Phase Transitions and

Self-Organization in Electronic
and Molecular Networks
FUNDAMENTAL MATERIALS RESEARCH
Series Editor: M. F. Thorpe, Michigan State University
East Lansing, Michigan
ACCESS IN NANOPOROUS MATERIALS
Edited by Thomas J. Pinnavaia and M. F. Thorpe
DYNAMICS OF CRYSTAL SURFACES AND INTERFACES
Edited by P. M. Duxbury and T. J. Pence
ELECTRONIC PROPERTIES OF SOLIDS USING CLUSTER METHODS
Edited by T. A. Kaplan and S. D. Mahanti
LOCAL STRUCTURE FROM DIFFRACTION
Edited by S. J. L. Billinge and M. F. Thorpe
PHASE TRANSITIONS AND SELF-ORGANIZATION IN ELECTRONIC AND MOLECULAR
NETWORKS
Edited by J. C. Phillips and M. F. Thorpe
PHYSICS OF MANGANITES
Edited by T. A. Kaplan and S. D. Mahanti
RIGIDITY THEORY AND APPLICATIONS
Edited by M. F. Thorpe and P. M. Duxbury
SCIENCE AND APPLICATION OF NANOTUBES
Edited by D. Tomnek and R. J. Enbody
A Continuation Order Plan is available for this series. A continuation order will bring delivery of each new
volume immediately upon publication. Volumes are billed only upon actual shipment. For further information
please contact the publisher.
Phase Transitions and
Self-Organization in Electronic
and Molecular Networks
Edited by
J. C. Phillips
Lucent Technologies
Bell Labs Innovations
Murray Hill, New Jersey
and
M. F. Thorpe
Michigan State University
East Lansing, Michigan
KLUWER ACADEMIC PUBLISHERS
NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW
eBook ISBN: 0-306-47113-2
Print ISBN: 0-306-46568-X
2002 Kluwer Academic Publishers
New York, Boston, Dordrecht, London, Moscow
Print 2001 Kluwer Academic / Plenum Publishers
New York
All rights reserved
No part of this eBook may be reproduced or transmitted in any form or by any means, electronic,
mechanical, recording, or otherwise, without written consent from the Publisher
Created in the United States of America
Visit Kluwer Online at: http://kluweronline.com
and Kluwer's eBookstore at: http://ebooks.kluweronline.com
SERIES PREFACE
This series of books, which is published at the rate of about one per year, addresses
fundamental problems in materials science. The contents cover a broad range of topics
from small clusters of atoms to engineering materials and involve chemistry, physics,
materials science, and engineering, with length scales ranging from ngstroms up to
millimeters. The emphasis is on basic science rather than on applications. Each book
focuses on a single area of current interest and brings together leading experts to give an
up-to-date discussion of their work and the work of others. Each article contains enough
references that the interested reader can access the relevant literature. Thanks are given to
the Center for Fundamental Materials Research at Michigan State University for supporting
this series.
M.F. Thorpe, Series Editor
E-mail: thorpe@pa.msu.edu
East Lansing, Michigan, September 2000
v
This page intentionally left blank
PREFACE
The problem of phase transitions in disordered materials is quite old, but until recently
it has seemed too complex a subject for formal study. The advent of computers has
changed matters in two important ways. First, it has become possible to implement formal
methods for microscopic study of phase transitions in ordered materials, even in the
quantum limit, in great detail. This work has been so successful that few qualitative
mysteries remain, and many microscopic details have been measured experimentally and
derived theoretically from first principles.
The second radical change brought about by computers is that scientists have been
forced to recognize that even today phase transitions in disordered materials are very
poorly understood. Apart from the inherent statistical problems raised by disorder, it is
becoming clear that new fundamental concepts are needed to explain qualitatively new
phenomena that arise in disordered materials that were absent in ordered crystalline
materials, or even in such materials with disordered sublattices.
This workshop addresses the need for fundamentally new concepts in three areas of
physical science. The first is network glasses, simple mechanical systems in which
important new phenomena (the intermediate phases, the reversibility window) have been
discovered as a result of exploring stiffness transitions both experimentally and in
numerical simulations made possible by new computer algorithms.
The considerable progress made here is most encouraging, but surprisingly it has turned
out that these new mechanical phenomena are closely paralleled by new electronic
phenomena. These are discussed for the second area, the metal-insulator transition in
semiconductor impurity bands, in which an intermediate phase has also been identified.
The third area is (mostly cuprate) perovskites, where an intermediate phase occurs which
can have superconductive transition temperatures well above 100K. It appears very likely
that the electronic intermediate phases exist because of disorder, and that the electronic
phase diagrams closely parallel the mechanical phase diagrams of network glasses.
On a microscopic level, minimization of the free energy of a disordered system at
moderate temperatures, followed by some kind of (mild) quenching, can produce self-
organization. There are many indications of this in network glasses, but of course life itself
is self-organized. Proteins can be described as self-organized disordered networks, and
they are discussed briefly here, and in a special issue of Journal of Molecular Graphics and
Modelling (edited by L.A. Kuhn and M.F. Thorpe, to appear early 2001). It turns out that
several constraint-based concepts that have been developed for network glasses apply
equally well to the apparently unrelated subject of protein folding.
This focused workshop was held at Hughes Hall, Cambridge, England, July 10-14,
2000. We are grateful to Dr. Martin Dove for assistance with local arrangements, and Ms.
Janet King and Mr. Mykyta Chubynsky for extensive editorial assistance.
J.C. Phillips
M.F. Thorpe
East Lansing, Michigan, September 2000
vii
This page intentionally left blank
CONTENTS
I. Some Mathematics
Mathematical Principles of Intermediate Phases
in Disordered Systems................................................................................................
J.C. Phi l l i ps
1
Reduced Density Matrices and Correlation Matrix..............................................................
A. John Coleman
The Sixteen-Percent Solution:
Critical Volume Fraction for Percolation.................................................................
Richard Zallen
The Intermediate Phase and Self-Organization
in Network Glasses...................................................................................................
M.F. Thorpe and M.V. Chubynsky
II. Glasses and Supercooled Liquids
Evidence for the Intermediate Phase
in Chalcogenide Glasses...........................................................................................
P. Boolchand, W.J. Bresser, D.G. Georgiev, Y. Wang, and J. Wells
Thermal Relaxation and Criticality of
the Stiffness Transition.......................................................................................................
Y. Wang, T. Nakaoka, and K. Murase
23
37
43
65
85
Solidity of Viscous Liquids................................................................................................
J.C. Dyre
Non-Ergodic Dynamics in Supercooled Liquids................................................................
M. Dzugutov, S. Simdyankin, and F. Zetterling
Network Stiffening and Chemical Ordering in Chalcogenide Glasses:
Compositional Trends of T
g
in Relation to
Structural Information from Solid and Liquid State NMR ........................
Carsten Rosenhahn, Sophia Hayes, Gunther Brunklaus,
and Hel l mut Eckert
101
111
123
ix
Glass Transition Temperature Variation as a Probe for
Network Connectivity............................................................................................
M. Micoulaut
Floppy Modes Effects in the Thermodynamical
Properties of Chalcogenide Glasses.......................................................................
Gerardo G. Naumis
The Dalton-Maxwell-Pauling Recipe for Window Glass ..................................................
Richard Kerner
Local Bonding, Phase Stability and Interface Properties of
Replacement Gate Dielectrics, Including Silicon Oxynitride Alloys
and Nitrides, and Film Amphoteric Elemental Oxides and Silicates ......
G. Lucovsky
Experimental Methods for Local Structure
Determination on the Atomic Scale .......................................................................
E.A. Stern
Zeolite Instability and Collapse..........................................................................................
G.N. Greaves
III. Metal-Insulator Transitions
Thermodynamics and Transport Properties of Interacting
Systems with Localized Electrons..........................................................................
A.L. Efros
The Metal-Insulator Transition in Doped Semiconductors:
Transport Properties and Critical Behavior............................................................
Theodore G. Castner
Metal-Insulator Transition
in Homogeneously Doped Germanium..................................................................
Michio Watanabe
IV. High Temperature Superconductors
Experimental Evidence for Ferroelastic Nanodomains in
HTSC Cuprates and Related Oxides......................................................................
J. Jung
Role of Sr Dopants in the Inhomogeneous Ground State
of La
2-x
Sr
x
CuO
4
.......................................................................................................
D. Haskel, E.A. Stern, and F. Dogan
143
161
171
189
209
225
247
263
291
311
323
x
Universal Phase Diagrams and Ideal High Temperature Superconductors:
J.L. Wagner, T.M. Clemens, D.C. Mathew, O. Chmaissem
B. Dabrowski, J.D. Jorgensen, and D.G. Hinks
Coexistence of Superconductivity and Weak Ferromagnetism
in Eu
1.5
Ce
0.5
RuSr
2
Cu
2
O
10
.........................................................................................
I. Felner
Quantum Percolation in High Tc Superconductors............................................................
V. Dallacasa
Superstripes: Self Organization of Quantum Wires
in High T
c
Superconductors....................................................................................
A. Bianconi, D. DiCastro, N.L. Saini, and G. Bianconi
Electron Strings in Oxides..................................................................................................
F.V. Kusmartsev
High-Temperature Superconductivity
is Charge-Reservoir Superconductivity..................................................................
John D. Dow, Howard A. Blackstead, and Dale R. Harshman
Electronic Inhomogeneities in High-T
c
Superconductors
Observed by NMR..................................................................................................
J. Haase, C.P. Slichter, R. Stern, C.T. Milling, and D.G. Hinks
Tailoring the Properties of High-T
c
and Related Oxides:
From Fundamentals to Gap Nanoengineering........................................................
Davor Pavuna
V. Self-Organization in Proteins
Designing Protein Structures..............................................................................................
Hao Li, Chao Tang, and Ned S. Wingreen
List of Participants..............................................................................................................
Index...................................................................................................................................
..............................................................................................................331
341
357
375
389
403
413
431
441
447
451
xi
This page intentionally left blank
MATHEMATICAL PRINCIPLES OF INTERMEDIATE PHASES IN
DISORDERED SYSTEMS
J. C. PHILLIPS
Bell Laboratories, Lucent Technologies (Retired)
Murray Hill, N. J. 07974-0636 (jcphillips@lucent.com)
INTRODUCTION
Intermediate phases are found in disordered systems that for a long time were
supposed to exhibit simple connectivity transitions, similar to dilute magnetic transitions.
The latter can be modeled by percolation on a lattice. The paradigmatic disordered off-
lattice systems that exhibit intermediate phases are network glasses, impurity bands in
semiconductors [the metal-insulator transition (MIT)], and high-temperature doped
(pseudo-)perovskite superconductors. The first two (relatively simple) examples show that
self-organization of the flexibility inherent in disorder is what creates intermediate phases,
and that these must be described by finite-size scaling methods. The third (very complex)
example shows that high temperature superconductivity (HTSC) itself depends on glassy
dopant disorder, and only indirectly on the crystalline matrix with its long-range order.
The mathematical principles underlying the filamentary or percolative theory of such
internally organized systems are fundamentally different from those of theories based on
the effective medium approximation (EMA) or fully disorganized (randomly) diluted lattice
connectivity transitions. These principles have been developed only in the last hundred
years and are little known to most scientists. The counting methods used in the filamentary
theory bear a striking resemblance to those used to prove Fermats Last Theorem or to
factor efficiently large numbers using quantum computers. Examples of the intermediate
phase for these three classes of materials are given that specifically identify the internal
self-organized complexity that is responsible for the remarkable physical properties of each
case.
There is a growing realization that the physics of complex disordered systems differs
qualitatively from that of simple crystalline systems with long-range order, especially in the
vicinity of connectivity transitions. In this workshop both experimental and theoretical
work illustrating this theme are discussed for a wide range of subjects, with special
emphasis on three topics: network glasses, impurity band MITs, and HTSC. In each case
we find that the single connectivity transition to which we are accustomed in simple
Phase Transitions And Self-Organization in Electronic and Mol ecul ar Networks
Edited by J. C. Phi l l i ps and M. F Thorpe, Kl uwer Academic/Plenum Publ i shers, 2001 1
systems is replaced by two transitions of quite different character. The first resembles the
continuous transition expected from percolation theory, but with much simpler exponents,
while the second is a first-order transition with catastrophic character. Between these two
transitions we find an intermediate phase which always has novel properties that are indeed
qualitatively different from those of simple dilute lattice systems. In some cases these
novel properties are of enormous technological value (window glass), and the study of
intermediate phases has for the first time enabled us to understand quantitatively why these
properties occur. In other cases, such as HTSC, the intermediate phase has properties that
are so novel and so unexpected that so far almost all theories have failed to develop beyond
the macroscopic or phenomenological stage.
Within the physical sciences the level of interest in cuprate HTSC has greatly
surpassed that of any other subject, with the sole exception of semiconductive materials
(such as Si) basic to modern electronics. Initially the interest was stimulated by amazingly
large values of the superconductive transition temperatures T
c
, typically five (ten) times
larger than the highest values found in compound (elemental) metals [1]. As expected,
there were other anomalies as well: high sensitivity to doping by non-magnetic oxygen, and
very little sensitivity to the presence of magnetic rare earths [2], both anomalies reversing
the situation found in metallic superconductors. In the normal-state to superconductive
phase transitions of metals, the superconductive properties are generally rather insensitive
to the normal-state behavior, but in the cuprates the normal-state transport at high
temperatures itself is anomalous. The anomalous behavior becomes most
characteristic at just those compositions that maximize T
c
, even in cases where
This tells us that a new electronic theory is needed to describe such strange metals [3].
This new theory must be very different from the Fermi liquid or Landau-Ginzburg theories
used to describe normal metals, which are based on the effective medium approximation
(EMA). The EMA cannot be even qualitatively correct here [4], as the Fermi liquid phase
is separated from the intermediate phase by a first-order phase transition.
Unfortunately, although the need for an alternative to Fermi liquid or Landau-
Ginzburg theory is widely recognized [2,3], only the authors own filamentary or
percolative theory [5] avoids the EMA. This theory relies essentially on set-theoretic
methods derived from number theory to establish quantitative results, and these methods
are largely unknown to physical scientists. These methods have long been regarded as
rather esoteric, even by most mathematicians, but their true significance, as a way of
unifying results from algebra, analysis, geometry and topology, has become apparent
recently from the proof [6] of Fermats Last Theorem (FLT). Several popular discussions
of set theory and FLT are available, but the connections with network glasses, impurity
bands, and perovskite superconductivity are so simple and so direct that this paper will
provide them as a matter of convenience to busy readers. We will then show how these
novel mathematical ideas match the results of several recent decisive experiments in great
detail.
DISCRETE INTEGER AND CONTINUOUS REAL NUMBER FIELDS: FLT
Physical scientists without a strong background in modern mathematics will find an
excellent introduction to the subject, which carries them from its beginnings right through
to an outline of the steps that led to the proof of FLT, in [7], amusing, anecdotal and
thoroughly entertaining. For a long time number theory was regarded as a collection of
strange and rather accidental results of no general significance, but in the late 1800s
Cantor invented set theory and established an essential difference between integers and real
numbers. Although both sets are infinitely large, the number (or order or cardinality) of
2
real (irrational) numbers is larger than that of the integers and rational numbers, which
have the same cardinality. He also hypothesized that there is no cardinality intermediate
between those of the rationals and irrationals (continuum hypothesis); later Godel
established a series of equivalent statements, including the axiom of choice, which showed
that this axiom is independent of the rest of mathematics. These ideas are important in the
present context because they highlight the idea that the methods of effective medium
theories are fundamentally limited because they apply only to simple continuum systems
represented by real number fields. If the number field of interest is the integers or rationals,
or a combination of these with the reals, then special methods need to be developed to
prove theorems or derive results that do not exist for real number fields only.
The nature of these special methods is dramatically illustrated in the proof of FLT,
which states that integer triples exist which satisfy the Pythagorean or Euclidean
metric only for n = 2. Physical scientists, familiar with the example
find Fermats conjecture quite plausible, especially as it has been confirmed
by computer searches up to n = four million, but of course these searches do not constitute
a proof. The proof involves two abstract mathematical tools, elliptic curves and modular
functions.
An elliptic curve is not an ellipse: it is the set of solutions to a cubic polynomial in
two variables, usually written in the form y
2
= x
3
+ Ax
2
+ Bx + C. For number theory x and
y are integers. Modular functions are periodic and are a kind of integral generalization of
sines and cosines. One can conjecture that all elliptic curves are modular. It then turns out
that if this conjecture is valid, FLT follows.
The proof of the latter began with a counter-example (Freys curve), which shows that
should such an exist, it would generate an elliptic function with anomalous properties,
in the sense that it would not be modular, as it is for integer triples with n = 2. To prove
that this relation between elliptic and modular functions is necessary, Wiles counted the
number of both and showed that the two numbers were the same; thus the essential step
was this counting [7].
Counting is a set-theoretic integral process. It is essential to our filamentary model of
network glasses and the semiconductor impurity band transition [8-10] and to our
filamentary model of cuprate superconductivity [5]. In all cases the number of basis
functions associated with cyclical vibrational states, or current-carrying states (or Cooper-
paired current-carrying states in the superconductive case) is actually counted, as part of
their separation from localized states in the neighborhood of the stiffness or metal-insulator
transitions. Within the EMA and real number fields only, so far counting methods appear
not to be feasible, and have not been used to discuss either the metal-insulator transition
(MIT) or HTSC. All the EMA results that have been obtained are based on analytic
(continuum) methods alone, which we believe are not well suited even to impurity band
metal-insulator transitions and to the anomalous electronic properties of cuprates in the
normal or superconductive states. It is obvious that in the network glass case continuum
methods cannot identify floppy modes, which are obtained only by numerical solutions of
matrix equations.
BROKEN SYMMETRY, QUANTUM COMPUTERS, AND SHORS ALGORITHM
The essential idea of our filamentary or percolative theory of random metals near the
metal-insulator transition (MIT) is that in the limit such metals develop a new kind
of broken symmetry even in the normal state. Electronic motion tangential ( ) to
percolative filaments is phase-coherent, just as in normal metals, but normal to the
filaments the motion is diffusive, as it is on the insulating side of the transition. This
3
fractal behavior is what makes it possible for the MIT to be continuous, even in the
presence of long-range Coulomb interactions, which could render the MIT first order in the
EMA, for example the MIT or Wigner transition of electrons in a box.
The presence of limited filamentary phase coherence makes many kinds of novel
effects possible. Consider, for example, hypothetical quantum computers, which have
attracted great interest recently among computer scientists [11]. These process complex
integers (amplitude and phase) rather than merely real integers. Such hypothetical
computers consist of quantum cells (cubits) connected by quantum wires which transmit
amplitude and phase information and thus are exactly the same as the quantum filaments
discussed above. With such hypothetical computers Shor showed [11] how to factor large
numbers in polynomial rather than exponential times by making use of fast Fourier
transforms and matrix methods, which rely on taking advantage of interference effects
which occur conveniently for complex numbers but not for real numbers.
COUNTING IN NETWORK GLASSES: AN EXAMPLE
Counting is essential to our understanding of the remarkable properties of network
glasses. Unlike the electronic cases, where the analysis is greatly complicated by both
long-range Coulomb interactions and phase effects associated with complex wave
functions, the properties of network glasses can be modeled by simple point mass-and-
spring systems. On the one hand, these matrix models with short-range forces and no
quantum effects have proved to be (relatively) easily solved, compared to the electronic
models (apparently insoluble). On the other hand, all the effects predicted by the network
glass models are gradually being observed experimentally, and especially the properties of
the intermediate phase are astonishingly similar to those observed for electronic materials.
Thus, while it would have been easy to be skeptical of these mathematical analogies alone,
it is apparent that they capture most, if not all, of the essential properties of intermediate
phases. Recent work on intermediate phases in network glasses is discussed here by
Boolchand, Thorpe, and Kerner, and the details can be found in their papers.
Apart from the pivotal importance of counting in understanding the properties of
network glasses, there is a second, and equally important, analogy between the methods
Wiles used to prove FLT and the constraint theory of network glasses. The proof relies on
establishing the connection between two sets, modular functions and elliptic curves, that at
first seem to be unrelated, except that their numbers are the same.
In constraint theory one compares the number of spatial degrees of freedom of the
system to the apparently unrelated number of Lagrangian constraints associated with
bonding interactions with localized vibrational frequencies (intact constraints).
These constraints may involve n-body forces with (such as bond-bending forces, n =
3). In conventional continuum treatments, the relevant number is the number of
interparticle forces (off-diagonal elements of the dynamical matrix, each of which
contributes a different interaction line in diagrammatic perturbation theory). Constraint
theory has shown that the relevant number is the number of intact potential interactions in
potential space, not the number of forces implied by real-space derivatives of these
potentials. In other words, spatial coordinates and interaction potential coordinates are
treated as separate and distinct sets. The mean-field condition for the glass stiffness
transition is that the numbers of elements in the two (apparently unrelated) sets are equal.
This point is illustrated in Fig. 1, where the number of vibrational modes with zero
frequencies (cyclical modes) is plotted [12] against average coordination number r in as
glassy network with bond-stretching and bending forces. At r = 2.40 the number of
constraints equals the number of degrees of freedom, and the extrapolated number of
4
Figure 1. The number of floppy modes as a function of r in bond-diluted models of three-dimensional glassy
networks with stretching and bending forces [12]. The inset shows a blowup of the critical region. The second
derivative of these curves shows a peak. This peak resembles the specific heat of a second-order phase
transition, which shows that incomplete relaxation of the models generates the largest number of hidden linear
dependencies of constraints very near the connectivity transition.
cyclical modes is zero. (The smoothing of the kink at r = 2.40 is probably due to the fact
that in the numerically simulated random network the topology is not ideally random.
This leads to redundancies among the constraints.)
INTERMEDIATE PHASES: THERE ARE TWO STIFFNESS TRANSITIONS!
For a long time we have all believed that in percolative problems there is only one
connectivity transition. The first doubts began to appear in Boolchands measurements of
the critical coordination number, which was nearly always close to 2.40. However, even in
cases where there was no evidence for nanoscale phase separation in the critical region, in
other words, in cases where the theory should have worked, there were small discrepancies.
Indeed, the numerical simulations shown in Fig. 1 predict small discrepancies, with a shift
of the critical coordination number to 2.38 or 2.39. Experimentally, the shifts were in the
other direction, to
Those not familiar with constraint theory would probably say that such small
discrepancies are insignificant - after all, in non-equilibrated glasses one should not expect
better agreement between theory and experiment. But to us these discrepancies seemed
significant. In particular, Boolchands ultraprecise Raman data also began to show
evidence for a first-order transition, whereas all percolative models predict a continuous
transition. The problems took definite form in 1998 workshop papers, where Boolchand et
al. showed (see Fig. 2) that an apparent jump in the Raman frequency associated with
corner-sharing tetrahedra in (Ge, Se) glassy alloys occurs at r = 2.46. This is the same
critical value of r as occurs in the density and non-reversible part of the glass-transition
enthalpy (discussed in more detail below). Yet still other Mossbauer experiments showed
that some kind of transition, probably continuous, was happening at r = 2.40. I also found
5
Figure 2. Composition dependence in glasses (r = 2 + 2x) of corner-sharing Raman frequency, non-
reversible enthalpy of glass transition, and molar volume.
6
some marginal evidence for two transitions in elasticity data, with the compressibility
transition occurring at r = 2.40, and the Youngs modulus transition occurring at r = 2.46.
Today we are quite certain that these two transitions mark the boundaries of a new
kind of phase in disordered systems, that we call the intermediate phase. To see what
causes the intermediate phase, suppose we prepare an underconstrained network by bond
dilution. Next we add bonds to the network at random until we reach the first connectivity
transition. At this point the backbone begins to percolate from one face of the sample to the
opposite face. It percolates as a pure filament that neither branches nor intersects itself.
As we continue to add more bonds, two things can happen: we may get new pure
filaments, or one of the old filaments can branch or cross. At the branching or crossing
points locally the network is overconstrained and this increases the strain energy
anharmonically compared to growing new filaments. Therefore the enthalpy, and initially
the free energy, can be reduced by adding bonds selectively to avoid branching and
crossing (smart bonds), and creating new filaments. However, as we add more and more
bonds, and more and more filaments, at a certain point adding one more bond will lead to
crossing or branching, no matter where it is added. This is the upper density limit for the
second transition. It is intuitively plausible, and it is confirmed by numerical simulations,
that the first transition is continuous, and the second is first-order (M. F. Thorpe et al., this
volume). [It should be remarked that it is not surprising that the intermediate phase was
overlooked for so long. It occurs only because the glassy network is not confined to a
lattice. Whenever percolation occurs on a randomly diluted lattice with short hops (short-
range forces), there is only one transition, and it is continuous. It is the off-lattice selective
relaxation character of the glassy network that makes smart bonds and a first-order
transition possible.]
COUNTING IN QUANTUM PERCOLATION THEORY: ANOTHER EXAMPLE
In a d-dimensional sample with N
d
randomly distributed impurities the formation of
phase-coherent ballistic states is blocked (in the sense of Lagrange) by
constraints [13,14]; note that from a counting viewpoint the microscopic nature of these
constraints, orbital or spin, external or electron-electron interactions, is irrelevant; all that
matters is their number. The central result of the filamentary theory of the MIT is the
existence theorem, which states that these filaments can exist providing that the following
condition on the log of this number is satisfied (Eqn. (7) of [8]):
This is a very amusing equation because of the way that it combines real and complex
numbers. On the left hand side we have the exponent that represents the number of
transverse degrees of freedom of our complex, current carrying basis states. Had these
states been real standing waves, the factor 2 would have been absent. The first term on the
right hand side is the number of real constraints generated by randomness. The second
term is the (transverse) areal density of current-carrying filaments, an observable which of
course is also real. Thus the left hand side measures complex quantum dimensionalities,
while the right hand side measures real observable dimensionalities. In a simple but very
fundamental way this equation describes all the implications of the quantum theory of
measurement for the transport properties of random metals [14]. Quantum percolation
theory explains all the experimentally observed critical exponents and prefactor sign
reversals which are observed [15,16] in uncompensated random metals near the MIT such
as Si:P and Ge:Ga. It applies to the scaling phase, which exhibits power-law behavior over
7
a range 100 times larger than that found for magnetic critical phenomena. Very close to the
continuous MIT, even the purest materials exhibit the effects of compensation, and the
residual resistivity associated with scattering from compensating impurities dominates the
transport properties, which revert to the conventional behavior predicted by the EMA.
COMPARISON WITH SCALING THEORY
The dependence of (1) on dimensionality is strongly suggestive of scaling theories that
have been developed to describe critical behavior of magnetic phase transitions [17]. It
seems, however, that the results given in [8,9] differ fundamentally from those in the
magnetic literature in two important respects. The latter rely on Bose, not Fermi, statistics,
and hence contain no destructive interference effects. The interparticle forces of the latter
are of short range, while electron-electron interactions are long range. Whether these two
differences are necessary or sufficient to explain the large qualitative differences between
the properties of random magnets and random metals has been unclear for a long time. It
was even suggested [18] that the experimental data [15] may have been in error, a
suggestion which has recently been laid to rest [16].
Detailed comparison of filamentary quantum percolation theory with magnetic lattice
percolation theory [17,18] has shown [10] that both Fermionic destructive interference and
long-range forces are necessary and sufficient to produce a consistent and successful theory
of impurity band random metals such as Si:P and Ge:Ga. The destructive interference
suppresses the divergence of the specific heat which is otherwise a characteristic of quasi-
one dimensional ( d* = 1 ) Fermionic systems embedded in a d-dimensional matrix. This
destructive interference is represented mathematically by a non-crossing condition for
semiclasical percolative paths similar to one that is already known for the integral quantum
Hall effect at large n. (The elliptic curves which play an essential part in the proof of
Fermats Last Theorem also satisfy a non-crossing condition [19], which suggests that
arithmetic algebraic geometry [briefly, modern arithmetic] may have a lot to offer in
treating problems involving many Fermions.)
In the limit Fermion statistics combined with long-range interactions cause d
to be replaced by d + d* = d + 1. One can identify d* with the fluctuations of the
component of the internal electric field that is locally tangential to the filament. Or one can
introduce local times for Fermi-energy electrons moving along the filament. Then Newton
is replaced by Einstein, and because of internal fields the filamentary paths fluctuate
dynamically not only in space, but also in their local time It is amusing that the
concepts of special relativity, originally developed to explain non-linear aspects of the
Doppler effect, should reappear in the context of critical fluctuations in random metals.
If the intermediate phase has a distinct topological character that is associated with
off-lattice disorder, then one can immediately infer that it can occur in any disordered
system that either has long-range forces, or can self-anneal. So I reasoned, and this led me
to re-examine all the experimental data on strongly disordered systems near a connectivity
transition. Two such systems immediately spring to mind: the simpler system is the metal-
insulator transition in semiconductor impurity bands, such as Si:P; following Shockleys
rule, we discuss it first.
There are two kinds of data on the metal-insulator transition in Si:P, both taken some
20 years ago. At that time everyone assumed that there was only one transition. This
transition was supposed to occur continuously in both the conductivity and the
coefficient of the linear term in the specific heat at the same value of uncompensated
dopant density n and with related exponents, as both observables were supposed to be
continuous functions of n.
8
The trouble with this assumption was that when it was applied to the experimental
data, the two sets of data, transport and thermal, did not lie on smooth, power-law curves
(see Fig. 3 ). However, as everyone at that time was certain that there was only one
transition, this problem passed unnoticed. Naturally, when I re-examined the by-now-all-
but-forgotten thermal data some twenty years later [14], I immediately saw that the critical
density for the thermal transition was actually much larger than that for the transport
transition (by almost a factor of 2).
All the effects of the intermediate phase that we have been discussing are connected
with filamentary coherence and finite-size scaling. When a few minority dopants are
present in the sample, two things happen, both based on interruption of coherence. Very
close to the low-density continuous MIT, the conventional incoherent MIT occurs, which is
smoother and has larger density exponents than the coherent transition. This conventional
transition is of little interest here, except that it shows how different the intermediate phase
is from a Fermi liquid.
The second point of interest is that the presence of the minority dopants creates a
natural length scale, associated with the average minority-minority spacing, that is larger
than the majority-majority spacing. Physically this larger length scale is that associated
with the residual resistance [8], a quantity that does not arise in many modern scaling
theories of metallic behavior. The residual resistance is, of course, a very important
quantity, as it eliminates divergences in the conductivity as It also provides a
natural platform for filamentary counting. One constructs Voronoi polyhedra around each
minority impurity, and counts the number of filaments crossing such polyhedra in various
limits. Thus here counting shows up as a very basic operation. The omission of this
construction has led some authors to the conclusion that the residual resistance of metals is
not associated with impurities at all, but depends on interactions of the electromagnetic
field with the environment, which is nonsense [8].
Counting has important implications for scaling in general. In continuum scaling
theory critical densities are irrelevant constants, and only the critical exponents are
universal for a given class of interactions, independent of coupling strength, at least over
limited ranges. Moreover, these critical exponents are in general irrational numbers. In
Figure 3. The electronic specific heat coefficient for Si:P, showing both the continuous transport
transition at n
c
and the first-order thermal transition at n
cb
. The dashed line shows the value expected for a
Fermi liquid. The transition from the Fermi liquid to the filamentary metal occurs at n = n
cb
, and this
transition is first-order [14].
9
network glasses, for simple alloys, there are only a few classes of intact constraints, and the
condition that their average number/atom be integral automatically leads to average
coordination numbers that are simple small fractions (such as 12/5). The existence of such
magic fractions is a direct consequence of treating the bonding interactions as a set
separate from the set of spatial coordinates in other words, the bonding interactions form
a space different from real space. By connecting these two separate spaces we can
identify one, and possibly even both, of the connectivity transition compositions. The
process of connecting two apparently different spaces to prove a certain arithmetic result is
much the same as in the proof of FLT.
It has often been conjectured that the results obtained from continuum lattice scaling
theories are universal. Specifically for given classes of interactions, for example, the
critical densities for different diluted lattices cubic, hexagonal, etc., will differ, but the
critical exponents will be the same. This is no doubt correct, but it does not include the
effects associated with intermediate phases in disordered solids, which are a new
phenomenon that lies entirely outside the framework of continuum lattice scaling theory.
The new phase transitions and the intermediate phase cannot even be described properly in
terms of diluted lattices with their single connectivity transitions. A more flexible and more
abstract description is required, that uses the methods and concepts of modem mathematics.
In particular, one must be satisfied to describe the properties of sets, as the presence of
disorder makes it impossible to describe fully the properties of the individual elements of
the sets.
BASIS FUNCTIONS IN FILAMENTARY METALS
Suppose we have an impurity band in the intermediate phase. In this phase the metallic
states are centered on arrays of filamentary, non-crossing paths that extend from electrode
to electrode. These paths are similar in some respects to the Self-Avoiding random Walks
(SAWs) that are used in statistical mechanics to describe the mathematical properties of
diluted magnetic lattices. There are also important differences. In the magnetic case the
SAWs are closed loops with pseudovector symmetry, whereas the electrical paths have
vector symmetry. In the magnetic case we are concerned with minimizing the free energy
associated with magnetic susceptibility, essentially an equilibrium property. In the
electrical case, the metallic conductivity contributes to dielectric screening of internal
electric fields, thus it also can be varied to minimize the free energy. Because of quantum
mechanics, the kinetic energy associated with transverse localization of charge carriers on
filaments increases as the filaments become more closely packed, eventually delocalizing
the electrons and leading to the transition to the Fermi liquid state at higher densities. This
does not occur in the spin case, as spins are already localized objects with no intrinsic
kinetic energy.
The generalization of Fermi liquid wave functions, indexed by the continuous
momentum p and represented by the EMA wave functions to the discrete case
of filamentary arrays, is not difficult. One assumes that the real-space centers of density of
each filament j are known, and denotes the corresponding path by Longitudinal wave
vectors are oriented parallel to the local tangent to the path. There are no transverse
wave vectors, only a local transverse localization length. In the normal state in the absence
of a magnetic field these wave vectors can be used to construct basis functions for each
filament. The actual wave functions near the Fermi energy will be time-dependent linear
combinations of individual filamentary wave functions that minimize the free energy by
screening internal electric fields.
10
BROKEN SYMMETRY IN THE SUPERCONDUCTIVE AND NORMAL STATES
The key feature of the BCS theory of metallic superconductivity is the formation of
Cooper pairs, which become the Landau-Ginzburg order parameter with 1 =
and As the volume of the system tends to the number of possible
choices for 2 (even when these are restricted by the isoenergetic constraint ) also
tends to but from this infinitely large set only one state, the time-reversed one, is used to
form the Cooper pair. In other words, the cardinality of the set of states consisting of the
time-reversed state is lower than that of the set consisting of all the isoenergetic states.
This is a characteristic feature of continuum or effective medium models in which
scattering by some kind of disorder is added after the basis states have been chosen.
In strongly disordered systems, such as random metals near the MIT, the situation is
quite different. In order to explain the critical properties of such systems one must select
the correct filamentary basis states at the outset. This is done variationally, and it affects
both normal-state and superconductive properties in many ways that are radically different
from normal metals, where all the isoenergetic states are essentially equivalent, as in
Landau Fermi liquid theory. The special properties of filamentary metals are the result of
atomic relaxation that leads to preselection of basis states even in the normal state, where
there are only two states per filament, and with In other words,
Figure 4. There are two components in the filamentary model, the CuO
2
planes (A), and the impurity bridges
combined with secondary metallic planes (B). This Figure shows the density of electronic states near the
Fermi energy, for with the energy scale set by the resonant bridging impurity width W
R
.
There is a strong peak in due to the impurity resonance pinned by electron-ion polarization
energies and the anti-Jahn-Teller effect, and a strong dip in N
A
(E) due to electron-electron Coulomb
interactions. The product has peaks near At the optimal composition
there are only extended states for The scattering rates are also shown for the optimal
composition. They are much larger for the localized states, than for the extended states,
11
for the one-dimensional filaments the Fermi surface collapses to two points, and this has
happened in the normal state because of quantum percolation.
Another extremely unexpected feature of filamentary states is that they maximize
(minimize) the conductivity (resistivity). This variational property means that the
dynamically optimized filamentary states already include all many-electron interaction
effects, including those of electron-electron scattering which give rise to T
2
resistivities in
Landau Fermi liquid theory. That such many-electron interactions are absent in impurity
band random metals has been shown by a filamentary analysis [8,9] of critical exponents in
Si:P [15] and neutron-transmutation-doped, isotopically pure Ge:Ga [16]. The dominant
remaining source of scattering in the cuprates is thermal excitation into localized states
with energies outside the W
R
resonance region [5] shown in Fig. 4. (This disappearance of
the Coulomb interaction between discrete and dynamically optimized filamentary current-
carrying states is analogous to the existence of zero-frequency floppy modes in the
intermediate phase of network glasses.)
