Vous êtes sur la page 1sur 18

8

5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE?
ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Abstract. We continue investigations of forcing notions with strong ccc prop-
erties introducing new methods of building sweet forcing notions. We also
show that quotients of topologically sweet forcing notions over Cohen reals are
topologically sweet while the quotients over random reals do not have to be
such.
0. Introduction
One of the main ingredients of the construction of the model for all projective sets
of reals have the Baire property presented in Shelah [8, 7] was a strong ccc property
of forcing notions called sweetness. This property is preserved in amalgamations and
also in compositions with the Hechler forcing notion D and the Universal Meager
forcing UM (see [8, 7]; a full explanation of how this is applied can be found in
[3]). Stern [10] considered a slightly weaker property, topological sweetness, which
is also preserved in amalgamations and compositions with D and UM. We further
investigated the sweet properties of forcing notions in [5, 4], where we introduced
a new property called iterable sweetness and we showed how one can build sweet
forcing notions. New examples of iterably sweet forcing notions can be used in
constructions like [8, 7], [7], but it could be that there is no need for this the
old forcing notions could be adding generic objects for all of them. In [6] we proved
that this is exactly what happens with the natural examples of sweet forcing notions
determined by the universality parameters as in [5, 2.3]: a sequence Cohen real
dominating real Cohen real produces generic lters for many of them.
In the present paper we show that sweetness is not so rare after all and we give
more constructions of sweet forcing notions. In the rst section we present a new
method of building sweet forcing notions and we give our rst example: a forcing
notion Q
T
sc
associated with scattered subtrees of 2
<
. We do not know if the
iterations of old forcing notions add generic objects for Q
T
sc
, but in Proposition
1.8 we present an indication that this does not happen. In the second section we
use our method to introduce two large families of sweet forcing notions, in some
sense generalizing the known examples from [5]. This time we manage to show
that some of our forcing notions are really new by showing that we have too many
dierent examples (in Theorems 2.10, 2.15).
In the last section of the paper we investigate the preservation of topological
sweetness under some operations. We note that a complete subforcing of a topolog-
ically sweet separable partial order is equivalent to a topologically sweet forcing (in
Date: October 2005.
1991 Mathematics Subject Classication. 03E40.
Both authors acknowledge support from the United States-Israel Binational Science Foundation
(Grant no. 2002323). This is publication 856 of the second author.
1
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


2 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Proposition 3.5). We also show that the quotient of a topologically sweet forcing
notion by a Cohen subforcing is topologically sweet (Theorem 3.7), but quotients
by random real do not have to be topologically sweet (Corollary 3.10).
0.1. Notation. Our notation is rather standard and compatible with that of clas-
sical textbooks (like Jech [2] or Bartoszy nski and Judah [1]). In forcing we keep the
older convention that a stronger condition is the larger one. Our main conventions
are listed below.
(1) For a forcing notion P,
P
stands for the canonical Pname for the generic
lter in P. Other Pnames for objects in the extension via P will be denoted
with a tilde below (e.g.,

, X

).
The weakest element of P will be denoted by
P
(and we will always assume
that there is one and that there is no other condition equivalent to it).
(2) The complete Boolean algebra determined by a forcing notion P is denoted
by BA(P). For a complete Boolean algebra B, B
+
is B 0
B
treated as
a forcing notion (so the order is the reverse Boolean order). Also, for a
formula , the Boolean value (with respect to B) of will be denoted by
[[]]
B
.
(3) Ordinal numbers will be denoted be the lower case initial letters of the Greek
alphabet (, , , . . .) and also by i, j (with possible sub- and superscripts).
Cardinal numbers will be called , , .
(4) For two sequences , we write whenever is a proper initial segment
of , and when either or = . The length of a sequence is
denoted by lh().
(5) The quantier (

n) is an abbreviation for (m )(n > m).


(6) The Cantor space 2

and the Baire space

are the spaces of all functions


from to 2, , respectively, equipped with the natural (Polish) topology.
0.2. Background on sweetness. Let us recall basic denitions related to sweet
forcing notions.
Denition 0.1 (Shelah [8, Def. 7.2]). A pair (P,

E) is model of sweetness whenever:
(i) P is a forcing notion,
(ii)

E = E
n
: n < ), each E
n
is an equivalence relation on P such that P/E
n
is countable,
(iii) equivalence classes of each E
n
are
P
directed, E
n+1
E
n
,
(iv) if p
i
: i P, p
i
E
i
p

(for i ), then
(n )(q p

)(q E
n
p

& (i n)(p
i
q)),
(v) if p, q P, p q and n , then there is k such that
(p

[p]
E
k
)(q

[q]
En
)(p

).
If there is a model of sweetness based on P, then we say that P is sweet.
Denition 0.2 (Stern [10, Def. 1.2]). A model of topological sweetness is a pair
M = (P, B) such that P = (P, ) is a forcing notion, B is a countable basis of a
topology on P and
(i)
P
is an isolated point in ,
(ii) if a sequence p
n
: n < ) P is converging to p P, q p and W is a
neighbourhood of q, then there is a condition r P such that
(a) r W, r q,
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 3
(b) the set n : p
n
r is innite.
If there is a model of topological sweetness (P, B), then the forcing notion P is
topologically sweet.
Lemma 0.3 (See [5, Lemma 4.2.3]). Assume that (P, B) is a model of topological
sweetness.
(1) If p, q P, p q and q U B, then there is an open neighbourhood V of
p such that
(r V )(r

U)(r r

).
(2) If m , p U B, then there is an open neighbourhood V of p such that
any p
0
, . . . , p
m
V have a common upper bound in U.
Denition 0.4 (See [5, Def. 4.2.1]). Let B be a countable basis of a topology on
a forcing notion Q. We say that (Q, B) is a model of iterable sweetness if
(i) B is closed under nite intersections,
(ii) each U B is directed and p q U p U,
(iii) if p
n
: n ) U and the sequence p
n
: n < ) converges to p

