
! <!

1.4 U1 ^ : : : ^ Un

S.144

!

TFS-370 w.X /

TFS-1.1
x1 ; : : : ; xn I O1 ; : : : ; On
TFS-1.1 X1 : : : Xn

TFS-1.2
X T

TFS-1.2 .Z/!

S.493
.Z/

S.493
Index
A centered family, T.058, T.280

absolute Borel set of class , 1.4, 323, 324, chain, 1.1, 068
373375 character at a point, 1.2, 054, T.092
almost closed set, 1.3 character of a set, T.222, T.487, T.489
almost disjoint family of sets, 1.5, 053 character of a space, 131, 158, 444, T.489
analytic space, 1.4, 334340, 352, 353, 360, clopen set, T.063, T.126, T.203, T.219, T.298,
361, 363, 366, 368371, 385387, 392, T.309, T.371
395, 397, 399, 460, T.132, T.341, T.363 closed discrete set, T.126, T.205, T.219, T.285,
antichain, 1.1 T.316, T.455
Aronszajn tree, 1.1, 068 closed cover, 1.3, 373
Aronszajn coding, 1.1, 068 closed map, 206, 245, 316, 478, 493, 494,
T.246, T.493, T.494
club (closed unbounded subset of an ordinal),
B 1.1, 064, 065, T.069
Baire property, 263265, 406, 441, T.046, cofinal set, 1.1, 097, T.285
T.371 collectionwise normality, 139, 140, 437 041,
Baturovs theorem, 269 045, 058, 060062, 072076, 082083,
Booth lemma, 052 085, 090096, 099, 118120, 128131,
Borel set, 1.4, 322, 330334, 339342, 354, 134, 201204, 211, 220, 225, 233, 237,
368, 372, T.322, T.333, T.339, T.372 241, 248, 254, 259, 260, 262, 273, 279,
287, 294299, 306, 347, 348, 355357,
359, 365, 375, 377, 381, 383389,
C 391396, 398, 413, 454, 462, 468,
caliber, 1.3, 275294, 299, T.281, T.285 482, 495
Cantor set (see also space K), 348, 353, 376, T.041, T.045, T.062, T.073, T.082,
T.250 T.090, T.098, T.131, T.211, T.218,
cardinal function, 1.2, 1.5, 145151, 405 T.222, T.227, T.229, T.235, T.250,
cardinal invariant, 1.1 T.270, T.272, T.298, T.309, T.346,
category of a set in a space, 057, T.351, T.371 T.357, T.372, T.377, T.384, T.385,
ccc property, 1.1 T.391, T.395, T.465, T.468, T.487,
Cech-complete space, 137, 138, 210, 232, 272, T.489, T.493
273, 324, 402, 443, T.210, T.272, T.385, compactification, 1.3, 233, T.224
T.500 compact-valued map, 1.3, 1.4., 240242, 249,
CechStone compactification, 042, 238, 286, 388
371, 386, 387, 403, 439, 491 T.126, compatible elements of a partially ordered set,
T.131, T.244, T.322, T.371, T.385 1.1
C -embedded subspace, T.218, T.455, T.468 complete family of covers, 1.4
578 Index
completely metrizable space, 315, 316, 373, density degree, 037, 095, 177, 188, 216, 405,
419, 493, 494, 498, 499, T.132, T.313, 457, 483, 487, T.285
T.333, T.348, T.357, T.368 depending on a set of coordinates, map, T.109,
completely regular space, T.139 T.268, T.298
condensation, 045, 077079, 102, 341, diagonal of a space, 028, 029, 087, 091, 203,
354359, 392, T.139, T.250, T.357, 235, 290, 293298, 300, 396, T.019,
T.363 T.020, T.021, T.028, T.029, T.030,
connected space, 1.3, T.309, T.312 T.062, T.078, T.081, T.087, T.089,
consistency with ZFC, 1.1, 047 T.098, T.173, T.203, T.294
continuous map, 094, 104, 105, 121123, 133, diagonal number, 028, 029, 091, 178, 179, 180,
148, 150, 201, 206, 243, 245249, 253, 449, T.087, T.173, T.235
254, 277, 305, 315317, 332, 338, 360, diagonal product of maps, T.266
361, 363, 364, 390, 463, 467473, diameter of a set, T.055, T.348, T.358, T.368
479482, 486, 491, T.063, T.069, T.104, diamond principle (}), 1.1, 069, 070, 073,
T.109, T.131, T.132, T.139, T.237, 079, 289
T.245, T.250, T.252, T.266, T.268, Dieudonn complete space, 430
T.294, T.298, T.316, T.318, T.333, disconnected space, 1.3, T.219
T.354, T.368, T.372, T.384, T.459, discrete family of sets, T.132, T.217, T.373
T.468, T.500 discrete space, 424, 443, 486, 492 494,
Continuum Hypothesis (CH), 1.1, 039, 040, 498500, T.371
041, 042, 046, 047, 069, 089, 097-100, discrete subspace, 188, 401, T.007, T.098,
237, 238, 298, 300 T.219, T.500
convex hull, 1.2, 104, T.104 discrete union, T.219, T.250
convex set, 1.2, T.104 domain of a map, 1.4
cosmic space, 107111, 192, 195, 198200, domination by irrationals, T.391
218, 225, 228, 244, 263, 270, 300, 346, double arrow space, 383
363, 364, 395, 451, T.109, T.250, T.270, Dugundji system, 103, T.104
T.300, T.363, T.368
countably additive class (or property), 1.5, 405,
406, 441, 442, 445450 E
countably compact space, 092, 133, 204, embedding 250, 303, 322, 370, 371, 376378,
218, 226, 417, T.126, T.203, T.235, 387, T.019, T.132, T.250, T.298, T.333,
T.391 T.372, T.385
countably paracompact space, 141 extent, 1.3, 1.5
cozero set (see also functionally open set), external base, T.092
T.080, T.252 extremal disconnectedness, T.219
D F
-system, 038 face of a product, T.109, T.110, T.268, T.298,
-system lemma, 038 T.415, T.455, T.500
-root, 038 factorization of a map defined on a subspace of
dense subset of a partially ordered set, 1.1 a product, T.109, T.268
dense subspace, 009, 010, 039, 040, 072, 080, faithfully indexed set, T.089
237, 239, 278, 368, 385387, 420, 431, filter, T.372
495 filter on a partially ordered set, 1.1
dense subspace T.058, T.063, T.074, T.078, finite intersection property (see the entry for
T.080, T.081 centered family)
dense subspace (see also the previous page), finitely additive class (or property or cardinal
T.109, T.187, T.309, T.312, T.333, function), 1.5, 401403, 406, 407
T.349, T.351, T.368, T.406, T.416, finite-to-one map, 1.5, 498, 499, T.498
T.421, T.500 first category set, 057, T.351
dense-in-itself, space, 057, 358, T.045, T.046, first countable space, 099, 202, 329, 401, 408,
T.219, T.272, T.358 415, T.045, T.205, T.351
Index 579
Fodors lemma, 067 hereditarily weakly Whyburn space, 215, 220