FILAMENTARY MODEL FOR HTSC
Even without measurements of the properties of the cuprates it was clear to crystal
chemists and materials scientists that these multinary compounds would be extremely
unusual from a structural point of view [1]. In addition to containing rare earths and
oxygen, these pseudoperovskites nearly always contain Cu, an element that in one
oxidation state shows a greater diversity in its stereochemical behavior than any other
element. This observation, together with the extremely anomalous transport properties,
certainly bodes ill for any microscopic theory of the cuprates which is based on the EMA,
as that approximation ignores the flexible material properties of Cu, and the ferroelastic
properties of the perovskite family, altogether. (All materials with unusual properties, from
Si to conjugated hydrocarbons to DNA, conform to the central principle of organic
chemistry, structure is function.) Undeterred by these inescapable ground rules, almost
all the theories developed so far, such as [2] and [3], are based on the EMA, augmented
only by good intentions and wishful thinking; given the richness and complexity of
materials science, this is unlikely to suffice.
The filamentary theory of HTSC diverges from EMA theories in two basic ways: it
incorporates an extensive knowledge of the experimental data [1], and it has a sound
mathematical foundation in the filamentary theory of the MIT in semiconductor impurity
bands [14], which supersedes inadequate EMA theories of the Fermi liquid or Landau-
Ginzburg type [2,3]. Because of interlayer ferroelastic interactions the metallic CuO
2
planes are partitioned into metallic nanodomains separated by semiconductive domain
walls. A specific filamentary path was envisaged [20] that connects metallic CuO
2
planes
with secondary metallic planes (CuO chains or BiO planes) via resonant impurity states
located in semiconductive planes (such as BaO) sandwiched between the metallic planes,
as shown in Fig. 5. In addition to the bridging impurity points most samples contain two
kinds of extensive defects which act as blocking lines or layers. Blocking macroscopic ab
planar layers explain the usually semiconductive c-axis resistance; this aspect of the data
has received too much attention [3], as these blocking layers are essentially extrinsic and
can be avoided in some cases, by relieving interlayer strain energies [21], or by overdoping
[22]. The interesting extensive defects are intraplanar semiconductive nanodomain lines in
the metallic planes; these form grids, and each cell of a grid is connected to a square in an
adjacent metallic plane by resonant (metallic) tunneling through a bridging impurity. The
experimental evidence for the existence of such buckled cellular grids is discussed here by
Jung. It is difficult to obtain evidence for spatial inhomogeneities on this scale, but the
evidence available has been growing steadily if slowly.
12
Figure 5. The variationally optimized percolative filaments, shown in cross section, follow planar locally
metallic CuO
2
layers until they approach a domain wall which is locally insulating. The zigzag metallic path
is continued by resonant tunneling through a state pinned at the Fermi energy associated with a defect, such as
an oxygen vacancy. The next segment of the path is that of a chain, and this segment terminates at a
chain O vacancy, where the zigzag path is continued by resonant tunneling back to a CuO
2
layer, and so on.
This model is designed for YBCO; in LSCO the tunnel paths simply connect CuO
2
layers. Such filamentary
paths should never be confused with stripe phases or pinned charge density waves, which are incidental
minority insulating phases.
There is also dramatic evidence for the existence of nanoscale spatial inhomogeneities
in Debye-Waller factors measured by ion channeling, which are very sensitive to a few
large out-of-plane atomic displacements and show striking precursive anomalies at T
c
;
these are absent from neutron diffraction data which measure EMA properties [23,4]. It is
the electronic structure associated with Fermi-level pinning defects which experimentalists
tune when they adjust oxygen concentrations, refine by annealing or observe as aging or
quenching effects. To understand transport properties one must understand the topological
connectivity of these states, which is scarcely possible within the EMA.
Within a single-particle picture the Fermi-level pinning metallic states can be
represented as a narrow band of resonant states of width few meV the valence
band width as shown in Fig. 4. Ordinarily one might expect that such states
would be unstable against a Jahn-Teller distortion, and indeed it has been stated [3] without
proof or citation that this is always the case. In fact, it is easy to find exceptions where
such peaks are located self-consistently (with respect to both electronic and atomic
coordinates) at E
F
, for example, in many total energy calculations. Moreover, Fermi-level
pinning by impurity, surface or interfacial states at metal-insulator junctions (Schottky
barriers) is one of the basic principles of semiconductor device physics. The error [3] arose
from simplistic inclusion of only one-electron interactions and neglect of both core-core
and non-local electron-electron interactions. An amusingly similar error (sometimes called
the Wentzel mistake), also on the subject of instabilities and superconductivity, led
Wentzel to suggest [24] that the Bardeen-Frohlich attractive electron-phonon interaction
was not the correct mechanism for simple metallic superconductivity.
What is needed for the cuprates is a general mechanism for frustrating the Jahn-Teller
effect. This is provided by a self-screening atomic relaxation mechanism which involves
13
long-range Coulomb interactions not representable as local single-particle energies [25].
The attractive self-screening energy of the already narrow band of resonant states is
maximized by further narrowing the band and centering it on E
F
. This anti-Jahn-Teller
effect has the congenial feature that it is expected to be especially effective in strongly
ionic materials where the long-range Coulomb forces are only weakly screened by a few
metallic electrons. This is exactly the situation in the cuprates, which are close to a metal-
insulator transition; it is also the case for impurity band random metals [8-10], where this
weak screening is responsible for the anomalously small exponent [15,16] which lies
below the limit expected from scaling theory [18] with Boson statistics and short-
range forces only, and far below the value of 3/2 predicted by some one-electron EMA
theories.
The narrow resonance region of width W
R
was previously portrayed [5] as a peak in
the density of electronic states centered on E
F
, but this need not be the case. All that is
necessary is that in this region the scattering rate be extremely low compared to that at
higher energies outside this region. Thus the picture we have now is that shown in Fig.4.
The density of extended states is depressed relative to that of the localized states by
Coulomb interactions, as happens for the random metal on the insulating side of the MIT
[16,26]. This density of states would go to zero at E = E
F
and T = 0 were it not for the anti-
Jahn-Teller effect, which leaves a residue of carriers at T = T
c
which is about half that for T
= T
0
[27]. The pinning of the most polarizable filamentary states to E
F
by Coulomb
interactions is similar to the energy-level reordering responsible for the pseudogap in
random metals [26].
NORMAL-STATE TRANSPORT
In a normal metal electron-phonon interactions typically contribute a temperature-
dependent term to the resistivity proportional to with For large crystalline
disorder, as in metallic glasses and thin films quenched at low temperatures, electron-
electron scattering is strong and In certain cuprates, notably those without secondary
metallic planes involving metallic elements other than Cu (Bi or Hg), the ab planar normal-
state resisitivity is linear in T approximately from T
0
to T
h
, where T
0
is close to T
c
and T
h
is the high-temperature limit of compositional stability [5]. This is a very
remarkable result, as in cases where T
c
is low, the ratio T
h
/T
0
has been observed to be as
large as 100. However, it holds only for those samples whose composition corresponds to
a maximum in T
c
; increasing doping causes to cross over to the Fermi liquid (strong
electron-electron scattering) value of 2. This means that a satisfactory theory should
contain some continuously tunable factor which will alter both anomalies at the same time,
and this factor should be responsible for the MIT as well. As the reader will realize, these
demands are very severe. He will probably not be surprised to learn that they are met by
the authors filamentary theory [5], but not by any theory based on the EMA, such as [2] or
[3].
The limitations of EMA models becomes obvious when one examines the field
theories developed by various authors [28] to implement Andersons suggestion that
electric (holon) and magnetic (spinon) effects are somehow separated in the cuprates [3]. It
is clear that such a separation is essential if the magnetic moments of the rare earths are not
to quench superconductivity, but Anderson gives no microscopic explanation of how this
can happen; it is merely one of his axioms, or, as he prefers, dogmas. Given this dogma,
one is able to explain microscopically why normal-state transport anomalies exist and are
loosely correlated with the optimization of T
c
. However, one is unable to derive any
functional form for the temperature dependence of the resistivity, much less to explain why
14
without assuming what was to be derived. Even the temperature scale ratio T
h
/T
0
is
merely the ratio of an inelastic high-temperature scattering rate to T
c
, which is yet another
assumption, which turns out to be incorrect, as we shall see.
The separation (spin bags [2]) of magnetic and electrical effects, moreover, need not
be axiomatic. It is derivable in the filamentary model simply by observing that the
magnetic states are all localized, and that the separation of the extended current-carrying
states from the localized states [14,5] automatically separates spin and Cooper-pair-forming
states. Note, however, that this separation cannot be carried out correctly within the EMA
because in that picture one is unable to count [6,7,19] the states that are being separated.
The importance of counting is illustrated convincingly by the much simpler case of random
impurity band metals, where the EMA has failed in calculating critical exponents [8,9], By
contrast, the dogmatic holon-spinon separation [3] becomes, in the filamentary model, what
one would naturally expect in optimized HTSC because of the success of the filamentary
model for the closely related impurity band MIT. It is nothing more than intralayer
nanoscale phase separation, driven by interlayer ferroelastic misfit forces.
Because the CuO
2
planes are divided into an irregular checkerboard or grid of
nanodomains by intraplanar domain lines which have semiconductive gaps
currents can flow only along filamentary paths passing through interplanar resonant
tunneling centers (impurity bridges). In YBCO, for example, such centers might be
represented by the much-studied apical oxygen sites between Cu atoms in the CuO
2
planes
and the chains, selectively associated with vacancies on the later. For optimal
doping there are two such centers per CuO
2
planar domain, one source and one drain.
When there are fewer than two, the sample is underdoped, and when there are more than
two, it is overdoped (see Fig. 2 of [5]). Thus the average integral (bridge/domain) ratio is
the continuously tunable factor mentioned above; there it was shown that when this factor
is two all the normal-state anomalies are explained, as is the optimization of T
c
. It was also
explained why overdoping depresses T
c
and increases from 1 to 2, at the same time
producing the observed anomaly in the Hall resistance [29].
An important historical point is that the fact that all the normal-state transport
anomalies can be explained by the existence of a narrow, high-mobility band pinned to E
F
was first explained in [29]. At that time the explanation was not generally accepted
because it was not accompanied by a specific structural model that explained the origin of
the narrow band. Such a model is shown in Fig. 5, and it is the only such model that has
been advanced. The narrow high-mobility band itself is the only way of explaining the
normal-state transport anomalies, so that together with Fig. 5 it may be taken as the only
satisfactory, perovskite-specific model for HTSC.
What happens to underdoped samples? In the YBCO case, underdoping produces
more O vacancies on the chains that generate the crystalline orthorhombic
symmetry. These chains are almost surely responsible for the phase shifts at twin
boundaries where the chain orientations rotate by which experimentalists and many
EMA theorists often like to describe as d-wave superconductivity. This is an EMA (or
Fermi liquid) misnomer that is entirely inappropriate for the non-Fermi liquid intermediate
phase. It implies a fundamental significance of what amounts to a non-bulk edge or surface
effect, which is seen to be trivial as soon as one realizes that the b-axis chains are
essentially involved in constructing filamentary paths, so that the observed phase shift is
unavoidable. The chain segments become shorter as x increases, and although probably
only short chain segments are needed to bypass intraplanar CuO
2
domain walls, it is clear
that the ab planar resistance will increase as the chain segments shorten.
Aside. Many experiments have shown that residual states exist within the pseudogap;
these residual states are also often described as the result of d-wave superconductivity.
In fact, the observed residual states are very similar to those predicted by [26]. Thus if
15
there are dopants in semiconductive domain walls that generate a pseudogap, then this
would easily account for the experimental observations, without nonsensically using Fermi
liquid terminology to describe the non-Fermi liquid intermediate phase.
In fact, we expect increasing phonon-assisted currents across oxygen vacancies within
the chains. The importance of these will increase with x and (the orthorhombic
plateau in T
c
(x)), it is possible that virtual phonon exchange at these vacancies will provide
a stronger attractive interaction for forming Cooper pairs than phonon exchange at the c-
axis impurity bridge resonances does. The width and strength of the density-of-states peak
of the latter may not depend on the oxygen mass, as they may well be associated with
collective relaxation and optimization of many internal coordinates. This would explain
the disappearance of the oxygen isotope effect for small x [30]. On the other hand, the
electron-phonon interactions at chain vacancies promote phonon-enhanced coherent
currents across these micro-weak links, enhancing the local energy gap. When this effect is
linearized with respect to vibrational amplitudes, it may still be equivalent to a local
Bardeen-Frohlich interaction and may thus give rise to what resembles a normal isotope
effect for large
After the above was written, a very important paper [31] appeared concerning the
relaxation of T
c
in YBCO after abrupt release of pressure. The relaxation was found to
follow the form of a stretched exponential, with The key parameter
of interest is the dimensionless stretching fraction which turns out to be highly
informative. The stretched exponential can be derived from a microscopic model. The
model involves diffusion of excitations in a configuration space of dimension d*
p
to
randomly distributed traps. As time passes, all the excitations near the traps have
disappeared, and only excitations distant from the traps remain. The latter must diffuse
further and further. This leads to the stretched exponential and to
At first, it might appear that all that has been done is to replace one empirical
parameter, with another, d*
p
. In fact,
for homogeneous glasses. (The dopants in a well-annealed and homogeneous HTSC
presumably form a glassy array.) Here d = 3 is the dimensionality of Euclidean space.
The key factor now is p. Comparison with experiment and several very accurate MDS
showed that for homogeneous glasses p is nearly always 1 or 2; it measures the number pd
of interaction channels involved in diffusion of excitations in d dimensions. In metals
where phonon scattering dominates the resistivity, one of these channels is always e-p
interactions. However, if other classes of interactions are present, there may be other
diffusive interaction channels as well. It is easy to see that adding channels increases the
stretching factor, which is
Mathematically the simplest and most rigorous example with p = 2 is provided by
quasicrystals, where the Euclidean coordinates r become and the Penrose projective
coordinates are Motion in space (the first d channels) involves phonons and
produces relaxation, while motion in space (the second channels) involves phasons,
which only rearrange particles without diffusion or relaxation. In the ideally random quasi-
crystal a given hop may tale place along either or Thus
16
where f
p
measures the effectiveness of hopping in pd channels, only d of which is
associated with relaxation. For an axial quasicrystal, which is quasi-periodic only in the
plane normal to the axis, the calculation is somewhat more complex. There are five
channels, three in space, and two in space, so that fp = 3/5, and d*p = 9/5. Thus
in excellent agreement with MDS
15
which give
From the value one can rigorously infer that p = 1, and thus only electron-
phonon interactions can cause HTSC. The proof is based on grouping the interactions
involved in diffusive relaxation into classes of interactions that are effective (such as
electron-phonon interactions) and ineffective (such as electron-magnon interactions) [32].
Because only the electron-phonon interaction is needed to explain all other
interactions (such as electron-[magnon, plasmon, any-old-whaton] interactions) are
excluded by experiment [31,33]. The remarkable aspect of this experiment and theory is
that the conclusion transcends all the details of structure and large-scale relaxation around
dopants in these complex materials. Note that once again, the success of this approach
rests on identifying two different sets, interaction space and real space, and then
connecting them, just as in constraint theory and the proof of FLT.
CHAIN LENGTHS AND LOW TEMPERATURE CUFOFF T
0
In the filamentary theory there is a close relation between average chain lengths L and
the lower cutoff (or pseudogap) temperature T
0
in for optimally and
underdoped samples:
(see (1,2) of [5]); here This relationship gave good results for the YBCO T
0
(x)],
which are linear in x with T
0
(0.1) 150K for T
c
optimized 90K. It should be mentioned
here that many samples appear to give T
0
less than T
c
, but these are probably
inhomogeneously overdoped. For optimally doped (x = 0.1) samples the value T
0
(0.1)
150K has been confirmed for single crystals and for thin films grown by several methods,
and even fine-tuned with low-energy electron irradiation [27].
In a large magnetic field T
c
is suppressed, unmasking or exposing the normal-state
resistivity at temperatures lower than T
c
(H) for H = 0 in relatively low T
c
cuprates ( T
c
40K), such as [35], The central unmasked single-crystal results stressed by
[34] are that for underdoped compositions p
ab
(T) increases and becomes semiconductive
below T
0
. Even more significant, however, is the disappearance of the pseudogap (the dip
in normal-state resistivity between T
c
and T
0
) in large magnetic fields. This disappearance
is not discussed at all, as it is virtually inexplicable within the EMA. In terms of our
topological model of the intermediate state, the explanation is immediate. The large
magnetic field replaces the self-organized, non-crossing coherent filamentary basis states
with vector symmetry by quasi-circular orbital states with pseudovector symmetry, thereby
restoring Fermi liquid character, including strong electron-electron scattering, to states near
E
F
.
Because chains represent the ideal local structure for filamentary currents, microscopic
probes of the local chain structure in untwinned samples of YBCO are of great interest.
Recently two experiments have done this, with results that cannot be explained by EMA
models based on bulk energy bands and bulk phonons. The first experiment [36] revealed
systematic changes in scattering strength with doping of longitudinal optic (LO) phonons
17
propagating in the basal plane normal to the chain direction. These phonon spectra contain
a pseudogap which can be explained [33] as the result of short-range ordering of chain
segments that alternate in oxygen filling factors. The changes in scattering strength are
more interesting, as they turn out to be direct measures of phonon coherence along
filamentary paths, and they change abruptly in as the composition passes
through the metal-insulator transition near x = 0.4. There is also a second abrupt change in
scattering strengths centered on x = [the transition between the T
c
= 60K and 90K
plateaus] as the smaller filling factor passes through ; this change is just what one would
expect from percolation theory, and from it one can successfully predict the change in the
T
c
ratios of the two plateaus.
In the second experiment [22], the longitudinal magnetoresistance in slightly
overdoped untwinned YBCO was studied; in these samples the c-axis resistivity is linear in
T, just as the ab planar resistivity is, which means that the coherent percolative paths are 3-
dimensional. Again the results show strong anisotropy that is connected with coherent
current flow along the chains.
NMR RELAXATION AND THE SPIN PSEUDOGAP
Anderson has discussed (see [3], Fig. 3.27) the observation of anomalous non-
Korringa planar Cu relaxation in various cuprates, and states that this anomalous
relaxation seems to be one of the common features of the high-T
c
. state, but it is actually
relatively more pronounced for somewhat lower T
c
materials., so is not closely related to
superconductivty. This author concurs, and would add that in his opinion in most cases
(LSCO is an exception) all that spin-scattering experiments are measuring is spin relaxation
in pockets of insulating material with compositions which can be quite different from those
of the superconductive bulk. For example, this effect is quite small in optimal
with but is much larger in optimal Note that the oxygen diffusivity is
very high in YBCO, but not the Sr diffusivity in LSCO, so that while it is possible to make
very nearly homogeneous samples of the former, this is not possible for the latter. Thus in
general there is no connection between T
0
and the spin pseudogap T
sg
, nor should we expect
to find one, except for materials like LSCO, where magnetic microphase inclusions may be
absent. In that case both T
0
and the spin pseudogap T
sg
can be related to the resonance
width W
R
.
COMPOSITION DEPENDENCE OF THE ENTROPY OF THE VORTEX PHASE
TRANSITION
The physics of magnetic vortices in the mixed state is extremely complex, and it is
nearly always treated from the point of view of the EMA, although it is clear that if the
sample is spatially inhomogeneous the vortices will nearly always localize preferentially in
regions of lower T
c
. In the authors view this is the most natural explanation of the step -
kink - peak phenomena in vortex lattice melting which have attracted much attention from
experimentalists [37]. However, two-phase models have many adjustable parameters and it
would appear that this greatly limits what can be gained from the analysis of such
phenomena.
Thus most of the discussion of these phenomena by experimentalists has focused on
the fact that the entropy of vortex lattice melting is much larger than would be expected if
the vortices are treated merely as point objects [38]. This can be easily explained by taking
account of changes in the nonlocal structure [39] of the vortices near T
c
. There is, however,
18
one very puzzling feature of these data which is explained quite easily by the present
theory. This theory is a counting theory, and thus it is well-suited to studying the entropy
of the vortex phase transition. In Bi
2
Sr
2
CaCu
2
O
y
single crystals it is observed [37] that this
entropy is several times larger for overdoped than for optimally doped samples, especially
at low T. Ordinarily in the EMA one would expect that T
c
would reach its maximum value
at the composition where N(E
F
), the density of electronic states in the normal state, has its
maximum value, that is, at optimal doping E
F
coincides with a peak in N(E). In such a case
the entropy should reach its maximum at the same optimally doped composition. However,
in the present model N(E
F
) is larger in the overdoped state than in the optimally doped
state, so giving a larger entropy of melting. The transition temperature decreases in the
overdoped state because the electrons at the Fermi energy spend more time in Fermi-liquid-
like states in the CuO
2
planes, where the electron-phonon coupling is weak, and less time at
the discrete resonant tunneling centers, where it is very strong. See Eqn. [3] and Fig. 2(c)
of [5].
INTERMEDIATE PHASES AND FIRST-ORDER PHASE TRANSITIONS
The mathematical character of intermediate phases is characterized by some discrete
features (the filaments) and some continuous features (the off-lattice space in which both
the network glasses and the spatially disordered impurities of the electronic examples are
embedded). One felicitous consequence of mixing discrete and continuous features is that
the lower-density (or first) transition from the disconnected (or insulating) phase to the
intermediate phase is continuous, while the higher-density (second) transition from the
filamentary phase to the effective medium (overconstrained, or Fermi liquid) phase is first
order. This asymmetry is quite striking, and it cannot be explained by an entirely discrete
lattice model, or by an entirely continuous effective medium model. Both of the latter
contain only one transition, and it was the similarity in this respect that has led many
people to suppose (mistakenly) that there is a universal character of phase transitions that
can be independent of the discrete/continuous dichotomy.
In this workshop both Boolchand and Thorpe discuss these two phase transitions in
convincing detail. In Figs. 3 and 6 the two transitions are sketched for impurity bands in
Si:P. The two critical densities are separated by a factor of 2, and there is no doubt that the
first one is continuous, and the second is not.
The layered cuprates that form HTSC are complex multinary compounds, and
preparing samples that are microscopically homogeneous is not easy. Of course, unless
such homogeneity is achieved, the second transition will be greatly broadened, and it will
be difficult to show that the second transition is first-order. So far, three successful studies
have reported first-order phase transitions: (1) near x = 0.21 in (after
annealing for several months [41] at high T and constant O partial pressure), (2) near =
0.19 in (also after annealing at constant composition [42]), and near x =
0.95 in (carefully designed chemical and thermal history, including slow
cooling [43]). Normally ones observes only parabolic T
c
(x)s. Self-organization is not
easily achieved, that is why only in a few experiments are the two HTSC transitions
separated to give trapezoidal T
c
(x)s.
19
Figure 6. A sketch of the thermal data on Si:P, showing both and in relation to the transport
transition [40].
GENERAL CONCLUSIONS: ANALYTICITY AND CARDINALITY
The essence of filamentary percolation theory is that it replaces the analyticity of field
theory, which has been an excellent guide to the physics of the old metallic
superconductors, by the countability, or cardinality of modern set theory. The
justification for this in HTSC is an internally consistent theory of the crucial experimental
properties, notably the normal state transport properties, as functions of both temperature
and composition. Analytic models [3] can be constructed which account qualitatively for
the temperature dependence, for example, but when these are examined in detail it soon
becomes apparent that they are not capable of accounting precisely for the observed
functional forms, either the resistivity linearity in T or the linearity and magnitude of the
composition dependence of the low-temperature pseudogap linear resistivity cutoff T
0
(x) in
YBCO
7-x.
The filamentary theory exhibits many similarities between the MIT in impurity bands
and that in the cuprates, especially YBCO, which is the most homogeneous and
macroscopic-defect free of the cuprates, because of its chains. In this sense the theory is
self-proving (self-testing), because one would expect those similarities to be most
pronounced for the best material. The disappearance of many of these properties from
LSCO and (Y,Ca)BCO alloys shows that the theory is both selective and incisive, for it
successfully differentiates those properties which EMA models cannot explain because
they are microscopic, from those which it cannot explain because they are macroscopic.
Both electronic theories are reduced to their simplest form in the intermediate phases of
network glasses. The overall similarity of the three systems shows that it is their shared
20
(discrete/continuous) topology that is responsible for their remarkable properties, from the
reversibility window to HTSC itself.
THE BIG, BIG PICTURE
The differences between pure continuum models (or the effective medium
approximation, EMA) and discrete off-lattice models (embedded in a continuum) are huge,
not only conceptually, but also psychologically. Scientists who have been educated to
think only in terms of continuum models, and who have developed their own concepts in
that framework, often find themselves enslaved by that framework. (A concrete analogy,
close to home, is that of the experimenter married to his equipment, or the computer
scientist married to his software.)
In this volume one can find many examples of problems and principles that certainly
go beyond the limits of the continuum approach. Perhaps it is not surprising that many of
these examples occur in the context of network glasses, as these materials are obviously
unsuited to continuum treatments. On the other hand, that nanodomains should exist in
perovskites and pseudoperovskites should come as no surprise, as this family of materials
has been known to be ferroelastic (and prototypically so) for more than 50 years. Yet Jan
Jungs elegant survey in this volume of these nanodomains reaches conclusions that are
politically very unpopular, as anyone who has attended one of the numerous conferences on
HTSC can testify.
One can ask oneself just why such an obvious consequence of one of the most general
principles in the materials properties of oxides should be deemed politically incorrect.
More than 10 years ago, when the subject of HTSC was still in an embryonic stage, I
believed that most people were still being conservative; perhaps there was not enough
direct evidence for the existence of nanodomains, and their dimensions remained to be
determined. As Jung shows, this is certainly not the case today.
Another explanation is that most people feel that if their own experiments do not
directly exhibit nanodomain features, then those features are not needed to explain their
results. This is not so nave as it sounds: in the theory of critical exponents of continuous
phase transitions, some very sophisticated theorists have postulated that (loosely speaking),
all such transitions are equivalent. (This is the concept of universality.) Such on-lattice
transitions never exhibit an intermediate phase. Thus, it is the existence of intermediate
phases in self-organized disordered systems that causes the breakdown of universality.
We are now at the crucial point, both conceptually and psychologically. It is widely
admitted, even by almost all those who still adhere to continuum descriptions of HTSC,
that the intermediate (often called non-Fermi liquid) phase is responsible for HTSC. If
this is the case (and all the experimental evidence so indicates), then the fact that no
microscopic continuum model is known that produces an intermediate phase between the
insulating and Fermi liquid phase, becomes decisive. There is such a model in discrete
theories, and it is very successful in describing intermediate phases in network glasses and
semiconductor impurity bands, as discussed elsewhere in this volume. It follows that a
discrete network model is the only practical model for HTSC. Of course, a successful
continuum model based on the EMA may be developed someday for HTSC, about the time
that there is pie in the sky.
REFERENCES
1. J. C Phillips, Physics of High-T
c
Superconductors, Academic Press, Boston 1989.
2. J. R. Schrieffer, X. G Wen and S. C Zhang, Phys. Rev. B 39, 11663 (1989).
21
3. P. W. Anderson, Theory of Superconductivity in the High-Tc Cuprates, Princeton Univ. Press, Princeton
1997.
4. J. C. Phillips, Physica C252, 188 (1995).
5. J. C. Phillips, Proc. Nat. Acad. Sci. 94, 12771 (1997).
6. A. Wiles, Ann. Math. 141, 443 (1995).
7. A. D. Aczel, Fermats Last Theorem, Four Walls Eight Windows, New York, 1996; S. Singh and K. A.
Ribet, Scien. Am. (11): 68 (1997); D. Mackenzie, Science 285, 178 (1999).
8. J. C. Phillips, Proc. Nat. Acad. Sci. 94, 10528 (1997).
9. J. C. Phillips, Proc. Nat. Acad. Sci. 94, 10532 (1997).
10. J. C. Phillips (unpublished).
11. A. Ekert and R. Jozsa, Rev. Mod. Phys. 68, 733 (1996); C. H. Bennett, Physics Today 48 (10), 24
(1995).
12. H. He and M. F. Thorpe, Phys. Rev. Lett. 54, 2107 (1985); M. F. Thorpe, D. J. Jacobs, N. V. Chubynsky
and A. J. Rader, Rigidity Theory and Applications (Ed. M. F. Thorpe and P. Duxbury Kluwer Academic
/ Plenum Publishers, New York, 1999), p. 239.
13. Y. Imry and S.-K. Ma, Phys. Rev. Lett. 35, 1399 (1975).
14. J. C. Phillips, Solid State Commun. 47, 191 (1983).
15. G. A. Thomas, M. A Paalanen,. and T. F. Rosenbaum, Phys. Rev. B 27, 3897 (1983).
16. K.M.Itoh, E. E.Haller, J. W. Beeman, W. L. Hansen, J.Emes, L.A. Reichertz, E. Kreysa, T. Shutt, A.
Cummings, W. Stockwell, B. Sadoulet, J. Muto, J. W. Farmer, and V. I. Ozhogin, Phys. Rev. Lett. 77,
4058 (1996).
17. M. F. Collins, Magnetic Critical Scattering, Oxford Univ. Press, Oxford (1989).
18. J. T. Chayes, L. Chayes, D. S. Fisher and T. Spencer, Phys. Rev. Lett. 57, 2999 (1986).
19. K. A. Ribet, and B. Hayes, American Scientist 82, 144 (1994).
20. J. C. Phillips, Phys. Rev. B 41, 8968 (1990).
21. X. D. Xiang, W. A. Vareka, A. Zettl, J. L. Corkill, M. L. Cohen, N. Kijima, and R. Gronsky, Phys. Rev.
Lett., 68,530(1992).
22. N. E. Hussey, H. Takagi, Y. Iye, S. Tajima, A. I. Rykov, and K. Yoshida, Phys. Rev. B 61, R64 (2000).
23. R. P. Sharma, F. J. Rotella, J. D Jorgensen,. and L. E. Rehn, Physica C 174, 409 (1991).
24. G. Wentzel, Phys. Rev. 83, 168 (1951).
25. J. C. Phillips, Phys. Rev. B 47, 11615 (1993).
26. A. L. Efros and B. I. Shklovski, J. Phys. C 8, L49( 1975).
27. S. K. Tolpygo, J.-Y. Lin, M. Gurvitch, S. Y. Hou and J. M. Phillips, Physica C 269, 207 (1996).
28. N. Nagaosa and P. A. Lee, Phys. Rev. B 45, 966 (1992).
29. H. L. Stormer, A. F. J. Levi, K. W. Baldwin, M. Anzlowar, and G. S. Boebinger, Phys. Rev. B 38,
2472 (1988).
30. J. P. Franck, Physica C 282-287, 198 (1997); Phys. Scrip. T66, 220 (1996).
31. S. Sadewasser, J. S. Schilling, A. P. Paulikas and B. W. Veal, Phys. Rev. B 61, 741 (2000).
32. J. C. Phillips, Rep. Prog. Phys. 59, 1133 (1996); J. C. Phillips and J. M. Vandenberg, J. Phys.: Condens.
Matter 9, L251-L258 (1997).
33. J. C. Phillips (unpublished).
34. G. S. Boebinger, Y. Ando, A. Passner, T. Kimura, M. Okuya, J. Shimoyama, K. Kishio, K. Tamasaku,
N. Ichikawa, and S. Uchida, Phys. Rev. Lett. 77, 5417 (1996).
35. T. Ito, K. Takenaka, and S. Uchida, Phys. Rev. Lett. 70, 3995 (1993).
36. Y. Petrov, T. Egami, R. J. McQueeney, M. Yethiraj, H. A. Mook, and F. Dogan, LANL Cond-
Mat/0003414 (2000).
37. T. Hanaguri et al., Physica C 256, 111 (1996).
38. A. I. M. Rae, E. M. Forgan, and R. A. Doyle, Physica C 301, 301 (1998).
39. H. Darhmaoui and J. Jung, Phys. Rev. B 53, 14621 (1996).
40. J. C. Phillips, Solid State Commun. 109, 301 (1999).
41. H. Takagi, R. J. Cava, B. Batlogg, J. J. Krajewski, W. F. Peck, P. Bordet, and D. E. Cox, Phys. Rev.
Lett. 68, 3777 (1996); H. Y. Hwang, B. Batlogg, H. Takagi, J. Kao, R. J. Cava, J. J. Krajewski, and W.
F. Peck, Phys. Rev. Lett. 72, 2636 (1994).
42. J. Wagner (this workshop).
43. E. Kaldis, J. Rohler, E. Liarokapis, N. Poulakis, K. Conder, and P. W. Loeffen, Phys. Rev. Lett. 79,
4894 (1997).
22
REDUCED DENSITY MATRICES AND CORRELATION MATRIX
A. JOHN COLEMAN
Department of Mathematics and Statistics, Queens University,
Kingston, Ontario, K7L 3N6, Canada
colemana@post.queensu.ca
I tackle a herculean task - attempting to wean our imagination from the 1-particle picture
which, implicitly, we have all been using since our youth. I shall try to entice you to join a
crusade for the creation of new concepts and images needed for problems in which interac-
tion between 3 or more electrons is significant and which are appropriate for describing the
information encapsulated in the second order reduced density matrix. (2-matrix for short)
Perhaps the difficult part of our task is changing our language and mental images.
It was to this task that we were called by Charles Coulson in private conversation, and in
his speech [1] in Boulder in June 1959, urging us to look in the 2-matrix for correlation. Also,
by H. Froehlich [2] when he bemoaned the fact that we have failed to exploit the deep import
of the results [3] of C.N Yang on the 2-matrix. It is to this task that I have devoted much
of my time and interest since 1952 culminating in the publication of REDUCED DENSITY
MATRICES - Coulsons Challenge [4]. I shall refer to this book, by Coleman and Yukalov,
as CY.
When we wrote CY, although it was known that the order parameter of one of the
phases of He
3
had p-symmetry, we were unaware that the existence of s,p and d symmetry
has appeared in some of the high temperature superconductors. As a result there is only a
brief reference in CY to the correlation matrix. I have made a modest effort to redress this
lacuna in the present paper.
ENERGY AND N-REPRESENTABILITY
A reduced density operator (RDO), for a normalized pure state, (123... N), of a system
of N identical fermions or bosons can be represented as an integral operator. For example,
the kernel of a 1-RDO, D
1
, is the first order reduced density matrix (1-RDM):
Phase Transi t i ons And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 23
whereas, the 2-RDM is
Thus the operator, D
1
, acting on a symmetric or antisymmetric function f of N particles,
defines a function D
1
f such that
More generally, the p-RDO, D
p
, is an operator with unit trace such that
In the context of Second quantization, it is usual to employ RDOs with somewhat dif-
ferent normalization introduced by Dirac and defined by
and
Thus,
and
As far as I am aware, it was Dirac [5] who first made effective use of RDMs. But he con-
sidered only states described by a single Slater determinant formed from N orthonormal
spin-orbitals in which case
As can be easily verified, this operator is merely the Identity operator on the linear space,
spanned by the N functions Dirac showed that all the physical properties of the Slater
state, including the p-matrices, can be obtained from a knowledge of or,equivalently, of
It is astounding that so much physics, including our understanding of the Periodic Table, has
been built with what would seem to be a trivial tool - the identity operator on a linear space
of dimension N. As we all know, the physics consists of a skillful choice of the spin-orbitals,
or rather of It is precisely the purpose of Hartree-Fock theory to lead us to the best
possible choice. A large thriving industry and much of the wealth of the pharmaceutical
companies is based on the simple equation (7). Such is the power of mathematics!
It was Husimi [6] who, apparently, first discussed the more general RDMs in (1) and
(2) .Indeed, he considered p-RDMs for arbitrary p. When the hamiltonian is represented in
the form
as a sum of N one-particle terms and two-particles terms, it is easy to see that, for
both fermions and bosons, the exact energy, E, of the state is given by
where the reduced hamiltonian is defined by
24
In my view, these last two formulas are absolutely basic for understanding the quantum me-
chanics of many-particle systems in which interaction among the particles plays a significant
role. From the form of (10) it appears that as N increases the relative importance of interac-
tion becomes increasingly significant! Unfortunately, hitherto little attention has been given
to the eigenstates of K and the role of N in determining its eigenvalues. I regard this as a
key challenge for any analyst who is interested in making a significant contribution to the
N-body problem. - cf. pp. 11 and 257 of CY.
I discovered (9) in 1952 while trying to understand Frenkels exposition of so-called
Second Quantization. Husimi had seen it at least ten years earlier! I immediately applied
(9) to calculate the ground-state energy of Li by assuming a simple ansatz A for
D
2
such that
I did extremely well, indeed too well! The result was about 20% below the observed
value! This was impossible and forced me to realize that imposing fermion statistics was
more subtle than I had imagined. This led me to invent the concept of N-representability:
The 2-matrix of a pure state, must be representable in the form (2) in order to satisfy
fermion or boson statistics. Analogously for a p-matrix.
So, in 1952 I proudly announced to a group of able physicists at Chalk River that I
had reduced the N-body problem to a body problem - we now merely had to solve the
N-representability problem, which I assumed would be childs play, and using (9) find the
ground states using Rayleigh-Ritz. After 48 years there is no easy practical way of doing
this in general. However, Carmela Valdemoro made a big break-through in 1992 which was
quickly followed up Hiroshi Nakatsuji and then by David Mazziotti. They have devised an
effective method of calculating the energy levels, which I have dubbed the VNM method and
which has been described [7] as wave mechanics without wave functions. For atoms and
molecules with as many as 20 electrons, the VNM method competes favourably with FCI
calculations of equal accuracy.