(in
the topology generated by B), then there is a condition p U such that
(n )(p
n
p).
Proposition 0.5 (See [5, Proposition 4.2.2]). If P is a sweet forcing notion in
which any two compatible conditions have a least upper bound, then P is iterably
sweet.
1. swclosed families and scattered trees
In this section we present a new method of building sweet forcing notions. This
method is, essentially, a generalization of that determined by the universality pa-
rameters of [5, 2.3].
Denition 1.1. (1) A tree is a family T of nite sequences such that for some
root(T) T we have
( T)(root(T) ) and root(T) T T.
Elements of a tree T may be called nodes of T.
(2) If is a node in the tree T then
succ
T
() = T : & lh() = lh() + 1 and
T
[]
= T : .
(3) For a tree T, the family of all branches through T is denoted by [T], and
we let
max(T)
def
= T : there is no T such that
and
split(T)
def
= T : [succ
T
()[ 2.
(4) A tree T is normal if max(T) = and root(T) = ).
Denition 1.2. Suppose that T is a family of normal subtrees of
<
. We say
that T is swclosed
1
whenever
1
sw comes from sweet. The closure property introduced here is instrumental for the sweetness
of the associated forcing notion Q
T
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


4 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(1) if T
1
T , T
2
T
1
and T
2
is a normal tree, then T
2
T ,
(2) if T
1
, T
2
T , then T
1
T
2
T , and
(3) if T
n
: n ) T is such that
_
n <
__
T


n
= T
n

n
_
,
then

n
T
n
T .
Denition 1.3. For a family T of normal subtrees of
<
we dene a forcing
notion Q
T
as follows.
A condition in Q
T
is a pair p = (N
p
, T
p
) such that N
p
< and T
p
T .
The order
Q
T of Q
T
is given by
p
Q
T q if and only if
N
p
N
q
, T
p
T
q
and T
q

N
p
= T
p

N
p
.
The relation between the forcing Q
T
and the family T is similar to that in the
case of the Universal Meager forcing notion UM and nowhere dense subtrees of
2
<
. Note that Q
T
does not have to be ccc in general, however in many natural
cases it is.
Proposition 1.4. Assume that T is an swclosed family of normal subtrees of
<
such that every T T is nitely branching. Then Q
T
is a sweet forcing notion in
which any two compatible conditions have a least upper bound (and consequently
Q
T
is iterably sweet).
Proof. One easily veries that Q
T
is indeed a forcing notion and that any two
compatible conditions in Q
T
have a least upper bound.
For an integer n < let E
n
be a binary relation on Q
T
dened by
q E
n
p if and only if
N
q
= N
p
and T
q

N
q
+n
= T
p

N
q
+n
,
and let

E = E
n
: n < ). We claim that (Q
T
,

E) is a model of sweetness.
Conditions 0.1(iiii) should be clear. To verify 0.1(iv) suppose that p
i
Q
T
for
n i are such that p
i
E
i
p

(for i < ). Thus, for n i < , N


pi
= N
p
and
T
pi

N
pi
+i
= T
p

N
p
+i
.
Put N = N
p
and T =

T
pi
: n i . It follows from 1.2(3) that T T ,
and plainly q = (N, T) Q
T
, q E
n
p

and (i n)(p
i
q), nishing justication
of 0.1(iv).
Finally, to check 0.1(v) suppose that p, q Q
T
, p q and n < . Let k = N
q
+n.
It should be clear that (p

[p]
E
k
)(q

[q]
En
)(p

).
Now we are going to present our rst example of an swclosed family: the family
of scattered subtrees of 2
<
.
Denition 1.5. (1) For a closed set A 2

, let rk(A) be the CantorBendixson


rank of A, that is
rk(A) = min <
1
: A

= A
+1
,
where A

denotes the
th
CantorBendixson derivative of A.
(2) We say that a tree T 2
<
is scattered if it is normal and [T] is countable.
The family of all scattered subtrees of 2
<
will be denoted by T
sc
.
(3) For a scattered tree T 2
<
, let g
T
: [T] rk([T]) and h
T
: [T]
be such that for each [T] we have
g
T
() = min < rk(T) : / [T]
+1

8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 5
and
h
T
() = minm < :
_
[T]
__
m = m ( = g
T
() < g
T
()
_
.
Proposition 1.6. Let T 2
<
be a normal tree. Then T is scattered if and only
if there is a mapping : T
1
such that
()
0
,T
_
, T
__
() ()
_
, and
()
1
,T
_
split(T)
__
(

0)) < () (

1)) < ()
_
.
Proof. It should be clear that if there is a function : T
1
such that ()
0
,T
+
()
1
,T
holds true, then the tree T contains no perfect subtree and hence T is
scattered.
We will show the converse implication by induction on rk([T]).
Suppose that T is a scattered tree. Let S = h
T
() : [T]. Clearly, for
every [T] there is n < such that n S, so the set
F
def
=
_
S : ()( / S)
_
is a front of T. Choose

: < n [T], n < , such that F =

h
T
(

) : < n
and let
A
def
=
_
T :
_
< n
__
h
T
(

) < lh() & (lh() 1)

&

__
.
Note that if < n, [T]

and h
T
(

) =

h
T
(

), then g
T
() < g
T
(

).
Hence
_
A
__
rk([T
[]
]) < rk([T])
_
, so by the inductive hypothesis for each A
we may choose

: T
[]

1
such that ()
0
,T
[]
+ ()
1
,T
[]
holds true. Put

= sup

() : A <
1
, k

= maxh
T
(

) : < n + 1 and let : T


1
be dened by
() =
_

+k

lh() if no initial segment of belongs to F, and

+ 1 if an initial segment of belongs to F


but no initial segment of belongs to A, and

() if A and .
One easily veries that the function (is well dened and) satises ()
0
,T
+()
1
,T
.