free sequence, 198, T.198 hereditary cardinal function, 146, 147
free union (see discrete union) hereditary density, 004, 008, 012, 014, 017,
FrchetUrysihn space, 1.5, 120, 131, 018, 020, 021, 024, 030, 032, 036, 039,
134136, 186, 204, 205, 210, 214, 384, 040, 043, 059, 169, 171, 172, 173, 174,
398, 402, 450, 464466, T.045 405, 428, 458, T.020, T.029, T.030,
F -set, 1.4, 323, 379, 420, T.422 T.036, T.166, T.173
hereditary Lindelf number, 001, 005, 007,
011, 014, 015, 017, 018, 020, 021, 023,
G 029, 033, 035037, 039, 040, 043, 059,
Generalized Continuum Hypothesis (GCH), 166, 167171, 190, 193, 194, 405, 429,
1.1 458, T.021, T.029, T.030, T.036, T.166,
generating topology by a base, 1.3 T.173, T.368, T.490
generating topology by a closure operator, 1.3 hereditary property, 420, 421, 433, T.455
generating topology by a family of local bases, Hewitt realcompactification, 145
1.3 HFD space, T.040
generating topology by a family of maps, 1.3 homeomorphic spaces, 313, 330, 347353,
generating topology by a linear order, 072 386, 424, 432, 500, T.132, T.217, T.219,
generating topology by a metric 419 T.250, T.313, T.322, T.348, T.351,
generating topology by a subbase, 1.3, T.363 T.363, T.415, T.457
generating topology by an interior operator, 1.3 homeomorphism, T.217, T.333, T.349
Gerlits property, 1.5, 463, 464 homeomorphism (see also previous column),
GerlitsPytkeev theorem, 465 T.357, T.371, T.372, T.421, T.498,
G -diagonal (or diagonal G ), 028, 029, 087, T.500
235, 300, T.087, T.235 homogeneous space, 1.5, T.371
G -set, 055, 091, T.041, T.062, T.090, T.333, Hurewicz space, 1.2, 1.3, 132, 188, 216
T.500, T.429 Hurewicz space (see also the previous page),
G -set, 1.2, 001, T.078 217, T.132, T.188
H I
Hausdorff space, T.098, T.372 identity map, T.357
hedgehog space, 019, 020, 021 induced topology, 072, T.357
height of a tree, 1.1 invariance under operation, 1.3, 254, T.250
height of an element of a tree, 1.1 irrationals, 4, 313, 317, 328, 329, 341, 347,
hereditarily analytic space, 400 352, 358, 359, 365, 367, 370, 371, 388,
hereditarily Cech-complete space, 272, 273 390, T.132, T.346, T.395
hereditarily irresolvable space, T.219 irreducible map, 492, T.246, T.493, T.494
hereditarilyK-analytic space, 400 irresolvable space, T.219
hereditarily k-space, 214 isomorphism of linearly ordered sets, 072
hereditarily Lindelf space, 001, 005, 007, i -weight, 177, 178, 244, 425, 451, 471
010, 011, 014, 015, 039, 074, 076, 085,
086, 099, 100, 190, 198, T.073, T.270
hereditarily normal space, 002, 090, 142, 201, J
202, T.090 Jensens axiom (see diamond principle (}))
hereditarily p-space, 272
hereditarily realcompact space, 404, 425, 451
hereditarily separable space, 1.1, 004, 008, K
012, 014, 040, 060, 073, 077079, 082, K-analytic space, 1.4, 343346, 388, 389, 390,
088, 089, 098, 198, 237, T.040, T.082, 391, 393395, 397400, 460, T.250,
T.099 T.346, T.391
hereditarily sequential space, 214 -Aronszajn tree, 1.1
hereditarily stable space, 200 -modification of a space, 1.2, 128
580 Index
-monolithic space, 1.2, 113, 114, 116, 117, locally compact space, 1.3, 098, 132, 434,