Since RDMs are initial values of Greens Functions, a similar condition must be satis-
fied by GFs. This has been generally unnoticed until quite recently and still has not been
really absorbed by main-line physicists. However, the book [8] by Parr and Yang about Den-
sity Functional Theory (DFT) contains an early Section pointing out that N-representability
is a dark cloud hovering over the validity of DFT. The usual methods of dealing with this
problem is either not to be aware of it or to hope that it will go away! The latter method is
not satisfactory. For small N, my experience with Li shows it is risky; whereas, for large N,
the theorem of Hugenholtz [9] that, for an interacting system in the limit as N gets arbitrarily
large, a single Slater determinant is orthogonal to the true wavefunction is rather dramatic.
Perhaps this was in Froehlichs mind [2] when he spoke to David Peat!
BCS, MATTHIAS AND ALL THAT
I would be the first to admit that the BCS theory has been extraordinarily successful,
making a contribution of immense value to Condensed Matter theory. Even so, as with the
remarkable success of a single Slater determinant, I have always been amazed how the origi-
nal BCS simple theory, managed to change and persist so long.
In the cold light of current knowledge we now realize that the simple BCS theory had
only two essential ingredients
(i) The choice of a trial wavefunction formed with the same material as is needed to
characterize one antisymmetric 2-particle function, or geminal.
25
(ii) An extremely simple ansatz for the potential as a step function exercising a positive
attraction between electrons with energy close to the fermi energy.
Of these, I certainly consider (i) as more important. The BCS wavefunction is a Fock
space equivalent of the wavefunction considered by Schafroth [10] and which, by a stroke
of luck, is as Yang proved [3], the type of wavefunction most likely to give rise to a large
eigenvalue of the 2-matrix. To my mind this is the explanation of the success of the BCS
model.
As for the nature of the force involved, we were told that the positive isotope effect
definitely proved that it was phonon-mediated. So in my innocence, as a naive mathematician
when a negative isotope effect was observed, I immediately inferred that this proved that the
force could not be phonon mediated. But no! Since by this time the idea that the force
was phonon-mediated had become firmly implanted in our collective consciousness, it was
soon proved by an able theoretician that a negative isotope gave us even added evidence
of our - by now - blind faith that the force was phonon-mediated! Also the myth was firmly
established that Cooper pairs consist of two electrons with opposite spins. Apparently
many phyicists still believe that this is essential to BCS theory. As suggested in Chapter 4 of
CY, this is not necessarily the case.
For every new observation that contradicted the currently accepted theory our faith was
saved by a small add-on or by a major or minor modification of the current formulation. The
evolving BCS theory became more and more complex and subtle. But at that period, during
which I had the rare privilege of meeting and challenging Bernd Matthias every winter at
Sanibel until his death, I developed the feeling that BCS theory had become, like Ptolemaic
astronomy, a system of epicycles piled on epicycles!
Bernd proudly proclaimed that he was anathematized by all theoretical physicists be-
cause for every new version of the theory proposed, he would go into the Bell Lab and emerge
with a counter-example! Perhaps because he was a polite Swiss being kind to a Canadian or
perhaps because he took pity on me as an innocent mathematician wandering among chemists
and physicists, he carefully stroked my ego by stating that the ideas re. superconductivity that
I advanced were not contradicted by any known observation. I will pursue this below!
However, in private conversation and in his lectures [11] at McGill in 1968, Matthias
insisted that the truly interesting theoretical question is why do nearly all substances manifest
a form of Long Range Order (LRO) at sufficiently low temperature. He asserted that even
gold would become a superconductor! I very much regret that he died before the discovery
of HTSC. He would have so much fun bating theoreticians re. anyons, RVB and the other
exotic ideas that have been bruited!
When I asked John Harrison, the former Editor of JLTP and my colleague in Physics
at Queens, to explain Matthiass observation, his response was immediate. Its really not
so surprising. At absolute Zero the entropy will vanish so we should expect total order.
Indeed, this is true and proves that my knowledge of thermodynamics is almost nil or I would
have made this point to Matthias. So the interesting question becomes, not why there is LRO,
but rather why is the LRO of the nature that actually occurs in a particular substance? I do
not pretend to have a detailed answer to this question. I do think that I offer the basic set of
the ideas essential for its answer.
Another beef that I have with current physics practice is the error which Whitehead
[12] calls The Fallacy of Misplaced Concreteness exhibited in such terms as p-electrons or
Cooper pairs. If you understand the meaning of the word fermion or if you believe in democ-
racy you know that all electrons are equal. They do not live in George Orwells Animal Farm
in which all electrons are equal but some are more equal that others. Only occasionally,
have I noticed momentary indications of a bad conscience by chemists or physicists about
this misuse of language. If challenged, as I am doing now, they excuse themselves with the
same remark Bourbaki often uses This is merely an innocent abus de language. Whereas,
I regard it as a noxious avoidance of our proper task of instilling in the minds of students a
26
set of valid concepts with which to explore the inner riches of Quantum Theory.
It is not an electron which has p-symmety but a spin-orbital. In fact, it is a partially
occupied eigenfunction of the 1-matrix! There are no such things as Cooper pairs, even if
we think of them in the charming image, due I understand to Schrieffer, as partners dancing
to Rock so that they can be at far ends of the floor yet fully synchronized, rather than breast-
to-breast in a gentle Strauss waltz. To even propose such an image almost makes the concept
absurd. The functional unit is not a pair of electrons it is a spin-geminal. In fact, the key
concept which we must learn to deploy is that of a partially occupied eigengeminal of the
2-matrix.
CORRELATION MATRIX AND ORDER INDICES
I read somewhere that the nobellist, C.N. Yang, regarded the paper [3], in which he
associated the onset of superconductivity with the appearance of a large eigenvalue in the 2-
matrix, as the most important paper of his distinguished career. My initial conjecture [13] was
the obvious generalization of his observation and asserts that every type of LRO is associated
with a large eigenvalue ot the 2-matrix.
This was refined [14] by the definition of order indices and the correlation matrix. It
is known that the least upper bound for the eigenvalues of the 2-matrix of a boson system
is N(N 1) and for a fermion system [15] the unattainable such bound is N. If we call
the occupants of geminals pairons then we can say that the l.u.b. for the pairon occupation
of a natural geminal is N(N 1) for a system in which the constituent identical particles are
bosons and N for a system if they are fermions. If a Cooper pair is anything it is a pairon.
But the term pairon is more general and is not necessarily associated with superconductivity.
Yang argued that superconductivity in a metal is triggered when has an eigenvalue
of order N. Bloch [16] connected such an eigenvalue with flux quantization related to carri-
ers with charge 2e, confirming Yangs theory. From energy considerations sketched below,
I inferred that eigenvalues of proportional to N were associated with eigengeminals de-
scribing a correlation which extends throughout the substance.- in other words, a Long Range
correlation.
The order index was then defined [17] as the largest value of such that has a
finite non-zero value, in the thermodynamic limit. The correlation matrix is defined as
By the above-mentioned [14] result, for systems of fermions For one or
more eigenvalues of is proportional to N so LRO is present. For systems of bosons,
could be as large as 2. As long as we conjecture that some form of mesoscopic [17]
or local order is present. If we compare this with a percolation model for the onset of a new
phase of matter, corresponds to the critical value of p when the diameter of an open
cluster is infinite, close to 0 corresponds to the first moments at which nuclei of the new
phase are present. As increases these nuclei become more widespread and larger. I am
thinking of the small bubbles becoming more widespread and larger which appear in water as
it approaches the boiling point, or the complex systems of fjords of superconducting phase
penetrating the whole of a cylindrical block of material which I had the privilege of viewing
via polarized light as it was cooled by Martin Edwards in his Low Temperatre Lab at the
Royal Military College of Canada many years ago.
CONJECTURE. For all mono-particle systems, the appropriate order parameter (OP)
is the correlation matrix
From the structure of the 2-matrix it immediately follows that the order parameter can
have spin character s, p or d and any combination of these. This is consistent with recent
observations [18] that the order parameters of some HTSCs exhibit s, p or d spin-symmetry
or a combination of these - a phenomenon, which, apparently, BCS has difficulty accommo-
27
dating. My conjecture is that is the appropriate order parameter for all types of order in
many-particle systems of one type of identical particles. Thus, I am making a bold general-
ization of Yangs observation from superconducting to many other order transitions. I am
encouraged by the fact that this conjecture is consistent with observations on the symmetry
of the OP for HTSC and also for He
3
in which p-type order occurs in at least one phase. I
assume that for helical magnetism and many other types of order it will be necessary to study
not only spin-symmetry but the total symmetry of the eigengeminals.
I have called the above a conjecture rather than a theorem because a proof has not been
obtained. This is because we do not yet have a sufficient understanding of the relation of the
eigenvalues of to the occupation numbers of eigenstates of the reduced hamiltonian K. I
regard this as an important urgent issue for theoretical research. Another is to explore the
dependence on the Order Index, of the physical properties of substances near the critical
point.
Note that if fermi-pairons were bosons, their occupation numbers could go to N(N 1).
This differs from the actual limit of N by a factor of (N 1) which is infinite in the thermody-
namic limit. Thus the universal practice in text-books, and in articles by writers who should
know better, of saying that superconductivity arises as a result of a bose condensation of pairs
is misleading talk which brings comfort, by creating the illusion that we know what we are
talking about, but prevents us from coping with the real task of forging a set of meaningful
concepts with which to understand condensed physics.
It is known that when Fock space is displayed with respect to a basis of a finite number,
r, of orthonormal orbitals (i.e. 1-particle functions), the highest possible value for the eigen-
values n
2i
of is This is attained(CY,Chapter 3) only if the wavefunction is an
antisymmetrized power of a single geminal - an AGP function and if the eigenvalues n
1i
of
are equal. In this case(CY, p. 137) there is one large eigenvalue and the rest are equal to
2 N( N 2 ) / r(r2). If we relax the condition that n
1i
be equal, it is possible to arrange that
for an AGP function has several eigenvalues which are proportional to N. and thus model
a variety of other situations including the co-existence of superconductivity and magnetic
ordering.
It is perhaps worth recalling here that r = N is a necessary and sufficent condition that
the wave function be a Slater determinant. In this case all the eigenvalues of are equal to
2 which is a long way from N. This corresponds to the fact that HF is accurate if and only if
the effective hamiltonian has no 2-particle terms.
We introduce some essential notation by recalling that in CY. Denote the eigenfunctions
of D
p
by with corresponding eigenvalues so
Setting
we obtain
and
28
Further, if the reduced hamiltonian (10) has eigengeminals, g
i
, such that
then the total internal energy
where
Suppose that the are so numbered that they increase monotonically with i, and the
numbering of n
2i
so that they decrease. Then in the ground state the system will choose
so that p
i
for small i, and especially for i = 1, are as large as possible consistent with
N-representability. The largest occupation of a natural geminal is n
21
. By the familiar
theory of separation of eigenvalues of hermitian operators, 2 (N 1) In particular,
2(N 1) p
1
= n
21
if and only if the eigenfunction g
1
of K coincides with the first natural
geminal, of the state. Exact coincidence is highly unlikely, but there will be a strong
tendency towards this so it is possible that p
1
will be of order N. In this case, we would
expect that g
1
describes a 2-particle correlation which extends throughout the sample, that is
a LRO.
For N electrons in a lattice if we neglect spin, we are led to study the hamiltonian
where i and j refer to electrons and k to nuclei; Z
k
is the charge on an ion at s
k
. For neutral
systems This implies, in the notation of (8), that
Notice that, though we mentioned electrons in a lattice, if k assumes only one value, (23)
would describe the hamiltonian of an N-electron atom with nuclear charge Z
k
= N, whereas,
if k takes two values, a diatomic molecule. And so on. In fact, almost anything.
By (10), associated with (23) is the Reduced Density Operator, K. However, for reasons
which will become apparent, we introduce an additional parameter, t, and define K (t) by
If we divide by N
2
, set and N
2
U (t) = K(t), and replace Nr
i
by r
i
and Ns
k
by s
k
,
then (25) takes the form
29
For a neutral system, For fixed N the spectrum of U(t) will depend continuously
on t. A famous theorem [19] of Zhislin assures us that when the operator
(26) has an i nfi ni t e number of bound states with energy levels crowding up to the limit of the
continuous spectrum.
U (0) is a two-electron hamiltonian which approaches the hamiltonian of H

as N in-
creases to infinity. On the other hand, for t = 0, and N = 2, (25) is the hamiltonian of the
helium atom. According to (18) and (22) it is the spectrum of K = K(1) which is of real
interest in the study of energy levels of N-particle systems. Since K = N
2
U ( l ) , it follows
that the spectrum of K is obtained from that of U (1) by scaling by the factor N
2
.
It is known [20] that H

has only one bound state. It is a

1
S state slightly below the
continuum which accounts for an absorption line in the solar spectrum. The two lowest states
of the helium atom are a
1
S and a
3
S state. For a fixed system, (25) depends continuously on
t so we expect that as t varies from 1 to 0 a correspondence will be established between the
spectra of U (1) and U (0). However, while U (0) is an atomic hamiltonian, we shall expect
the spectrum of U ( t ) , when t > 0, to be a series of bands, possibly narrow, each of which
collapses, when and which could be named by an energy level of the atomic system
which U (0) describes. If spin is neglected then it would be reasonable to expect that all
levels of the lowest band would be
1
S.
In this case, for a system manifesting long-range order at low temperature, we would
anticipate that the correlation matrix will depend on the eigenfunctions corresponding to
the levels of the lowest band weighted by a distribution function depending on the inverse
temperature, Unfortunately, little study has been made of the spectrum of K for solids or
other condensed matter even though the fact that it must play a key role in understanding the
energetics of condensed matter has been obvious for forty or fifty years.
We noted above that the late Bernd Matthias, who probably discovered more supercon-
ductors than any three other experimentalists together, constantly insisted that an important
task for theoretical physics was to explain why nearly all fermion systems manifest long-
range order of some type at sufficiently low temperatures - superfluidity, superconductivity,
ferro- or antiferro-magnetism, charge density waves, coexistence of superconductivity and
helical spin density waves, etc. To properly describe the electrons in condensed matter, our
hamiltonian (23) would need to be supplemented by terms describing LS coupling, spin-
spin effects, motion of the ions etc. However, the electric forces described by (18) would
probably dominate the energy.
If in fact the spectrum of K is similar to that of H

in having one eigenvalue, or a band of

eigenvalues significantly below all others, then that level would tend to be occupied as fully
as possible consistent with the statistics, the inter-particle forces and the temperature. For
fermions, n
21
could be of order N which, if it occurred, would manifest itself as long-range
order. The nature of the particular LRO would be characterized by the correlation matrix.
ANTISYMMETRIZED GEMINAL POWER
A theorem attributed [21] to Zumino states that a fermion geminal, in other words, an
antisymmetric two-particle function, can be transformed by a unitary transformation into a
canonical form in which each orbital is a member of a unique pair of orbitals.
The reader should be aware that it is my custom to denote by the word orbital a func-
tion of a single particle including all relevant coordinates. Thus, depending on context, the
word may denote the classical meaning of a chemist(if the particle is without spin), or what
a chemist means by spin-orbital, or a function of spatial co-ordinates and two dichotomic
variables for spin and isotopic spin.
Thus if
30
is such that the normalized function g (12) = g (21), with then r = 2s is even,
and by a unitary transformation it is possible to find an orthonormal basis
i
with respect to
which
In (27), r is the rank of g and also the rank of the matrix c
i j
. It is well-known that the
rank of an antisymmetric matrix is even. The antisymmetrized power of an orbital is always
zero. Thus f(1)f(2) f ( 2) f ( 1) = 0. However, the antisymmetrized power of a geminal,
g, to obtain a function of N particles will vanish if and only if the rank, r, of g is less than
N. When N = r, this N-particle function is a single Slater determinant formed with a basis of
g. Perhaps inadvisedly, I have adopted the symbol g
N
to denote a normalized N particle
function obtained by antisymmetrizing an appropriate power of g.
Several persons, of whom Nakamura [22] may have been the first, showed that the pro-
jection of the BCS function(which is a coherent ensemble in Fock space of functions of all
possible particle number) onto a subspace of Fock space of particle number N produces an
AGP function of rather special type. The importance of the AGP function is signalled by
the fact that it has appeared in a variety of contexts with different names such as: Schaftroth
condensed pair function; projected BCS function; correlated pair function; pairiing func-
tion; Generalized Hartree-Fock function. The mathematical concept goes back to Hermann
Grassmann in the 1840s since it arises naturally in Grassmann algebra. I prefer to use a
name which suggests its mathematical nature and does not place it in a misleading context.
For applications to physics and chemistry there is no need to insist that the occupants
of a geminal are particles with opposite spin. Any kind of fermion geminal forces a natural
pairing. Therefore it can be cogently argued that the apparent pairing in BCS is not forced
by the physics but rather appears as a mathematical artifact forced by the assumption that the
wavefunction is AGP. I realize that there is such a widespread commitment to the religious
belief that Cooper pairs are real that there is a high probability that I shall be accused of
blasphemy, tried, condemned and burned at the stake!!
Since the whole of Chapter 4 of CY is devoted to AGP, here I shall restrict myself to
quickly mentioning what every young person should know about Grassmann algebra and
fermions.
1) If is an N-particle fermion function and is an orbital such that the Grassmann
product then there is an (N l)-particle function, such that Further,
these equivalent conditions are necessary and sufficient that be a natural orbital of with
occupation unity.
2) Suppose that an AGP function formed from an arbitrary geminal, g, then if r
is the rank of a) r < N implies that b) r = N implies that is a Slater determinant,
c) r > N implies that can be expressed as a linear combination of ( ) Slater determinants
consisting only of paired orbitals, where r = 2s, and N = 2m.
3) If N is even and the natural orbitals of are evenly degenerate, with occupation
strictly less than unity, then there exists a geminal g such that This remark-
able result, proved around 1965 by Erdahl, Kummer and myself, implies that for a many-
particle fermion state satisfying these conditions, all one-particle properties can be exactly
described by an AGP function.
4) Further, suppose that N = p + q where q is even and precisely p natural orbitals have
occupation unity, then where S is a Slater determinant containing the indicated
p natural orbitals and g is a geminal. Such functions have been called Generalized AGP
functions.
31
5) On p. 139 of CY we indicate that an AGP function might have the possibility of
having a 2-matrix with a finite number of large eigenvalues. It is therefore conceivable
that observed co-existence of superconductivity and helical magnetism could be modelled by
GAGP.
6) If N is even and if is of rank N + 2, then by making use of the so-called Hodge
Correspondence between subspaces of dimension 2 and those of dimension N in a space of
dimension N + 2, we can prove that is an AGP function.
As the cranking model, AGP proved useful in nuclear theory. At first a theory with the
fanciful sobriquet superconducting nuclei was introduced using the BCS coherent ensemble
equation subject to a condition that the expected value of the number operator be N. However,
Nogami and others soon noticed that it was more accurate to use a projected BCS function,
that is an AGP function. Chemists also found that AGP, as an ansatz for the wave-function,
was more successful in modelling the dissociation of diatomic molecules than Haertree-Fock.
It was observed that the Random Phase Approximation is self-contradictory i f , as is common,
a single Slater is taken as the initial ground state, whereas the most obvious contradictions
are avoided if the ground state is assumed to be a Generalized AGP.(For this, cf. p. 140 of
CY).
In view of these properties and the fact that GAGP can be a single Slater modeling a
fermion system with no correlation or, on the other hand, a system with a with the largest
GRASSMANN AND THE FERMI SURFACE
I come now to a little-known theorem of Grassmann for which I will present a proof,
partly because I do not want to disappoint your expectation that proving theorems is my main
purpose in life, as a mathematician, but also because the result is unexpected and may be
the real reason why we must replace fermions by fermi pairons in our thinking and
therefore the ultimate reason that Cooper pairs proved so serviceable.
I announced this result [23] without giving the proof in 1961. In fact the proof was
rather easy making use of the Hodge correspondence between sub-spaces of dimension p
and those of dimension n p in a linear space of dimension n. With a more complicated
proof, the same result was proved later in the RMP by a theoretical physicist, but I have
lost the reference. In fact, Whitehead [24] provides an almost trivial proof, attributing it to
Grassmann - presumably from the 1840s!
Here in two equivalent forms, first Grassmanns and secondly mine, is the
Theorem
(i) A homogeneous element of order n in a Grassmann algebra of rank n + 1, is elemen-
tary.
(ii) If is a pure N-particle state, then the rank of is not N + 1.
Proof. I shall state the argument in the language of antisymmetric wavefunctions. Sup-
pose the basis has N + 1 orbitals. Associated with a Slater determinant of order N is a unique
subspace of dimension N spanned by the N vectors of the determinant or by any N linearly
independent vectors in the same subspace. Changing these vectors does not change the sub-
space but may multiply the Slater by a constant. Suppose that is a linear combination of
two Slaters. By a basic theory about subspaces, the dimension of the intersection of the sub-
spaces associated to the two Slaters is N + N (N + l ) = N l. This intersection is common
to both subspaces and is characterized by a Slater S. of order N 1. Adjoin vectors and
32
possible eigenvalue and therefore modelling the highest possible correlation, it is apparent
that the GAGP ansatz is of great scope and could be used to provide insight into a wide
variety of fermion systems.
to the intersection so that S and are N-th order Slaters respectively characterizing
the two subspaces Then there are constants a and b such that
which is a single Slater of rank N. By induction we see that any linear combination of Slaters
of order N, in a space spanned by N + 1 orbitals, is again a Slater(i.e , in the language of
Grassmann algebra, elementary) of N orbitals. It follows easily that version (ii) is implied
by version (i) of the statement of the theorem.
Hence if is not a Slater it must have rank at least N + 2. But it could have that rank
as follows from Item 6 of the previous Section. The discussion of the energy of an AGP
state in Section 4.6 of CY was used to estimate the change in energy if a Slater state of rank
N is changed to a state of rank N + 2 by replacing two orbitals each of occupancy 1 by four
orbitals each with occupancy 1/2. It was found (CY, p.155) that the change in energy of the
state was
where and name distinct pairs of orbitals. The number denotes the interaction
energy where and and has fixed phase. Whereas, denotes
the interaction energy and has adjustable phase which was used to arrange
the negative contribution in (29). Thus the Fermi surface is unstable with respect to pair
formation unless is positive and numerically greater than If the so-called pairing
hamiltonian is used, automatically so that for the pairing hamiltonian, the Fermi
surface is always unstable. This seems to contradict the commonly expressed view that the
existence of superconductivity requires an attractive force which was part of the rational for
the existence of Cooper pairs.
It is now widely recognized that HTSC is usually associated with phase separation. In
the next section we find that a sufficiently strong repulsive Coulomb force is required to
account for phase separation.
PHASE SEPARATION AND SUPERCONDUCTIVITY
In his well-known survey [25], published in 1989, of the properties of HTSC, Phillips
has 11 references to Lattice instabilities, a topic to which he devotes several pages at various
points in the book. He even went so far as to suggest that lattice instability is the only factor
that causes HTSC. I do not admit this since in 1991 Yukalov [26] surveyed 508 experimental
and theoretical papers which dealt with evidence bearing on the incidence of phase transitions
of what he called heterophase fluctuations. Yukalov, who is a Senior theoretical physicst
in the Joint Institute for Nuclear Research in Dubna, is a remarkably competent and careful
authority on Quantum Statistics . He agrees that instability of the lattice can be important
but there are other significant factors. During the past ten years more evidence [26] has
accumulated similar to the impressive collection which he assembled. In his basic paper
referenced above, are foreshadowed many ideas that have recently become current.
Chapters 5 and 6 of our book were laregly due to Yukalov since I know so little of the
nitty-gritty of physics. In particular this is the case for Section 6.2 of CY in which we attempt
to work out in a form relevant to HTSC the theory developed in his paper [26] when there are
only two phases interpenetrating. Here I merely sketch the course of our argument directing
the interested reader to Section 6.2 of CY and the references in Notes 2 and 3.
We posit a situation which can be thought of as microsopic or mesoscopic nuclei of
one phase (e.g. superconducting) scattered randomly through a host phase (e.g. normal)
Experimental observation of this possibility was recently provided by a group from Dubna
33
with associates in a paper [28] entitled Microscopic phase separation in induced
by the superconducting transition. We propose a simple model which takes into account the
three interrelated factors: Coulomb interaction, phase separation, and lattice softening. We
give a detailed analysis of the dependence of the critical temperature on parameters related to
the attractive and repulsive interactions and to the superconducting phase fraction w.
Since the Hartree-Fock-Bogolubov - essentially AGP - approximation is used, our for-
mulas look very much like those in usual presentations of BCS theory.. However, they have
quite different meaning because they involve the parameter w in an intricate manner.We as-
sume that the interaction between electrons is the sum of two components
- a direct part which is taken as a Debye-type shielded Coulomb force. and
- an indirect part for which we assume the conventional Froehlich phonon term.
Additional parameters are introduced by which it is possible to vary (i) the phonon
frequency, (ii) electron-phonon coupling, and (iii) the strength of the direct interaction. The
resulting equations were solved numerically by Dr. E. Yukalova. The results are exhibited in
CY as twelve graphs portraying the superconducting critical temperature against w, in three
groups corresponding to weak, moderate and strong softening of the lattice.
We were pleasantly surprised by the wide variety of shapes of these graphs. Despite
the rough approximations assumed for our model, the behaviour of the critical temperature in
Figs. 3,4,7 and 8, has striking similarity to some corresponding experimental curves observed
for HTSC. Even though one might expect the relation between doping intensity and w to be
monotone, the actual relation is not known so a detailed comparison of our results with those
of experiment is not possible.
We were led to the following conclusions:
- The presence of repulsive interaction is a necessary condition for mesoscopic
phase separation.
- Phase separation favours superconductiviy making it possible in certain het-
erophase samples when it would not occur in a pure sample.
- The critical temperature as a function of the relative fraction of superconductive
phase can exhibit the nonmonotonic behaviour characteristic of HTSC.
FINAL REMARKS
1) I conclude that we need to develop a habit of thinking more comfortably about the
second order reduced density matrix its eigenvalues and its eigengeminals.
2) For a system of a large number of identical particles which is all that I discussed,
the large component of the 2-RMD, denoted by is proposed as the appropriate order
parameter. If there is no order present so this could correspond to what we usually
call normal. For fermion systems Long Range Order corresponds to and for bosons,
to I have not expatiated on my conviction that neither bose nor fermi condensation,
as normally understood, in the simple-minded sense derived from London(whose memory I
honour!), actually occur in an interacting system and that we poison the innocent minds of
our students if we persist in suggesting that they do.
3) More imporrtant, it is my view that because for the earliest discovered superconduc-
tors, T
c
was so low and the isotope effect had a simple explanation, we were misled into
thinking that the origin of superconductivity is exotic and/or subtle. However, I take seri-
ously Matthiass observation that LRO at sufficiently low temperature is universal and there-
fore should have a robust explanation.. Further the behaviour of various substances near the
34
critical temperature seems to have much in common. When charged particles are involved, I
conclude that Coulomb forces are the real culprit. So I claim that the secret for this univer-
sal phenomenon is to be found in the second order reduced hamiltonian. The isotope effect
implies that interaction with the lattice must play a role. My musings at the end of Section 3,
suggest that contributions by the static coulomb interaction, specified by Zhislins theorem,
involve quite minute energy differences for K jbetween the continuum level and a narrow
energy band which is almost at the continuum limit. This means that lattice dynamics, L
.
S
coupling and other spin-effects could also play a significant role accounting for the known
variation of T
c
across the Periodic Table which was noted by Matthias [11] in his McGill
lectures.
4) I fully realize that some of my heterodox opinions are anathema to many. I shall try
to face this with the equanimity of old age, welcoming all comments, questions and coun-
terexamples at my email address on the title-page.
NOTES
1. Charles Coulson, in conversation with graduate students and Coleman in Oxford June 1975. Also in a
speech in Boulder, Colorado: Rev. Mod. Phys. 32, 175 (1960).
2. H. Froehlich, shortly before his death, in private conversation with David Peat.
3. C.N. Yang, Rev. Mod. Phys. 34, 694 (1962).
4. By A.J. Coleman and V.I. Yukalov, Vol 72 in Series published by Springer in Lecture Notes in Chemistry,
April, 2000.
5. P.A.M. Dirac, Proc. Cam. Ph. Soc. 26, 376 (1930); 27, 240 (1931).
6. K. Husimi, Proc. Phys. Math. Soc. Japan 22, 264 (1940).
7. Section 7.3 of CY.
8. R.G. Parr and W. Yang, Density Theory of Atoms and Molecules, Oxford University Press, 1980.
9. N.M. Hugenholtz, Physica 23, 481 (1957); L. van Hove, Physica 25, 849 (1958).
10. M.R. Schafroth, Phys. Rev. 96, 1149, 1442 (1954); 100, 463, 502 (1955); 111, 72 (1958).
11. B. Matthias, Three Lectures, in Superconductivity, Proc. Ad. Summer Study Institute, June, 1968, at
McGill University, ed. P.R. Wallace, Gordon and Breach, New York.
12. A.N. Whitehead, p. 64 Science and the Modern World, Cambridge U.P., 1933; p.11 Process and Reality,
Macmillan Comp.,1929.
13. A.J. Coleman, Can. J. Phys. 42, 226 (1964).
14. A.J. Coleman, V.I.Yukalov, Nuovo Cimento B 108, 1377 (1993).
15. A.J. Coleman, Rev. Mod. Phys. 35, 668 (1963).
16. F. Bloch, Phys. Rev. A 137, 787 (1962).
17. A.J. Coleman, Jl. Low Temp. Phys. 74, 1 (1989).
18. H. Srikanth et al., Phys. Rev. B 55, R14 733 (1997); K.A. Kouznetsov et al., Phys. Rev. Lett. 79,
3050 (1997); in Physica C 317-318, 410 (1999), van Hartington claims unambiguous determination of
d-wave symmetry in HTSC cuprates; in Nature 396, 658 (1998), Ikeda et al. observe p-wave symmetry
in a second HTSC.
19. G.M. Zhislin, Trudi Mosk.Mat. Obsc. 9, 81 (1960), Th.III, p.84.
20. R.N. Hill, J. Math. Phys. 18, 2316(1977).
21. B. Zumino, J. Math. Phys. 3, 1055 (1963); see also Thm.6, Coleman, Bull. Can. Math. Soc. 4, 209
(1961).
22. K. Nakamura, Progr. Theor. Phys. (Kyoto) 21, 273 (1959).
23. A.J. Coleman, Can. Math. Bull. 4, 209 (1961), Thm.7.
24. A.N. Whitehead, Universal Algebra, Cambridge University Press, 1898.
25. J.C. Phillips, Physics of High-T
c
Superconductors, Academic Press, 1989.
26. V.I. Yukalov, Phase Transitions and Hetrophase Fluctuations, Physics Reports 206, 395488(1991).
27. Phys. Rev. B 54, 9054 (1996); Phys. Rev. Lett. 76, 439 (1996).
28. V.Yu. Pomjakushin et al., Phys. Rev. B 58, 12 350 (1998).
35
This page intentionally left blank
THE SIXTEEN-PERCENT SOLUTION:
CRITICAL VOLUME FRACTION FOR PERCOLATION
RICHARD ZALLEN
Department of Physics, Virginia Tech
Blacksburg, VA 24061
INTRODUCTION
The English call it value for money (vfm). The American equivalent is bang for
the buck. The idea is simple: to provide a rough measure of the ratio of benefit to cost. For
an author of scientific papers, one possibility for a vfm-type measure of benefit (impact)
to cost (time and effort) is this:
vfm = (number of citations)/(papers length in printed pages).
In my case, the vfm winner is clear. It is a two-page paper by Harvey Scher and myself,
published quietly as a note in J. Chem. Phys. [1], which has been cited over 350 times.
Later work related to the central idea of that paper has also been widely cited [2, 3]. That
idea is the concept of a critical volume fraction for site-percolation processes.
NOSTALGIA
One afternoon in mid-May of 1970, at my desk in the research building of the Xerox
complex near Rochester, NY, I was poring over experimental Raman spectra, searching for
significant peaks with my spectroscopists eye [4]. I was not having much luck, and I
needed a break. So I left my office, walked down the hall, and went into the office of a
colleague, Harvey Scher. Harvey was, as usual, good-natured and patient about the
interruption of his own work, and he took the opportunity to describe an interesting
problem that he was working on. A very approachable resident theorist, Harvey had been
consulted by a technology group working on photosensitive layers in which
photoconductor particles were dispersed in a resin. Their measurements had shown a
dramatic threshold in the dependence of photosensitivity on photoconductor concentration.
Elliott Montroll, then a frequent visitor to Xerox, had suggested to Harvey that he look at
the literature on percolation theory. Harvey had assimilated that literature and made use of
it, and he introduced me to percolation theory that afternoon. I was fascinated by this stuff,
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 37
and when I got back to my office I did not return to the strip-chart recordings (no PCs in
1970). Instead, I worked on some geometry problems related to ideas that we had kicked
around, and I became enthusiastic about writing up a short paper reporting what we had
found.
Ten days later the paper was circulated internally within Xerox, and it was submitted
for publication in mid-June. This speed was then, and is now, uncharacteristic of both
authors. The reason for the choice of J. Chem. Phys. is somewhat obscure. We did not want
to send it to a math or math-phys journal, and we had seen a short paper in J. Chem. Phys.
that mentioned percolation. It turns out that the referee for our paper was almost certainly a
mathematician! He (or she) chided us for the empirical and approximate nature of our
critical density. (We knew it was approximate, and we were proud of empirical!) But he
(or she) nevertheless pointed out to us an additional result for (an exact value for the
two-dimensional Kagom lattice) which fit our ideas very well. We added it to the Table in
our paper.
The anecdote described above, dealing wi t h the fruitfulness of an afternoon schmooze
session at the Xerox lab in Webster, NY, was characteristic of a period now remembered
by some as a golden age of industrial research [5]. The scientific issue arose in the
context of a technological setting, which is of course a familiar tradition in condensed-
matter physics [6]. The atmosphere was one in which it was OK to spend time on scientific
issues as well as on product-development and engineering ones. In the year 2000, that era
is history and opportunities to do science are rare in present-day corporations.
Globalization is sometimes given as the reason (or excuse) for this, but human herd-instinct
considerations also enter: Everybody did it then (corporations supported research) because
everybody else did it; nobody does it now because nobody else does it. This is a
cooperative phenomenon, so perhaps we can hope that a phase transition can happen again.
CRITICAL VOLUME FRACTION
In three dimensions, the percolation threshold for site-percolation processes varies
from lattice to lattice by more than a factor of two [7]. For two-dimensional lattices,
varies by more than a factor of 1.5 [7, 8]. The Scher-Zallen construction for the critical
volume fraction associates with each site a sphere (or circle, in 2d) of diameter equal to
the nearest-neighbor separation. Spheres surrounding filled sites are taken to be filled. At
the critical value of the site-occupation probability p , the fraction of space occupied
by the filled spheres is taken to be the critical volume fraction The key point is this:
From lattice to lattice (in a given dimensionality), is nearly constant, varying by just a
few percent. It is an approximate dimensional invariant. In three dimensions, is close to
0.16; in two dimensions, is close to 0.45 [1, 3].
The relationship between and is where f is the filling factor of the
lattice when viewed as a sphere packing. The values forming the basis of the 1970
paper correspond to familiar crystal structures. A structure that is not crystalline but is
experimentally well defined is random close packing (rcp). The rcp structure corresponds
to the atomic-scale structure of simple amorphous metals [3]. Since f is known for the rcp
structure, predicts the value for this structure. Experiments carried out to
determine the conductivity threshold (insulator-to-metal transition) of rcp mixtures of
insulating and metallic spheres are in good agreement with this prediction [9, 10, 11].
One way to view is as an expression connecting the ease-of-percolation
with the connectivity of the underlying structure. For bond percolation, such an
38
(approximate) connection had been found earlier, in 1960 [12, 13]. A reasonable measure
for the ease-of-percolation for a given structure is (1/ pc
) , the reciprocal of the percolation
threshold. For bond percolation, (1/p
c
) is very close to (2/3)z in three dimensions and
(l/2)z in two dimensions. Here z is the average coordination number of the lattice. The
proportionality between ease-of-percolation and coordination number shows that, for bond-
percolation processes, the coordination number is the appropriate measure of the
connectivity of the lattice. This, of course, makes sense. But for site-percolation processes,
z does not work. Instead, shows that (1/p
c
) is proportional to f. This reveals
that, for site-percolation processes, the sphere-packing filling factor is the appropriate
measure of the connectivity of the underlying structure. This insight is a byproduct of the
work on the critical volume fraction.