Proposition 1.7. T
sc
is an swclosed family and consequently Q
T
sc
is iterably
sweet.
Proof. Plainly T
sc
satises the conditions (1) and (2) of 1.2.
To verify 1.2(3) suppose that T
n
: n ) T
sc
is a sequence of scattered trees
such that
_
n <
__
T

2
n
= T
n
2
n
_
. Let T =

n
T
n
. We are going to
show that T is a scattered tree, and for this we have to show that [T] is countable.
Note that if n < , 2
<
T

and lh() n, then / T


n
. Therefore, if
2
<
T

then [] [T]

[T
n
] : n < lh(), so [] [T] is countable. Hence
[T] [T

] is countable and thus (since [T

] is countable) so is [T].
The consequently part follows from 1.4 (remember that members of T
sc
are
subtrees of 2
<
so nitely branching).
Recall that a forcing notion P has
1
caliber if for every uncountable family
T P there is a condition p P such that

_
q T : q p
_

=
1
(see Truss [11]).
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


6 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Proposition 1.8. (1) If a forcing notion P has
1
caliber, then in V
P
there
is no tree T 2
<
such that
(a) for every <
1
there is a countable closed set A 2

coded in V
such that rk(A) = and A [T], and
(b) T includes no perfect subtree from V.
Consequently, P does not add generic object for Q
T
sc
.
(2) If b >
1
, then neither the Hechler forcing notion D nor its composition
D C with the Cohen real forcing add generic objects for Q
T
sc
.
Proof. (1) Suppose towards contradiction that P has an
1
caliber, p P and T

is a Pname for a subtree of 2


<
such that the condition p forces that both (a)
and (b) of 1.8(1) hold true for T

. Then for each <


1
we may choose a scattered
tree T

2
<
and a condition p

P such that (T

V and)
rk([T

]) = and p p

and p


P
T

.
Since P has an
1
caliber we nd a condition p

P such that the set


Y
def
= <
1
: p

is uncountable. Put T

Y
T

. Clearly T

is a non-scattered tree and (T

V
and) p

, contradicting (b).
Concerning the consequently part it is enough to note that if T

sc
is the canon-
ical Q
T
sc
name for a subset of 2
<
such that

Q
T
sc
T

sc
=
_
_
T
p
: p
Q
T
sc
_
,
then T

sc
is a tree satisfying 1.8(1)(a,b) .
(2) If the unbounded number b is greater than
1
, then both D and D C have
the
1
caliber, so part (1) applies.
Remark 1.9. The forcing notion Q
T
sc
is somewhat similar to the universal forcing
notions discussed in [5, 2.3] and [6]. However it follows from 1.8(2) that if MA
holds true, then the composition CDC does not add generic real for Q
T
sc
. This is
somewhat opposite to the result presented in [6, Theorem 2.1] and it may indicate
that the answer to the following question is negative.
Problem 1.10. Can a nite composition (or, in general, an FS iteration) of the
Hechler forcing notions add a generic object for Q
T
sc
?
2. More sweet examples
In this section we will present two classes of swclosed families of trees, producing
many new examples of sweet forcing notions. Let us start with extending the
framework of universality parameters to that of swclosed families.
The sweet forcing notions determined by the universality parameters were in-
troduced in [5, 2.3]. In [6] we showed that, unfortunately, the use of them may
be somewhat limited because the composition of, say, the Universal Meager forcing
notions adds generic reals for many examples of the forcing notions determined by
universality parameters. However, as we will show here, families of universality pa-
rameters may determine forcing notions which cannot be embedded into the known
examples of sweet forcing notions.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 7
Let us start with recalling denitions concerning universality parameters and the
related forcing notions. We will cut down the generality of [5, 2.3] and we will
quote here the somewhat simpler setting of [6]. Let H be a function from to 2.
Denition 2.1. (1) We let T =

i<

j<i
H(j), Tn =

in

j<i
H(j).
(2) A nite Htree is a tree S TN with N < , root(S) = ) and max(S)

i<N
H(i). The integer N will be denoted by lev(S).
(3) An innite Htree is a normal tree T T.
Denition 2.2. A simplied universality parameter p for H is a pair ((
p
, F
p
) =
((, F) such that
() elements of ( are triples (S, n
dn
, n
up
) such that S is a nite Htree and
n
dn
n
up
lev(S), (), 0, 0) (;
() if: (S
0
, n
0
dn
, n
0
up
) (, S
1
is a nite Htree, lev(S
0
) lev(S
1
), and
S
1
(Tlev(S
0
)) S
0
, and n
1
dn
n
0
dn
, n
0
up
n
1
up
lev(S
1
),
then: (S
1
, n
1
dn
, n
1
up
) (,
() F

is increasing,
() if:
(S

, n

dn
, n

up
) ( (for < 2), lev(S
0
) = lev(S
1
),
S is a nite Htree, lev(S) < lev(S

), and S

(Tlev(S)) S (for
< 2),
lev(S) < n
0
dn
, n
0
up
< n
1
dn
, F(n
1
up
) < lev(S
1
),
then: there is (S

, n

dn
, n

up
) ( such that
n

dn
= n
0
dn
, n

up
= F(n
1
up
), lev(S

) = lev(S
0
) = lev(S
1
), and
S
0
S
1
S

and S

(Tlev(S)) = S.
Remark 2.3. (1) The function F in 2.2 is to allow more examples of universal
parameters. In condition 2.2() it creates room for taking care of some
extra nodes which we may be forced to take to S

because S can be a
proper superset of (S
0
(Tlev(S))) (S
1
(Tlev(S))).
(2) Note that 2.2() implies that ( is closed under end extensions.
Denition 2.4. Let p = ((, F) be a simplied universality parameter for H. We
say that an innite Htree T is pnarrow if for innitely many n < , for some
n = n
dn
< n
up
we have
_
T (T(n
up
+ 1)), n
dn
, n
up
_
(. The family of all
pnarrow innite Htrees will denoted by T

(p, H).
Proposition 2.5. If p is a simplied universality parameter, then T

(p, H) is an
swclosed family (of nitely branching normal trees). Consequently, Q
T