120122, 144, 152, 154, 157, 190192, T.203, T.223, T.357, T.434
197, 199, 296, 426, 468, 473 locally convex space, 1.2, 104, T.105, T.131
-scattered space, 1.2, 133, 187, 477 locally finite family, 103, 115, T.104, T.244,
-simple space, 1.2, 127, 129, 130, 157 T.354
-small diagonal, 1.3, 290, 293, 298, 300, locally pseudocompact space, 435
T.294, T.298, T.300 lower semicontinuous map, 1.4, 315
-Souslin tree, 1.1, 070, 071, 073, 074, T.073 L-space, 1.1, 039, 059, 074, 099
-stable space, 1.2, 106, 108112, 118, Luzin space, 1.1, 043046, 063, T.046, T.063
123126, 143, 152154, 156, 192, 195,
200, 266268, 478, T.112, T.237
K -space, 1.3, 250, 261, 262, 362, 367, M
T.250, T.262, T.377 Martins Axiom (MA), 1.1, 047, 048063, 071,
k-space, 1.2, 131, 210, 214, 230, 402, 465, 083, 088, 099, 140, 197200, 288, 382,
466, T.131, T.210 395 T.050, T.063, T.395
-tree (for a regular cardinal ), 1.1, 068, 070, maximal almost disjoint family, 053
071, 074, T.073 maximalelement of a partially ordered set,
Kowalsky hedgehog J./, 019021, T.019 T.074
maximal family with a property P , T.058
maximal space, T.219
measurable map, T.363, T.368, T.384
L
metacompact space, 1.5, 437440
large inductive dimension, 1.4, 308
metric space, T.055, T.105, T.313, T.333,
Lavrentieff theorem, T.333
T.348, T.368, T.373
left-separated space, 1.1, 004, 007, 009, 037,
metrizable space 062, 083, 090094, 099,
T.004, T.078
101106, 203, 217, 219, 221, 229, 239,
lexicographic order, 1.4
272, 285, 295299, 315, 316, 348350,
limit of a sequence, 054, 379, 389, T.055,
357, 359, 373, 374, 392, 395, 396, 401,
T.131, T.246, T.316, T.384, T.493
402, 412414, 416419, 446, 451, 455,
limit of a family of sets, 1.5
493499 T.062, T.132, T.203, T.235,
limit of a transfinite sequence, 1.3, T.298
T.285, T.300, T.357, T.385
Lindelf number (degree), 001, 128, 240, 269,
monolithic space, 1.2, 107, 115117, 120, 122,
405, 456 T.490
152, 154, 155, 266, 297
Lindelf p-space, 1.3, 223, 231, 244246, 250,
Mrowka space, 407, T.130
252, 253, 255, 260, 261, 271
multiplicative class of Borel subsets of a space,
Lindelf property (see also Lindelf space),
1.4
1.2
multiplicative class of absolute Borel sets, 1.4
Lindelf -space, 1.3, 223228, 230, 231,
233, 234, 236, 239, 242, 243, 248, 249,
253, 254, 256259, 261, 263, 265270,
300, 459, T.227, T.237, T.270, T.399 N
Lindelf space, 076, 089, 098, 112, 127, 128, natural projection, T.109, T.110, T.250, T.298,
129, 135137, 189, 199, 234, 264, 294, T.455, T.500
306, 422, 438, 452, 454, T.112, T.217, network, 235
T.223, T.268 network weight, 096, 107111, 192, 197, 199,
linear homeomorphism, T.217 200, 218, 225, 228, 244, 260, 263 270,
linear map, T.132 300, 346, 363, 364, 395, 405, 451, 457,
linear topological space, 1.2, 104, T.104, 470, 488, T.109, T.250, T.270, T.300
T.105, T.131 network with respect to a cover, 1.3, T.229
linearly ordered space, 072076, T.073, normal space, 002, 100, 139141, 234, 308,
T.075 309, 311, 407, 438, 453455, T.201,
linearly homeomorphic vector spaces, T.203, T.217, T.245, T.311, T.372
T.217 nowhere dense set, 057, 058, T.039, T.042,
local base, T.416 T.045, T.089, T.219, T.351
Index 581
O Pressing-Down Lemma, 067