LIMITATIONS
Thanks to the piece of information provided by the unknown referee, we knew
immediately that is only approximately invariant. The site-percolation threshold is
known exactly for two two-dimensional lattices, the triangular lattice (2d close packing)
and the Kagom lattice, so that is exactly determined for each. The two values differ by
2%. A few people in the critical-phenomena community took an instant dislike to It
wasnt exact. It wasnt rigorous. It wasnt even an exponent, so why care about it? [One
can imagine one of them having the following reaction to the experimental discovery of a
new superconductor: So T
c
is 450 K, so what? What are the exponents? But maybe thats
unfair.]
The value of can be estimated from a plot of (1/p
c
) versus f [3]; the slope is
Here a question arises at the low end of the plot, where the proportionality
between the ease-of-percolation and the filling factor has to eventually fail because (1/ p
c
)
cannot be less than 1. This consideration is unimportant in three dimensions in which
(1 / p
c
) does not closely approach unity; the values cluster in the region from about 2.3 to
5.0. In two dimensions, typical (1/p
c
) values are closer to 1.0, lying between 1.4 and 2.0.
Within this region, the proportionality of (1/p
c
) to f holds very well [1]. However,
Suding and Ziff [8] have recently considered very-low-connectivity two-dimensional
l at t i ces wi t h (1/ p
c
) values down to 1.24. Their results show that at these very low
connectivities, the deviation from becomes appreciable. Suding and Ziff
offer a revised, nonlinear relation between p
c
and f that improves the fit in the very-low-
connectivity region. Most structures of physical interest are far from this region.
APPLICATIONS
The notion of a critical volume fraction insensitive to the details of local structure, as
suggested in the 1970 paper, is an attractive one. But it is heuristic, empirical, approximate.
It had been my original plan for this paper to review its success (or failure) in relation to
experimental literature on metal/insulator composites. This has turned out to be too
mammoth an undertaking for the presently available space and time, and will have to be
deferred. The experimental literature is vast; one extensive compilation can be found in a
1993 article by Ce-Wen Nan [14]. The experimental studies span an enormous variety of
systems and differ greatly in depth and quality.
39
Figure 1. The conductivity threshold in graphite/boron-nitride composites [19].
At a later time I may attempt a plot of frequency-of-occurrence versus value, but
here only some less-than-satisfactory observations will be offered. For three-dimensional
composites a value close to 0.16 is very often encountered, and it is interesting that this
occurs for some of the most carefully studied systems. Examples are the carbon-
black/polymer composites studied by Heaney and co-workers [15, 16, 17] and the
graphite/boron-nitride composites studied by Wu and McLachlan [18, 19]. Figure 1
displays the very clean experimental results of Wu and McLachlan, showing a conductivity
threshold spanning many orders of magnitude. Graphite and boron nitride are structural
and mechanical isomorphs, but differ in conductivity by a factor of 10
18
. The points are
measured values; the curves are scaling-law fits that closely determine (0.15 for this
system).
But there are many systems for which is quite different from 0.16; this value is not
universal. The reason is unclear, though different classes of topology have been suggested.
One of these is the Swiss-cheese void-percolation topology analyzed by Halperin and co-
workers [20] and studied experimentally by Lee et al. [11]
ACKNOWLEDGMENTS
I wish to thank Wantana Songprakob for crucial help in preparing this paper. I also
wish to thank Harvey Scher for thirty years of friendly interaction.
40
REFERENCES
1. Scher, H. and Zallen, R. (1970) Critical density in percolation processes, J. Chem. Phys. 53, 3759.
2. Zallen, R. and Scher, H. (1971) Percolation on a continuum and the localization-delocalization
transition in amorphous semiconductors, Phys. Rev. B. 4, 4471.
3. Zallen, R. (1998) The Physics of Amorphous Solids, John Wiley and Sons, New York. pp. 183-191.
4. I first heard this apt term mentioned in a talk given by Manuel Cardona.
5. In the seventies, the Xerox lab in Palo Alto was the site of some now-famous computer-science
examples: Hiltzik, M.A. (1999) Dealers of Lighting, Harper, New York.
6. Harvey Scher, now at the Weizmann Institute, has commented on technology as a rich source of
scientific questions in his recent Festschrift article: Scher, H. (2000) Reminiscences, J. Phys. Chem. B
104, 3768.
7. Reference [3], p. 170.
8. Suding, P.N. and Ziff, R.M. (1999) Site percolation thresholds for Archimedean lattices, Phys. Rev. E
60, 275.
9. Fitzpatrick, J.P., Malt, R.B., and Spaepen, F. (1974) Percolation theory and the conductivity of random
close packed mixtures of hard spheres, Physics Letters 47A, 207.
10. Ottavi, H., Clerc, J.P., Giraud, G., Roussenq, J., Guyon, E., and Mitescu, C.D. (1978) Electrical
conductivity of conducting and insulating spheres: an application of some percolation concepts, J.
Phys. C: Solid State Phys. 11, 1311.
11. Lee, S.I., Song, Y., Noh, T.W., Chen, X.D., and Gaines, J.R. (1986) Experimental observation of
nonuniversal behavior of the conductivity exponent for three-dimensional continuum percolation
systems, Phys. Rev. B 34, 6719.
12. Domb, C. and Sykes, M.F. (1960) Cluster size in random mixtures and percolation processes, Phys.
Rev. 122, 170.
13. Shklovskii, B.I. and Efros, A.L. (1984) Electrical Properties of Doped Semiconductors, Springer-
Verlag, Berlin, p. 106.
14. Nan, C.W. (1993) Physics of inhomogeneous inorganic materials, Prog. Mater. Sci. 37, 1.
15. Viswanathan, R. and Heaney, M.B. (1995) Direct imaging of the percolation network in a three-
dimensional disordered conductor-insulator composite, Phys. Rev. Letters 75, 4433.
16. Heaney, M.B. (1995) Measurement and interpretation of nonuniversal critical exponents in disordered
conductor/insulator composites, Phys. Rev. B 52, 12477.
17. Heaney, M.B. (1997) Electrical transport measurements of a carbon-black/polymer composite, Physica
A 241, 296.
18. Wu, J., and McLachlan, D.S. (1997) Percolation exponents and thresholds obtained from the nearly
ideal continuum percolation system graphite/boron-nitride, Phys. Rev. B 56, 1236.
19. Wu, J., and McLachlan, D.S. (1997) Percolation exponents and thresholds in two nearly ideal
anisotropic continuum systems, Physica A 241, 360.
20. Halperin, B.I., Feng, S., and Sen, P.N. (1985) Differences between lattice and continuum percolation
transport exponents, Phys. Rev. Letters 54, 2391.
41
This page intentionally left blank
THE INTERMEDIATE PHASE AND SELF-ORGANIZATION IN NETWORK
GLASSES
M.F. THORPE and M.V.CHUBYNSKY
Department of Physics and Astronomy, Michigan State
University, East Lansing, MI 48824
INTRODUCTION
The study of the structure of covalent glasses has progressed steadily since the initial
work of Zachariasen [1] in 1932 that introduced the idea of the Continuous Random
Network (CRN). Zachariasen envisaged such networks maintaining local chemical order,
but by incorporating small structural distortions, having a topology that is non-crystalline.
This seminal idea has met some opposition over the years from proponents of various
microcrystalline models, but today is widely accepted, mainly as a result of careful
diffraction experiments from which the radial distribution function can be determined.
The CRN has been established as the basis for most modem discussions of covalent
glasses, and this has occurred because of the interplay between diffraction experiments and
model building. The early model building involved building networks with ~ 500 atoms
from a seed with free boundaries in a roughly spherical shape [2]. Subsequent efforts have
refined this approach and made it less subjective by using a computer to make the
decisions and incorporating periodic boundary conditions. The best of these approaches
was introduced by Wooten, Winer and Weaire [3] and consists of restructuring a crystalline
lattice with a designated large unit supercell, until the supercell becomes amorphous. The
large supercell contains typically ~5000 atoms. Both the hand built models and the
Wooten, Winer and Weaire model are relaxed during the building process using a
potential. The final structure is rather insensitive to the exact form of the potential and a
Kirkwood [4] or Keating [5] potential is typically used.
Despite this success in understanding the structure, some concerns remain. Perhaps
the most serious of these is that the network cannot be truly random. Even though bulk
glasses form at high temperatures where entropic effects are dominant, it is clearly not
correct to completely ignore energy considerations that can favor particular local structural
arrangements over others. A simple example of this is local chemical ordering, where, for
example, bonding between certain same-type atoms is unfavorable. This can lead to
chemical thresholds that appear at certain concentrations, at which unfavorable bonding
can no longer be avoided. A more interesting and subtle effect of interest to us here is how
the structure itself can incorporate non-random features in order to minimize the free
Phase Transitions And Self-Organization in Electronic and Mol ecul ar Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 43
energy at the temperature of formation. Such subtle structural correlations, which we refer
to as self-organization, will almost certainly not show up in diffraction experiments, but
may have other manifestations, as discussed in the paper of P. Boolchand in this volume.
Here we focus on the mechanical properties and critical mechanical thresholds, as this is
where it is easiest to make theoretical progress at this time.
How can such an idea be developed theoretically? A proper procedure might be to
consider a very large supercell and use a first principles quantum approach, like that of Car
and Parrinello [6], to form the glass. The problem with this is that the relaxation times at
the appropriate temperatures are very large, so full equilibration is impossible. The
structure thus obtained would be unreasonably strained. This situation is made worse as
only small supercells with about 100 atoms can be used at present and in these the periodic
boundary conditions produce unacceptably large internal strains. Using the fastest linear-
scaling electronic structure methods or even molecular dynamics with empirical potentials
is still much too slow. We therefore need to look at other ways of generating self-
organizing networks. One promising approach is that of Mousseau and Barkema [7] who
explore the energy landscape of a glass by moving over saddle points. In network terms,
this corresponds to selective (thus non-random) bond switching. In these lecture notes, we
look at even more simplified approaches that show what kinds of effects self-organization,
and the resulting non-randomness, can lead to.
The layout of this paper is as follows. In the next section we review ideas of rigidity
percolation that lead to a mechanical threshold in networks as the number of bonds per
atom in them (related to the mean coordination number) changes. Then we describe our
model of self-organization of glassy networks, which has two thresholds instead of one and
thus exhibits an intermediate phase, and study the properties of this model. We also
consider a similar model for random resistor networks, based on the analogy between the
usual (connectivity) percolation and rigidity percolation.
Throughout much of this paper we focus on both central force networks in two
dimensions and bond-bending networks in three dimensions that have central and non-
central forces as these are more relevant to glasses. The reader should be aware that we do
flip back and forth between these model systems, as appropriate, in order to illustrate
various points.
RIGIDITY PERCOLATION
Going back more than a century, Maxwell was intrigued with the conditions under
which mechanical structures made out of struts, joined together at their ends, would be
stable (or unstable) [8]. To determine the stability, without doing any detailed calculations
(that would have been impossible then except for the simplest structures), Maxwell used
the approximate method of constraint counting. The idea of a constraint in a mechanical
system goes back to Lagrange [9] who used the concept of holonomic constraints to reduce
the effective dimensionality of the space.
The problem under consideration is a static one given a mechanical system, how
many independent deformations are possible without any cost in energy? These are the
zero frequency modes, which we prefer to refer to as floppy modes because in any real
system there will usually be some weak restoring force associated with the motion.
Sometimes it is convenient to look at the system as a dynamical one, and assign
potentials or spring constants to deformations involving the various struts (bonds) and
angles. It does not matter whether these potentials are harmonic or not, as the
displacements are virtual. However it is convenient to use harmonic potentials so that the
system is linear. It is then possible to set up a Lagrangian for the system and hence define a
dynamical matrix, which is a real symmetric matrix having real eigenvalues. These
eigenvalues are either positive or zero. The number of finite (non-zero) eigenvalues defines
44
the rank of the matrix. Thus our counting problem is rigorously reduced to finding the rank
of the dynamical matrix. The rank of a matrix is also the number of linearly independent
rows or columns in the matrix. Neither of these definitions is of much practical help, and a
numerical determination of the rank of a large matrix is difficult and of course requires a
particular realization of the network to be constructed in the computer. Nevertheless the
rank is a useful notion as it defines the mathematical framework within which the problem
is well posed.
The rigidity of a network glass is related to how amenable the glass is to continuous
deformations that require very little cost in energy. A small energy cost will arise from
weak forces, which are always present in addition to the hard covalent forces that involve
bond lengths and bond angles. These small energies can be ignored because the degree to
which the network deforms is well quantified by just the number of floppy modes [10]
within the system. This picture of floppy and rigid regions within the network has led to
the idea of rigidity percolation [11,12]. When new constraints are added to an initially
floppy network and it crosses the rigidity percolation threshold, a single rigid region
percolates through the network and it becomes stable against external straining (elastic
moduli become non-zero).
There are two important differences between rigidity and connectivity percolation.
The first difference is that rigidity percolation is a vector (not a scalar) problem, and
secondly, there is an inherent long-range aspect to rigidity percolation. These differences
make the rigidity problem become successively more difficult as the dimensionality of the
network increases. In two dimensions, Figure 1(a) shows four distinct rigid clusters
consisting of two rigid bodies attached together by two rods connecting at pivot joints.
Now the placement of one additional rod, as shown in Figure l(b), locks the previous four
clusters into a single rigid cluster. This non-local character allows a single rod (or bond) on
one end of the network to affect the rigidity all across the network from one side to the
other.
Using concepts from graph theory, we have set up generic networks where the
connectivity or topology is uniquely defined but the bond lengths and bond angles are
arbitrary. A generic network does not contain any geometric singularities [13], which occur
when certain geometries lead to nul l projections of reaction forces. Null projections are
caused by special symmetries, such as, the presence of parallel bonds or connected
collinear bonds. Rather than these atypical cases, their generic counterparts as shown in
Figures 1(b) and (c) will be present. This ensures that all infinitesimal floppy motions carry
over to finite motions [13-15].
Figure 1. The shaded regions represent 2D rigid bodies. The (closed, open) circles denote pivot-joints that
are members of (one, more than one) rigid body. (a) A floppy piece of network with four distinct rigid
clusters. (b) Three generic cross links between two rigid bodies make the whole structure rigid. If the bonds
were parallel, the structure would not be rigid to shear. (c) A set of three non-collinear connected rods
connecting across a rigid body is generic and contains one internal floppy mode. If they were collinear (along
the dashed line), then there would be two infinitesimal (not finite) floppy motions, and under a horizontal
compression buckling would occur.
By considering generic networks, the problematic geometric singularities are
completely eliminated. Therefore, the problem of rigidity percolation on generic networks
leads to many conceptual advantages because all geometrical properties are robust.
45
Moreover, real glass networks have local distortions, and are modeled better by generic
networks.
Constraint Counting
The genius of Maxwell [8] was to devise the simple constraint counting method that
allows us to estimate the rank of the dynamical matrix and hence the number of floppy
modes.
The number of floppy modes in d dimensions is given by the total number of degrees
of freedom for N sites (equal to dN ) minus the number of independent constraints. A
dependent (redundant) constraint does not change the number of floppy modes. It can only
add additional reinforcement and it cannot be accommodated without changing the natural
bond lengths and angles of the network, so stressed (over-constrained) regions would be
created. A key quantity is the number of floppy modes, F , in the network, or normalized
per degree of freedom, f = F/dN. By defining the total number of constraints per degree
of freedom as n
c
and the number of redundant constraints per degree of freedom as n
r
, we
can write quite generally,
It is straightforward to find the total number of constraints (and consequently n
c
) for each
given network. Neglecting redundant constraints [n
r
in Eq. (1)] as first done by Maxwell
[8], we come to Maxwell counting:
Now the idea is to associate the rigidity percolation transition with the point where f
M
goes to zero. The Maxwell approximation gives a good account of the location of the phase
transition and the number of floppy modes, but it ultimately fails, because some constraints
are redundant and also because, as we will see soon, there are still some floppy pockets
inside an overall rigid network.
We now describe Maxwell counting for specific cases.
Central Force Network in Two Dimensions. The elastic properties of random
networks of Hooke springs have been studied over the past 15 years [11,16-20]. This
system can be viewed as a network of Hooke springs in 2 dimensions, which is built from a
regular (say, triangular) lattice, whose bonds are represented by springs, by removing the
bonds at random, so each one is present with probability p (bond dilution). The site diluted
version of the problem was also considered (see, e.g., [21]).
For constraint counting it is convenient to introduce the mean coordination as an
average number of bonds stemming from a site. It is given by where p is the
probability of the bond being present, z is the coordination of the underlying regular
lattice (6 for the triangular lattice, for example). If the total number of sites is N , the
number of bonds is Each of these bonds represents one constraint, as
always in central force networks, and therefore the number of constraints per degree of
freedom is given by
46
Therefore, according to Eq. (2), Maxwell counting gives
This quantity goes to zero at which we associate with the rigidity percolation
transition. For the triangular lattice this corresponds to p
c
=2/3.
Bond-Bending Glassy Networks in Three Dimensions. We start by examining a
large covalent network that contains no dangling bonds or singly coordinated atoms. We
can describe such a network by the chemical formula Ge
x
As
y
Se
1xy
, where the chemical
element, Ge, stands for any fourfold bonded atom, As for any threefold bonded atom and
Se for any twofold bonded atom. Each atom has its full complement of nearest neighbors
and we consider the system in the thermodynamic limit, where the number of atoms
There are no surfaces or voids and the chemical distribution of the elements is not
relevant, except that we assume there are no isolated pieces, like a ring of Se atoms. The
total number of atoms is N and there are n
r
atoms with coordination r (r = 2, 3 or 4), then
and we can define the mean coordination
We note that (where ) gives a partial but very important description of the
network. Indeed, when questions of connectivity are involved the average coordination is
the key quantity.
In covalent networks like Ge
x
As
y
Se
1xy
, the bond lengths and angles are well
defined. Small displacements from the equilibrium structure can be described by a
Kirkwood [4] or Keating [5] potential, which we can write schematically as
The mean bond length is l, is the change in the bond length and is the change in the
bond angle. The bond-bending force is essential to the constraint counting approach
for stability, in addition to the bond stretching term The other terms in the potential
are assumed to be much smaller and can be neglected at this stage. If floppy modes are
present in the system, then these smaller terms in the potential will give the floppy modes a
small finite frequency. For more details see Ref. [16]. If the modes already have a finite
frequency, these extra small terms will produce a small, and rather uninteresting, shift in
the frequency. This division into strong and weak forces is essential if the constraint
counting approach is to be useful. It is for this reason that it is of little, if any, use in metals
47
and ionic solids. It is fortunate that this approach provides a very reasonable starting point
in many covalent glasses.
To estimate the total number of zero-frequency modes, Maxwell counting was first
applied by Thorpe [16], following the work of J.C. Phillips [22,23] on ideal coordinations
for glass formation. It proceeds as follows. There are a total of 3N degrees of freedom.
There is a single central-force constraint associated with each bond. We assign r/2
constraints associated with each r-coordinated atom. In addition there are constraints
associated with the angular forces in Eq. (7). For a twofold coordinated atom there is a
single angular constraint; for an r-fold coordinated atom there are a total of 2r3 angular
constraints. The total number of constraints is therefore
Using Eqs. (5) and (6), their fraction n
c
can be rewritten as
thus, according to Eq. (2),
Note that this result only depends upon the combination which is the relevant
variable. When (e.g. Se chains), then f
M
= 1/3 ; that is, one third of all the modes
are floppy. As atoms with higher coordination than two are added to the network as cross-
links, f
M
drops and goes to zero at and network goes through the rigidity
percolation transition. This mean field approach has been quite successful in covalent
glasses and helps explain a number of experiments. Also in later sections, we discuss the
results of computer experiments and show that they are rather well described by the results
of this subsection.
We note that Eq. (8) holds only when there are no 1-fold coordinated atoms. Their
presence leads to the threshold being shifted down [24-26].
The Pebble Game
Until recently it has not been possible to improve on the approximate Maxwell
constraint counting method, except on small systems with up to ~ 10
4
sites using brute
force numerical methods. Now a powerful exact combinatorial algorithm, called the Pebble
Game, has become available. This algorithm, first suggested by Hendrickson [13] and
implemented by Jacobs and Thorpe [12,27,28], allows systems containing more than 10
6
sites to be analyzed in two-dimensional generic central-force networks and in three-
dimensional networks with both central forces and bond-bending forces.
The crux of the Pebble Game algorithm in two dimensions is based on a theorem by
Laman [14] from graph theory. We note first that if for a two dimensional network
Maxwell counting gives less than 3 floppy modes (3 modes are always there, as they
correspond to rigid motions of the network), the counting cannot be exact and thus a
redundant bond (or bonds) are present. One says that the Laman condition for the network
is violated in this case. But if the opposite is true (the Laman condition is satisfied), this is
not sufficient for redundant bonds to be absent, as the network can have more than 3 floppy
modes and redundant bonds simultaneously. The statement of the theorem is that non-
48
violation of the Laman condition for every subnetwork is sufficient for not having
redundant bonds. This statement does not generalize to dimensions higher than two.
We do not go into details of the algorithm, which can be found elsewhere [12,27,29].
For this consideration it is enough to know that one starts from an empty lattice (having
no bonds, only sites) and adds bonds one at a time. Each newly added bond is tested for
independence and each independent bond decreases the number of floppy modes by one.
Besides providing exact constraint counting, the algorithm is able to identify all rigid
clusters (and thus whether or not rigidity percolation occurs) and find all the regions, in
which redundant bonds introduced stress (over-constrained regions).
Figure 3. The topology of a typical section from a bond-diluted generic network at p = 0.62 (below
percolation) and at p = 0.70 (above percolation). A particular realization would have local distortions (not
shown), thus making the network generic. The heavy dark lines correspond to over-constrained regions. The
open circles correspond to sites that are acting as pivots between two or more rigid bodies.
Sections of a large network on the bond-diluted generic triangular lattice are shown in
Figure 2 after the pebble game was applied. Below the transition the network can be
macroscopically deformed as the floppy region percolates across the sample. Above the
rigidity transition, stress will propagate across the sample. However, below the transition
there are clearly pockets of large rigid clusters and over-constrained regions, while above
the transition there are pockets of floppy inclusions within the network.
While Lamans theorem does not generally apply to three dimensions, it is possible to
generalize the Pebble Game algorithm for a particular class of networks, namely, the bond-
bending networks with angular forces included as in a Kirkwood or Keating potential [Eq.
(7)]. Fortunately, the bond-bending model is precisely the class of models that is applicable
to the study of many covalent glass networks. A longer discussion of the three dimensional
Pebble Game is given in Refs. [28,30].
Two Dimensional Central Force Network. In this subsection, we review some
results for central-force generic rigidity percolation on the triangular net. A more detailed
account can be found in Ref. [27].
We begin by finding the number of floppy modes and comparing it to the Maxwell
counting result. The exact value of f is very close to f
M
far enough below the mean-field
estimate for the rigidity transition but then starts to deviate significantly and
does not reach zero (until full coordination, is reached). The quantity f looks
quite smooth, but the second derivative of it with respect to (shown in the insert) does
in fact have a singularity. This singularity corresponds to the rigidity percolation threshold,
as can be checked by detecting the percolating rigid cluster directly. Using finite-size
49
scaling, the position of the transition was found to be This is
amazingly close to the mean-field value of 4.
The behavior of the second derivative suggests that the number of floppy modes is an
analogous quantity for rigidity and connectivity percolation. In the case of connectivity
percolation, the number of floppy modes is simply equal to the total number of clusters,
which corresponds to the free energy [16,31-33]. It would be nice if a similar result holds
for rigidity percolation. It turns out that the second derivative of the total number of
clusters changes sign across the transition, thus violating convexity requirements. Noting
that typically rigid clusters are not disconnected, it was suggested that the number of
floppy modes generalizes as an appropriate free energy [16,31-33]. With this assumption,
the exponent is estimated in the usual context of a heat capacity critical exponent, even
though no temperature is involved here.
Again analogously to connectivity percolation, the fraction of bonds in the percolating
rigid cluster serves as the order parameter for this system. The critical exponent is
defined as the rigid cluster size critical exponent. Another order parameter is also possible,
namely, the fraction of bonds in the percolating stressed cluster, which is defined as a
percolating stressed subset of the percolating rigid cluster. It was found (and this is an
important point) that both and go to zero at the same point the percolation
transition. This will be different in the next chapter on self-organization and will lead to the
existence of the intermediate phase, and two phase transitions.
The results of study of this model [27] lead to the conclusion that the rigidity
transition in this system is second order, but in a different universality class than
connectivity percolation.
It has been suggested by Duxbury and co-workers [21] that the rigidity transition
might be weakly first order on triangular networks. While we think this is unlikely, it
cannot be completely ruled out at the present time.
Three Dimensional Bond Bending Networks. It can be shown [28] that the only
floppy element in a three dimensional bond-bending network is a hinge joint. Hinge joints
can only occur through a central-force (CF) bond and are always shared by two rigid
clusters allowing one degree of freedom of rotation through a dihedral angle. Note that in
two dimensional central force generic networks, sites that belong to more than one cluster
act as a pivot joint, and more than two rigid clusters can share a pivot joint. Because of this
difference between CF and bond-bending networks, the order parameters analogous to
and of the previous subsection, have to be defined as a fraction of sites in respective
percolating clusters and not bonds, as bonds can be shared between a percolating and a
non-percolating clusters.
For purposes of testing rigidity in generic three-dimensional bond-bending networks,
it is only necessary to specify the network topology or connectivity of the CF bonds, since
the second nearest neighbors via CF bonds define the associated bond-bending constraints.
Here, we have considered two test models. In the first model, a unit cell is defined from our
realistic computer generated network of amorphous silicon [34] consisting of 4,096 atoms
having periodic boundary conditions. Larger completely four-coordinated periodic
networks containing 32,768, 262,144 and 884,736 atoms are then constructed from the
amorphous 4,096-atom unit cell.
The four-coordinated network is randomly diluted by removing CF bonds one at a
time with the constraint that no site can be less than two-coordinated. That is, a CF bond is
randomly selected to be removed. If upon removal either of its incident atoms becomes less
than two coordinated, then it is not removed and another CF bond is randomly selected
from the remaining pool of possibilities. The order of removing CF bonds is recorded. This
process is carried out until all remaining CF bonds cannot be removed, leading to as low an
50
average coordination number as possible. All CF bonds that were successfully removed are
marked. This method of bond dilution gives a simple prescription for generating a very
large model of a continuous random Ge
x
As
y
Se
1-x-y
type of network. For comparison, a
second test model, a diamond lattice, was diluted in the same way and contained 32,768,
262,144 and 10
6
atoms.
The results of simulations of both models are qualitatively similar to those for 2D
central-force networks. Both have a rigidity transition slightly below the Maxwell counting
estimate of 2.4. Again, the rigidity transition can be accurately found from the sharp peak
in the second derivative of the fraction of floppy modes. In particular, for the diamond
lattice and for a-Si, Remarkably, the Maxwell
counting estimate is accurate to about 1% in locating the threshold in both cases.
A more detailed account of the results can be found in Ref. [35].
SELF-ORGANIZATION AND INTERMEDIATE PHASE
Self-Organization in Rigidity Percolation
Description of the Model. We have mentioned that starting from an empty lattice
(without bonds) and adding one bond at a time, we can use the pebble game to analyze
whether the bond we are adding is independent of those already in the network or
redundant. We also know that redundant bonds create stressed (over-constrained) regions.
Thus within the present approach we have a rather unique opportunity to construct stress-
free networks without a huge computational overhead.
The idea is to start, as before, from an empty lattice and add one bond at a time to it,
applying the pebble game at each stage. If adding a trial bond would result in that bond
being redundant and hence create a stressed region, then that move is abandoned. Thus the
network self-organizes in such a way that there is no stress in it at all. Note that the pebble
game now serves not only as a tool to analyze the network, as before, but also as a
decision-making mechanism when building the network.
It is not possible to keep adding bonds beyond a certain point, without introducing
stress (this is considered in more detail below). How should we proceed then? While going
on with some sort of self-organization would be reasonable (as some bonds would create
less stress than others), it is impossible to analyze this within our model, so we start
inserting bonds completely at random, once avoiding stress becomes impossible.
General Properties. First of all, how long is it possible to keep adding bonds to a
network without introducing stress? It is certainly impossible to have more independent
constraints then there are degrees of freedom in the network. Now recall that in the
Maxwell counting approximation, the rigidity transition occurs when the numbers of
constraints and of degrees of freedom balance. Thus it is certainly not possible to have an
unstressed network with the mean coordination above where Maxwell counting predicts
the transition (that is, above for central-force networks in 2d and for
glassy networks in 3d). This provides an upper limit (still not always reachable, as we will
see) for the unstressed networks. Note, though, that since the Maxwell counting percolation
limit is not exact, this does not mean that rigid networks are necessarily stressed! The
actual rigidity transition may occur below the point where Maxwell counting puts it. This
is a very important point that leads to possibility of an intermediate phase, as described
below.
Secondly, we know that the Maxwell counting result for the number of floppy modes
would be exact if all constraints in the network were independent. But this is exactly what
we have in our case! Thus the number of floppy modes in Maxwell counting is exact for as
51
long as we are able to keep the network unstressed. Hence we follow the Maxwell result
for the number of floppy modes in the floppy and intermediate phases.
We now analyze some specific cases in more detail.
Intermediate Phase in 2D Central-Force Networks. Let us first prove that it is
indeed possible to reach the Maxwell counting limit without any stress in this case
(and for any CF networks), provided that the fully coordinated (undiluted) network has no
floppy modes (which is the case for triangular networks). As we have seen before,
generally speaking, we should distinguish carefully between constraints and bonds. A
constraint can be thought of as one algebraic relation for the coordinates of atoms; stress
appears whenever one or more of such relations are not satisfied. A bond can have several
associated constraints, as in bond-bending networks. In the case of CF networks, though,
each bond has only one associated constraint (the distance between the sites it connects), so
bonds and constraints are identical. Recall once again that every single constraint can
be either independent (in which case it reduces the number of floppy modes of the network
by 1), or redundant (so it does not change the number of floppy modes). At the point where
stress becomes inevitable any trial bond would cause stress (be redundant). So all the
bonds, which will be subsequently inserted, are redundant. Thus f will remain constant up
to the very end (which is the full lattice), therefore f = 0 at this point. Since Maxwell
counting is still exact there, the proof is complete. We would like to emphasize that
equivalence of bonds and constraints was essential for this proof (we used these terms
interchangeably). See the next subsection for comparison.
Secondly, it is possible to establish a relation between the self-organized networks and
those obtained by usual completely random insertion (to which we for simplicity refer as
random in contrast to self-organized in what follows). Indeed, assume we are using the
same random list of M bonds to build a random network and a self-organized one, trying to
insert bonds as they are listed. For the random network, all the M bonds will get in; for the
self-organized network, some of them will be, generally speaking, rejected, so that
will be inserted. The bonds rejected in the self-organized network will be
redundant in the random one; they do not influence the number of floppy modes, the
configuration of rigid clusters (and thus whether or not rigidity percolation occurs) and the
redundancy or independence of all the subsequently inserted bonds. Thus all these
characteristics will be identical for the two networks. The consequence is that there is a
correspondence between self-organized and random networks having the same number of
floppy modes; in particular, rigidity percolation occurs at the same number of floppy
modes.
This analysis allows us to make a very important conclusion. Since in random
networks rigidity percolates at a non-zero f and the same has to be true for self-organized
networks (because of the just mentioned consequence), yet stress appears exactly at f = 0,
we conclude that there exists an intermediate phase, which is rigid (i.e. the infinite rigid
cluster exists), but unstressed (so, evidently, there is no stress percolation). This is different
from the situation with random insertion, where the rigidity and stress percolation
thresholds always coincide (see Figure 3).
It could be possible that stress does not percolate immediately after it is introduced;
we will see from simulation results that this is not the case, so the upper boundary of the
intermediate phase (the stress transition) may be defined as either the point where stress
first appears, or equivalently, the point where it percolates. As is seen from our
consideration, it lies at
As we have mentioned, the fractions of bonds in percolating rigid and stressed clusters
(denoted and respectively) can serve as order parameters. Now, since there is an
intermediate phase where rigidity percolates, while stress does not, these two parameters
52
turn zero at different points, between which the intermediate phase lies. Besides, since the
number of floppy modes is zero above the stress transition, the whole network is rigid, and
thus is identically 1. These facts are illustrated in Figure 3.
Figure 3. Order parameters and for self-organized and random triangular networks. It is seen that
the intermediate phase (shaded) is formed in the self-organized case, extending from 3.905 to 4, while in the
randomcase the two thresholds coincide and there is no intermediate phase. All results are averages over two
realizations on 400400 networks.
Given the discussion of the floppy modes in the random and self-organized networks,
it is tempting to suggest that the same relation holds for the just defined rigidity order
parameter. The subtlety is that the relation is defined in terms of sites (i.e., same sites are in
the percolating cluster and same sites are pivot joints on its border), while the order
parameter is defined in terms of bonds. Of course, there is no direct correspondence
between bonds, as there are different numbers of bonds in related random and self-
organized networks. Still it might be safely assumed that the rigid cluster size critical
exponents are the same for rigidity percolation in random and self-organized networks.
Other critical exponents may be different, though.
It is interesting to note that since f given by Maxwell counting is exact in the whole
unstressed region, in both the floppy and the intermediate phase f is a perfect straight line
and the rigidity transition does not show up in f .
Results of our simulations of this model are shown in Figures 3 and 4. The
simulations were done for networks with periodic boundary conditions in both directions.
There are several facts to be inferred (besides confirming all the results we have obtained
so far). We see that stress percolates immediately after it appears at (this fact was
mentioned above). Second, the cluster size critical exponent for the stressed cluster is quite
small (smaller than the one for the rigid cluster). In random networks, the stressed cluster
exponent is larger than the rigid cluster exponent, which is because the stressed percolating
cluster is smaller than the rigid cluster (the former being a subset of the latter) and the two
thresholds coincide.
53
Figure 4. Number of floppy modes per degree of freedom for self-organized and random triangular
networks. Thresholds are shown with different symbols. The intermediate phase in the self-organized case is
shaded. Note that rigidity percolation occurs at the same f in the random and self-organized cases. The self-
organized plot is strictly linear up to and coincides with Maxwell counting.
Intermediate Phase in 3D Bond-Bending Networks. In case of glassy networks
there is a slight problem with implementing our general algorithm of self-organization. In
the CF case we were starting from an empty lattice to ensure that it had no stress initially.
In the present case the initial dilution can only go as far as to the point where any further
dilution would create a 1-coordinated site. At this limit there are no bonds with both ends
being sites of coordination 3 and higher, so that further dilution is impossible. It is
generally not true that this final network is unstressed. For smaller networks (~10
4
sites and
less), it is possible to pick those that are unstressed; for larger ones such cases are rare, and
it is reasonable to assume that the fraction of constraints that are redundant is a constant in
the thermodynamic limit. This constant seems to be very low, though (in our simulations,
typically about 0.05% of constraints were redundant). Besides, the number of redundant
constraints does not grow when new bonds are inserted according to our algorithm (up to
the stress transition), so this problem is largely irrelevant.
Unlike the case of CF networks, BB networks have more than one constraint
associated with each bond. When a new bond is added, not only the distance between the
sites it connects is fixed, but the angles between the new bond and those stemming out of
the two sites at either end of that bond are fixed as well. Any bond that has at least one
redundant constraint associated with it would cause stress. Some of the stress-causing
bonds have only part of the associated constraints redundant and the rest independent, and
such a bond will change the number of floppy modes. This makes some of our conclusions
made for CF networks invalid in this case.
Firstly, this invalidates the proof of the reachability of the Maxwell counting limit
( in this case). This is because even when at the upper reachable limit all the as
yet uninserted bonds would cause stress, some of these bonds may further decrease the
number of floppy modes and thus this number is not necessarily zero at this point.
Secondly, the nice relation between random and self-organized networks no longer
holds, because out of the redundant bonds by which the two differ, some (namely, the
partially redundant ones) change f , rigidifying the network and changing the
54
configuration of rigid clusters. Still the equality of critical exponents for rigid cluster
sizes in random and self-organized cases probably holds.
At the same time, some facts are unchanged. In particular, f given by Maxwell
counting is still exact in the unstressed region. Most importantly, the intermediate phase
still exists.
The results of simulations done for the diluted diamond lattice are given in Figures 5
and 6. As in the previous subsection, we use periodic boundary conditions in all directions.
We note in addition to the graphs that, as in the CF case, stress percolates immediately
after it appears. The intermediate phase extends from to 2.392 (not reaching
2.4). Again, the stress transition is sharper than the rigidity transition. Our results are
consistent with the second order transition with the very small critical exponent
or a first order transition is more likely.
Another feature of the plot in Fig. 5 is that the rigidity order parameter is not exactly
unity in the stressed phase (which is expected, as some floppy modes remain in the stressed
phase) and the second transition shows up as a kink in the rigidity order parameter.