(p,H)
is an
iterably sweet forcing notion.
Proof. It is should be clear that T

(p, H) satises 1.2(1,2). The proof of 1.2(3) is,
basically, included in the proof of [5, Proposition 4.2.5(3)].
The examples of simplied universality parameters include the following.
Denition 2.6 (Compare [6, Denition 1.7, Example 1.9(2)]). Suppose that the
function H is increasing and g

is such that (i )(0 < g(i) < H(i)). Let


A []

. We dene (
g,A
H
as the family consisting of (), 0, 0) and of all triples
(S, n
dn
, n
up
) such that
() S is a nite Htree, n
dn
n
up
lev(S), A [n
dn
, n
up
] ,= , and
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


8 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
() for some sequence w
i
: i A [n
dn
, n
up
]) such that w
i
[H(i)]
g(i)
(for
i A [n
dn
, n
up
]) we have
_
max(S)
__
i A [n
dn
, n
up
)
__
(i) w
i
_
.
Proposition 2.7. Assume that H, g, A are as in 2.6, and F

is an increasing
function such that
_
n <
__

in
H(i)

[n, F(n)) A

_
.
Then p
g,A
H
def
= ((
g,A
H
, F) is a simplied universality parameter (and it even is a
regular universality parameter in the sense of [6, Denition 1.14]).
The universality parameters p
g,A
H
from 2.7 are related to the strong PPproperty
(see [9, Ch VI, 2.12*], compare also with [4, 7.2]). Note that an innite H
tree T is p
g,A
H
narrow if and only if there exist sequences w = w
i
: i A) and
n = n
k
: k < ) such that

_
i A
__
w
i
H(i) & [w
i
[ g(i)
_
, and
n
k
< n
k+1
< for each k < , and

_
[T]
__
k <
__
i A [n
k
, n
k+1
)
__
(i) w
i
_
.
It should be clear that the intersection of a family of swclosed sets of normal
trees is swclosed. So now we are going to look at the intersections of the families
of p
g,A
H
narrow trees.
Denition 2.8. Let H, g be as in 2.6 and let ,= B []

.
(1) Put T (B) = T
g
H
(B)
def
=
_
T

(p
g,B
H
, H) : B B and P
B
= Q
T (B)
.
(2) Let T

B
be a P
B
name such that

PB
T

B
=
_
_
T
p
: p
PB
_
.
(3) For a set A []

put
S
A
=
_
T :
_
i lh() A
__
(i) = 0
__
.
Lemma 2.9. Suppose that H, g are as in 2.6.
(1) Let A, C []

. Then the tree S


A
is p
g,C
H
narrow if and only if A C is
innite.
(2) Let , = B []

. Then, in V
PB
, T

B
is an innite Htree which is p
g,B
H

narrow for all B B, and
(a) if T V is an innite Htree which is p
g,B
H
narrow for all B B,
then there is an n < such that
_
T

B
__
T
__
n = lh() < lh()

[n, lh()) T

B
_
,
(b) if an innite Htree T V is not p
g,B
H
narrow for some B B, then
_
n <
__
T
__
lh() > n & (

i<n
H(i))(

[n, lh()) / T

B
)
_
.
Theorem 2.10. Suppose that P is a ccc forcing notion,
P
2
0
= , < 2
2

0
.
Then there is a family B []

such that P does not add a generic object for the


(iterably sweet) forcing notion P
B
.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 9
Proof. Note that if | is a uniform ultralter on , then (A, B |)([A B[ = )
and hence, by 2.9(1), for every A | and every B |, the tree S
A
is p
g,B
H
narrow.
Also by 2.9(1), for every A []

the tree S
A
is not p
g,\A
H
narrow.
Now, if |

, |

[]

are two distinct uniform ultralters on , then we may


pick A []

such that A |

and A |

. Then the tree S


A
is p
g,B
H
narrow for every B |

, but
is not p
g,\A
H
narrow, A |

.
Therefore, by 2.9(2), the interpretations of the names T

U
, T

U
by the correspond-
ing generic lters must be dierent. Since there are 2
2

0
ultralters on we easily
get the conclusion.
Corollary 2.11. There exists an iterably sweet forcing notion Q which cannot be
embedded into the forcing notion constructed in [8, 7].
Let us present now a dierent class of swclosed families of normal trees and
corresponding forcing forcing notions.
Denition 2.12. The swclosure cl
sw
(T ) of a family T of normal subtrees of
<
is the smallest family T

of subtrees of
<
which includes T and is swclosed.
Clearly, cl
sw
(T ) is well dened for any family T of normal subtrees of
<
.
Lemma 2.13. (1) Suppose that T

is a normal subtree of
<
and let T

be the family of all normal subtrees of T

. Then T

is swclosed. Conse-
quently, if T T

, then cl
sw
(T ) T

.
(2) Assume that T is an swclosed family of normal subtrees of
<
and
A

is a non-empty closed set. Let


T

(A) =
_
T T : [T] A is nowhere dense in A
_
.
Then T

(A) is swclosed.
(3) If T is a family of normal subtrees of
<
, T
<
is a normal tree and
_
T

T
__
[T] [T

] is nowhere dense in [T]


_
,
then T / cl
sw
(T ).
Proof. (1) Should be clear.
(2) Clearly T

(A) is closed under nite unions. Assume now that T
n
, T

T

(A)
are such that
_
n <
__
T


n
= T
n

n
_
and let T =

n
T
n
. We want
to show that T T

(A). Since T is swclosed we see that T T , so we need to
show that [T] A is nowhere dense in A. To this end let S
<
be a normal
tree such that A = [S] and suppose that S. Since T

T

(A), we may nd

0
S such that
0
and
0
/ T

. Then, by our assumptions on T


n
: n ),
also for each k lh(
0
) we have
0
/ T
k
. Since T
n
T

(A) (for n < lh(
0
)), the
set

n<lh(0)
[T
n
] A is nowhere dense in A and hence we may nd S such that

0
and /

n<lh(0)
T
n
. Then we also have S and / T.
(3) Follows from (2).
Denition 2.14. (1) For a set A []

let T
A
be the collection of all normal
subtrees T of 2
<
such that
_
split(T)
__
lh() A
_
.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