!-cover, 1.5, T.188, T.217, T.464 product space, 050, 109, 114, 117, 254, 255,
!-monolithic space, 1.2, 116, 117, 120, 122, 256, 268, 280282, 302, 333, 335, 343,
190192, 197, 199, 296, 468, T.468 493, 494499, T.050, T.089, T.109,
open mapping, 326, 415, 476, 477, 498, 499, T.110, T.112, T.132, T.250, T.266,
T.110 T.268, T.298, T.363, T.415, T.455,
open-separated sets in a topological space, T.500
T.309 pseudocompact space, 1.4, 093, 119, 130, 131,
operator on a family of subsets of a set, 1.3 138, 205, 435, 452, 495497, T.132,
operator on a set, 048, 049, 068, 072, T.004, T.205, T.391, T.465, T.494
T.005, T.074 pseudocomplete space, 1.5, T.498
ordinal space, 064, 065069, 273, 491, T.069, pseudocharacter of a space, 179, 236, 401, 409,
T.211 448
oscillation of a function, T.368, T.384 pseudometric, 1.4
!-simple space, 1.2, 127 pseudo-open map, 1.3, 251
!-stable space, 1.2, 106, 112, 119, 125, 126, pseudoradial space, 1.3, 211, 212
145, 192, 195, 267, 268, T.112, T.237 p-space, 1.3, 221, 222224, 230232
outer base, 396 p-space (continued from the previous page),
244247, 250, 251, 252, 253, 255, 260,
271, 272, T.222, T.223, T.224, T.245
P P -space, 1.1, 1.2, 112, 127, 135, 137, T.112
paracompact space, 203, 217, 314, 315, 422, -weight, 131, 406, 442, T.089, T.158, T.187
437, 452, T.104, T.203, T.217, T.244,
T.245, T.246, T.422
partially ordered set, 048, 049, 068, T.058 Q
-character, 158, 402, 442, T.131, T.158, quotient image, 149, 251, 274, 275
T.298
P-directed family of sets, 1.4
P-dominated space, T.391 R
P ./-monolithic space for a property P , 1.2, radial space, 1.3, 209, T.211
139, 140, 146151, 158186, 189, 190, rational numbers, the space of, 349351, 356,
193, 194, 427429, 475 T.309, T.349, T.351
perfect image (or preimage), 1.2, 243, 245, real line, 342, 355, 380, 382, T.349
249, 252, 304, 325, 390, 487492, realcompact space, 391, 404, 423, 425, 430,
T.245, T.489, T.490, T.492 436, 451, T.436, T.492
perfect map, 326, T.266 resolvable space, T.219
perfect space, 001, T.331, T.363, T.368 restriction map, T.080, T.217, T.218, T.368,
perfectly disconnected space, T.219 T.455, T.500
perfectly normal space, 003, 061, 075, 080, retract, 1.2, 123, 316, 500, T.132, T.217, T.500
081087, 089, 095, 096, 142, 202, retraction, 1.2
T.080, T.081, T.087 right-separated space (this coincides with the
point-countable family of sets, T.203 concept of scattered space), 1.1, 005,
point-finite cellularity, 175, 284, 405, 485, 006, 008, T.005
T.491 Rosenthal compact space, 1.4, 383385, 387,
pointwise bounded subset of function space, T.385
T.384 R-quotient map (or image), 147, 149, 150, 181,
pointwise countable type, T.222 183, 184, 469, T.139, T.268
Polish space, 1.4, 319, 320, 321, 325, 326330,
338340, 347, 351, 358, 365, T.322,
T.333, T.339, T.357, T.358, T.384, S
T.385 SA axiom, 1.1, 036, 086, 193, 195, 196
P -point, 1.1, 042 scattered space, 1.1, 006, 098, 099, 128130,
precaliber, 1.3, 275280, 283, 284, 286, 288, 133, 134, 136, 213, 272, 273, T.130,
289, T.050, T.280 T.272
582 Index
-compact space, 132, 216, 226, 274, 323, space D (see also Cantor cube), 039, 040,
351, 352, 354, 355, 362, 364, 366, 367, 089, 303, T.040, T.298, T.372