In conclusion to this section, we would like to mention that it is possible within our
approach to establish a hierarchy of stress-causing bonds (by the number of associated
redundant constraints) and when stress becomes inevitable, first put those having one
redundant constraint, then those having two, and so on. Exactly at only those
bonds having no associated independent constraints will remain uninserted. It is unlikely,
though, that there is a good correlation between the number of redundant constraints and
the actual increase in stress energy, as the distribution of stresses caused by different bonds
is quite wide, so this complication seems unreasonable.
Figure 5. The order parameters and for the self-organized diluted diamond lattice. The intermediate
phase is shaded. Circles are average over 4 networks with 64,000 sites, triangles are averages over 5 networks
with 125,000 sites. The dashed lines are the power law fit below the stress transition and for guidance of the
eye above. Note the break in the slope at the stress transition.
Elastic Properties of Self-Organized Networks. So far our study of self-organized
networks was limited to their geometrical properties. Of course, this work becomes really
meaningful when we turn to what the physical consequences of self-organization are. The
simplest quantity to look at is the elasticity of the networks of springs. Unfortunately, the
55
pebble game, being concerned with the geometric properties only, is unable to help us find
the numerical values of elastic constants, so we have to do a usual relaxation using, for
example, the conjugate gradient method [36] and consider particular configurations, and
not just the connectivity. So far in this preliminary study, we have only considered the 2d
case.
The first and quite surprising fact is that in case of periodic boundary conditions in all
directions the elastic constants are exactly zero in the intermediate phase, regardless of the
size of the supercell and despite the existence of the percolating rigid cluster. Indeed,
periodic boundary conditions mean that positions of images of same site in different
supercells are fixed with respect to each other. The network is built stressless with these
additional constraints taken into account. The exact specification of these constraints
beyond stating what sites are involved is determined by the particular size and shape of the
supercell, but is never taken into account (just as particular bond lengths never matter in
determination of stressed regions). So straining the network by changing this size and
shape leaves it stressless. The important thing here is that straining does not add any new
constraints. We confirmed this result numerically by doing exact diagonalization of the
dynamical matrix (similar to [37]), rather than by relaxation, which ensures better
precision.
Figure 6. The fractions of floppy modes per degree of freedom for the diluted diamond lattice (both self-
organized and random cases). Different thresholds and the Maxwell prediction for the rigidity threshold are
shown with different symbols. The intermediate phase in the self-organized case is shaded. The Maxwell
counting line is seen only above the stress transition point in self-organized networks, as below this point it
coincides with the self-organized line.Note that the rigidity transition in the two cases no more occurs at the
same f . Instead, the values of are close, which is probably coincidental.
Of course, for different boundary conditions the elastic constants may be non-zero for
finite samples, but are expected to vanish in the thermodynamic limit. We consider the
busbar geometry, in which busbars are applied to two opposite sides of the network and it
is strained perpendicular to the busbars. The network is built assuming open boundaries at
the busbars and periodic boundary conditions parallel to the busbars. The first and the last
rows of sites are assumed belonging to the respective busbar (i.e., attached rigidly to it). In
addition, when building the network, we consider the sites belonging to each busbar as
being fixed with respect to each other, connecting them with fictitious bonds and
considering these bonds as belonging to the network. This makes the open boundaries less
open and eliminates certain boundary effects, as will be clear from an analogy in the next
56
section with connectivity percolation. The arguments of the previous paragraph do not
apply here, as the network is built not assuming a fixed distance between the busbars (as if
it is allowed to relax) and straining changes and fixes it thus imposing an additional
constraint.
Figure 7. An example of the triangular self-organized network 150150 in the intermediate phase (at
). The thickest bonds belong to the applied-stress backbone, those of medium thickness are in the
percolating rigid cluster (but not in the backbone), the thinnest ones are not in the percolating cluster. The
busbars are shown schematically.
When introducing the boundary conditions as described above, we will have non-zero
stress when an external strain is applied, and some of the bonds will be stressed. These
bonds are said to belong to the applied stress backbone [21] (which we refer to as simply
backbone in what follows). It can be found easily by the pebble game using a method
proposed by Moukarzel [38], which in our case consists in putting an additional bond
across the network emulating the external strain, and finding those bonds in which stress is
induced. A typical result is shown in Figure 7. It is seen that the backbone has filamentary
structure. We note that stress in this backbone was created by putting just one extra bond
and thus it is enough to take any one bond out of the backbone for it to be destroyed, so it
is extremely fragile. Also, since the backbone always has only one redundant bond (when
the bond across is added), it does not grow throughout the intermediate phase after it
appears at the rigidity transition, because growth can only occur by adding new redundant
bonds. This means that for any given sample the elastic constants are the same throughout
the intermediate phase (here we mean finite samples, of course, as in the infinite limit the
elastic constants are zero).
57
Figure 8. The elastic modulus c
11
for self-organized triangular networks. Each point corresponds to one
sample (their linear sizes are specified by different symbols). The intermediate phase is shaded. The dashed
line is the mean-field linear dependence, reaching 1 at the full coordination.
We found the elastic modulus c
11
numerically in both the intermediate and stressed
phases. The triangular lattice was distorted by random displacement of atoms. For
displacements along each axis uniform distribution on an interval (0.1; 0.1) in units of
the lattice constant was chosen, but the results are only slightly sensitive to the width of the
distribution. Equilibrium lengths of springs were chosen equal to the distance between the
atoms they connect, so the initial network is unstressed. Thus subtraction of two large
energies when finding elastic constants is avoided. The results are shown in Fig. 8. Pre-
determining the applied stress backbone speeds up the relaxation greatly, as was first
pointed out in Ref. [21]. Still, we were unable to reach full relaxation in the intermediate
phase in all but the smallest samples (up to 3030). The values in the intermediate phase
are very low and are assumed to go to zero in the limit of large samples. We are currently
doing finite size scaling to test this. Above the stress transition, the modulus seems to grow
linearly, but, of course, it is hopeless to try and determine the critical exponent with
reasonable precision from our data.
Self-Organization in Connectivity Percolation
The model. It is interesting and useful to see if similar phenomena are possible in the
more familiar case of connectivity percolation, especially as connectivity percolation is
easier to study and understand.
The essence of our algorithm of building self-organized networks in the rigidity case
is rejecting stress-causing bonds (or those having redundant constraints). As we have seen,
in the CF case, when bonds and constraints are the same, we may equivalently
formulate this as rejecting redundant or irrelevant bonds. In bond connectivity percolation
we also can build the networks by inserting bonds one by one; most importantly, there is a
clear analog to redundant bonds. The relevant property now is connectivity, by which we
mean the presence or absence of paths connecting any two sites of the network. Redundant
bonds are those which connect sites already connected, that is would close a loop in the
network. Thus the analog of self-organization is building loopless networks.
There are other equivalent ways to draw this parallel. The first is based on the fact that
connectivity percolation can be considered as rigidity percolation with the sites having one
58
degree of freedom regardless of the lattice dimensionality. Each site thus has one
coordinate and each bond is a relation between the coordinates of the sites it connects.
Then the concepts of rigid clusters and clusters in the usual connectivity sense coincide.
The number of floppy modes f is now the number of clusters. A redundant bond in the
rigidity sense is the one that does not change f, it is also stress-causing, as it would
introduce a relation between coordinates that cannot generally be satisfied. On the other
hand, viewed from the connectivity perspective, such a bond connects the sites belonging
to the same cluster and closing a loop, and our model is again recovered. Yet another way
is to recall that rigidity percolation with angular constraints in 2D (or with angular and
dihedral constraints in 3D) is equivalent to connectivity percolation. Then stresslessness is
equivalent to looplessness.
Connectivity percolation and related phenomena were studied so extensively in all
imaginable flavors that it would be strange if this and similar models were not studied
before. Indeed exactly this model was proposed as far back as 1979 [39] and rediscovered
in 1996 [40]. Besides, there was an extensive study of loopless graphs (trees) in relation to
various phenomena ranging from resistance of a network between two point contacts
(considered by Kirchhoff in mid nineteenth century [41]) to river networks [42] to certain
optimization problems [43,44]. In many of these and other papers the algorithm for
building trees was equivalent to ours. Still, we consider this model from a different
perspective.
Given that connectivity percolation can be considered as rigidity percolation with one
degree of freedom per site, we can apply the usual two-dimensional pebble game with the
simple modifications.
Of course, the essence of our self-organization algorithm is still the rejection of bonds
that are not independent. The pebble game allows the determination of all analogs of the
quantities considered for rigidity.
The Intermediate Phase. In this section we carry out the same kind of analysis as
was done for rigidity percolation.
First of all we describe Maxwell counting, as this, although simple, is rarely discussed
in relation to connectivity percolation. For a network with N sites the number of degrees of
freedom is now simply N, the number of constraints is, as before, so the number
of floppy modes per site is and this becomes zero at
Since, as we have seen, connectivity percolation is nothing but a kind of rigidity
percolation on a CF network with 1 degree of freedom per site, all of the general analysis
for CF networks in the previous section is valid. Specifically, Maxwell counting is exact in
the unstressed (this now means loopless) phase; the limit is reachable without
creating loops; the relation between random and self-organized networks also holds.
The order parameters are defined analogously to the rigidity case. The first parameter
is (by analogy) the size of the percolating (connectivity) cluster. However, the difference is
that now the clusters (including the percolating one) can be defined in terms of either
bonds or sites (there are no pivot joints that would be shared between several clusters).
Therefore, there is a possibility to define this order parameter as the fraction of sites
(instead of bonds) in the percolating cluster. This makes the relation between the order
parameters of self-organized and random networks with the same number of floppy
modes (clusters) exact. Yet, to be consistent, we ignore this possibility and define the
order parameters as fractions of bonds, not sites. The second order parameter is, logically,
the fraction of stressed bonds (bonds in loops).
We do not show the results of simulations (which were done for the square lattice) as
they are very similar to those in rigidity case, except that the stressed cluster critical
exponent is larger, not smaller than the connected cluster exponent. Existence of the
59
intermediate phase is confirmed in the range from to 2 for the square net. The
lower transition coincides with the result obtained in Ref. [40].
Conductivity. Similarly to the elasticity case, we consider the busbar geometry here
in two variants, with and without fictitious bonds making sites at the busbars rigid with
respect to each other (we refer to these two cases as boundary conditions A and B
respectively). As in rigidity, it is possible to find the conductivity backbone by
Moukarzels method [38] (for B all the fictitious busbar bonds have to be put prior to
placing the bond across, while they are already in the network in case A). Two examples
corresponding to A and B are shown in Fig. 9. For A the backbone consists of just one
path, while for B it is tree-like with branching near the busbars. Analog of these boundary
effects in B is what was eliminated in study of elasticity, when the boundary conditions
analogous to A were chosen.
The simulations for conductivity in 2D can be done very efficiently with the Frank-
Lobb algorithm [45], whose only limitation is that it is applicable for the open boundary
conditions only.
Figure 9. Examples of self-organized square networks in the intermediate phase with boundary conditions A
(left panel) and B (right panel), as described in the text. The thickest bonds are in the conducting backbone,
those of medium thickness are in the percolating cluster (but not in the backbone), the thinnest are not in the
percolating cluster. The busbars are shown schematically.
It is known from work on a river network model built in the same way as our network
[42] that the backbone branches (in fact, all network branches) are fractal and the fractal
dimension is The only essential difference between the river network model and
our one is that they consider spanning trees (i.e., all sites are in the connecting cluster),
which in our case corresponds only to This should not matter, though, since it is
the dimensionality of the network that the cluster actually spans (i.e., of the connecting
cluster) that is important and this dimensionality is 2 everywhere in the intermediate phase.
Thus we come to a conclusion that the fractal dimension of backbone branches is the same
throughout the intermediate phase and equals 1.22. We confirm this fact in our simulations.
We note that this differs from both the random walk result (d = 2) and that for self-
avoiding random walks (d = 4/3) in our case branches are more straight than both of
these walks. Then for boundary conditions A it is obvious that the fraction of bonds in
the backbone is and the conductance is Our simulations confirm
this result, for both variants of boundary conditions. The effective conductivity in 2D is
equal to the conductance. Thus we come to the conclusion that the conductivity does
indeed go to zero in the thermodynamic limit for the intermediate phase.
60
The results in both the intermediate and the stressed phase are shown in Fig. 10. Just
as for elastic constants, the dependence in the stressed phase is linear, but now much larger
sizes are available, so this linearity may be exact, but we know of no reason for this to be
so. Note the finite value of the conductivity in the intermediate phase, which is a finite size
effect. This value would be constant for boundary conditions A, as the conducting
backbone consists of just one stem not changing across the intermediate phase. Here this
value changes slightly across the intermediate phase.
We mention here briefly that our preliminary results in 3 dimensions show that the
conductivity is also zero in the thermodynamic limit in the intermediate phase, and it goes
to zero with increasing size even faster than in two dimensions.
Figure 10. Conductivity for resistor networks with present bonds having resistance R
1
= 1 and missing
having resistance (diamonds); superconducting networks (R
1
= 0, R
2
= 200, circles); mixed
networks (R
1
= 1, R
2
= 200, triangles). All results are averages over 10 square networks 100100 with open
boundary conditions parallel to the busbars, the busbar sites are treated as in case B (see text).
Superconducting Networks. We have seen that in the thermodynamic limit the
conductivity is zero in both the disconnected and intermediate phases (just as elastic
constants were zero in both the floppy and intermediate phases in the rigidity case). These
results make us wonder if the lower transition shows up in any physical quantities for
infinite networks. One possibility is to consider superconductor networks instead of
resistor networks. In this model all the existing bonds are replaced with conductors of zero
resistance (superconductors), while all the absent bonds are equal resistors with finite
resistance.
It turns out that the same kind of correspondence between random and self-organized
networks with the same f we had for clusters is valid for the conductance in this case.
Indeed, these networks differ by redundant bonds that connect sites already connected. All
the connected sites have zero potential difference (as they are connected with
superconductors), so putting redundant bonds does not change the distribution of the
potential and thus does not influence the conductance.
It is known [46] that in the random case the resistivity is zero above the threshold and
non-zero below it, with the critical exponent the same as for the conductivity of resistor
networks (1.30). Thus in the self-organized case the resistivity will turn zero in the point
related to the percolation threshold of random networks by the above relation, i.e., at the
61
lower transition. The critical exponent will be the same as in the random case (1.30), but
this is now different from the value for of resistor networks
Mixture of Two Sorts of Resistors. We can now combine the resistor and
superconductor models by introducing two sorts of resistors, with resistances R
1
and R
2
,
R
1
< R
2
, and putting R
1
resistors in place of present bonds and R
2
resistors in place of
missing bonds. Assume now that R
1
<< R
2
. Below the lower transition there are no lower
resistance percolating paths, thus essentially all the potential drop is on the higher-
resistance bonds connecting lower-resistance clusters. This means the potential is almost
constant within these clusters, and they may be considered as consisting of
superconducting bonds. Thus we will have a smeared near-singularity in the conductivity
at the lower transition. There will be a monotonic rise in conductivity throughout the
intermediate phase, but it will remain low, as the low-resistance filaments are irrelevant by
themselves in the thermodynamic limit. Above the second transition the low-resistance
backbone by itself gives finite conductivity, so the high-resistance part of the network is
irrelevant. Unlike the first transition, the second one is sharp, which is an artifact of change
in the bond-insertion algorithm. Figure 14 shows the result for this model for
R
2
/R
1
= 200.
Stability
We have already mentioned the extreme fragility of the backbone and thus of the very
fact of percolation in the intermediate phase: removal of a single bond from it can destroy
it. In fact, if at some point in the intermediate phase we start removing bonds at random,
then since there are more than O(l) bonds in the backbone, percolation will be destroyed
immediately. But maybe upon insertion of bonds it is re-created as easily? We can ask a
more proper question. Namely, our networks turn out to be strongly biased [47]: at any
given not all networks satisfying the condition of stresslessness are equiprobable.
What if we build the networks, which are truly random with the only constraint of
stresslessness? This question was studied for connectivity by variety of methods [47-49],
and, formulating the result in our terms, regrettably, the intermediate phase is destroyed.
We did not study this question in rigidity case yet, but it is clearly important and more
work needs to be done on this issue.
SUMMARY
We have discussed the rigidity of random and self-organized networks. We find that
there is a single transition from floppy to rigid in random networks, but an intermediate
phase intervenes in the self-organized networks. This intermediate phase is rigid but
contains no redundant bonds and so is stress-free.
Some of this work, in the form of more extensive lecture notes has been presented at a
NATO-ASI in Czech Republic in the summer of 2000, and will appear in the NATO-ASI
series. We should like to thank D.J. Jacobs for his many contributions to the pebble game
algorithm and to the US National Science Foundation for support under grant numbers
DMR-0078361 and CHE-9903706.
62
REFERENCES
1. Zachariasen, W.H. (1932) The atomic arrangement in glass, J. Am. Chem. Soc. 54, 3841-3851.
2. Polk, D.E. (1971) Structural model for amorphous silicon and germanium, J. Non-Cryst. Solids 5, 365-
376.
3. Wooten, F., Winer, K., Weaire, D. (1985) Computer generation of structural models of amorphous Si
and Ge, Phys. Rev. Lett. 54, 1392-1395.
4. Kirkwood, J.G. (1939) Skeletal modes of vibration of long-chain molecules, J. Chem. Phys. 7, 506-
509.
5. Keating, P.N. (1966) Theory of the third-order elastic constants of diamond-like crystals, Phys. Rev.
145, 674-678.
6. Car, R. and Parrinello, M. (1985) Unified approach for molecular dynamics and density functional
theory, Phys. Rev. Lett. 55, 2471-2474.
7. Mousseau, N. and Barkema, G.T. (1998) Traveling through potential energy landscapes of disordered
materials: The activation-relaxation technique, Phys. Rev. E57, 2419-2424; Barkema, G.T. and
Mousseau, N. (1996) Event-based relaxation of continuous disordered systems, Phys. Rev. Lett. 77,
4358-4361.
8. Maxwell, J.C. (1864) On the calculation of the equilibrium and stiffness of frames, Philos. Mag. 27,
294-299.
9. Lagrange, J.L. (1788) Mcanique Analytique, Desaint, Paris.
10. Guyon, E., Roux, S., Hansen, A., Bideau, D., Trodec, J.-P. and Crapo, H. (1990) Nonlocal and
nonlinear problems in the mechanics of disordered systems application to granular media and rigidity
problems, Rep. Progr. Phys. 53, 373-419.
11. Feng, S. and Sen, P. (1984) Percolation on elastic networks new exponent and threshold, Phys. Rev.
Lett. 52, 216-219.
12. Jacobs, D.J. and Thorpe, M.F. (1995) Generic rigidity percolation: the pebble game, Phys. Rev. Lett.
75, 4051-4054.
13. Hendrickson, B. (1992) Conditions for unique graph realizations, SIAM J. Comput. 21, 65-84 and
private communications.
14. Laman, G. (1970) On graphs and rigidity of plane skeletal structures, J. Engrg. Math. 4, 331-340.
15. Lovasz, L. and Yemini, Y. (1982) On generic rigidity in the plane, SIAM J. Alg. Disc. Meth. 3, 91-98.
16. Thorpe, M.F. (1983) Continuous deformations in random networks, J. Non-Cryst. Solids 57, 355-370.
17. Cai, Y. and Thorpe, M.F. (1989) Floppy modes in network glasses, Phys. Rev. B40, 10535-10542.
18. Feng, S., Thorpe, M.F. and Garboczi, E.J. (1985) Effective-medium theory of percolation on central-
force elastic networks, Phys. Rev. B31, 276-280.
19. Day, A.R., Tremblay, R.R. and Tremblay, A.-M.S. (1986) Rigid backbone: A new geometry for
percolation, Phys. Rev. Lett. 56, 2501-2504.
20. He, H. and Thorpe, M.F. (1985) Elastic properties of glasses, Phys. Rev. Lett. 54, 2107-2110.
21. Moukarzel, C. and Duxbury, P.M. (1995) Stressed backbone and elasticity of random central-force
systems, Phys. Rev. Lett. 75, 4055-4058.
22. Phillips, J.C. (1979) Topology of covalent non-crystalline solids. 1. Short-range order in chalcogenide
alloys, J. Non-Cryst. Solids 34, 153-181.
23. Phillips, J.C. (1981) Topology of covalent non-crystalline solids. 2. Medium-range order in
chalcogenide alloys and a-Si(Ge), J. Non-Cryst. Solids 43, 37-77.
24. Boolchand, P. and Thorpe, M.F. (1994) Glass-forming tendency, percolation of rigidity, and onefold-
coordinated atoms in covalent networks, Phys. Rev. B50, 10366-10368.
25. Boolchand, P., Zhang, M. and Goodman, B. (1996) Influence of one-fold-coordinated atoms on
mechanical properties of covalent networks, Phys. Rev. B53, 11488-11494.
26. Dhler, G.H., Dandoloff, R. and Bilz, H. (1981) A topological-dynamical model of amorphycity, J.
Non-Cryst. Solids 42, 87-96.
27. Jacobs, D.J. and Thorpe, M.F. (1996) Generic rigidity percolation in two dimensions, Phys. Rev. E53,
3682-3693.
28. Jacobs, D.J. (1998) Generic rigidity in three-dimensional bond-bending networks, J. Phys. A.: Math.
Gen. 31, 6653-6668.
29. Jacobs, D.J., Hendrickson, B. (1997) An algorithm for two-dimensional rigidity percolation: The
pebble game, J. Comput. Phys. 137, 346-365.
30. Jacobs, D.J., Kuhn, L.A. and Thorpe, M.F. (1999) Flexible and rigid regions in proteins, in M.F.
Thorpe and P.M. Duxbury (eds.), Rigidity Theory and Applications, Kluwer Academic / Plenum
Publishers, New York, pp. 357-384.
31. Fortuin, C.M. and Kasteleyn, P.W. (1972) On the random-cluster model. 1. Introduction and relation to
other models, Physica 57, 536-564; Kasteleyn, P.W. and Fortuin, C.M. (1969) Phase transitions in
lattice systems with random local properties, J. Phys. Soc. Japan 26 sup, 11-14.
32. Essam, J.W. (1980) Percolation theory, Rep. Prog. Phys. 43, 833-912.
63
33. Duxbury, P.M., Jacobs, D.J., Thorpe, M.F. and Moukarzel, C. (1999) Floppy modes and the free
energy: Rigidity and connectivity percolation on Bethe lattices, Phys. Rev. E59, 2084-2092.
34. Djordjevic, B.R., Thorpe, M.F. and Wooten, F. (1995) Computer model of tetrahedral amorphous
diamond, Phys. Rev. B52, 5685-5689.
35. Thorpe, M.F., Jacobs, D.J., Chubynsky, N.V. and Rader, A.J. (1999) Generic rigidity of network
glasses, in M.F. Thorpe and P.M. Duxbury (eds.), Rigidity Theory and Applications, Kluwer
Academic/Plenum Publishers, New York, pp. 239-278.
36. Press, W.H., Teukolsky, S.A., Vetterling, V.T. and Flannery, B.P. (1992) Numerical Recipes in
FORTRAN, 2nd ed., Cambridge University Press, New York.
37. Franzblau, D.S. and Tersoff, J. (1992) Elastic properties of a network model of glasses, Phys. Rev. Lett.
68, 2172-2175.
38. Moukarzel, C. (1998) A fast algorithm for backbones, Int. J. Mod. Phys. C9, 887-895.
39. Straley, J.P. (1979) Phase transitions in treelike percolation, Phys. Rev. B19, 4845-4846.
40. Manna, S.S. and Subramanian, B. (1996) Quasirandom spanning tree model for the early river
network, Phys. Rev. Lett. 76, 3460-3463.
41. Kirchhoff, G. (1847) ber die Auflsung der Gleichungen, auf welche man bei der untersuchung der
linearen verteilung galvanischer Strme gefhrt wird, Ann. Phys. Chem. 72, 497-508.
42. See, e.g., Cieplak, M., Giacometti, A., Maritan, A., Rinaldo, A., Rodriguez-Iturbe I. and Banavar, J.R.
(1998) Models of fractal river basins, J. Stat. Phys. 91, 1-15.
43. Prim, R.C. (1957) Shortest connection networks and some generalizations, Bell Syst. Tech. J. 36, 1389-
1401.
44. Cieplak, M., Maritan, A. and Banavar, J.R. (1994) Optimal paths and domain walls in the strong
disorder limit, Phys. Rev. Lett. 72, 2320-2323; Cieplak, M., Maritan, A. and Banavar, J.R. (1996)
Invasion percolation and Eden growth: Geometry and universality, Phys. Rev. Lett. 76, 3754-3757.
45. Frank, D.J., Lobb, C.J. (1988) Highly efficient algorithm for percolative transport studies in two
dimensions, Phys. Rev. B37, 302-307.
46. Stauffer, D. and Aharony, A. (1992) Introduction to Percolation Theory, 2nd ed., Taylor & Francis,
London.
47. Braswell, W.D., Family, F., Straley, J.P. (1984) Treelike percolation in two dimensions, Phys. Rev.
A29, 254-256.
48. Wu, F.Y. (1978) Absence of phase transitions in tree-like percolation in 2 dimensions, Phys. Rev. B18,
516-517.
49. Straley, J.P. (1990) Treelike percolation, Phys. Rev. A41, 1030-1033.
64
EVIDENCE FOR THE INTERMEDIATE PHASE
IN CHALCOGENIDE GLASSES
P. BOOLCHAND, W.J. BRESSER, D.G. GEORGIEV, Y. WANG AND J.
WELLS
Department of Electrical, Computer Engineering
and Computer Science
University of Cincinnati
Ci nci nnat i , OH 45221-0030
INTRODUCTION
Progressive cross-linking of polymeric networks, whether crystalline or non-
cryst al l i ne in nature, alters their physical behavior in remarkable ways. In crystalline solids
the effect is strikingly illustrated if one compares the physical properties of group VI
elemental solids (S, Se and Te that possess a coordination number, r, = 2) with those of
group IV ones (C, Si, Ge that have r = 4). Thus, although the Pauling single-bond strengths
[1] of a Se-Se bond (44kcal/mole) and of a Ge-Ge bond (37.6kcal/mole) are nearly the
same, the thermal (melting temperature and heat of fusion), elastic (Young's modulus) and
plastic (hardness) behavior of crystalline Ge (r = 4) completely overwhelms that of trigonal
Se (r = 2) understandably because of its higher connectivity.
More profound effects occur in non-crystalline solids. In the Ge
x
Se
1-x
binary glass
system, for example, one has the luxury to change the network connectivity or mean
coordination number in a continuous fashion by composition (x) tunning. And
one finds that glass transition temperatures [2], T
g
(x), and bulk elastic constants,
progressively increase with or the degree of cross-linking. Furthermore, random
networks in contrast to crystalline networks can self-organize [3] with remarkable
consequences on physical properties of glasses. In this review, we shall provide
experimental evidence for self-organization in glasses. The broad consequences of these
observations on percolative transitions in condensed matter, in evolutionary biology and in
protein folding, are discussed in the contributions of J.C. Phillips [4] and M.F. Thorpe [5]
in this volume.
Self-organization in random networks apparently occurs when the global network
stress nearly vanishes. The generic condition for a 3d-network to be stress-free is that the
number of Lagrangian bonding constraints/atom, n
c
, exhausts the three available degrees of
freedom per atom,
Phase Transi t i ons And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kl uwer Academi c/ Pl enum Publ i shers, 2001 65
Historically, the counting algorithm (1) was first introduced [6] to describe
optimization of the glass forming tendency in a covalently bonded random networks, and
onset of rigidity [7] in covalently bonded random networks. These two conditions
apparently coincide with the stress-free requirement of such networks. Glasses that are
under- (n
c
< 3) and over- (n
c
> 3) constrained possess stress which appears to be,
respectively, entropic and enthalpic in origin. Mean-field predictions of the rigidity
transition and existing numerical simulations at T=0, treat enthalpic rigidity [8]. As ideas
on entropic rigidity [9] evolve, one may be able to understand stress in under-constrained
networks.
The count of Lagrangian bonding constraints (n
c
) in covalent systems is rather
straightforward. In such systems the distinction between the intact nearest-neighbor (bond-
stretching and bond-bending constraints and the broken more distant (dihedral
angle and van der Walls) ones is transparent because of the hierarchical nature of the
bonding interactions. On the other extreme, in metallic systems, the impediment to apply
these ideas on constraint counting results from the long range nature of bonding
interactions between mobile electrons and the positive ion cores that contribute to the
cohesive energy of such solids. The case of ionic solids such as the oxides is a particularly
fascinating one and is a subject of current discussions. The hierarchical separation between
intact and broken Lagrangian bonding constraints for pure Silica has been facilitated by
structure results. Although the bonding angle about cation atoms (Si) is close to the
tetrahedral angle, the bonding angle about the anion atoms (O) displays wide excursions
(100-180). These structure results suggest that while both the bond-stretching and
bond-bending constraints on Si appear to be intact, it is only the r-constraints on O that
is intact. Indeed that separation permits one to understand [10] why silica, the archetypal
glass former satisfies equation (1) exactly, and thus forms a stress-free continuous random
network. The counting algorithm(s) also show that window glass, a pseudo-ternary alloy
of and in the approximate molar ratio 75:15:10 could well represent [11]
yet another example of a stress-free oxide network.
In this article, we review recent optical and thermal experiments that shed light on the
existence of the self-organized state in network glasses. The self-organized state apparently
occurs in between the rigid phase and the floppy phase and is therefore also termed as the
intermediate phase. Our experiments with chalcogenide glasses [12] reveals that glass
transitions (T
g
) become almost completely thermally reversing in character in the self-
organized state. The thermally reversing character of Tgs can be probed by a recently
developed variant of DSC, known as T-modulated Differential Scanning Calorimetry
(MDSC). The non-reversing heat flow term near T
g
accessed in an MDSC measurement
provides a direct measure of network stress [13] in a glass, and represents a significant
recent development in the field. Raman Vibrational spectroscopy in the bond-stretching
regime provides a uniquely powerful probe of optical elasticity which displays specific
power-laws as a function of in the floppy, intermediate and stressed rigid phases of
network glasses. Before describing the results of these experiments, we provide an
overview of principle theoretical ideas that have evolved in the field of enthalpic rigidity in
the next section.
PREDICTION OF A RIGIDITY TRANSITION
Mean-field Theory in Random Networks
66
(1)
In 1788, J.L. Lagrange [14] wrote the celebrated book "Mcanique Analytique", and
thus laid the foundations of Lagrangian mechanics as we know it today. In particular he
introduced the notion of generalized coordinates and forces that can act as constraints
(Lagrangian multipliers) in solving mechanical problems. About a century later, J.C.
Maxwell [15] used the notion of constraints to examine mechanical stability of
macroscopic structures such as trusses and bridges.
Phillips-Thorpe Rigidity Threshold. In 1979, a new beginning to develop a
microscopic theory of glass formation in covalent solids was made [6]. It was suggested
that valence forces such as bond-stretching and bond-bending forces can serve as
atomic constraints in assembling covalent networks. It was conjectured that glass formation
would be optimized when the number of constraints/atom equal 3, the degrees of freedom
of an atom in a 3d network (Table 1). These ideas were advanced further in 1983 when a
normal mode analysis of covalent networks showed [7] that the number of zero frequency
solutions of the dynamical matrix, f (= floppy modes), equal n
c
3. These developments
led to the recognition that a glass network will become elastically rigid when the number of
zero frequency solutions f vanish, i.e., n
c
= 3. For 3d networks in which atoms possess a
coordination number the number of and constraints/atom, are given by
The onset of rigidity in such networks it then follows, would occur in a mean-field sense,
when
or
or the mean coordination,
The Phillips-Thorpe mean-field rigidity threshold, has been refined by
numerical simulations [8] using the standard model of a covalent glass as a random
network. These calculations are discussed in the contribution of M.F. Thorpe et al., in this
volume. For completeness we summarize the principal results of these calculations here.
The numerical simulations [8] start with a base network at r = 4, such as a diamond lattice
or a 4096-atom computer generated amorphous Si network with periodic boundary
conditions. Bonds are depleted at random to produce networks of lower in the
range insuring that one-fold coordinated atoms are not produced. A Kirkwood or Keating
potential is used to describe the and forces between atoms. The number of floppy
modes/atom are then calculated as a function of by performing a normal mode analysis at
each The results show (Fig. 1) a linear decrease in in the range (as
shown by mean-field theory) which is followed by a long exponential tail in the
approximate range 2.35 The onset of rigidity transition is then localized by
the second derivative, that reveals (Table 1) a sharp peak at for the
case of diamond lattice and at for the case of a-Si base network. These
results are remarkably close (within 1%) to the mean-field result of for the onset
of rigidity in random networks estimated by constraint counting.
One-fold Coordinated Atoms (OFC, r=1). The Phillips-Thorpe mean-field estimate
for the onset of rigidity is restricted to covalent networks in which and
constraints are intact, and all atoms possess a r of OFC atoms (r = 1) have to be
handled separately because such atoms do not produce constraints but can only erase
them according to the 2r 3 estimate. Many glass systems include OFC-atoms, such as
67
(2)
(3)
(4)
68
Figure 1. Results of numerical simulations for the fractions of floppy modes per degree of freedom for the
bond-depleted diamond network in the random case and the self-organized case. The filled circle gives the
stress and rigidity transition for the random case. The open square gives the mean-field predicted rigidity
transition at The open circle and triangle give, respectively, the rigidity and stress transitions for
the self-organized case. The shaded region gives the intermediate phase. Figure is taken from ref.3a.
halogens in the chalcohalides, hydrogen in hydrogenated a-Si, and alkalis in alkali-silicates.
In 1984 Boolchand and Thorpe [16] extended mean-field constraint counting algorithms to
include OFC-atoms. They showed that equation (3) takes on a correction term in the
presence of a finite fraction, n
1
/N, of such atoms, and rigidity percolates in such networks
when
In general, OFC-atoms play no role if the base glass network (without 1-fold
coordinated atoms) is optimally constrained, on the other hand, these atoms will soften [10]
an overconstrained base network, and conversely stiffen an underconstrained base glass
network. The latter is illustrated in Fig. 2 which shows a plot of the negative of f,
henceforth labeled h (= hardness index) as a function of the mean coordination of the
base glass network. The base glass network is the network remaining once all 1-fold
coordinated atoms have been plucked. For the case when a glass network has no 1-fold
coordinated atoms, i.e., n
1
/N = 0, we recover the usual circumstance, h = -1 at r = r = 2 and
h = 4 at r = 4, and rigidity onsets at h = f = 0 when In the presence of a finite
concentration of 1-fold coordinated atoms, the slope of the line
progressively decreases from its maximum value of 5/6 with increasing n
1
/N, with all
plots passing through the point h = 0, Such a behavior constitutes evidence of a
softening of an overconstrained network or conversely stiffening of an
underconstrained network in the presence of OFC-atoms. These ideas provide a
means to understand [17] the linear variation of the nano-indentation hardness results of
diamond-like films as a function of as discussed elsewhere. Furthermore, these ideas
have also provided a basis to understand the glass forming tendency in a wide variety of
chalcohalide glasses as recently discussed in a review by Mitkova and Boolchand [18].
In dealing with networks possessing OFC-atoms, two circumstances need to be
distinguished [19]. Case 1, when OFC-atoms bond to an atom possessing and case 2
when OFC-atoms bond to an atom possessing an r = 2. Equation 5 is valid for case 1
69
(5)
Figure 2. Graphical representation of hardness index h vs. mean coordination of base network showing
the straight-line loci for different values of OFC-atom concentration n1/n=0.025 and 0.50. Figure is taken
from ref. 17.
above. The more complicated case 2 has also been treated. In case 2, the rigidity threshold
for the base glass network does not coincide with that of the (complete) glass network. For
these reasons, in general, it is safer to always calculate mean-field constraints of the
complete glass network with the OFC-atoms, rather than truncate it and work with the base
glass network.
Bond-Bending Forces and Broken Constraints. Valence forces form a hierarchy
with force strength exceeding force strength by a factor of 3 or more. And it is
possible that as glass transition temperatures T
g
increase, the constraint associated with the
weaker forces intrinsically break. The structure manifestation of the broken -constraint
is that the underlying bond-angle distribution becomes wide. Presumably this results
because thermal energies (kT
g
) can overwhelm the underlying strain energies that would
require a unimodal bond angle distribution. The circumstance is encountered in the
chalcogenides [17] and oxide glasses [18]. It is useful to examine the underlying mean-
field estimate of the rigidity transition in the presence of broken -constraint.