10 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
(2) For a family / []

let T
A
= cl
sw
_
T
A
: A /
_
.
Theorem 2.15. Suppose that P is a ccc forcing notion,
P
2
0
= , < 2
2

0
.
Then there is a family / []

such that P does not add the generic object for the
(iterably sweet) forcing notion Q
TA
(see 1.3).
Proof. Let us start with some observations of a more general character.
Claim 2.15.1. Assume that T is an swclosed family of subtrees of 2
<
such that
() for every 2

and T T we have
T
2

def
= 2
<
: +
2
(lh()) T T .
Let T

T
be a Q
T
name such that

Q
T T

T
=
_
_
T
p
: p
Q
T
_
.
Then, in V
Q
T
, T

T
is a subtree of 2
<
such that
(1) for every T T there is an n < such that
if
1
T

T
2
n
and T, lh() > n, then
1

[n, lh()) T

T
,
(2) for every normal tree T 2
<
such that T / T , T V, we have
_
n <
__
T
__
lh() > n & ( 2
n
)(

[n, lh()) / T

T
)
_
.
Proof of the Claim. (1) Suppose that p Q
T
and T T . Let

: < 2
N
p
)
list all elements of 2

which are constantly zero on [N


p
, ). It follows from our
assumption () that
_
< 2
N
p__
T
2

T & T
p

T
_
.
Since T is swclosed we may now conclude that (by 1.2(2))
T
0
def
=
_
<2
N
p
(T
2

)
_
<2
N
p
(T
p

) T ,
and hence also (by 1.2(1))
T
1
def
= T
0
: (lh() N
p
& T
p
) (lh() > N
p
& N
p
T
p
) T .
Now, letting N
q
= N
p
and T
q
= T
1
we get a condition q Q
T
stronger than p and
such that
q
_

1
T

T
2
N
q __
T
__
lh() > N
q

1

[N
q
, lh()) T

T
_
.
(2) Now suppose that p Q
T
, n < and T 2
<
is a normal tree which does
not belong to T . Let N = N
p
+n and let

: < 2
N
) list all elements of 2

which
are constantly zero on [N, ). It follows from () that T
0
def
=

<2
N
(T
p

) T
and since T / T we may conclude by 1.2(1) that T T
0
,= . Pick T T
0
and
note that necessarily lh() > N n. Letting N
q
= lh() and T
q
= T
0
we get a
condition q Q
T
stronger than p and such that
q ( 2
n
)(

[n, lh()) / T

T
)
_
.

8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 11
Claim 2.15.2. If T is a collection of normal subtrees of 2
<
such that the demand
in 2.15.1() holds for T , then also cl
sw
(T ) satises this condition. Consequently,
for each / []

, () of 2.15.1 holds true for T


A
.
Proof of the Claim. Should be clear.
Claim 2.15.3. Suppose that A []

and / []

are such that


_
B /
__
[A B[ = ).
Then T
A
T
A
.
Proof of the Claim. Let T = 2
<
: (n < lh())((n) = 1 n A).
Plainly T T
A
. Also, for every B / and T

T
B
the set [T] [T

] is nowhere
dense in [T], so by 2.13(3) T / cl
sw
_
T
B
: B /
_
= T
A
.
Now choose a family 1 []

of almost disjoint sets, [1[ = 2


0
.
Suppose that /, B 1, / ,= B, say A / B. Then (B B)([A B[ = ) and
hence (by Claim 2.15.3) we get T
A
T
B
, so we have a normal tree T T
A
T
B
.
Now look at Claim 2.15.1 by 2.15.2 it is applicable to Q
TA
, Q
TB
and we get from
it that if T
A
, T
B
2
<
are trees generic over V for Q
TA
, Q
TB
, respectively, then

_
n <
__
T
A
2
n
__
T
__
lh() > n

[n, lh()) T
A
_
,

_
n <
__
T
__
lh() > n & ( 2
n
)(

[n, lh()) / T
B
_
.
Hence T
A
,= T
B
. Since P satises the ccc and
P
2
0
= and < 2
2

0
, we may
nd a family T of subsets of 1 such that [T[ = and

P
for no / 1 with / / T, there is a Q
TA
generic lter over V .

One should note that the examples of sweet forcing notions which cannot be
embedded into the one constructed in [8, 7] which we gave in this section are not
very nice it may well be that the parameters /, B needed to dene them are not
denable from a real. Even the candidate for a somewhat denable example from
the previous section, the forcing notion Q
T
sc
, is not Souslin. Thus the following
variant of [5, Problem 5.5] may be of interest.
Problem 2.16. Is there a Souslin ccc iterably sweet forcing notion Q such that no
nite composition of the Universal Meager forcing notion adds a Qgeneric real?
Such that the forcing of [8, 7] does not add Qgeneric real?
3. Subforcings, Quotients and likes
Topological sweetness, as dened in 0.2, is the property of a particular repre-
sentation of a forcing notion. It is only natural to ask if a forcing notion having
a topologically sweet dense subforcing is topologically sweet, or, in general, if a
forcing notion equivalent to a topologically sweet one is topologically sweet. We
start this section with some results in these directions.
Denition 3.1. We say that a forcing notion P has a GLBproperty provided that
for every q
0
, q
1
P, there is q P such that
() q q
i
for i = 0, 1, and
() if q

P satises (i < 2)(q

q
i
), then q

q.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


12 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Remark 3.2. (1) GLB stands for Greatest Lower Bound . Note the direction
of the inequality on the denition of GLB and remember that p q means
q is stronger than p.
(2) If B is a Boolean algebra, then B
+
is a forcing notion with the GLB
property. Also the forcing notions R and A dened in 3.8 later have this
property.
Proposition 3.3. Suppose that a forcing notion P has the GLBproperty and Q
P is its dense subforcing. If Q is topologically sweet, then so is P.
Proof. Let (Q, B) be a model of topological sweetness and let be the topology on
Q generated by B. For sets U
0
, . . . , U
k
B, k < , dene
W(U
0
, . . . , U
k
) = p P : (i k)(q U
i
)(p q),
and let
B