368, 466, 482, T.132, T.203, T.372, space I, 133, 143, 144, 305, T.436, T.500
T.395 space I , 354, 369, T.250, T.298
-countably compact space, 132 space J./ (see Kowalsky hedgehog)
-discrete family, 373, T.229, T.235 space K (see also Cantor set), 1.4, 318, 376,
-discrete network, 228, 235, T.229 378, T.250, T.348
-disjoint base, T.412 space L./, 227, 402, 440, T.220
second category set, T.371 space P (see irrationals)
second countable space, 046, 055, 057, 080, space Q (see rational numbers)
084, 102, 107, 109111, 229, 248, 249, space R (see real line)
252, 254, 271, 306, 318, 320, 321, space R , 117, 330, 354, 360, 361, 367, 371,
323, 324, 329, 332, 333, 341, 351, 381, 382, 424, 492, 494, 500, T.019,
362365, 367, 368, 376379, 403, 411, T.132, T.217, T.312, T.372, T.379,
455, T.055, T.062, T.063, T.080, T.089, T.384, T.385, T.399, T.434, T.455,
T.092, T.131, T.132, T.220, T.250, T.500
T.298, T.320, T.322, T.341, T.351, space .A/, 1.2, T.312
T.354, T.363, T.368, T.377, T.379, space .A/, 1.2
T.384, T.455 space !1 , 064069, 491, T.069
separable space, 039, 044, 046, 073076, 081, space !1 C 1, 273, T.211
084, 088, 089, 217, 282, 287, 385, 397, spread, 007, 008, 013, 014, 016, 019, 022, 028,
412, 493, 495, T.045, T.073, T.074, 031, 034, 036, 037, 039, 040, 060062,
T.081, T.082, T.087, T.089 078, 079, 160, 161, 163, 164, 189, 190,
separable metrizable space (in this book this is 191, 192, 193, 195, 196, 197, 199, 405,
the same as second countable space), 427, 458, T.007, T.015, T.019, T.028,
106, 219, 493, 497499 T.036, T.078, T.082, T.087, T.098,
separating points by a family of maps, T.354 T.160
separating subsets by a family of sets, 1.3, -product, 1.2, T.110, T.268
233 -product, 1.2, T.110, T.268
separating subsets by disjoint open (or Borel) -pseudocompact space, 132, 435
sets, T.309, T.339 -space, 228, 235
separation axioms, T.098 -space, 1.3, 221, 223, 224, 226, 228, 229,
sequential space, 041, 131, 210, 211, 214, 402, 234236, 238, T.223, T.229
465, 466, T.041, T.316 S-space, 1.1, 040, 059, 098, 099
sequentially compact space, T.384 stable space, 1.2, 108111, 118, 124, 152154,
Shanin condition, 282 156, 200, 268
simple space, 1.2, 129, 130 stationary set, 1.1, 065, 066, 067, 069
-locally compact space, 132 strictly weaker topology, 072
small diagonal (see also -small diagonal), 1.3, strong L-space, 1.1, 059, 099
290, 293, 294, 295, 296298, T.298, strong S-space, 1.1, 059, 098, 099
T.300 strong -space, T.223, T.229
Sorgenfrey line, 227 strongly dense subspace, 1.5
Souslin line, T.074 strongly -monolithic space, 1.2, 135, 157, 158
Souslin number, 075, 275, 405, 485, T.050, strongly monolithic space, 1.2, 101, 136
T.089 strongly zero-dimensional space, 1.4, 306312,
Souslin property, 050, 058, 288, 289, T.039, 314316, T.306, T.311
T.046, T.050, T.073 subbase of a topology, 1.1
Souslin tree (see -Souslin tree) submaximal space, 1.3, 208, T.219
space A./, T.203, T.219, T.223 submetrizable space, 1.4, 392
space !, 042, 238, 371, 386, 387, 403, 439, subparacompact space, T.223, T.235
491, T.042, T.131, T.322, T.385 subtree, 1.1
space D./ (see discrete space) supremum of a family of topologies, T.139
Index 583
T upper semicontinuous map, 1.3, 240242, 249,