Consider a glass network possessing finite concentration (m
r
/N) of r-fold
coordinated atoms that have their -constraint broken. The rigidity transition in such a
network will, in a mean-field sense, occur at
where m
2
/N, m
3
/N, m
4
N, . . . represent the fraction of 2, 3, 4,...fold coordinated atoms in
the network that have their -constraints broken. A special case of equation (6) is that of a
glass network possessing a finite concentration of 2-fold coordinated atoms that
70
have their -constraints broken. In binary Ge
x
Se
1-x
glasses, although the -constraint
associated with 2-fold coordinated Se-atoms is intact in a Se-glass at x = 0, upon cross-
linking with Ge, T
g
s rapidly increase, and the underlying constraint is apparently broken
as x increases to 0.20. For this reason, the rigidity transition [20] invariably shifts to higher
values of due to the presence of the first correction term of equation (5), i.e.,
where m
2
/N designates the fraction of Se atoms in Se
n
-chain fragments in Ge
x
Se
1-x
glasses
near x = 0.23. A second illustrative example of broken constraints determining the
stiffness transition occurs in the Ge
1-x
Sn
x
Se
2
ternary, where Mssbauer spectroscopy
reveals [21] evidence of Sn atoms to become all tetrahedrally coordinated when x increases
to 0.35. The overconstrained tetrahedra, when admixed with the
underconstrained tetrahedra, drive the network optimally constrained. Equations
(7) predicts [22,23] the transition to occur at x = 2/5, indeed quite close to the observed
value of 0.35.
Two Rigidity Transitions in Self-Organized Networks
In 1999, numerical simulations were extended to self-organized networks for the first
time, and one found [3] rigidity to onset in two steps at r
c
(1), r
c
(2). The first of these
transitions (r
c
(1)) is quite similar to the one described earlier for random networks in which
rigid regions percolate at At r > r
c
(1), if the additional crosslinks are selectively
placed in the floppy domains of the network, network stress will not accumulate. This
situation must be contrasted to the usual case when crosslinks placed randomly drive some
of the isostatically rigid-regions stressed-rigid (redundant bonds) thus building up network
stress. One has found that such selective placement of crosslinks saturates at r = r
c
(2) =
2.395, when it is no longer possible to avoid forming redundant bonds. At (2) the
network makes a first order transition to a stressed rigid phase (Fig. 1). The term isostatic
or stress-free can be used to describe the self-organized state of a network in the r
c
(1) < r <
r
c
(2) phase, residing in between the floppy (r < r
c
(1)) and stressed rigid (r > r
c
(2)) phase.
The unstressed rigid phase is also termed as the intermediate phase. Calculations [24] show
that the size of the first order jump in elasticity at r
c
(2) is controlled by the concentration of
small rings (6-membered rings or smaller) where rigidity apparently nucleates.
Raman scattering experiments on chalcogenide glasses provide evidence of two elastic
thresholds (r
c
(1), r
c
(2)), as we shall describe in the next section. And even though there
persists a systematic difference between theory and experiments on the values of r
c
(1) and
r
c
(2), these new numerical simulations represent a significant advance. They not only
narrow the gap between theory and experiments, but also provide a physical basis to
characterize the order of the transitions and the self-organized state prevailing in between
the two transitions.
GLASS TRANSITION VARIATION AND NETWORK CONNECTIVITY
Glasses are distinguished from amorphous materials in that they display a glass
transition temperature, T
g
, that represents a softening or melting transition of the solid
glass into a metastable liquid glass. Across T
g
, viscosity, thermal expansion, entropy,
molar volumes and specific heat change [25] qualitatively. The basic nature of glass
transition continues to be a subject of profound current discussions. There are undoubtedly
71
kinetic and sample thermal history effects associated with T
g
. But, there are also much
larger structure or network connectivity related effects that apparently control the
magnitude of T
g
s in glass forming materials.
T-Modulated Differential Scanning Calorimetry (MDSC)
In a conventional DSC measurement, the signature of softening of a glass is an
endothermic heat flow usually measured with respect to an inert reference, as the
temperature of the glass and reference is swept at a fixed scan rate. By programming a
sinusoidal temperature variation over the linear T-ramp, it is possible to deconvolute [26]
the heat flow into two components, one that tracks the sinusoidal T-variation and is
therefore called the reversing heat flow and the remainder that does not track the periodic
T-variation and is called the non-reversing heat flow Figure 3 provides MDSC
scans [23] of a GeSe
2
glass and Ge
0.23
Se
0.77
glass illustrating the deconvolution of heat
flow rates into the non-reversing and reversing components. Two noteworthy features
become apparent from these scans, first the apparent glass transition temperature
deduced from the total heat flow is, in general, lower than the glass transition temperature
(T
g
) deduced from the reversing heat flow. Second, is found to be miniscule for the
x = 0.23 sample but it is an order of magnitude larger for the stoichiometric glass, GeSe
2
.
The shift, is only 3C for the x = 0.23 glass sample when is nearly
vanishing, but it is 12C for GeSe
2
glass when increases by almost an order of
magnitude. The presence of a sizeable term in a glass will lower in relation to
T
g
due to kinetic heat-flow effects.
Figure 3. MDSC scans (a) Ge
0.23
Se
0.77
and (b) GeSe
2
glasses revealing the deconvolution of the heat flow
into the reversing and non-reversing components. Note the qualitative reduction in for the optimally
coordinated glass. See text for details. Figure is taken from ref. 23.
72
Network Stress and Non-Reversing Heat Flow. Important insights into the
physical origin of the term in glasses have emerged from compositional trends. The
case of the Ge
x
As
x
Se
1-2x
ternary is particularly significant because for this system
compositional trends in activation energies of Kohlrausch relaxation of an external
stress were established in flexural studies [27]. Recently, compositional trends in
for this ternary were measured [13], and the comparison of the two sets of results taken in
two different laboratories, at different time frames, shows a remarkable similarity (Fig. 4).
The results suggest a specific interpretation of the term; the heat-flow term provides a
measure of internal stress in a network glass, while the activation energies, a
measure of relaxation of an external stress.
T
g
and Network Connectivity. The T
g
deduced from the reversing flow provide a
very useful datum characterizing the connectivity of a glass. At the outset, it should be
mentioned that T
g
s deduced from the reversing heat flow are virtually independent
of sample thermal history, although these will shift up in T as scan rates are increased.
Therefore if T
g
s are measured at the same low scan rate at different compositions in a glass
system, one can obtain useful structure information. The latter represents a relatively
recent development in the field.
Figure 4. variation in Ge
x
As
x
Se
1-2x
glasses showing the thermally reversing window in the
< 2.42 range (a). Activation energy for Kohlrausch relaxation of an external stress in flexural studies (b).
Figure in top panel taken from ref. [13], while the figure in bottom panel is taken from ref. [27].
73
Agglomeration theory [28, 29] developed by R. Kerner and M. Micoulaut provides a
quantitative means to analyze compositional trends in T
g
in the stochastic regime. One
can, for example, predict the T
g
(x) variation in binary Ge
x
Se
1-x
and As
x
Se
1-x
glasses near x
when cross-linking of Se
n
-chain fragments by 4-fold Ge and 3-fold As atoms
proceeds in a stochastic fashion, i.e. randomly. These predictions are well supported by
experiments. The predictions also provide a quantitative means to distinguish domains of
stochastic from non-stochastic network formation in these binary glasses [2].
In the Ge
x
Se
1-x
binary glass system, the prediction of agglomeration theory is that the
slope
where T
0
is the glass transition of Se glass. In the experiments one observes (Fig. 5a) T
g
(x)
to increase linearly with x, with a slope of T
0
/ln2 = 44C/10 at % of Ge up to x = 0.08
defi ni ng the stochastic regime. At x > 0.08, the experimental T
g
s exceed the predicted T
g
s,
suggesting onset of extended range structures beyond Ge(Se
1/2
)
4
tetrahedra and Se
n
-chain
segments, and signaling the onset of a non-stochastic regime (Fig. 5a). It is in this
compositional range that
129
I Mssbauer spectroscopy measurements reveal [30, 31] that
the oversized Te-tracer atom ceases to replace available Se-sites in a random fashion. Thus
the onset of non-stochastic behavior at x > 0.08 in these thermal measurements is in
excellent agreement with
129
I Mssbauer spectroscopy results published nearly two
decades ago.
RAMAN ELASTIC THRESHOLDS AND THE INTERMEDIATE PHASE
Although Raman scattering has been used as a probe of glass structure [32-35] for
the past three decades, the application of the optical method as a probe of rigidity
transitions in network glasses is a recent development [20, 36-39]. In many cases when
vibrational bands are resolved, it is possible to quantitatively follow mode frequency
changes with glass composition and deduce power-laws describing optical elasticity
changes. And although the scales of mode frequencies are set by the strength of and
forces, variations in mode-frequencies with glass compositions result due to inter-
tetrahedral couplings, i.e. network connectivity.
Because of the presence of light-induced effects in glasses, the need to excite Raman
scattering at low-laser power is paramount. For this reason, resonant enhancement of the
scattering by tunning the laser energy close to the optical bandgap of the glass, is
particularly desirable. It results in a large signal to noise ratio with very modest laser
exciting power. Raman scattering results on several IV-VI glass systems have now been
performed and provide evidence of the intermediate phase. The principal results on the
Ge
x
Se
1-x
binary glass system are presented next.
Bulk Ge
x
Se
1-x
Glasses. Glasses in the titled binary at several compositions in the 0
< x < 1/3 range were synthesized the usual way and characterized by MDSC [2, 20, 38].
The observed T
g
s measured from the reversing heat-flow serve an important check on glass
compositions. Raman scattering excited by 647.1nm radiation from a Kr
+
ion laser was
studied in a conventional back scattering set up using a T64000 triple monochromator
system with a CCD detector. The laser beam was brought to a loose focus of about 500m
spot size, to keep the photon flux low (10
17
photons/cm
2
/sec), in a set up [38] usually
described as macro-Raman scattering.
74
Macro-Raman Scattering
Figure 6 shows Raman spectra of the glasses obtained at selective compositions.
The lineshapes show evolution of modes of corner-sharing (CS) tetrahedra (200 cm
- 1
) and
edge-sharing (ES) Ge(Se
1/2
)
4
tetrahedra (215 cm
-1
) at the expense of the Se-Se stretch
mode of Se
n
-chains (CM) at about 250cm
-1
. In general CS and ES modes are symmetric
and could be analyzed in terms of a Gaussian each. This was not the case of the CM,
however, and it required at least 3 Gaussians for an adequate deconvolution at all x. Figure
7 provides the mode frequency variation for the CS and ES mode from a comprehensive
lineshape deconvolution of the Raman spectra [38].
In Fig. 7a, one can discern a regime of a linear variation in v(x) in the 0.08 < x <
0.20 interval, and a regime of a power-law variation in the 0.26 < x < 1/3 interval. In
between these two regimes, lies a transition region wherein the variation
in v( x) is sub-linear. To extract the underlying power-laws, we have made a plot of
against and obtained the slope p
CS
of the resulting
Fig. 5(a). Observed Tg(x) variation in indicated binary glasses. The Tgs are deduced from the inflection
point of the reversing heat flow. (b) variation in Ge
x
Se
1-x
glasses deduced from the non-reversing
heat flow showing a global minimum near x = 0.225. Figure taken from ref. 2.
75
Figure 6. Macro-Raman scattering in bulk Ge
x
Se
1-x
glasses at indicated glass compositions x. Note the
growth in scattering of corner-sharing (CS) and edge-sharing (ES) tetrahedra at the expense of Se
n
-chain
mode (CM). See text for details.
line (Fig. 8a) and get a value of p
CS
= 1.54(6). A parallel analysis for the transition region
(Fig. 8b) using x
c
(1) = 0.21, yields a slope p
t
= 0.75(15). For the ES mode, our
results suggest that the linear variation in at low x (0.10 < x < 0.20) smoothly connects
to the power-law variation at higher x (x > 0.21). Figure 8c shows a plot of
against which yields a straight line with a slope p
ES
= 1.32(06),
defining the underlying power-law associated with elasticity due to ES units. The rational
for extracting the Raman elastic power-laws is discussed next.
Stressed Rigid Phase. On general grounds one expects rigidity in the present
glasses to nucleate at moities that are overconstrained such as the ES and CS
units. Changes in and result from local or optical elasticities and can be expected
to display a power-law variation with once exceeds a critical value when the
backbone becomes rigid, i.e.,
The justification for such a variation is suggested by the numerical simulations [40, 41] on
random networks constrained by and forces (Keating potential) that show the
elasticity to display a power-law power p = 1.4 or 1.5.
The power-law variation of (Fig. 7a), starting at x = x
c
(2) = 0.26 with p
CS
=
1.54( 10) constitutes direct evidence for onset of stressed rigidity. The power-law p
CS
= 1.54
is in excellent agreement with numerical estimates [40, 41] of elasticity in random
networks. Furthermore, the stressed nature of the backbone at x > x
c
(2) follows from the
rapidly increasing heat-flow term once x > x
c
(2) as illustrated in Fig. 5a. The small
76
Figure 7(a). CS- and (b) ES-Raman mode frequency variation with x in Ge
x
Se
1-x
glasses displaying the
various phases as discussed in text. Figure is taken from ref. 38.
jump in v
c
between x = 0.25 and 0.26 of suggests that the rigidity transition near
x
c
(2) is first order.
Intermediate Phase. Our Raman experiments also show that rigidity first onsets at
x = x
c
(l) = 0.20(1) in which both ES and CS units partake. This is suggested by the kink in
and the onset of a power-law behavior in both starting near x
c
(l). The optical
elasticity associated with the CS and ES units yields a lower power-law of p
CS
= 0.75(15)
and P
ES
= 1.38(10). The sublinear power-law for the CS units, P
CS
(=0.75(15)) is
reminiscent of the finite-size scaling result [42] which is 2/d = 2/3 for d = 3 (3d-network).
To obtain a power-law p < 1, one must invoke large-scale or long-range fluctuations as are
discussed in equilibrium-scaling theory [43]. The composition x
c
(l) coincides with the
lower bound of the global minimum in suggesting that in the x
c
(l) < x < x
c
(2)
interval glasses are largely unstressed. Furthermore, the transition at x
c
(l) appears to be
second order as revealed by the continuous variation of mode frequencies and
These signatures of the experimental results constitute evidence for glasses in the x
c
( 1) < x
< x
c
(2) composition range to be unstressed rigid, and to be self-organized, constituting the
intermediate phase.
77
Figure 8. Log-log plots of Raman mode frequency squared against glass compositions yielding power-laws
for (a) CS mode in the stressed rigid phase (b) CS mode in the intermediate phase (c) ES mode in the rigid
phase. See ref. 38
78
Floppy Phase. The present experiments [38] also suggest that at x < x
c
(1), glasses
are floppy. Both and display a linear variation with x in the 0.10 < x < 0.20
range with a small slope. However at x < 0.10 in the stochastic regime, the slope of
with x actually vanishes. Numerical simulations [40 ,41] using the bond-depleted diamond
structure reveal the elasticity in the floppy region to vanish. Thus, theory and experiment
agree rather well in the stochastic regime. At x > 0.10, both T
g
(x) variation and Mssbauer
site occupancy variation [30] with x, signal the emergence of a non-stochastic network
consisting of extended rigid inclusions formed in a floppy background. Here the
experiments show the elasticity to slowly increase as It would be of interest to
see whether such an observed variation can be reproduced in numerical simulations on
non-stochastic floppy networks.
In summary, the present Raman and MDSC results on Ge
x
Se
1-x
glasses provide
evidence of two rigidity transitions, a second order transition at x
c
(1) = 0.20 from a floppy
to an unstressed rigid phase, and a first order transition at x
c
(2) = 0.26 from an unstressed
rigid to a stressed rigid phase. These two rigidity transitions represent the lower (x
c
( 1)) and
upper (x
c
(2)) bounds of an intermediate phase that separates the floppy (x < x
c
(1)) from the
stressed rigid (x > x
c
(2)) phase in the present glasses.
Complete Raman and MDSC measurements are now available on the companion
IV-VI binary glass system, Si
x
Se
1-x
[36,37]. The results reveal strikingly parallel details of
the intermediate phase in this binary glass system with x
c
(l) = 0.20 and x
c
(2) = 0.27.
Taken together, the close similarity of results on the width and centroid of the intermediate
phase in the two independent IV-VI binary glass systems provides for some self-
consistency in the underlying physics of the rigidity transition.
PHOTOMELTING OF THE INTERMEDIATE PHASE
Historically, the rigidity transition in Ge
x
Se
1-x
binary glasses was first investigated
in micro-Raman measurements [20]. In these measurements, the exciting laser beam
(647. 1nm) was brought to a sharp focus using a microscope attachment. This had
the consequence that the photon flux (N
f
) of 10
22
photons/cm
2
/sec, exceeded that in the
macro-Raman measurements (N
f
= 10
17
photons/cm
2
/sec) by about 5 orders of magnitude.
Changes in N
f
reflect the 2 orders of magnitude smaller versus laser spot size
in the micro-Raman measurements. In both experiments, the exciting laserbeam
wavelength of 647.1nm or E
ph
= 1.92eV was kept the same. For glass compositions in
the intermediate phase, possessing an optical gap E
g
> 2.0eV, the exciting beam is thus
weakly absorbing and penetrates to a depth of about 100m. In sharp contrast to the
macro-Raman results, the micro-Raman measurements revealed [20] a solitary rigidity
transition from a floppy phase to a rigid phase near x = x
c
= 0.23(1).
To facilitate a comparison, we have reproduced variation in Ge
x
Se
1-x
glasses
from the macro-Raman (Fig. 9a) with the micro-Raman (Fig. 9b) measurements together.
One is led to the suggestion that while the macro-Raman measurements probe the intrinsic
rigidity transitions of the glasses, the micro-Raman ones probe a light-induced modification
of the transitions. Certain features become transparent from the comparison of these
results. The two transitions at x
c
(1) = 0.20 and x
c
(2) = 0.26 coalesce to a point at the
centroid x
c
= (x
c
(1) + x
c
(2))/2 = 0.23 location in the micro-Raman measurements.
Furthermore, the centroid location x
c
= 0.23 coincides with the global minimum in
heat flow term (Fig. 1b).
79
Figure 9. Raman mode frequency variation of CS sharing Ge(Se
1/2
)
4
tetrahedra taken observed in (a) macro-
Raman and (b) micro-Raman measurements compared. The intermediate phase collapses to a solitary point
at its centroid in the micro-Raman measurements due to the higher photon-flux of the exciting 647.1nm
radiation. Figure is taken from ref. 38.
Our interpretation of these Raman results is that the collapse of the intermediate
phase to a solitary point is that it represents a photo-structural effect. In the presence of a
high flux of near band-gap radiation photomelting of the self-organized
state to a random network takes place. The underlying physical process probably consists
of a transient self-trapped exciton [44, 45] that recombines non-radiatively leading to a
light-induced dynamic state in which rapid bond-switching and diffusion occurs. The
photodiffusion process, apparently is optimized when the network stress vanishes and
photomelting is the consequence. It is for this reason that the photomelted state is easily
formed at the global mi ni mum of the term (x
c
= 0.23), when network stress vanishes.
Parallel physical effects have recently been observed in Brillouin scattering
measurements [45] on these glasses. In these experiments, one found the longitudinal
acoustic (LA) mode to soften reversibly and by a gigantic amount (25%) as a function of
laser excitation power, but only for glass compositions near the center of the intermediate
phase, i.e. x=0.225.
These Raman measurements on Ge
x
Se
1-x
glasses provide an essential link between
the two extremal descriptions of the rigidity transitions in glasses. One observes two
80
transitions [38] when the backbone is intrinsically self-organized in the melt-quenched
glasses and one observes a solitary rigidity transition [20] characteristic of a random
network in the photomelted state. And one expects the solitary transition to occur in
connectivity space when corresponding to the counting algorithm(s).
INTERMEDIATE PHASE AND GLASS STRUCTURE
The discovery of the intermediate phase in several chalcogenide glasses [12] is
beginning to shed light on its connection to glass structure. Apart from to the IV-VI
glasses, there are several other glass systems [13, 39, 46, 47] in which the phase has now
been observed (Fig. 10). In table 2 we provide a summary of results on the subject.
Perusal of the results in table 2 shows that the width of the intermediate phase changes
by almost an order of magnitude in going from the Ge-S-I ternary [39] to
the Ge-As-Se ternary with the results on binary glasses of Ge-Se, and As-Se
lying in between 0.08(1)) these two extremes (Fig. 10). The Ge-S-I ternary
appears [39] to be a paradigm of a random network. In this ternary, replacement of
S in the base Ge
0.25
S
0.75
glass by 1-fold coordinated I, proceeds to randomly depolymerize
the network. Enumeration of constraints reveals that only one building block (m=l) out
of the five possible (m=0, 1, 2, 3, 4) mixed GeSe
4-m
I
m
tetrahedra is optimally constrained,
and thus contributes to the formation of the intermediate phase. In Raman scattering, one
can observe modes of these mixed tetrahedra and thus establish details of glass
structure. Furthermore, the Raman mode frequency of Ge(S
1/2
)
4
tetrahedra, v
0
(x), is found
to vary linearly with Iodine concentration (x) to show a kink (change in slope) at the
rigidity transition, which coincides with the sharp minimum in The transition,
according to mean-field theory, is predicted to occur at Remarkably, the present
experiments (Fig. 10 show the transition to occur at
Table 2 Experimental Results for the Intermediate Phase in Network Glasses, r
c
( 1) designates onset of
81
Figure 10. variation as a function of glass composition revealing thermally reversing windows in
GexSe1-2x(O), GexAsxSe1-2X ( ) and bulk glasses. Note the large variation in the window
width from 0.15 for the Ge-As-Se ternary to < 0.01 for the Ge-S-I ternary.
In sharp contrast, the Ge
x
As
x
Se
1-2x
ternary, often described as a prototype of a random
network, turns out to be one of the best example of a glass system in which a high degree of
self-organization occurs [13] as reflected by the extremely large width of the
intermediate phase. We believe the result is intrinsically related to the multiplicity of
optimally constrained units and combination of over- and under-constrained units that can
form as part of the stress-free backbone. Some of these units include pyramidal As(Se
1/2
)
3
,
quasi-tetrahedral Se=As(Se
1/2
)
3
, in addition to a combination of Ge(Se
1/2
)
4
tetrahedra (both
CS and ES) with Se
n
-chain fragments. And it is possible that molecular dynamic
simulations, in conjunction with the Raman elasticity power-laws can impose stringent
restrictions on possible structures prevailing in this special region, for these to be identified
in future.
CONCLUSIONS
Raman scattering and T-modulated Differential Scanning Calorimetry
measurements on several families of chalcogenide glasses have been performed.
Comprehensive results are now available on the Ge
x
Se
1-x
and Si
x
Se
1-x
binary glass
systems, where rigidity is found to onset in two steps; a second-order transition at x
c
(l) =
0.20 in both binaries from a floppy to an unstressed rigid phase, and a first-order transition
at x
c
(2) = 0.26 for Ge-Se, = 0.27 for Si-Se binary from an unstressed rigid to a stressed
rigid phase. The two transitions (x
c
(1), x
c
(2)) define the bounds of an intermediate phase
that separates the floppy from the stressed rigid phase. The near absence of the non-
reversing heat-flow, constitutes evidence for the stress-free nature of the backbone of
glass compositions in the intermediate phase. Light-induced melting of the intermediate
phase in micro-Raman measurements on the Ge-Se glass system has also been observed.
82

The observation constitutes evidence for photomelting of the self-organized state to a

random network. These results confirm that the breakdown of mean-field theory, results
due to an intrinsic self-organization of the backbone, and provide the link between the two
extremal descriptions of the rigidity transition in network glasses modeled in numerical
simulations.
ACKNOWLEDGEMENTS
It is a pleasure to acknowledge discussions with Mike Thorpe and Jim Phillips
during the course of this work. This work is supported by NSF grant DMR-97-02189.
REFERENCES
1. Pauling, L. (1960) Nature of the Chemical Bond, Cornell University Press, Ithaca, NY pp 85.
2. Boolchand, P. and Bresser, W. J. (2000) Structural Origin of Broken Chemical Order in GeSe Glass.
Phil. Mag B 80, 1757 1772.
3. a. Thorpe, M.F., and Chubynsky, M.V. (2000), Rigidity and Self-Organization of Network Glasses and
the Intermediate Phase in M. F. Thorpe (ed.) Properties and Applications of Amorphous Materials,
Kluwer Academic Publishers, Dordrecht, (in press).
b. Thorpe, M.F., Jacobs, D.J., Chubynsky, M.V., Phillips, J.C. (2000) Self Organization in Network
Glasses, J. Non-Cryst Solids 266-269, 859-866.
4. Phillips, J.C. Mathematical Principles of Intermediate Phases in Disordered Systems, present volume.
5. Thorpe, M.F., and Chubynsky, M.V. Rigidity and Self-Organization of Network Glasses and the
Intermediate Phase, present volume.
6. Phillips, J.C. (1979) Topology of covalent non-crystalline solids I: Short-range order in chalcogenide
alloys, J. Non-Cryst. Solids 34, 153-181.
7. Thorpe, M.F. (1983) Continuous deformation in random networks, J. Non-Cryst. Solids 57, 355-370.
8. Thorpe, M.F., Jacobs, D.J., Djordjevic, B.R. (2000) The structure and rigidity of network glasses, in P.
Boolchand (ed), Insulating and Semiconducting Glasses, World Scientific Press, Singapore, pp. 95-
145.
9. Jos, B., Plischke, M., Vernon, D.C., and Zhou, Z. (1999) Entropic Rigidity in M.F. Thorpe and P.M.
Duxbury (eds.), Rigidity Theory and Applications, Kluwer Academic/Plenum Publishers, New York
pp 315-328.
10. Zhang, M. and Boolchand, P. (1994) The central role of broken, bond-bending constraints in promoting
glass-formation in the oxides, Science 266, 1355-1357.
11. Kerner, R., and Phillips, J.C. (2000) Quantitative Principles of Silicate Glass Chemistry Solid State
Comm. (in press).
12. Boolchand, P., Selvanathan, D. Wang, Y., Georgiev, D.G., and Bresser, W.J. (2000) Onset of Rigidity
in steps in Chalcogenide Glasses The intermediate Phase in M. F. Thorpe (ed). Properties and
Applications of Amorphous Materials, Kluwer Academic Publishers, Dortrecht, (in press).
13. Wang, Y., Boolchand, P. and Micoulaut, M. Glass structure rigidity transitions and the intermediate
phase in the Ge-As-Se ternary, Europhys. Lett, (in press).
14. Lagrange, J. L. (1788) Mecanique Analytique, Paris.
15. Maxwell, J. C. (1864) On the calculation of the equilibrium and stiffness of frames, Philos, Mag. 27,
294-299.
16. Boolchand, P., and Thorpe, M.F. (1994) Glass-forming tendency, percolation of rigidity, and one-fold-
coordinated atoms in covalent networks, Phys. Rev. B 50, 10366-10368.
17. Boolchand, P., Zhang, M. and Goodman, B. (1996) Influence of one-fold-coordinated atoms on
mechanical properties of covalent networks, Phys. Rev. B 53, 11488-11494.
18. Mitkova, M. and Boolchand, P. (1998) Microscopic origin of the glass forming tendency in
chalcohalides and constraint theory, J. Non-Cryst. Solids 240, 1-21.
19. Boolchand, P., Zhang, M., Goodman, B. (1997) One-fold coordinated atoms, constraint theory and
nanoindentation hardness, in M.F. Thorpe and M. Mitkova (eds.) Amorphous Insulators and
Semiconductors, Kluwer Academic Publishers, Dortrecht, pp. 339-348.
20. Feng, X.W., Bresser, W.J. and Boolchand, P. (1997) Direct evidence for stiffness threshold in
chalcogenide glasses, Phys. Rev. Lett. 78, 4422-4425.
21. Stevens, M., Grothaus, J., Boolchand, P. and Hernandez, J.G. (1983) Universal structural phase-
83
transition i n network glasses, Solid State Comm. 47, 199-202.
22. Phillips, J.C. (1983) Realization of a Zachariasen glass, Solid Stale Comm. 47, 203-206.
23. Boolchand, P. (2000) Vibrational excitation i n glasses: Ri gi di t y t ransi t i on and Lamb-Mssbauer
factors, in P. Boolchand (ed.) Insulating and Semiconducting Glasses, World Scientific Press,
Singapore, pp. 369-414.
24. Thorpe, M.F., Jacobs, D.J., Chubynsky, M.V. and Rader, J.A. (1999) Generic rigidity of network
glasses in M.F. Thorpe and P. M. Duxbury (eds.) Rigidity Theory and Applications, Kluwer
Academic/Plenum Publishers, New York, pp. 239-277.
25. Zallen, R. (1983) The Physics of Amorphous Solids, John Wiley and Sons, New York, pp.3.
26. Modulated DSC
TM
Compendium (1997) Reprint #TA-210, TA Instruments, Inc., New Castle, DE
http://www. tainst. com/
27. Bhmer, R. and Angell, C.A. (1992), Correlations of the non-exponentiality and state dependence of
mechanical relaxation with bond-connectivity in Ge-As-Se supercooled liquids. Phys. Rev. B 45, 1091-
1094.
28. Kerner, R. and Mi coul aut , M. (1994) A theoretical-model of formation of covalent binary glasses. 1.
general setting, J. Non-Cryst. Solids 176, 271-279.
29. Micoulaut, M. and Naumis, G.G. (1999) Glass transition temperature variation, cross-linking and
structure i n network glasses: a stochastic approach, Europhys. Lett. 47, 568-574. Also see contribution
of M. Micoulaut Glass Transition Temperature variation as a probe for network connectivity in this
volume.
30. Bresser, W.J., Boolchand, P., and Suranyi, P. (1986), Rigidity Percolation and Molecular Clustering in
Network glasses. Phys. Rev. Lett. 56, 2493-2497.
31. Bresser, W.J., Boolchand, P., Suranyi, P. deNeufville, J.P. (1981), Direct evidence for intrinsically
broken chemical ordering in melt-quenched glasses 46, 1689-1692.
32. Galeener, F.L. (1990) The structure and vibrational excitations of simple glasses, J. Non-Cryst. Solids
123, 182-190. A symposium was held in 1995 to honor the late Frank L. Galeeners contributions to
glass science. The interested reader may want to look at the proceedings of this symposium published
in J. Non-Cryst. Solids 182, 1-212.
33. Lucovsky, G. (1979) Chemical effects on the frequencies of Si-H vibrations in amorphous solids, Solid
State Commun. 29, 571-576.
34. Griffiths, J.E., Espinosa, G.P., Remeika J.P., et al. (1982) Reversible quasi-crystallization in GeSe
2
glass, Phys. Rev. B 25, 1272-1286.
35. Murase, K. (2000) Vibrational excitations in glasses: Raman scattering, in P. Boolchand (ed.),
Insulating and Semiconducting Glasses, World Scientific Press, Inc., Singapore, pp. 415-463.
36. Selvanathan, D., Bresser, W.J., Boolchand, P., and Goodman, B. (1999) Thermally reversing window
and stiffness transitions i n chalcogenide glasses, Solid State Commun. 111, 619-624.
37. Selvanathan, D., Bresser, W.J., Boolchand, P. (2000) Stiffness transitions in Si
x
Se
1-x
glasses from
Raman scattering and temperature-Modulated Differential Scanning Calorimetry, Phys. Rev. B61,
15061-15076.
38. Boolchand, P., Feng, X., Bresser W.J. (2000) Ri gi di t y transition in binary Ge-Se Glasses and the
intermediate phase, J. Non-Cryst. Solids (submitted).
39. Wang, Y., Wells, J., Bresser, W.J., Boolchand, P. (2000) Stiffness Transition in a Zachariasen glass:
Theory and experiment (unpublished).
40. Franzblau, D.S., Tersoff, J. (1992) Elastic properties of a network model of glasses, Phys. Rev. Lett. 68:
2172-2175.
41. He, H., Thorpe, M.F. (1985) Elastic properties of glasses, Phys. Rev. Lett. 54: 2107-2110.
42. Josephson, B.D. (1966) Relation between the superfluid density and order parameter for superfluid He
near T
c
Phys. Lett. 21, 608-609.
43. Chayes, J.T., Chayes, L., Fisher, D.S., et al. (1986) Fubute-size scaling and correlaton lengths for
disordered systems, Phys. Rev. Lett. 57, 2999-3002,
44. Fritzsche, H. (2000) Light induced structural changes in Glasses in P. Boolchand (ed.) Insulating and
Semiconducting Glasses, World Scientific Press, Inc., pp. 653-690.
45. Gump, J., Finkler I., Xia, H., Sooryakumar, R., Bresser, W.J., Boolchand, P. (2000) Direct evidence for
photomelting of the Intermediate Phase in Network glasses (submitted to Phys. Rev. Lett.).
46. Georgiev, D.G., Boolchand, P., Micoulaut, M. Rigidity transitions and molecular structure of As
x
Se
1-x
glasses, Phys. Rev. B (in press).
47. Georgiev, D.G., Mitkova, M., Boolchand, P., Brunklaus, H., Eckert, H., Micoulaut, M. Molecular
Structure, Glass transition temperature variation, agglomeration theory, and network connectivity of
binary P-Se glasses, Phys. Rev. B (submitted).
84
THERMAL RELAXATION AND CRITICALITY OF
THE STIFFNESS TRANSITION
Y. Wang, T. Nakaoka, and K. Murase
Department of Physics, Graduate School of Science, Osaka
University, 1-1 Machikaneyama, Osaka 560-0043, Japan
INTRODUCTION
Covalent network glass system Ge-Se has been well studied and analyzed in terms of
testing theories on materials (glasses) science. Mean-field constraint theory [14] of net-
work glasses is proving to be a powerful tool in explaining numerous anomalous behaviors
around the critical composition of rigidity transition threshold at an average coordination
number, corresponding to x=0.20 for at which the number of constraints
per atom is equal to the degree of freedom per atom [511 ]. Short- and medium-range order
in glasses can be varied continuously to form both the floppy and rigid glasses
in the sense of the constraint theory, because that for the composition range
the coordination numbers of Ge and Se are 4 and 2, respectively. The character of a net-
work glass undergoes a qualitative change, from being easily deformable at to being
rigid at Recently, Boolchand et al. [5, 6] demonstrated that the results from Raman
scattering, modulated differential scanning calorimetry, molar volumes and Mssbauer spec-
troscopy work provide evidence of a multiplicity of stiffness transition, an onset point near
and a completion point near between which a transition region or
an intermediate phase separates the floppy glass from the rigid one. Kamitakahara et al. [7]
reported the dynamic density of state around 5 meV to prove the existence of the floppy
mode (zero-frequency mode in floppy glass) by studying the inelastic neutron spectra of
chalcogenide glasses, however, the structural explanation of the floppy mode is still under-
constructed yet.
In this paper, we report the Raman-scattering results for glasses in a temper-
ature range of 101000 K which covers glassy, super cooled liquid (SCL), crystalline and
liquid states. The thermal relaxations well below the glass-transition temperature, T
g
, in the
vicinity of T
g
and above T
g
differ critically between the floppy and rigid glasses. We propose
a micro-structural explanation of the floppy modes in glasses that structural units
Se
n
(n>2) play an important role in influencing the macroscopic properties. The criticality
of the stiffness transition is discussed from the point of view of structural admixture of Se
n
Phase Transi t i ons And Self-Organization in Electronic and Mol ecul ar Net works
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 85
Figure 1. The preparation of thin film. (a, b) Se element or bulk source of Ge
x
Se
1x
are held in the cell where
a narrow gap of is confined by two paralleled quartz plates. (c) After sealing the cell in a 10
6
Torr
vacuum, we heat up the cell to 200 C higher than the melting point of Ge
x
Se
1x
and make the temperature of
Ge
x
Se
1x
source slightly higer than the temperature in the narrow gap. Then, liquid Ge
x
Se
1x
penetrates into
the narrow gap. (d) After quenching, the thin film between the two quartz plates is obtained.
and GeSe
4/2
units. We demonstrate that reversible structural changes with temperature are
associated with a medium-range order in GeSe
2
glasses, involving significantly layered-like
crystalline GeSe
2
fragments and inhomogeneously distributed nanophases of GeGe and Se
Se homopolar bonds. The intrinsic GeGe bond that does break the chemical order assists
the glass to construct the layered-like crystalline fragments.
GLASS PREPARATION
Bulk glasses Ge
x
Se
1x
are prepared by quenching the melts after rocking the ampule
of mixed compounds sealed in for at least 48 hours. All the measured bulk samples are
picked from the as-prepared broken pieces with a typical dimension of 5x5x1 mm, except
the thin films for studying the temperature dependence at high temperature. The thin films
of Ge
x
Se
1x
, with thickness of about are prepared by quenching from melts held in
evacuated fused-silica cells with a narrow gap defined by two parallel plates. The preparation
procedure is sketched in Fig. 1. Confinement of the films in the cells takes advantage of
keeping the same composition of glasses even at the high temperature, at which considerable
chalcogenide compounds might vaporize if the bulk are held in vacuum. Furthermore, the
films are assumed to be quenched with less difference in cooling rate between the surface and
the inside than the situation in melt-quenched thick bulks.
RAMAN SCATTERING AND MEASUREMENT SETUPS
In noncrystalline materials, wavevector is not a good quantum number for phonons
because that losing the translational symmetry yields the break-down of the usual wave-vector
conservation law. Raman scattering processes are allowed to occur from essentially all the
vibrational modes of the material. Usually the intensity of scattered light is proportional
to the vibrational density of states
for Stokes process. describes the coupling of the vibrational modes of frequency
86
Figure 2. Polarized Raman spectra for Ge
x
Se
1x
at 80 K. The resolution of the spectra is better than 1 cm
1
.
to the light, and is the Bose factor.