=
_
W(U
0
, . . . , U
k
) : k < & U
0
, . . . , U
k
B
_

_
.
It should be clear that
B

is closed under nite intersections, and


it is a countable basis of a topology

on P, and

P
is an isolated point in

.
We are going to show that the topology

satises the demand of 0.2(ii). So


suppose that a sequence p = p
n
: n < ) P is

converging to p P and
q p and W is a

neighbourhood of q. Pick U
0
, . . . , U
k
B such that q
W(U
0
, . . . , U
k
) W and let q
i
U
i
(for i k) be such that q q
i
. Furthermore,
for i k, let V
i
n
: n < be a basis of neighbourhoods of q
i
Q such that
(n
0
< n
1
< )(q
i
V
i
n1
V
i
n0
U
i
).
Since p W(V
0
n
, V
1
n
, . . . , V
k
n
) B

(for each n < ) and the sequence p

converges to p, we may choose an increasing sequence m


n
: n < ) such that
_
n <
__
p
mn
W(V
0
n
, V
1
n
, . . . , V
k
n
)
_
. Then we may also pick p

n,i
(for n <
and i k) such that p
mn
p

n,i
V
i
n
. Fix i k and look at the sequence
p

i
= p

n,i
: n < ): clearly it converges to q
i
. Consequently, we may easily
choose (be repeated application of 0.2(ii) for ) conditions q

i
Q such that
q
i
q

i
U
i
for i k, and
(

n < )(i k)(p

n,i
q

i
).
Since P has the GLBproperty we may pick q

P such that
() q

q
i
for i k, and
() if r P is weaker than q

0
, . . . , q

k
, then r q

.
Then, plainly, q

W(U
0
, . . . , U
k
) and q q

and (

n < )(p
mn
q

).
Proposition 3.4. Assume that P is a topologically sweet forcing notion. Then
there is a model (P, B

) of topological sweetness such that all members of B

are
downward closed.
Proof. Let (P, B) be a model of topological sweetness. For U B put W(U) =
p P : (q U)(p q), and let B

= W(U) : U B. Note that if p


W(U
0
) W(U
1
) and p p
0
U
0
, p p
1
U
1
, then there is V B such that
p V and V W(U
0
) W(U
1
) (remember 0.3(1)). Hence we easily conclude that
B

is a base of a topology

on P. Similarly as in 3.3 one shows that (P, B

) is a
model of topological sweetness.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 13
Proposition 3.5. Assume that P is a topologically sweet and separative partial
order. Suppose also that B
0
is a complete subalgebra of the complete Boolean algebra
BA(P) determined by P. Then B
+
0
(as a forcing notion) is topologically sweet.
Proof. It follows from our assumptions on P that it is (isomorphic to) a dense subset
of BA(P)
+
and hence, by 3.3+3.4, there is a model (BA(P)
+
, B) of topological
sweetness such that all members of B are downward closed. By our assumptions
B
0
< BA(P), so let : BA(P) B
0
be the projection. We will use the symbol
to refer to the forcing order of B
+
0
(which is the reverse Boolean order). Put
B

= U B
+
0
: U B.
We claim that (B
+
0
, B

) is a model of topological sweetness. It is easy to verify


0.2(i), so let us only argue that 0.2(ii) holds true. To this end suppose that a
sequence p = p
n
: n < ) B
+
0
converges to p B
+
0
(in the topology generated
by B

) and let p q U B
+
0
, U B. Then also p converges to p in the topology
generated by B on BA(P)
+
, so we may nd r BA(P)
+
such that q r U and
(

n < )(p
n
r). Let r

= (r) B
0
. Then we have
q r

(as is the projection and q B


+
0
, q r),
(

n < )(p
n
r

) (as is the projection and p


n
B
+
0
),
r

U (as U is downward closed, r

r U).

Sweetness and topological sweetness are important properties because they are
preserved in amalgamations of forcing notions. Since the amalgamation can be
represented as the composition with the product of two quotients (see, e.g., [3] on
that), one may ask if sweetness is also preserved in quotients.
Denition 3.6. Let P, Q be forcing notions and suppose that Q < P. The quotient
(P : Q) is the Qname for the subforcing of P consisting of all p P such that p is
compatible (in P) with all members of
Q
. Thus for p P and q Q,
q
Q
p (P : Q) if and only if
(r Q)(q r r, p are compatible in P).
Theorem 3.7. Let C be the standard Cohen forcing notion (so it is a countable
atomless partial order). Suppose that (P, B) is a model of topological sweetness and
C < P. Let B

C
be the C-name for the family U (P : C) : U B. Then

C

_
(P : C), B

C
_
is a model of topological sweetness .
Proof. First note that, in V
C
, B

C
is a countable basis of a topology on (P : C),
and
(P:C)
=
P
is an isolated point in this topology. Thus the only thing that we
should verify is the demand in 0.2(1)(ii).
Suppose that C and Cnames p

i
: i < ), p

, q

and W

are such that



C
p

i
, p

, q

(P : C), W

C
, p

and
the sequence p

i
: i < ) converges to p

in the topology generated by B

Replacing by a stronger condition in C (if necessary), we may assume that for


some p, q P and W B we have

C
p

= p & q

= q & W

= W (P : C) .
Then also
C
p, q (P : C) and p q W. Let us choose a condition q
+
P
which is stronger than both q and , and let U B be a neighborhood of q
+
such
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


14 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
that any two members of U are compatible in P (remember 0.3(2)). Next, choose
W
+
B such that q W
+
W and every member of W
+
has an upper bound in
U (possible by 0.3(1)).
Pick V
i
B (for i < ) such that V
i
: i < forms a neighbourhood basis at p
(for the topology generated by B) such that for each i < :
() p V
i+1
V
i
,
() any i + 1 elements of V
i+1
have a common upper bound in V
i
.
[The choice is clearly possible; remember 0.3.]
Clearly
C
V
i
(P : C) : i < forms a neighbourhood basis at p (for the
topology generated by B