Talagrand space, 234 388, T.346
tightness of a space, 120, 176, 183185, 207,
234, 287, 296, 410, 447, 456, 461, 462,
463, T.041, T.173, T.298, T.384 V
topology of uniform convergence (this is the Vietoris topology, T.372
same as uniform convergence topology) Velichkos theorem, 030
tree, 1.1, 068, 070, 071, 074, T.073
T0 -separating family, T.270
W
T1 -separating family, T.203, T.270
weak functional tightness, 181, 182, 484
T1 -space, T.219
weaker topology, 072
T2 -space (see Hausdorff space)
weakly Whyburn space, 1.3, 206, 209, 211,
T3 1 -space (see Tychonoff space)
2 213, 215, 219, 220, T.220
T4 -space (in this book this is the same as weight, 091, 094, 096, 102, 105, 131, 187,
normal space) 244, 295, 303, 304, 403, 405, 411,
two arrows space (see double arrow space) 445, T.039, T.046, T.089, T.092, T.102,
Tychonoff cube (see space I ) T.105, T.109, T.158, T.187, T.250,
Tychonoff space T.098, T.139, T.205, T.219, T.268, T.285, T.322, T.357, T.372,
T.245, T.363, T.372 T.412, T.487, T.489, T.490
Whyburn space, 1.3, 204, 205212, 216219,
T.217, T.219
U
ultradisconnected space, T.219
ultrafilter, T.058, T.371 Z
uniform convergence topology, 431, T.357, zero-set (see also functionally closed set), 1.5,
T.379, T.421 T.080, T.252
uniformly dense set, 1.5, 456461, T.457, zero-dimensional space, 1.1, 034, 035,
T.459 301307, 309, 312, 313, 347, 348,
uniformly discrete set in a metric space, T.063, T.132, T.205, T.298, T.306,
T.373 T.309, T.311