Conventional backscattering configuration for incident and scattered lights is employed
in our case. For avoiding light-induced events and restricting the sample temperature differ-
ence from the environment to less than 3 C, a low power-density probing light of less than
3W/cm
2
(< 10W/cm
2
for measuring the spectra of liquid phase around 1000 K) is ap-
plied for the Raman measurements. The back-scattered light is collected and analyzed with
the triple grating polychromator (JOBIN YVON T64000) and charge-coupled-device (CCD)
detector in polarized and depolarized configurations. Generally, we use a 680 nm line from an
argon ion laser pumped Dye (DCM) laser due to the sensitivity of our analysis setup, mainly
determined by the monochromator and CCD detector, peaking around 600 nm. Insofar as
we eliminate the interference of Rayleigh wings in the low-frequency spectra, an 800 nm
line from argon ion laser pumped Ti:Sapphire laser is a better choice that extends our lower
limit to 3 cm
1
at room temperature. A longer wavelength of laser light used as the pumping
source yields a higher resolution for Raman spectrum in our experimental setups as well.
STRUCTURAL RELAXATION WELL BELOW T
g
Stokes Raman spectra of Ge
x
Se
1x
glasses for various composition at 80 K are shown
in Fig. 2. Raman spectra are reduced by and normalized to the maximum intensity,
for low Ge concentration the one at 255 cm
1
or at 200 cm
1
for high Ge concentration.
In g-Se, the dominant peak at 255 cm
1
and a shoulder at the lower frequency side are
relevant to the bond-stretching modes of Se
n
chains [12, 13]. The weight of Se-chains related
intensity decreases with increasing Ge concentration, which qualitatively reflects the changes
in the vibrational density of states. It follows the fact that the number of GeSe
4/2
tetrahedra
increases with increasing Ge content. A strong peak at lower (A
1
) and a weak one at higher
87
Figure 3. Raman spectra for Ge
8
Se
92
glass at various temperatures. The spectra are obtained in a T-increasing
sequence.
energy side in the vicinity of 200 cm
1
are classified to the breathing modes of GeSe
4/2
tetrahedra which are in the corner-sharing (CST) and edge-sharing (EST) configurations,
respectively [14]. It should be noted that even in the glass of Ge content of 8 percents a small
amount of tetrahedra shares the edges. For GeSe
2
and Ge
35
Se
65
, the GeGe bond related
band at 180 cm
1
increases with Ge content. In GeSe2 glass, the homopolar GeGe and
SeSe wrong bonds are assumed to release the exceeding tension or stress in the highly
distorted network, since its mean coordination number is 2.67 much larger than 2.40.
The SeSe vibrational bands are able to be observed at 260 cm
1
with the resonant Raman
condition as well [15]. Further discussion on the GeGe and SeSe wrong bonds in GeSe2
glass will be done in the following sections and elsewhere.
In Ge
x
Se
1x
glasses, the basic structural units are Se
n
chains and GeSe
4/2
tetrahadra.
The former are floppy units that build floppy regions in the network; the latter are rigid units
that construct rigid ones. Studying the temperature dependence of vibrational properties of
SeSe or GeSe bond for each unit will supply plenty information on the microstructure of the
network. Figure 3 shows Raman spectra for Ge
8
Se
92
glass from 20 K to room temperature.
The spectrum at the successive higher temperature is plotted with the base line raised by the
same value. To take systematic analyses, the line shape of the spectra in the range of 160300
cm
1
is fitted by one straight line and four Gaussian curves, corresponding to two breathing
modes of GeSe
4/2
tetrahedra and two Se-chains related modes.
Here, we focus on temperature dependence of peak positions of GeSe
4/2
breathing
mode, A
1
, at 200cm
1
and Se-Se bond-stretching mode at 255 cm
1
, as shown in Figs. 4(a
d). In the floppy glass of Ge
8
Se
92
, an anomalous peak shift of the A
1
mode, a positive shift
with temperature, is observed below 100 K. This anomalous feature is hard to be understood
without introducing a structural changes at such low temperature.
During we cool the glass down to 20 K (well below the T
g
s), in other senses which
can be regarded as a second kind of quenching process, each structural unit freezes at the
correspondent critical temperature below which the unit is forbidden to frequently change
the topology in the network. In Ge-Se network, first the GeSe
4/2
-unit rich region freezes, and
88
Figure 4. The temperature dependence of the A
1
breathing mode of tetrahedra and the Se-Se bond-stretching
mode. Both are scaled by the frequency of each mode at the lowest temperature we measured, (a) and (b)
display the typical case in floppy glasses. (c) and (d) the rigid glasses.
at this time (or temperature) the tetrahedra might not be constraint or stressed. If all the atoms
vibrate in limited spaces caused by the construction of volume with decreasing temperature,
a large number of stress or tension then should be induced since the glassy sample is not
an ideal glass. Appending such stress or tension on the local structural units of GeSe
4/2
tetrahedra and Se
n
chains, the structure of each unit must be distorted more or less because in
the microscopic region such kind of force will not be isotropic. The lowering frequency of A
1
mode during the cooling process suggests that the tetrahedra are departing from the regular
structure. The similar stress or tension should influence the Se-chain structure as well. The
reason why the Se-Se bond-stretching mode shows less temperature dependence than the A
1
mode does is probably because that the bending force of Se-Se bond is much weaker than its
stretching one. Most part of distortion to the Se
n
(n>2) chains causes a wide distribution of
SeSeSe bond angle. Unfortunately, we can only detect the rest stretching mode of Se-Se
bond.
Structural Origin of Floppy modes in Ge
x
Se
1x
In Ge
8
Se
92
glass, Se
n
chains are the dominant structural units and GeSe
4/2
tetrahedra
are probably isolated from each other (few tetrahedra binding together should enhance the
isolation). The structural units of Se
n
(n>2) chains, as shown in Fig. 5, are assumed as the
key units for the heavily distorted tetrahedra relaxing to the regular structure. We suppose
that the rotating motion of Se(2) atom on the axis through the Se(l) and Se(3) atoms will
completely freeze at 20 K. [For the Se(3) atom, if there is a Se(4) atom next to it, the same
discussion we did for the Se(2) atom holds.] With increasing temperature, the rotating mode
of Se(2) atom could be thermally excited, since the relative force determined by the structure
89
Figure 5. Schematic diagram for atom arrangements of Se
n
chains. The Se(2) atom rotates on the axis through
the Se(l) and Se(3) atoms. The rotating motion is constrained by two prior free dihedral angles consisting of
the next nearest neighbor forces. The same rotating mode for the Se(3) exists with the Se(4) bonding to it.
consisting of two priori free dihedral angles, is fairly weak. Such thermally excited rotating
motion supplies channels to leak the stress or tension by the way of changing the topology of
network. The heavily distorted network then relaxes and make the remaining (or observed)
A
1
mode strengthened. When temperature is approaching to 100 K, the rotating modes is
totally excited, and thereafter the vibrational peaks start to shift in negative behavior due to
anharmonic effects.
In our measured samples, the most remarkable positive shift of the A
1
mode occurs in
the Ge
8
Se
92
. For x<0.08, although the positive shift (with the same origin) is expectant, it
might be difficult to figure them up strictly because of a smaller population of tetrahedra to
form the vibrational density of states (VDOS) that brings big errors in fitting the line shape of
A
1
mode. Or, the network could settle the stress or tension through the large floppy regions
that consist of Se
n
(n>2) chains surrounding the GeSe
4/2
-rich region. While for Ge
15
Se
85
,
the floppy region that surrounds the rigid region of GeSe
4/2
units is not large enough to
provide space for the distorted tetrahedra relaxing. Nevertheless, we consider the rotating
mode in Fig. 5 to be a probable micro-structural explanation for the floppy modes in Se-
contained covalent network glasses. The same kind of rotating modes has been discussed in
the computing analysis for modeling the flexible and rigid regions in proteins [16]. The easy
excited dihedral angles that free the rotating motion of Se(2) and Se(3) in Fig. 5 are assumed
as the bananas graphs proposed by Thorpe et al. [4].
In Se-contained glass system, the floppy mode was claimed in the vicinity of 5 meV
from the neutron-scattering results of vibrational density of states. Weakly binding forces
shift the floppy modes from zero frequency to 5 meV. In figure 4(a), the most positive shift
of A
1
mode occurs at 60 K. The relation between the frequency of 5 meV for floppy modes
and the critical temperature of ~60 K for the thermal excitation of rotating motion contains
rich hints to build a structural picture for the network glasses. It is natural to regard the soft
constraint rotating modes as the floppy modes in the realistic floppy glasses. Up to now, most
of properties of covalent glasses are discussed at room temperature or higher, at which such
constraint (weakly binding forces) floppy modes are totally free. That is why the elegant
constraint theory works so good [511], although the theory settles the problem of networks
at 0 K.
Criticality of Stiffness Transition
With increasing x, the of floppy glass is approaching to the rigidity percolation
threshold of (x=0.20). The stiffness transition behavior clearly appears in the temper-
ature dependence of the vibrational modes between 100 K and room temperature where the
anharmonic effects determine the essential rules. As shown in Figs. 4(a) and (b), large drops
90
Figure 6. The slope of straight line obtained from the data of each modes in the temperature range of 100250
K (figure 4). The A
1
mode (square) has a fourth time large of the volume obtained from the SeSe stretching
mode(

) for While for they are similar. The solid lines are guide to the eye.
of the shifts for both the A
1
mode and the SeSe mode at room temperature (RT) are ob-
served in the floppy glasses. The RT is the temperature to which the samples were quenched
(it should be 0 C for an accurate discussion), and at which they rested for a long time before
the measurement. The lower temperatures (<RT) represent a second kind of quench, Thus
between 100 K and 250 K, we can draw straight lines through the data for both of the two
modes and get well linear fits. The slopes of lines as a function of x and mean coordination
number for both the A
1
mode and the SeSe mode are plotted in Fig. 6. A nice con-
vergence of the slopes, like the shifts themselves [Fig. 4(c)], comes to a common value at
x=0.22. This is the stiffness transition at which the number of degrees of freedom is equal to
the number of constraints [5]. Apparently, all the observed modes shift together at this tran-
sition, because the network becomes almost ideally random to realize a maximum mixing of
the different local modes.
For the rigid glasses, the fact that few Se
n
(n>2) units are included in the structure is
the reason of admixture of all the local modes. Probably the same mixing of SeSe bending
motion (it should be the SeSeGe bending motion) goes into both the A
1
and SeSe stretch-
ing modes. So although the local modes of the A
1
breathing and the Se-Se stretching have
different frequencies, 200 cm
1
and 255 cm
1
, respectively, the CHANGES in those frequen-
cies (scaled by their frequency at the lowest measured temperature) become the same. This
kind of mixing should occur in a long-range scale and well detected by the light-probe tech-
nique. It should be stressed that the long-range mixing of vibrational modes is very analogous
to critical fluctuations of random system in second-order phase transitions. Comparing the
modes mixing (in Fig. 6) in Ge
22
Se
78
and Ge
30
Se
70
glasses, we find that the two modes in
Ge3oSe
70
glass mix better than those in Ge
22
Se
78
. The slope of A
1
slightly differ from that of
Se-Se mode at (x=0.22). These results should have respect to the transition window
in which the glass is not ideally random due to a self-organization. Boolchand et al. has done
a systematic study of the width of the transition window and pointed out the completion point
( or x=0.23).
For the floppy glasses, the A
1
mode shifts much more than the SeSe bond-stretching
mode does. The probable reason is that the mixing of SeSe bending motion into the bending
of SeGeSe permits all the bending modes mixing in the floppy glasses. In other words,
the A
1
band is not a pure breathing mode because of the admixture of bending modes. Thus,
91
Figure 7. Dynamic phase diagram for Ge-Se glass system. The solid line is the liquidus taken from Ref. [18].
All the data are estimated from the Raman measurement.
the SeSe stretching mode shows less temperature dependence than the A
1
breathing mode
in which the soft bending modes are mixed, since the stretching forces are stronger than the
bending ones.
CRYSTALLIZATION TENDENCY
An extensive study of the temperature dependence of the spectral shape for the Ge
x
Se
1x
thin films [17] has been applied to understand the structural changes around and above the
glass transition temperature, T
g
. In general, the spectrum is accumulated for 10 minutes
at each measured temperature which is fixed within and the rate of increase of the
temperature between two successive measurements is 25 C/min. For each composition, all
the spectra are obtained within ~10 hours.
In the vicinity of stoichiometic composition, or the medium range struc-
tures of Se
n
chain or GeSe
2
fragments that are topologically similar to the layered crystalline
GeSe
2
, respectively, dominantly contribute to the changes. In the glasses at intermediate
composition, the two structural units mix well. With increasing temperature above the T
g
,
the low viscosity of the SCL enables us to detect the structural changes within our measure-
ment periods. Thus, we define the T
g
, obtained by Raman measurement, as the onset of the
changes of spectral shapes. Heating the glasses thereafter, characteristic lines of crystalline
(c-) Se or c-GeSe2 may appear in the spectra and disappear at higher temperatures. We term
the recrystallization temperature T
c
and the melting point T
m
as the temperatures at which
the crystalline peaks appear and disappear, respectively, regardless of whether the sample
is stoichiometic (x = 0, 0.33). The phase diagram of Fig. 7 summarizes the results for T
g
(square), and The T
g
(square) from Raman result is in a good agreement
92
with the result from DSC measurement [17]. The present values for T
m
agree with the
results of Ref. [18], where the liquidus line indicates the primary crystallization temperature
for c-GeSe2 or c-Se.
For x<0.04, as the small amount of GeSe
4/2
tetrahedra fails to prevent the crystalliza-
tion of Se, mixtures of c-Se and liquid Ge
x
Se
1x
appear between the liquid and super-cooled
liquid (SCL) states. While, for x>0.18, the large amount of GeSe
4/2
tetrahedra builds a
medium-range structure topologically similar to the layered c-GeSe
2
to promote the crystal-
lization of GeSe2. For since the time for the (Se)
n
chains forming the long-
range order is longer than the experimental time period, neither c-Se nor c-GeSe
2
appears.
Embryo of c-GeSe2 constructs in SCL state for x>0.10, however, only those of x>0.18 evolve
into nuclei and crystallize.
Here, we define a crystallization ability (CA) by a ratio of crystallizable temperature
range to SCL range,
using the Raman measurement results, and make the CA to be unit for the glass transferring
directly from glassy to liquid phase. The crystallization tendency derives from the glass form-
ing tendency (GFT), with the relation of Figure 8 indicates the crystallization
tendency, obtained by the Raman measurement, which is very similar to the graph of glass
forming difficulty summarized in Ref. [1]. It is worth to mention the recent result of GFT
in Ge
x
Se
1x
glasses from Boolchands group [5]. They synthesized homogeneous and strain-
free bulk glasses by cooling melts at a very slow rate of 2.5 C/minute in a T-programmed
resistive furnace. The GFT is apparently optimized in the composition range of 0.20<x<0.24
(very similar to the stiffness transition window they found out) where no eutectic effects can
be introduced. In accordance with the CA from the Raman result of heating glasses from
RT to liquid phase, we do not observe a minimum of CA in the stiffness transition window.
It should be noted that our samples are quenched from the temperature 200 C higher than
the T
m
inferred from the liquidus [18]. Thus, in our samples the local structure must quite
differ from the structure in the melts during cooling at the rate of 2.5 C/minute. The local
structure, which is restricted to a nanoscale, attracts wide attention, such as the nanophase
separation in GeSe
2
glass being discussed in the following sections.
Figure 8. Crystallization ability calculated form the dynamics phase diagram (Fig. 7) by equation 2.
93
Figure 9. Three types of typical temperature dependence of the Raman spectra in GeSe
2
glasses. Each spectrum
is normalized at the most intense peak around 200 cm
1
. Starting from the Raman spectra at RT, however, the
glasses crystallize into (a) the phase, a layered crystal involving both CST and EST units; or (b) the phase,
a 3D-network crystal consisting only CST unit; or (c) the phase, for which the Raman spectrum is the first
reported.
TRIFURCATED CRYSTALLIZATION AND IN HOMOGENEITY IN GeSe
2
GLASS
In this section, we concentrate on the structural changes in GeSe
2
glass prior to the
crystallization towards three crystalline phases (trifurcated crystallization). Thermally in-
duced structural changes are reversible even well below the T
g
of ~390 C. These reversible
changes provide an indication that the GeGe wrong bonds bearing a nanophase play an im-
portant role on the network stability. Strong correlation between the degree of the structural
change and the crystallized phase is discussed in terms of homogeneity in a nanoscale.
Structural changes and crystallizations
Free-standing bulk GeSe
2
samples are sealed in a silica tube in an argon gas atmosphere
(~360 Torr) to reduce evaporation and oxidation effects. Raman spectra are measured using
a 1.83 eV laser light with a very low power density of less than 1 W/cm
2
. The laser light
is focused onto the samples in a ~0.1x4 mm
2
rectangular region. Figures 9 (a), (b), and (c)
show the three typical temperature dependence of the Raman spectra of GeSe
2
glasses from
RT to the crystallization temperatures, T
c
. The intense Raman band located at 200 cm
1
at
RT represents an A
1
in-phase breathing vibration of the corner-sharing tetrahedra (CST) [9].
The A
1
companion band at 215 cm
1
is related to a breathing vibration quasi-localized
at edge-sharing tetrahedra (EST). The A
G
band at 176 cm
1
is interpreted as the stretching
mode of GeGe bonds involved in ethane-like units [19, 20]. No distinguishable difference
is observed among the Raman spectra of the glasses at RT because all of the measured GeSe2
glasses are prepared at the same conditions. However with increasing temperature, notable
spectral changes occur at 300 C which is well below the T
g
of 390 C.
It was reported that once a crystallization process starts, some of them result in the same
crystalline phase as they initially appeared, while some of them transform into other phases
after held at the same temperature for an additional long period (~ 60 hours) [21]. In this
work, we restrict the discussion of crystalline phase to the initially appeared crystalline phase.
We checked the phase transition with keeping temperature for the same as the glasses initially
94
Figure 10. Integrated intensity ratios of the A
G
band to the A
1
band (ac) and the band to the A
1
band (df)
as a function of temperature. The ratios are normalized to those at RT. The GeSe2 glasses crystallize into the
phase [(a) and (d)], the phase [(b) and (e)], and the phase [(c) and (f)]. Lines guide to the eye. The error
bars are within the symbols. Two measurement results for each case of crystallization are picked up from over
50 measurements (samples).
crystallizing and found that the phase (low temperature phase) is tending to transform to
the phase (high temperature one). Figure 10 depicts the temperature dependence of the
integrated intensity ratios of S(A
G
)/S(A
1
) and normalized to those at RT. The
GeSe
2
glass crystallizes into the phase (2D) when the intensity ratios decrease slightly
with temperature, and they start to increase above T
g
. The increases that relates to the glass
transition will be discussed later. The phase (3D) is obtained when the decreases of the
ratios are larger than the case of the phase occurring. The Raman spectra of the thermally
formed and phases are in agreement with those of previous works [9, 22].
An unreported Raman spectrum of a crystalline phase [Fig. 9(c)] appears when the ratios
show the largest decreases among the three types of the crystallization processes. This Raman
spectrum is similar to that of the except the low-frequency vibrational modes (see,
Fig. 11). We assign this crystalline phase to the phase which was reported to be appearing
when Se-rich Ge
x
Se
1x
glasses (0.15<x<0.32) were well annealed [23, 24]. According to
the similarity of the position of Raman band relevant to the vibrations of CST in crystals
around 200 cm
1
, we conclude that the phase is composed of CST units. The details
of the phase, such as the basic structural units, are future work. We confirmed that the
spectrum of Fig. 9(c) was also observed in the cases of Ge
28
Se
72
and Ge
30
Se
70
glasses being
well annealed. The two Se-rich glasses always crystallize into the phase in contrast to
the trifurcated crystallization behavior in GeSe
2
glasses. The crystallization to the phase
always follows a strong growth of a Raman band around 260 cm
1
(in Fig. 9(c) at 300 C for
example). This strong Raman band has been assigned to the stretching mode of SeSe bonds
[12, 13]. A weak growth of the same Raman band can be distinguished during the process
to the a crystallization. The strong growth of the SeSe mode is in accordance with the fact
95
Figure 11. Expanded view of the Raman spectra of thermally formed crystals shown in Fig. 9(ac). The
spectrum of the is similar to the around 200 cm
1
but is different in the low frequency
range; for example, the Raman peaks at 25 cm
1
and 60 cm
1
of the are very weak in the spectrum
of the and the peak at 90 cm
1
of the is hardly observed in the spectrum of the
A slight admixture of the different phases might happen.
that the phase is formed by annealing Se-rich glasses as well.
A mixture of the and phases appears after the changes of the intensity ratios whose
degree is between those towards the pure and -phases crystallization. Similarly, a mix-
ture of and phases is also observed. Thus, the crystallized phase is strictly determined by
the degree of the decreases of the integrated intensity ratios, and S(A
G
)/S(A
1
).
The close correlation between the changes of the integrated intensity ratios and the crystal-
lized phase suggests the creation and annihilation of structural units besides the deformation
of the units, which also decrease the Raman intensity due to breaking symmetry of the struc-
tural units.
Figure 12 shows temperature dependence of the full width at half maximum (FWHM) of
A
1
and bands. The FWHM for the AC band should be more valuable (interesting) to the
following discussion if we can measure the extremely weak A
G
band at high temperature in
a good quality. The width of the A
1
band first increases with temperature, and then decreases
while that of the band increases monotonically. The lower temperature at which the width
of A
1
band starts to decrease, the more the integrated intensity ratios decrease. Because the
degree of the decreases of the S(A
G
)/S(A
1
) is much larger than the in all the
samples, the breaking of GeGe wrong bonds should play a dominant role on the structural
changes. The GeGe wrong bonds are expected to be transformed by heating to energetically
more favored GeSe bonds. By this transformation, the CST units are formed from ethane-
like Ge-Ge units. We believe that a growth of the CST units at the expense of the GeGe
bonds promotes an ordering of the CST units to reduce the FWHM of the A
1
band. The
is consistent with the crystallization of the and phases after the larger decrease of the
S(A
G
)/S(A
1
) than the phase. The and phases, which are composed of only CST units,
should be formed though the higher degree of CST formation.
96
Figure 12. Temperature dependence of the full width at half maximum (FWHM) of the A
1
band (a), and the
band (b) in the GeSe
2
glasses which are crystallizing towards the phase the phase (

), and the
phase by further heating. Lines are added to guide the eye. The FWHM of the A
1
band increases with
temperature and then starts to decrease while that of the band increases monotonically.
Medium-range Order of GeSe
2
Glass
The phase is realized through the least structural changes among the three types of
the crystallization processes. The fact stresses that fragments topologically similar to the
crystalline phase are significantly involved in the network of GeSe
2
glasses. These fragments
are larger than the GeGe and CST units, and they are randomly oriented. When GeSe
2
glass
is heated over the T
g
, a lower viscosity yields the local structure to be rearranged more easily.
In Fig. 9(a), the increases of the intensity ratios above T
g
is assumed as the trend toward the
nucleation for crystallization of Thus, the structure of GeSe
2
glass is similar to the
phase, hence, the large degree of formation of CST units is needed to crystallize into the
or phases.
Boolchand et al. [25] have recently suggested that, in the Ge-rich Ge
x
Se
1x
glasses
the presence of GeGe signatures decreases the global connectivity, and it con-
stitutes part of an ethane-like units bearing nanophase formed separately from the GeSe
4/2
tetrahedra bearing backbone of the glass. Together with our results, it follows that the ethan-
like units involving GeGe bonds and like fragments coexist in the melt-quenched
GeSe
2
glass. We can observe the trifurcated behavior in GeSe
2
and Ge
35
Se
65
glasses, but, we
can never observe it in Ge
28
Se
72
and Ge
30
Se
70
glasses. Instead, the Se-rich glasses always
crystallize into the phase. Therefore, the GeGe nanophase seems to be the physical origin
of the trifurcated crystallization.
Generally, disordered materials are more or less inhomogeneous or heterogeneous. For
instance, the importance of the inhomogeneity to understand the dynamics at the glass tran-
sition has been indicated experimentally [26] and theoretically [27]. Here, we assume an
inhomogeneous distribution of the GeGe nanophases. The size of the GeGe phase and the
surrounding situation, such as the sorts of surrounding units and the surrounding stresses,
are nanoscopically inhomogeneous. Since the laser spot size is 0.1x4 mm
2
, no evidence of
inhomogeneity is able to be detected among the samples at RT. With increasing temperature,
97
the ethane-like GeGe units are selectively transformed to the CST units. A large size of
GeGe nanophase transforms to a large one of the CST region, and thereafter the CST re-
gion might grow at higher temperature. However, if the size of GeGe nanophase is smaller
than a critical volume, it is not transforming to the CST region. Because larger CST units is
more stable due to the gain in volume free energy of a phase composed of CST units at high
temperatures. Thus, the degree of formation of CST units spatially differs. When the gain
in volume free energy of the CST phase can compensate a larger CST region for the cost in
interfacial free energy, the CST formation will produce a nucleus to form the or phases.
The resultant crystalline phase, or seems to be determined by the degree of formation
of the Se-chain clusters. Note that a stronger increase of the Se-chain vibrational mode lead
the crystallization to the phase [Fig. 9(c)]. The difference in the degree of formation of the
Se-chain clusters is also due to the inhomogeneity in an initial distribution of SeSe bonds
and the surrounding situation. On the other hand, if the degree of the CST formation is low
and the size is smaller than the critical size above T
g
, the like fragments will grow
to form the phase. Thus, the appearance of three crystalline phases depends on the
degree of the formation of structural units of CST and Se-chain. The nature of the trifurcated
crystallization is attributed to the inhomogeneous network that consists of nanoscopically
phase-separated structural units.
Figure 13. Integrated intensity ratio of the A
G
band to the A
1
band (a, b) and the ratio of the band to the A
1
band (c, d) as a function of temperature. The ratios are normalized to those at RT. The as-quenched sample is
heated to and kept at 300 C for five hours, and thereafter, it is cooled to RT. (a, c) show the first cycle of the
heat treatment and (b, d) the second one. Lines are added to guide the eye. The error bars are less than the size
of the symbols.
The reason why our samples show such a clear trifurcated behavior is that our melt-
quench temperature of 960 C is 220 C higher than the liquidus [18, 23]. The percentage
of the GeGe units in GeSe
2
glasses increases with quench temperature [28], and the high
temperature quenching will increase the inhomogeneity in the distribution of GeGe units.
We should note that the existence of the GeGe nanophase and its inhomogeneity is intrin-
sic, because the annealed samples also show the trifurcated behavior. A higher temperature
quenching process just enhances such properties.
98
Reversibility and stability
The reversibility and stability of the GeSe
2
glasses are investigated by heating and
cooling free-standing bulk glasses as follow. As-quenched GeSe
2
sample is heated to and
kept at 300 C (below the T
g
) for five hours, and thereafter, it was cooled to RT. Figure 13
shows the normalized integrated intensity ratios, and S(A
G
)/S(A
1
), in the first
[Figs. 13(a,c)] and the second [Figs. 13(b,d)] heating cycles. The changes of the intensity
ratios with temperature are nearly reversible in the first cycle, and they are fully reversible in
the second one. The slightly irreversible parts in the first cycle are due to usual annealing ef-
fects. Changes of FWHMs are also reversible. The reversibility confirms that the thermally
transformed structures are stable at each fixed temperature during the experiment time scale
(15 minute). The stable structure depends on temperature, and it is realized within a minute.
At higher temperature, the structure similar to the phase or phase, which consists of
CST units, is more stable than that at lower temperature. The recovery of the S(A
G
)/S(A
1
)
ratio during the cooling process demonstrates that the network at lower temperature prefers to
involve a larger percentage of the GeGe wrong bond. In other words, the GeGe homopolar
bond in GeSe
2
glass may reduce the total free energy of the network at RT, thus, it could
not be wrong. It should be stressed that the temperature dependence of the S(A
G
)/S(A
1
)
ratio is qualitatively the same as that of the ratio. It could be understood in the
restricted sense of the local structure that rising the percentage of the GeGe wrong bond are
accompanied by a larger population of EST units. The layered fragments that consist of both
EST and CST units must be constructed in GeSe
2
glass with the help of the GeGe wrong
bond. Thus, the GeGe wrong bond plays an indispensable role on the stability of the glassy
structure.
The reversible structural change from the as-quenched structure to a quasi-three dimen-
sional structure composed of CST units can be caused by pressure [29]. The comparison of
the reversible changes (with temperature and pressure) between the amorphous-amorphous
phase transition observed in SiO
2
[30, 31], is interesting and it will be discussed elsewhere.
ACKNOWLEDGMENTS
It is a pleasure to acknowledge J.C. Phillips and P. Boolchand for helpful discussions and
suggestions. This paper draws the recent results developed in our laboratory of collaboration
and discussions with K. Inoue, O. Matsuda, M. Nakamura, and M.K. Nakamura. This work
was supported by Grants-in-Aid for Scientific Research (B)(No. 09440117), Encouragement
of Young Scientists (No. 11740173), and a grant for Scientific Research in the Priority Area
Cooperative Phenomena in Complex Liquids, from the Ministry of Education, Science and
Culture (Japan). Y.W. acknowledges support from the Inamori Foundation.
REFERENCES
1. Phillips, J.C. (1979) Topology of covalent non-crystalline solid I: short-range order in chalcogenide al-
loys, J. Non-Cryst. Solids 34, 153181.
2. Phillips, J.C. (1999) Constraint theory, stiffness percolation and the rigidity transition in network glasses,
in M.F. Thorpe and P.M. Duxxbury (eds.), Rigidity Theory and Applications, Kluwer Academic / Plenum
Publishers, New York, pp. 155172.
3. Thorpe, M.F. (1983) Continuous deformations in random networks, J. Non-Cryst. Solids 57, 355370.
4. Thorpe, M.F., Jacobs, D.J., and B.R. (2000) The structure and rigidity of network glasses, in
P. Boolchand (ed.) Insulating and semiconducting glasses, World Scientific Press, Singapore, pp. 95145.
5. Boolchand, P., Feng, X., Selvanathan, D., and Bresser, W.J. (1999) Rigidity transition in chalcogenide
glasses, in the book listed in Ref. [2], pp. 279296.
99
6. Selvanathan, D., Bresser, W.J., and Boolchand, P. (2000) Stiffness transitions in Si
x
Se
1x
glasses from
Raman scattering and temperature-modulated differential scanning calorimetry, Phys. Rev. B61, 15061
15076.
7. Kamitakahara, W.A., Cappelletti, R.L., Boolchand, P., Halfpap, B.L., Gompf, F., Neumann, D.A., and
Mutka, H. (1991) Vibrational densities of states and network rigidity in chalcogenide glasses, Phys. Rev.
B44, 94100.
8. Feng, X., Bresser, W.J., and Boolchand, P. (1997) Direct evidence for stiffness threshold in chalcogenides
glasses, Phys. Rev. Lett. 78, 44224425.
9. Murase, K. (2000) Raman Scattering in the book listed in Ref. [4], pp. 415463.
10. Wang, Y., Nakamura, M., Matsuda, O., and Murase, K. (2000) Raman-spectroscopy studies on rigidity
percolation and fragility in Ge(S,Se) glasses, J. Non-Cryst. Solids 266269, 872875.
11. Tatsumisago, M., Halfpap, B.L., Green, J.L., Lindsay, S.M., and Angell, C.A. (1990) Fragility of Ge-As-
Se glass-forming liquids in relation to rigidity percolation, and the Kauzmann paradox, Phys. Rev. Lett.
64, 15491552.
12. Nemanich, R.J., Connel, G.A.N., Hayes, T.M., and Street, R.A. (1978) Thermally induced effects in
evaporated chalcogenide films. I. Structure, Phys. Rev. B18, 69006914.
13. Lucovsky, G. and Galeener, F.L. (1980) Intermediate range order in amorphous solids, J. Non-Cryst.
Solids 35&36, 12091214.
14. Murase, K., Inoue, K., and Matsuda, O. (1993) Medium-range structure and relaxation in chalcogenide
glasses investigated by Raman scattering, in Y. Sakurai, Y. Hamakawa, T. Masumoto, K. Shirae, and
K. Suzuki (eds.), Current Topics in Amorphous Materials: Science and Technology, Elsevier, Amsterdam,
pp. 4758.
15. Nakaoka, T., Wang, Y., Matsuda, O., Inoue, K. and Murase, K. Reversible photoinduced structural
changes in GeSe
2
glass at low-temperature, in Proc. 25rd Int. Conf. Phys. Semicond., Osaka 2000, H.
Kamimura and T. Ando (eds.), (Springer Verlag, Berlin), to be published in January 2001.
16. Jacobs, D.J., Kuhn, L.A., and Thorpe, M.F. (1999) Flexible and rigid regions in proteins, in the book
listed in Ref. [2], pp. 357384.
17. Wang, Y, Matsuda, O., Inoue, K., Yamamuro, O., Matsuo, T., and Murase, K. (1998) A Raman scattering
investigation of the structure of glassy and liquid Ge
x
Se
1x
, J. Non-Cryst. Solids 232234, 702707.
18. Tronc, P., Bensoussan. M., and Brenac, A. (1973) Optical-absorption edge and Raman scattering in
Ge
x
Se
1x
glasses, Phys. Rev. B8, 59475956.
19. Bridenbaugh, P.M., Espinosa, G.P., Griffiths, J.E., and Phillips, J.C. (1979) Microscopic origin of the
companion A
1
Raman line in glassy Ge(S,Se)
2
, Phys. Rev. B20, 41404144.
20. Jackson, K., Briley, A., Grossman, S., Porezag, D.V., and Pederson, M.R. (1999) Raman-active modes of
a-GeSe
2
and a-GeS
2
: A first-principles study, Phys. Rev. B60, R14985R14989.
21. Sakai, K, Yoshino, K., Fukuyama, A., Yokoyama, H., Ikari, T., and Maeda., K, (2000) Crystallization of
amorphous GeSe
2
semiconductor by isothermal annealing without light radiation, Jpn. J. Appl. Phys. 39,
1058-1061.
22. Nakaoka, T., Wang, Y., Murase, K., Matsuda, O., and Inoue, K. (2000) Resonant Raman scattering in
crystalline GeSe
2
, Phys. Rev. B61, 15569-15572.
23. Azoulay, R., Thibierge, H., and Brenac. A. (1975) Devitrification characteristics of Ge
x
Se
1x
, glasses, J.
Non-Cryst. Solids 18, 33-53.
24. Stlen, S., Johnsen, H.B., Be, C.S., Grande, T., and Karlsen, O.B. (1999) Stable and metastable phase
equilibria in the GeSe
2
-Se system, J. Phase Equilib. 20, 1728.
25. Boolchand. P. and Bresser, W.J. (2000) The structural origin of broken chemical order in GeSe
2
glass,
Phil. Mag. B, to be published.
26. Moynihan, C.T. and Schroeder, J. (1993) Non-exponential structural relaxation, anomalous light scat-
tering and nanoscale inhomogenities in glass-forming liquids, J. Non-Cryst. Solids 160, 5259; Ci-
cerone, M.T. and Ediger, M.D. (1995) Relaxation of spatially heterogeneous dynamic domains in su-
percooled ortho-terphenyl, J. Chem. Phys. 103, 56845692.
27. Doliwa, B. and Heuer, A. (1998) Consequences of kinetic inhomogeneities in glassess, Phys. Rev. E54,
16521662.
28. Petri, I., Salmon, P.S., and Howells, W.S. (1999) Change in the topology of the glass forming liquid
GeSe
2
with increasing temperature, J. Phys.: Condens. Matter 11, 1021910227.
29. Murase, K and Fukunaga, T, (1984) Pressure induced structural change of clusters in chalcogenide
glasses, in P.C. Taylor and S.G. Bishop (eds.) Optical effects in amorphous semiconductors, A1P Conf.
Proc. No. 120 (AIP, New York), pp. 449456.
30. Poole, P.H., Grande, T., Angell, C.A., and McMillan, P.P. (1997) Polymorphic phase transitions in liquids
and glasses, Science275, 322323.
31. Lacks, D.J. (2000) First-order amorphous-amorphous transformation in silica, Phys. Rev. Lett. 84, 4629
4633.
100
SOLIDITY OF VISCOUS LIQUIDS
J.C. DYRE
Department of Mathematics and Physics (IMFUFA),
Roskilde University, Postbox 260, DK-4000 Roskilde,
Denmark
INTRODUCTION
The glass transition takes place when a liquid upon cooling becomes more and more
viscous and finally solidifies to form a glassy solid [1-7]. Most liquids that are able to form
glasses are supercooled and thus not in genuine thermal equilibrium, but this fact is probably
not important for understanding the dramatic increase in viscosity precedingglass formation.