C
) . Hence, without loss of generality, we may assume
that
C
p

i
V
i
(as we may change the names p

i
reecting a passage to a
subsequence). Let us x a list

: < of all conditions in C stronger than ,


and for every i, < let us pick p
i,
P such that

,
C
p

i
,= p
i,
. Note that
then p
i,
V
i
, so by clause () above we may choose p

i
V
i
such that for each
i > 0 we have
( i)(p
i+1,
p

i
).
The sequence p

i
: i < ) converges to p so (by 0.2(1)(ii) for (P, B)) there are a
condition r P and an innite set A such that
r W
+
and q r and (i A)(p

i
r).
By the choice of W
+
, the condition r has an upper bound in U and hence (by the
choice of U) r, q
+
are compatible in P. Therefore, as q
+
is stronger than (in P),
there is C stronger than such that
C
r (P : C) . Now the proof follows
from the following Claim.
Claim 3.7.1.
C
(

i < )(p

i
r) .
Proof of the Claim. If not, then we may nd

C stronger than and i

<
such that

C
(i i

)(p

i
r) . Let < be such that

and let i A
be larger than +i

+ 1. Look at our choices before - we know that:


(i) p

i
r,
(ii) p
i+1,
p

i
,
(iii)

,
C
p

i+1
,= p
i+1,
.
Therefore some condition

C stronger than

forces that p

i+1
r, contradict-
ing the choice of

(as i + 1 > i

).

In the rest of this section we are going to show that the result of 3.7 cannot be
very much improved: when taking a quotient over a random real forcing we may
lose topological sweetness. Let us start with recalling some notation and denitions,
which we will need later.
Denition 3.8. (1) The Lebesgue (product) measure on 2

is denoted by

Leb
, Borel(2

) is the eld of Borel subsets of 2

and L is the
ideal of Lebesgue null subsets of 2

. The quotient Boolean algebra B =


Borel(2

)/L is called the random algebra (it is a complete ccc Boolean


algebra).
(2) The random forcing notion R is dened as follows:
a condition in R is a closed subset of 2

of positive Lebesgue measure,


the order of R is the reverse inclusion.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 15
(3) The amoeba for measure forcing notion A is dened as follows:
a condition in A is a closed subset F of 2

such that
Leb
(F) >
1
2
,
the order of A is the reverse inclusion.
Of course, B = BA(R). Let us also recall that both R and A are topologically
sweet (see [10, 1.3.3]).
Proposition 3.9. (1)
B
A
V
is not topologically sweet .
(2)
B
R
V
is not topologically sweet .
Proof. (1) Suppose towards contradiction that
[[ there is a model of topological sweetness based on A
V
]]
B
,= 0
B
.
Since the random algebra is homogeneous, we may assume that we have Bnames
U

n
for subsets of A
V
such that
()
0

B
(A
V
, U

n
: n < ) is a model of topological sweetness .
For i < let m
i
=
i2
i
log
2
(12
2
i+1
)
+ 2, so
mi
2
i
>
i
log
2
(12
2
i+1
)
and thus
()
i
1
(1 2
2
i+1
)
mi/2
i
< 2
i
.
Let be the product Lebesgue measure on the space

i<
m
i
, and let

be the
corresponding outer measure.
Dene n
i
: i < ) by n
0
= 0, n
i+1
= n
i
+m
i
2
i+1
, and for i < , j < m
i
put
t
i
j
def
=
_
2
[n
i
, n
i+1
)
:
_
< 2
i+1
__
(n
i
+j 2
i+1
+) = 1
__
.
Note that
()
i
2
if j
0
< j
1
< . . . < j
k
< m
i
, then

t
i
j0
t
i
j1
. . . t
i
j
k

=
_
1 2
2
i+1_
k+1
2
mi2
i+1
.
For x

i<
m
i
let
Z
x
def
=
_
2

:
_
i <
__
[n
i
, n
i+1
) t
i
x(i)
__
and note that Z
x
is a closed set and
Leb
(Z
x
) >
1
2
, so Z
x
A. For each x

i<
m
i
and n < we may pick a Borel set B(x, n) 2

such that [[Z


x
U

n
]]
B
= [B(x, n)]
L
.
Next, for each k < (and x

i<
m
i
and n < ) choose a clopen set C(x, n, k)
2

such that
Leb
_
B(x, n) C(x, n, k)
_
< 2
k
. Now, for n < , consider the
binary relation
n
on

i<
m
i
given by
x
n
y if and only if
_
k, n
__
C(x, , k) = C(y, , k)
_
.
It should be clear that (for each n < )
n
is an equivalence relation on

i<
m
i
such that
()
n
3
x
n+1
y x
n
y (for each x, y

i<
m
i
), and
()
n
4

i<
m
i
/
n
is countable.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


16 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
Consequently we may pick x


i<
m
i
such that for each n < we have
lim

__
x

i<
m
i
: x = x

& x
n
x

__

__
x

i<
m
i
: x = x

__ = 1.
So now we may choose an increasing sequence
i
: i < ) such that for i <
we have

__
x

j<
m
j
: x
i
= x

i
& x
i
x

__
>
1
2

__
x

j<
m
j
: x
i
= x

i
__
,
and then for each i < we may choose v
i
m
i
and y
i
k
: k v
i
)

j<
m
j
such
that
()
i
5
[v
i
[ >
1
2
m
i
,
()
i
6
y
i
k

i
= x

i
, y
i
k
(
i
) = k and y
i
k

i
x

for k v
i
.
It follows from the denition of the relations
n
and from ()
i
6
that for each k v
i
and all i we have

Leb
_
B(x

, ) B(y
i
k
, )
_
< 2
1i
.
Thus, for each i < , we may pick a partition B
i
k
: k v
i
) of 2

into disjoint Borel


sets such that for all k v
i
we have
()
i,k
7

Leb
(B
i
k
) =
1
|vi|
, and
()
i,k
8

Leb
_
B
i
k
(B(x

, ) B(y
i
k
, ))
_
< 2
1i
/[v
i
[ for all i.
Let x

i
be a Bname for a member of V

j<
m
j
such that
_
k v
i
__
[[x

i
= y
i
k
]]
B
= [B
i
k
]
L
_
.
Claim 3.9.1.