(Problem Books in Mathematics) Vladimir V. Tkachuk (Auth.) - A Cp-Theory Problem Book - Special Features of Function Spaces-Springer International Publishing (2014) PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

(Problem Books in Mathematics) Vladimir V. Tkachuk (Auth.) - A Cp-Theory Problem Book - Special Features of Function Spaces-Springer International Publishing (2014) PDF

Transféré par

Droits d'auteur :

Formats disponibles

Problem Books in Mathematics

For further volumes:

A Cp-Theory Problem Book

Mathematics Subject Classification (2010): 54C35

Springer International Publishing Switzerland 2014

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

This is the second volume of the series of books of problems in Cp -theory

The author wants to emphasize that if a postgraduate student mastered the

Mexico City, Mexico Vladimir V. Tkachuk

1 Duality Theorems and Properties of Function Spaces . . . . . . . . . . . . . . . . . . . 1

List of Special Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 573

1.1. Tampering with Additional Axioms: Some Hereditary Properties

Versions of !-monolithity in Cp .X / implying that X is finite . . . . . . . Problem 132.

Absolute F - and absolute G -spaces . . . . . . . . . . . . . . . . . . . . . . . Problems 323324.

Sometimes, as we formulate a problem, we use without reference definitions and

V.V. Tkachuk, A Cp-Theory Problem Book: Special Features of Function Spaces, 1

1.1 Some Additional Axioms and Hereditary Properties

A space X is left-separated (right-separated) if there exists a well-order < on

043. Let X be a Luzin space. Prove that hl.X / D ! and hd.X / !1 .

1.2 Monolithity, Stability and Their Generalizations

As usual, all spaces are assumed to be Tychonoff. Given a cardinal , a space X

tx C .1 t/y 2 A for any number t 2 0; 1 . The convex hull conv.A/ of a subset A

101. Prove that any metrizable space is strongly monolithic.

(ii) A is a FrchetUrysohn space for any countable A Cp .X /;

149. Suppose that  and are cardinal functions such that

183. Given an infinite cardinal , prove that X is l ./-monolithic if and only if

1.3 Whyburn Spaces, Calibers and Lindelf -Property

space X if for any indexed family U D fU W < g .X /, there exists A

224. Prove that

(ii) if an infinite successor cardinal is a precaliber of a space X , then

293. Let be an uncountable regular cardinal. Prove that if is a caliber of Cp .X /,

1.4 A Glimpse of Descriptive Set Theory

All spaces are assumed to be Tychonoff. A space X is called zero-dimensional if it

A space is called Polish if it is second countable and completely metrizable.

(ii) any open subspace of an analytic space is an analytic space;

388. Prove that a space X is K-analytic if and only if X is an image of P under a

1.5 Additivity of Properties: Mappings Between

All spaces are assumed to be Tychonoff. A topological property P will be called

419. For an arbitrary space X and any f; g 2 C .X /, let

d.f; g/ D supfjf .x/ g.x/j W x 2 X g:

Prove that d is a complete metric on the set C .X / and the topology,

458. For an arbitrary space X and a uniformly dense Y Cp .X / prove that

1.6 Bibliographic Notes

The main text of Chap. 1 consists of problems of the following types:

V.V. Tkachuk, A Cp-Theory Problem Book: Special Features of Function Spaces, 45

volume by [TFS], so a reference Problem 225 of [TFS] says that we refer to

x 2 Z, so Z is left-separated by . Fact 1 is proved. t

so Z is right-separated by . Fact 1 is proved. t

Proof. Take any T Y which is right-separated by an order < and jT j D . There

Solution. If X is a right-separated space, take any well-order < on X which

Given a space Z we denote by IP .Z/ the set of isolated points of Z. It is clear

Observe that . / implies that the set D D fz W < g is discrete because

As a consequence, Y nLz U and therefore Uz [ U U is a subcover of Y of

cardinality which is a contradiction. This shows that H D ;, so our solution is

is a neighborhood of x with W \ Ly D ; so x Ly . The point x 2 Y nLy has

T.014. Suppose that s.X X / . Prove that hl.X / or hd.X / .

x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng

is called a standard open subset of Cp .X /. Standard open sets x1 ; : : : ; xn I O1 ; : : : ;

We claim that the well-order right-separates F ; to see this, take any f 2 F .

Therefore, for each f 2 F , there exists a standard open set Of 2 .f; Cp .X //

xg O.f /. Thus we have proved that

x1 ; : : : ; xn I O1 ; : : : ; On D ff 2 Cp .X / W f .xi / 2 Oi for all i ng

149. Suppose that and are cardinal functions such that

Proof. Assume that s.Cp .Z// D ; if we are given a point z D .z1 ; : : : ; zn / 2 Z n ,

For any < C and i 2 Mn take a function fi 2 C.Z; 0; 1 / such that

(**) hd.X n nn / for any n 2 N with n 2.

-many closed sets and therefore .X / .

have hl.F / by Fact 2. It follows from jF j and F D X nn .X / that

Since hl.X n nn / by . /, we have hl.X n / so our inductive step is

2 !1 with < for all 2 A. The function s W fg ! D defined by s./ D 1

Proof. The diagonal is closed in Z Z; if it is a G -subset of Z Z, then, by