B

_
n <
__

i <
__
Z
x
U

n
Z
x

i
U

n
_
.
Proof of the Claim. Note that for n, i < we have
[[Z
x

i
/ U

n
]]
B
=
_
_
kvi
B
i
k
B(y
i
k
, n)
_
L
,
and thus [[Z
x
U

n
& Z
x

i
/ U

n
]]
B
=
_
kvi
_
B(x

, n) B(y
i
k
, n)
_
B
i
k

L
. It follows
from ()
i,k
8
that (for n i < ) we have

Leb
_
_
kvi
_
B(x

, n) B(y
i
k
, n)
_
B
i
k
_
< 2
1i
.
Hence for, each n < ,

Leb
_

m<
_
i>m
_
_
kvi
_
B(x

, n) B(y
i
k
, n)
_
B
i
k
_
_
= 0,
so
[[(

i < )(Z
x
U

n
& Z
x

i
/ U

n
]]
B
= 0
B
,
and the Claim follows.
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


HOW MUCH SWEETNESS IS THERE IN THE UNIVERSE? 17
It follows from ()
0
and 3.9.1 that
[[
_
F A
V
__
F Z
x
& (

i < )(F Z
x

i
)
_
]]
B
= 1
B
,
and therefore we may nd F AV such that F Z
x
and a
def
= [[(

i < )(F
Z
x

i
)]]
B
,= 0
B
. For i < put
w
i
= k v
i
: F Z
y
i
k
and C
i
=
_
kwi
B
i
k
.
Plainly, a =
_
m<

i>m
C
i

L
so (as a ,= 0
B
)

i=1

Leb
(C
i
) = , and hence the set
I
def
= i < :
Leb
(C
i
) > 2
1i

is innite.
Fix i I for a moment. Then
2
1i
<
Leb
(C
i
) =

kwi

Leb
(B
i
k
) =
[w
i
[
[v
i
[
,
and thus (by ()
i
5
)
[w
i
[ > [v
i
[ 2
1i
>
1
2
m
i
2
1i
m
i
/2
i
.
Hence, by ()
i
1
, we get
_
12
2

i
+1
_
|wi|
< 2
i
2
i
. Now (for our i I) consider
the closed set Y
i
def
=

kwi
Z
y
i
k
and note that
Y
i

_
2

: (k w
i
)([n
i
, n
i+1
) t
i
k
)
_
.
Thus, by ()
i
2
, we may conclude that (for our i I)

Leb
(Y
i
)

kwi
t
i
k

2
m

i
2

i
+1
=
_
1 2
2

i
+1 _
|wi|
< 2
i
.
Since I is innite and for every i I we have F

kwi
Z
y
i
k
= Y
i
we may now
conclude that
Leb
(F) = 0, contradicting F A.
(2) The same proof as for (1) works here too.
Putting together 3.3 and 3.9 (and remembering that products of sweet forcings
are sweet) we may easily conclude the following.
Corollary 3.10. Both R R and A A are topologically sweet, but

R
no dense subforcing of (R R : R) ((R A : R), respectively)
is topologically sweet .
(Above, in (R R : R) we consider the quotient over the rst coordinate of the
product. Similarly in (R A : R).)
8
5
6


r
e
v
i
s
i
o
n
:
2
0
0
5
-
1
0
-
2
7







m
o
d
i
f
i
e
d
:
2
0
0
5
-
1
0
-
2
7


18 ANDRZEJ ROS LANOWSKI AND SAHARON SHELAH
References
[1] Tomek Bartoszy nski and Haim Judah. Set Theory: On the Structure of the Real Line. A K
Peters, Wellesley, Massachusetts, 1995.
[2] Thomas Jech. Set theory. Springer Monographs in Mathematics. Springer-Verlag, Berlin,
2003. The third millennium edition, revised and expanded.
[3] Haim Judah and Andrzej Ros lanowski. On Shelahs Amalgamation. In Set Theory of the
Reals, volume 6 of Israel Mathematical Conference Proceedings, pages 385414. 1992.
[4] Andrzej Roslanowski and Saharon Shelah. Norms on possibilities I: forcing with trees
and creatures. Memoirs of the American Mathematical Society, 141(671):xii + 167, 1999.
math.LO/9807172.
[5] Andrzej Roslanowski and Saharon Shelah. Sweet & Sour and other avours of ccc forcing
notions. Archive for Mathematical Logic, 43:583663, 2004. math.LO/9909115.
[6] Andrzej Roslanowski and Saharon Shelah. Universal forcing notions and ideals. Archive for
Mathematical Logic, 46:179196, 2007. math.LO/0404146.
[7] Saharon Shelah. 446 revisited.
[8] Saharon Shelah. Can you take Solovays inaccessible away? Israel Journal of Mathematics,
48:147, 1984.
[9] Saharon Shelah. Proper and improper forcing. Perspectives in Mathematical Logic. Springer,
1998.
[10] Jacques Stern. Regularity properties of denable sets of reals. Annals of Pure and Applied
Logic, 29:289324, 1985.
[11] John Truss. Sets having calibre
1
. In Logic Colloquium 76, volume 87 of Studies in Logic
and the Foundations of Mathematics, pages 595612. North-Holland, Amsterdam, 1977.
Department of Mathematics, University of Nebraska at Omaha, Omaha, NE 68182-
0243, USA
E-mail address: roslanow@member.ams.org
URL: http://www.unomaha.edu/logic
Einstein Institute of Mathematics, Edmond J. Safra Campus, Givat Ram, The Hebrew
University of Jerusalem, Jerusalem, 91904, Israel, and Department of Mathematics,
Rutgers University, New Brunswick, NJ 08854, USA
E-mail address: shelah@math.huji.ac.il
URL: http://shelah.logic.at

Vous aimerez peut-être aussi