Approaching the glass transition the viscosity becomes so large that molecular motion
is arrested on the time-scale of the experiment. The fascination of this phenomenon from a
theorist point of view lies in the fact that chemically quite different liquids involving ionic
interactions, covalent bonds, van der Waals forces, hydrogen bonds, or even metallic bonds
have a number of properties in common when they are in the very viscous regime, close to
the glass transition temperature T
g
. These general properties can be summarized into three
nons:
Non-Arrhenius temperature-dependence of the average relaxation time.
Non-Debye linear response.
Non-linearity of structural relaxation even with respect to quite small temperature changes
(i.e., 1% of T
g
).
The last non is not very interesting, because non-linearity is unavoidable whenever the
relaxation time is strongly temperature dependent. This is true independent of whether the
relaxation time is Arrhenius or not, and independent of whether there is just one relaxation
time or a whole relaxation time spectrum.
Our focus here is on the two first nons. Before proceeding, let us briefly remind
the reader of them. Almost all viscous liquids have viscosities and average relaxation times
that are non-Arrhenius. This is perhaps not very puzzling, given the fact that it is likely a
priori that not just one single energy barrier is involved in viscous liquid dynamics. What
is puzzling, however, is the sign of the non-Arrhenius behavior: Whenever a range of
activation energies is involved any naive model predicts an apparent activation energy which
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kl uwer Academic/Plenum Publishers, 2001 101
decreases as temperature decreases; this is because at low temperatures the system prefers the
reaction channels with lowest energy barriers. However, as is clear from Fig. 1, precisely
the opposite behavior is observed:
Figure 1. Angells famous plot of the viscosity of a number of viscous liquids [8], giving the logarithm of the
viscosity as a function of T
g
/T. If the viscosity is Arrhenius this plot gives a straight line (diagonal). Almost all
viscous liquids have apparent activation energies (= slope in the plot) which increase as temperature is lowered.
The non-Arrhenius behavior is one of the great mysteries of this research field. An-
other great mystery is the origin of the non-Debye linear response functions. Typical linear
response quantities are dielectric relaxation, frequency-dependent shear or bulk moduli, or
frequency-dependent specific heat. The Debye linear response is given by
Via the fluctuation-dissipation theorem this corresponds to a time autocorrelation function
that is a simple exponential.
Except for certain mono-alcohols Debye relaxation is never observed. Instead, one al-
ways observes loss peaks (imaginary parts of response functions) that are asymmetrically
deformed towards the high frequency side (Fig. 2): Below the loss peak frequency the re-
sponse function follows Eq. (1), above the loss peak frequency the loss typically decays as
where is often close to 0.5 [see, e.g., Ref. 9 and its references]. Despite many years of
research and thousands of measurements there is no consensus on the origin of the non-Debye
behavior. The puzzle is not so much why Debye behavior is rarely observed, but rather why
not much broader response functions are observed [10]. If one again naively thinks in terms
of energy barriers, the correct barrier distribution has a very sharp high energy cut-off and
is an exponential towards low energies with a width which is only 5-10% of the maximum
activation energy, the activation energy of the loss peak frequency itself.
In this paper we shall discuss the physics of viscous liquids from a (perhaps) simplistic
point of view. We shall argue that these liquids are more like solids than like ordinary liquids
102
Figure 2. Typical loss peaks for a viscous liquid. The figure shows the dielectric loss of triphenyl phosphite at
206.0 K, 208.0 K, and 210.0 K [10]. Below the loss peak frequency the behavior is as predicted by the Debye
expression Eq. (1) but above the peak the loss follows an -decay. At very high frequencies the
so-called relaxation is visible.
and that this solidity explains both the non-Arrhenius and the non-Debye behaviors. First,
we consider how a viscous liquid flows at all, by asking: Is a viscous liquid like an ordinary
liquid (e.g., water) or is it qualitatively different?
ARE VISCOUS LIQUIDS JUST LIQUIDS THAT ARE VISCOUS?
Any medium described by ordinary hydrodynamics in the long wave-length limit has a
shear viscosity, As shown by Maxwell, if denotes the instantaneous shear modulus the
shear relaxation time is related to the viscosity by
According to this equation ordinary liquids like water have relaxation times in the picosecond
range. In contrast, a liquid approaching the glass transition has a viscosity typically 10
15
times larger than that of water and thereby a relaxation time of 1000 seconds or longer. This
enormous difference justifies asking whether there are qualitative differences between the
two cases.
The decoupling of relaxation times from phonon times as viscosity increases is reflected
in a dramatic decoupling of diffusion constants: For ordinary liquids the molecular dif-
fusion constant D is of the same order of magnitude as the transverse momentum diffusion
constant, the dynamic viscosity of the Navier-Stokes equation is mass density].
With increasing viscosity D decreases as according to a simple Stokes-Einstein type
argument while increases. At the glass transition is about 10
30
times larger than D.
The physical interpretation of the very long relaxation times of viscous liquids is simple:
If s it takes roughly 100 s for a molecule to move one intermolecular distance. This
interpretation is somewhat modified by the existence of dynamic heterogeneities and of the
103
relaxation, but basically there is no reason to question this picture. How can molecular
movement be so slow? Again the answer is simple: Average velocities, of course, are deter-
mined by temperature, so molecular motion can only be slow because it is ineffective. It is
ineffective because almost all motion is vibrational.
The vibrations take place around a potential energy minimum in configuration space.
For the vibrations to persist for billions and billions of times before anything else happens to
the molecules, there must be rather large barriers for jumping into a neighboring minimum.
The jumps are referred to as flow events. It is the existence of these rare flow events which
eventually allows a viscous liquid to flow, but clearly the liquid looks like a solid on shorter
time scales.
It is interesting to note that this picture is almost as old as the research field itself. Thus
Kauzmann referred to flow events as jumps of molecular units of flow between different
positions of equilibrium in the liquids quasicrystalline lattice [1]. He thereby emphasized
the fact that a viscous liquid most of the time is in a state of mechanical equilibrium. The
flow event picture [11], which has never really been challenged, was recently confirmed by
extensive computer simulations of a highly viscous Lennard-Jones mixture [12].
The solid-like-ness of viscous liquids the fact that these liquids are virtually always
in a state of mechanical equilibrium we shall refer to as their solidity [13,14]. Solidity
expresses the simple fact that on a short time scale there is only vibration. For instance, sim-
ulating a liquid with an average relaxation time around 1 s on the fastest computer available
today would be a real disappointment: Only vibrations would be visible and one cannot pos-
sibly tell the difference between the viscous liquid and a solid. When a flow event finally
does take place, after a short time there is again mechanical equilibrium in the surroundings
and the molecules vibrate over and over, now just around a slightly different potential energy
mi ni mum [15]. The effects of one flow event in the surroundings is discussed in detail in
Refs. 13 and 14. There is a displacement field varying as where r is the distance to
the flow event. This result is found by a straightforward application of solid elasticity theory.
It is important to note that the solidity of viscous liquids has only a finite range [13].
Thus there is a solidity length l beyond which flow events effectively do not induce molec-
ular motions at all. To estimate l we note that elastic displacements propagate with the ve-
locity of sound, c. Consider a sphere with radius R. Within this sphere, if r
0
is the size of
one rearranging region, there are roughly possible locations for flow events. A
molecule at the center of the sphere only feels the full effects from a flow event within the
sphere if the following condition is obeyed: The displacement deriving from such a flow
event must propagate throughout the sphere and elastic equilibrium be reestablished before
another flow event takes place. Since is the average time between two flow events involving
the same molecules, the time between two flow events within the sphere is
This time must be longer than or equal to R/c. To estimate l we use equality for R = l and
note that c is the sound velocity of the glass, c
glass
. This leads to the following expression for
the solidity length l
The solidity length diverges slowly as When l is several thousand Angstroms.
SHOVING MODEL OF NON-ARRHENIUS BEHAVIOR
According to the solidity picture, molecules in a viscous liquid have to overcome a
relatively large energy barrier in order to move. The non-Arrhenius behavior means that the
barrier is temperature dependent. For simplicity, we shall not distinguish between energy
barriers and free energy barriers. If the barrier is denoted by and is a typical
prefactor (of order 1 picosecond), the average relaxation time is given by
104
Experiments (Fig. 1) imply that increases as temperature decreases. The challenge is
to understand why.
Elsewhere we have briefly reviewed the two standard models for non-Arrhenius behav-
ior, the entropy model and the free volume model, and critiques against these models [7].
The alternative shoving model is based on the following picture, which involves postulates
similar to those of the free volume model: Molecules in a liquid are closely packed, so in
order to be able to rearrange for a flow event to take place extra space is needed. This
is the first postulate. A further input to the model is the well-known fact that intermolecular
interactions are strongly anharmonic with harsh repulsions but only weak attractions. It is
these harsh repulsions which makes it unfavorable (i.e., too energy costly) to rearrange at
constant volume. To lower the barrier the molecules therefore shove aside the surrounding
molecules.
Let us quantify these ideas. Suppose the rearranging molecules constitute a sphere which
at the transition state has increased its radius by The surrounding liquid behaves entirely
like a solid during the flow event. Therefore, the energy needed for expanding the sphere
is quadratic in The energy barrier to be overcome inside the sphere is given by some
function so the total barrier is for some constant A given by
Minimizing this leads to Thus, the ratio between the shoving work and
the inner barrier, is given by
Because of the harsh intermolecular repulsions the logarithmic derivative of f must be nu-
merically much larger than one (reflecting the fact that a lot is gained by not rearranging at
constant volume). Consequently, 1 and the inner contribution to the activation energy
is small. The shoving model is characterized by ignoring this contribution.
The next step is to determine the shoving work. What happens when a sphere in a
solid is enlarged? Many people answer that the surroundings must be compressed in the
process, but this is wrong. The deformation of the surroundings is a pure shear deformation.
In fact, this is a standard exercise in elasticity theory, where one finds [16] that the radial
displacement field varies as We know from Coulombs law that this field has zero
divergence. Since the divergence of the displacement field is the relative volume change,
there is no compression. This means that the relevant elastic constant for the shoving work
is the shear modulus. Now, the shear modulus of any liquid is zero at zero frequency, but
on short time scales the liquid behaves like a solid with shear modulus This quantity
is crucial to the shoving model, because the activation energy is proportional to It is
well-known that in viscous liquids increases strongly as temperature is lowered, and this
is now our explanation for the non-Arrhenius behavior [7,17]. The final expression for
is:
Here, V
c
is a proportionality constant with dimension volume. In comparing to experiment V
c
is always of order the volume of one molecule. The mathematical expression Eq. (6) may be
traced back to 1943 (see, e.g., Refs. 5, 7, and 17).
As a model for the non-Arrhenius behavior the shoving model is not fundamental;
it does not allow a calculation of the energy barrier because V
c
is unknown. Instead, the
model is phenomenological in the sense that it links the energy barrier to another macroscopic
physical quantity. In this respect the model is like the entropy model. The shoving model,
105
however, has the unique feature of linking short time dynamics with long time dynamics:
reflects short time dynamics and may be measured by a fast experiment. Note also that
may be calculated as a simple statistical mechanical canonical average (of the square of
the fluctuating shear stress), so if the model is correct a considerable simplification has been
achieved.
Before comparing the shoving model to experiment, we briefly summarize its basic
postulates:
The main contribution to the activation energy is elastic energy.
This elastic energy is located in the surroundings of the reorienting molecules.
The elastic energy is shear energy.
It is easy to compare the model to experiment. One simply measures [or equiv-
alently, and to check Eq. (6). It is convenient to use an Angell type plot for
this. Instead of having T
g
/T as x-axis, however, we use normalized to one at
T = T
g
. Figure 3 gives two examples. The data confirm the model. More measurements are
needed, however, before the model can be said to be well-established.
Since the shoving model predicts that the rate of relaxation depends on the instan-
taneous shear modulus, the model should apply even for nonlinear relaxations. This has
recently been tested in aging experiments on a silicone oil [19]. The liquid was subjected to
sudden temperature jumps from well annealed states, and was continuously monitored
as a probe of the structural relaxation. By using the Tool-Narayanaswami formalism with a
reduced time definition based on Eq. (6) it was shown that the model is able to fully account
for the aging observed.
A DIFFUSION-TYPE MODEL OF NON-DEBYE RELAXATIONS
We now proceed to discuss relaxation phenomena in viscous liquids. Only linear relax-
ations, i.e., those governing how infinitesimal perturbations decay to equilibrium, are consid-
ered. Moreover, the discussion focusses on the so-called a relaxation, the slowest and the
dominant relaxation process. The relaxation process defines the average relaxation time
which is linked to viscosity by Eq. (2). How does one construct a model exhibiting the kind
of asymmetric relaxation seen in Fig. 2? We look for a model based on the solidity of viscous
liquids, predicting a sharp cut-off at long relaxation time and an high frequency decay
of the loss.
Two points of views may be taken towards linear responses of viscous liquids: 1) There
are a number of different response functions, all of which are more or less of equal status,
probing different autocorrelation functions. Alternatively, 2) some response functions are
regarded as more fundamental than others. We adopt the latter viewpoint: The condition of
mechanical equilibrium is a zero-divergence condition referring to the stress tensor. This puts
stress tensor fluctuations in focus. Consequently, we shall regard the frequency-dependent
shear and bulk moduli as the two basic linear response function (the bulk modulus exists
in two versions, the adiabatic and the isothermal here only the isothermal bulk modulus is
considered). Other response functions should somehow be derived from the mechanical ones.
For example, a rotating sphere model for dielectric relaxation links the dielectric response to
the frequency-dependent viscosity (or shear modulus).
Consider the shear modulus. The -decay of the shear loss at high frequencies,
which is to be reproduced by our model, means that at high frequencies one has (in dimen-
sionless units): Since where is the
shear stress, we find that at short times the stress autocorrelation function is given (again in
dimensionless units) by
106
Figure 3. Comparing shoving model predictions to experiments on molecular liquids (reproduced from Ref.
17). (a) shows our own data, while (b) replots old data of Barlow and coworkers [18]. For both data sets the
actual temperature dependence of the viscosity is plotted just as in Fig. 1. Also, the viscosity is plotted as a
function of the variable normalized to one at T = T
g
. The diagonal line gives the predictions of
the shoving model, starting at high temperature at a physically reasonable prefactor (lower left corner).
The model we adopt is a diffusion model. The idea is that the stress averaged over some
small volume of the liquid (containing, e.g., 100 molecules) although it changes abruptly
whenever a flow event takes place changes by only a small amount. If so, it might be a good
guess to try to describe the process by a diffusion equation in stress space. This space has 6
coordinates, the 6 independent values of the stress tensor.
The diffusion constant must be thermally activated if relaxation is to become strongly
temperature dependent. We know that when a large shear stress is applied to a set of molecules
107
they are much more likely to rearrange than otherwise (this is the mechanism behind fracture
in its initial stage). Thus a large shear stress lowers the barrier, so the barrier must depend on
the stress state. This means that in stress space the diffusion constant depends exponentially
on the stress state. Denoting the 6 stress coordinates collectively by our first guess at a
diffusion equation for the probability is
Here the diffusion constant is given by where is the stress-
dependent activation energy. All stress states are regarded as equally likely, but the allowed
stress states define only a finite region of stress space. This region is bounded by the states
with Diffusion out of this region is forbidden; in the model this is ensured by a
suitable boundary condition.
This model cannot possibly reproduce Eq. (7), however. For any activation energy
that varies down to zero at large stresses, the model gives precisely the kind of very broad
loss peaks that are naively expected (and not found in experiments) when a whole range of
activation energies are involved. However, we have not yet taken solidity into account. How
does this change things? Because of the mechanical equilibrium condition the solidity
stress changes at one place cannot take place without minor stress changes throughout the
viscous liquid. Both before and after a flow event there is mechanical equilibrium, so the
stress change must also satisfy the zero-divergence condition. As shown in Refs. 13 and 14
this implies that the stress change in the surroundings varies as and is thus long ranged.
Locally, stress may change either as a result of a local flow event or as a result of flow
events in the surroundings. The latter changes are small, but quite common, and they are
independent of the actual stress state. Thus, we find that the stress diffusion constant has two
contributions and is given by
When a large range of activation energies is involved, depending on the actual value of
this expression is dominated by either the first term or the last term. In stress space there
is a rather sharp edge separating these two cases. On one side of this edge (the slow part
of stress space) is almost stress independent while on the other side of the edge
varies rapidly and is generally much larger than D (the fast part of stress space). The
linear size of the slow part of stress space, L, defines a characteristic time by As
becomes clear below, is the average relaxation time of the relaxation.
How does the model with Eq. (9) reproduce Eq. (7) and the asymmetry of the loss peak?
First, we note that there is a cut-off at long relaxation times: Relaxation towards equilibrium
can at most last the time it takes to diffuse across the slow part of stress space, The square
root time dependence of the initial stress decay comes about in the following way: Stress
states close to the edge, but in the slow part of stress space, preferably relax by moving to the
edge and crossing it to utilize the fact that motion is fast on the other side. The limiting factor
is the time it takes to move to the edge. For systems a distance d away from the edge this
time is given by t = d
2
/D. When d is small the number of states a distance less than d away
from the edge is proportional to d. Thus the number of states relaxing within time t varies as
This gives us Eq. (7).
It is interesting to note that for the model to work properly, there must be relaxations
much faster than those of the a relaxation. We tentatively identify these fast relaxations with
the Goldstein-Johari relaxation.
108
SUMMARY
The solidity of viscous liquids is the key to understanding the physics of these liquids.
Solidity is a direct consequence of the extremely high viscosity. Thus the similarity between
chemically quite different liquids becomes less surprising. If this viewpoint is correct, under-
standing the physics of viscous liquids may be simpler than has hitherto been recognized.
REFERENCES
1. Kauzmann, W. (1948) The nature of the glassy state and the behavior of liquids at low temperatures,
Chem. Rev. 43, 219-256.
2. Harrison, G. (1976) The dynamic properties of supercooled liquids, Academic Press, New York.
3. Brawer, S. (1985) Relaxation in viscous liquids and glasses, American Ceramic Society, Columbus, OH.
4. Angell, C.A. (1991) Relaxation in liquids, polymers and plastic crystals - strong/fragile patterns and
problems, J. Non-Cryst. Solids 131-133, 13-31.
5. Nemilov, S.V. (1995) Thermodynamic and kinetic aspects of the vitreous state, CRC, Boca Raton, FL.
6. Debenedetti, P.O. (1996) Metastable liquids: Concepts and Principles, Princeton University Press,
Princeton.
7. Dyre, J.C. (1998) Source of non-Arrhenius average relaxation time in glass-forming liquids, J. Non-Cryst.
Solids 235-257, 142-149.
8. Angell, C.A. (1985) Strong and fragile liquids, in K.L. Ngai and G. B. Wright (eds.) Relaxations in
complex systems, U.S. G.P.O., Washington, DC, pp. 3-11.
9. Olsen, N.B., Christensen, T., and Dyre, J.C. (2000) Time-temperature superposition in viscous liquids,
e-print cond-mat/0006165.
10. Voit, P., Tarjus, G., and Kivelson, D. (2000) A heterogeneous picture of relaxation for fragile super-
cooled liquids, J. Chem. Phys. 112, 10368-10378.
11. Dyre, J.C. (1987) Master-equation approach to the glass transition, Phys. Rev. Lett. 58, 792-795.
12. Schrder, T.B., Sastry, S., Dyre, J.C., and Glotzer, S.C. (2000) Crossover to potential energy landscape
dominated dynamics in a model glass-forming liquid, J. Chem. Phys. 112, 9834-9840.
13. Dyre, J.C. (1999) Solidity of viscous liquids, Phys. Rev. E 59, 2458-2459.
14. Dyre, J.C. (1999) Solidity of viscous liquids. II. Anisotropic flow events, Phys. Rev. E 59, 7243-7245.
15. Goldstein, M. (1969) Viscous liquids and the glass transition: A potential energy barrier picture, J. Chem.
Phys. 51, 3728-3739.
16. Landau, L.D., and Lifshitz, E.M. (1970) Theory of elasticity, Pergamon Press, Oxford.
17. Dyre, J.C., Olsen, N.B., and Christensen, T. (1996) Local elastic expansion model for viscous-flow acti-
vation energies of glass-forming molecular liquids, Phys. Rev. B 53, 2171-2174.
18. Barlow, A.J., Lamb, J., Matheson, A.J., Padmini, P.R.K.L., and Richter, J. (1967) Viscoelastic relaxation
of supercooled liquids. I Proc. Roy. Soc. A 298, 467- 480.
19. Olsen, N.B., Dyre, J.C., and Christensen, T. (1998) Structural relaxation monitored by instantaneous shear
modulus, Phys. Rev. Lett. 81, 1031-1033.
109
This page intentionally left blank
NON-ERGODIC DYNAMICS IN SUPERCOOLED LIQUIDS
M. Dzugutov, S. Simdyankin and F. Zetterling
Department of Numerical Analysis
Royal Institute of Technology
100 44 Stockholm, SWEDEN
INTRODUCTION
The nature of the glassy state and the liquid-glass transition admittedly remains the
most interesting unsolved problem of the solid state theory [1,2]. The interest is focused
on two universal features observed in the liquids which remain in a (possibly metastable)
equilibrium when approaching the glass transition. In fragile glass-formers [3], a super-
Arrhenius increase of the relaxation time is observed below a characteristic temperature T
A
,
leading to an apparent solidification at a finite temperature T
g
. Simultaneously, these liquids
exhibit another feature, related to the first one, the stretched-exponential decay of the time
correlation functions.
Laboratory glass transition is defined as a point where the relaxation time of a super-
cooled liquid exceeds the experimentally accessible observation time. A liquid cooled be-
low the glass transition temperature thus remains, on the experimental time-scale, in a non-
equilibrium state, being unable to comprehensively explore the relevant area of its configu-
rational space. Viewed from this point, T
g
is commonly regarded as a crossover temperature
separating the ergodic and non-ergodic regimes of the supercooled liquid dynamics.
It is clear that this definition is not helpful for understanding the possible transformation
in the behaviour of a liquid undergoing the glass transition. Indeed, the observation time
limit affordable for an experimentalist involved in the study of the glass transition is not
a relevant quantity for the description of relaxation processes in a supercooled liquid, and,
therefore, both the glass transition and the ergodicity breaking as defined in the above way
may be regarded as just artifacts of observation. In order to address this issue, we need an
operational definition of ergodicity breaking that would indicate a criterion for discerning a
possible change in the phase-space behaviour of a liquid in the the glass transition domain
from the macroscopic observations. Moreover, this criterion should be formulated in terms
of equilibrium liquid parameters.
The ergodic aspects of the glass-transition can be considered in the context of a long-
standing problem known as the Kauzmann paradox [4]. The latter is an observation that
the extrapolation of experimental results for the specific heat in supercooled liquids leads
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 111
to zero value of the configurational entropy at finite temperature. This can be interpreted
as indicating an underlying phase transition [5]. The existence, the nature, and the possible
location of such a transition are key issues for the problem of glass transition.
Two possible scenarios can be considered. The mode-coupling theory [6] interprets the
characteristic dynamics observed in the supercooled domain, and the eventual glass transition,
as a purely dynamical equilibrium phenomenon arising as a result of an inherent feedback
mechanism - the so-called cage effect whereby the diffusive dynamics is coupled with the
structural relaxation. Liquid dynamics remains in equilibrium as it slows down under cooling
to complete structural arrest which in the idealized version of the theory occurs at a critical
temperature T
c
. In this context, the condition of ergodic equilibrium implies that the local
structural fluctuations relax sufficiently fast so that the liquid remains spatially homogeneous
on the time scale relevant for the diffusion. The theory thus circumvent the problem of
Kauzmann paradox.
A number of experimental observations, however, indicate a different scenario. It was
found that, in the vicinity of T
g
, structural relaxations in a number of glass-formers become
significantly retarded as compared with the diffusive dynamics, breaking the Stokes-Einstein
relation. This observation was interpreted as indicating that the liquid decomposes into sub-
systems which relax at different rate. Moreover, the slow dynamical states were concluded
to be confined to structurally distinct spatial domains with the life-time much exceeding the
time-scale of the diffusive dynamics. Formation of these domains, conceivably confining an
energy favoured structure, may be regarded as an indication of an underlying phase transition
which is possibly masked by the onset of the laboratory glass transition.
On the other hand, the conjecture of a long-lived spatial heterogeneity in supercooled
liquids implies that, on the time scale characterizing the relaxation dynamics of the faster
subsystems, the liquid falls out of ergodic equilibrium already above T
g
. In terms of the
phase-space behaviour, this can be viewed as decomposition of the entire phase-space region
corresponding to the thermodynamic equilibrium (the region of motion) into distinct com-
ponents which are connected only on a time scale long as compared with the characteristic
equilibration time within a component.
To a considerable degree, the controversy that arises in discussing the ergodic aspects
of the glass transition stems from the fact that there is currently no uniformly accepted def-
inition of ergodicity. It is clear that a meaningful discussion of ergodicity breaking must
refer to some qualitative change in the phase-space behaviour. In the following, we shall dis-
cuss a conjecture relating the characteristic anomalies arising in a supercooled liquid to the
anomalies in its phase-space behaviour as well as the possible ways of detecting these latter
anomalies from finite-time measurements of macroscopically accessible quantities.
AN OPERATIONAL DEFINITION OF DYNAMICAL ERGODICITY
A statistical-mechanical system is regarded as ergodic if the time-average of a dynam-
ical variable converges to its ensemble average as defined with respect of an invariant
measure
The invariant measure here specifies the region of motion - the subspace to which the system
is confined by the macroscopic constraints.
A major problem involved in the above definition of ergodicity is that it refers to an
ensemble of identical systems and infinite observation time, whereas we, in fact, need to
assess the ergodicity of a single system within finite time. Moreover, the definition (1) is
usually interpreted in the spirit of the original Boltzmanns ideas asserting that, to attain an
112
ergodic equilibrium, the systems phase trajectory has to cover with sufficient density the
whole relevant phase-space region. This picture of the ergodicity restoring relaxation process
is obviously wrong, as far as liquids are concerned. Indeed, the number of configurations the
phase trajectory has to explore, and, consequently, the conjectured relaxation time, increase
exponentially with the number of particles, whereas the relaxation times in real liquids are
not size-dependent [12].
This problem can be resolved if we redefine the concept of statistical ensemble. We can
do it using the fact that the entropy S of an equilibrium system of N particles is an exten-
sive quantity: The extensivity implies that there exist a finite correlation length
and a finite correlation time [12], such that regions separated by distances exceeding
evolve independently and produce statistically indistinguishable time averages for time in-
tervals exceeding An equilibrium liquid system can thus be viewed as an ensemble of
independently evolving subsystems which are confined to regions separated in space by the
distances exceeding
In dealing with systems of particles, it appears reasonable to consider effective ergodic-
ity based on the concept of mixing [12,13]. The latter requires that the measure of the points
of a region R of the phase space which happened to be in any other region R after sufficiently
long time t must be proportional to the volumes of these regions:
If a dynamical system satisfies this condition, its phase trajectory uniformly samples the
coarse-grained phase space. Notice that although the property of mixing is sufficient to ensure
ergodicity, it is not known whether this is also a necessary condition for ergodic behaviour
[12,13].
The approach to effective ergodicity can be monitored by a measure based on the idea
of statistical symmetry. The latter means that time-averages of the quantities associated with
independently evolving regions of a system (or its constituent particles) must become statisti-
cally indistinguishable when approaching ergodic equilibrium; this is an obvious result of the
independence principle. If f
i
is a quantity associated with particle i, the respective measure
of ergodic convergence for a system of N identical particles is defined as [14]
where is the ensemble average (average over all the particles of the system).
In the domain of Arrhenius liquid dynamics above T
A
, this measure, defined for the par-
ticle energy e, was found [14] to decay with time universally as By contrast, in
the glass-transition domain it asymptotically approaches a non-zero limit indicating ergod-
icity breaking. Thus, using the measure defined by Eq. 1 we can detect the divergence of
the structural relaxation time-scale in a strongly supercooled liquid which is an indicator of
the ergodicity breaking. Although such microscopic diagnostics can be useful in the analysis
of computer simulations, its major drawback is that it requires full information on the phase
trajectory which is not available in macroscopic experiments.
The mixing behaviour is realized for the systems with the property that two phase tra-
jectories, initially infinitely close, diverge exponentially with time [13,15]. For each phase
space coordinate x
i
, this divergence is quantified by the respective Lyapunov exponent
Notice that the connection between the exponential chaos and mixing, al-
though intuitively clear, has been rigorously proved only for a few simple cases, like the gas
of hard spheres [16]. Numerical simulations appear to be a natural way to extend these results
to real systems.
We discuss here a conjecture [17] that ergodicity of a dynamical system should be under-
stood as global chaotic connectivity of its region of motion. If exponential chaos is confined to
113
subregions connected only on a sufficiently long time-scale, two trajectories, arbitrary close
but belonging to different regions of chaotic behaviour, do not diverge exponentially. Obvi-
ously, such decomposition of a single chaotic domain results in (i) separation of the relaxation
time-scales and (ii) slowing down the relaxation dynamics. In the following, we present ar-
guments, supported by evidence from simulation, indicating that the discrepancy between the
volume of a single stochasticity region and the total volume of the region of motion can be
assessed by exploiting a recently found universal relation between the diffusion coefficient
and the entropy [18].
ERGODIC DIFFUSION
The process of diffusion in liquids is controlled by the so-called cage effect whereby
the diffusive motion of a particle surrounded by a cage of its neighbours is coupled with the
relaxation of the local structural environment [19,20]. If a liquid is regarded as an ensemble
of independently relaxing regions as we pointed out above, the rate of diffusion is determined
by the frequency at which these regions change their configurations.
The phase-space picture of structural relaxation in the liquid state can be conveniently
discussed in terms of the energy landscape paradigm [2]. The topography of the energy
landscape can be probed by mapping an instantaneous configuration of onto a local poten-
tial energy minimum by the steepest descent minimization [21]. The process of structural
evolution of an independently relaxing region in a liquid can be viewed as a sequence of
transitions between adjacent energy minima. Each point of the ensemble representing these
regions explores adjacent minima positions by performing random trial moves at a rate which
is determined by the general rate of the momentum and energy transfer. We assume that the
probability that a random move in the configurational space results in a successful transi-
tion to a new minimum is proportional to the number of adjacent minima which, in its turn,
is proportional to the total number of available minima. The latter scales with the thermo-
dynamic excess entropy as The excess entropy represents the difference between full
thermodynamic entropy and that of the perfect gas at the same conditions: S
ex
= S S
PG
[22].
It has been recognized by Enskog [23] that the momentum and energy transfer in a dense
hard-sphere fluid is mediated by binary collisions. The collision rate thus provides a
natural time-scale for the dynamics. It can be assessed from the value of the radial distribution
function, g(r), at the collision distance [23]:
where m is the particle mass, is the hard-sphere diameter and T is the temperature. At
the same time, the hard-sphere diameter is a natural unit of length. If we express the
diffusion coefficient in dimensionless form using these units, the diffusion coefficient in the
hard-sphere liquid can be expresses, according to the conjecture suggested above, can be
expressed as follows:
with the empirically found value of the scaling factor A = 0.049.
In real liquids, the hard-sphere diameter can be replaced by the position of the first
peak of g(r), which allows the calculation of In this way, we can express D in different
liquids in terms of universal units of time and length. Another approximation concerns cal-
culation of S
ex
. The latter can be expressed as an expansion in terms of n-particle correlation
functions [24]. In two-particle approximation, this gives:
114
It was found [18] that, with the use of these approximations, eq. (6) universally describes the
relation between S and D in a wide range of simple liquid systems.
The scaling relationship between the thermodynamic entropy and diffusion rate is based
on the assumption that the system explores its phase space at a rate proportional to The
rate at which the state of an exponentially chaotic system is delocalized in the phase space
can be quantified by the Kolmogorov-Sinai entropy [13]. This quantity, according to Pesins
theorem [25], can be presented as the sum of all positive Lyapunov exponents:
Thus, the scaling relation between D and S
ex
implies the existence of a respective re-
lation between h
KS
and S
ex
. Recently, it was indeed found [26] that the Kolmogorov-Sinai
entropy in simple liquids, expressed in terms of is uniquely and universally related to
the thermodynamic entropy. This implies that the collision frequency represents a univer-
sal time-scale for the liquid dynamics. Another conclusion, important in the context of the
present discussion, is that the observed relation between D and S
ex
can be regarded as an
indicator of exponentially chaotic behaviour of a liquid.
Eq. (6) can be compared with other theories of liquid dynamics involving entropy. In the
model of Adam and Gibbs [27] the central quantity is the minimum size of the cooperativity
rearranging region the configuration of which can be changed without interfering with the
environment. The model conjectures that the the relaxation time and the configurational
entropy S
c
are relates as:
Another model [28], suggested by Rosenfeld, relates the entropy and diffusion in liquids
in a different way:
Recently, Di Marzio [29] proposed a relation between the relaxation time and the configura-
tional free energy F
c
:
Notice that parameter A in Eq. (9) is the number of particles participating in a single
jump. If that relation is interpreted in terms of hard spheres where with an
additional assumption that A = 1 it becomes clearly consistent with (6). In that case, B would
be expressed in terms of the collision frequency.
An obvious advantage of relation (6) as compared with (8), (9) and (10) is that it does
not involve free parameters. Therefore, a deviation from it is an unambiguous signatuture of
a break down in the fundamental mechanism postulated for liquid diffusion which in the case
of other relations may be masked by fitting.
NON-ERGODIC DIFFUSION
A central postulate in the model of liquid diffusion in ergodic domain presented in the
previous section is that the probability of transition from a current energy minimum config-
uration to an adjacent one is entirely determined by the total number of these configurations
which can be expressed through the excess entropy. This postulate implicitly assumes that all
adjacent configurations available are accessible within the same time-scale characterizing the
local dynamics. The universal relation between the h
KS
and S
ex
also implies the validity of
this assumption. We now consider a situation where this assumption is not valid, and analyze
its possible consequences for the systems behaviour in the phase space. A connection will
be shown between this behaviour and the non-ergodic dynamics as it was defined in above,
115
Figure 1. A simple model demonstrating the impact of the valley structure of the phase-space on the relax-
ation dynamics. Energy barriers separating the valleys are depicted by solid lines. Squares and circles denote
configurations belonging to different valleys. The crossover point between the valleys is marked by the cross.
and its possible diagnostics in terms of the macroscopically measurable parameters will be
discussed.
It was found that the average energy of the potential energy minima which remains con-
stant above T
A
drops rapidly as the temperature decreases below T
A
[30]. This indicates that
in the supercooled domain, the liquid resides on a different part of the energy landscape than
the normal liquid, predominantly staying in deep valleys connected by narrow bottlenecks.
These connections are effectively used only on the long time-scale, while the short time-scale
dynamics of a supercooled liquid unfolds in a limited subregion of the total region of motion.
In order to illustrate the impact of such strongly profiled landscape on the liquid relax-
ation dynamics, we consider a simple model sketched in Fig. 1. The whole set of config-
urations comprising the region of motion is divided in two components (valleys), depicted
by squares and circles. The filled symbols denote the configurations which are energetically
forbidden. The components are separated by barriers, indicated by solid lines, and the single
connecting pass is marked by the cross. Consider the probability w
ij
of transition between
the configurations i and j. At high temperature, where the components separation is not rele-
vant, the average transition probability is entirely determined by the average probability that
the destination configuration is allowed, and, in this way, by the entropy:
Clearly, this leads to the relation (6) for the diffusion rate.
In the case when the dynamics unfolds in the valley landscape, the average probability of
transition can be estimated as where is the average probability that both
i and j belong to the same valley. Obviously, in this case and, therefore the the
diffusion rate is expected to show a negative deviation from the relation (6). This deviation,
indicating that the liquid dynamics is not any more related to the static properties, can be used
as a macroscopic diagnostics of the onset of the supercooled (non-ergodic) regime.
116
Figure 2. Reduced diffusion coefficient in the binary mixture of hard spheres as a function of packing fraction
(dots). Dashed line corresponds to the universal scaling law (6) relating the diffusion coefficient to the excess
entropy, in the pair approximation, S
2
, with A = 0.049.
EVIDENCE FROM MOLECULAR DYNAMICS SIMULATION
In order to test the above conjecture, we investigated a two-component hard-sphere liq-
uid simulated by molecular dynamics. The model consisted of 862 particles. The two species
of hard spheres comprising the model (A and B) are characterized by the ratio of diameters
and the ratio of the number densities The liquid phase of
this system was found [31] to lose its thermodynamic stability when compressed beyond the
critical value of packing fraction At higher densities, its stable phase was identified
as the AB
13
crystal, the unit cell of which includes 112 atoms.
Due to the complexity of its crystallization pattern, this system possesses a pronounced
glass-forming ability. In this simulation, it was found to remain in long-living metastable
equilibrium liquid state when compressed beyond the indicated critical packing fraction value.
The absence of crystalline nucleation was thoroughly verified by monitoring the pressure and
the diffusion coefficient, both of which remained constant during the simulation run.
In order to test relation (6) we calculated the diffusion coefficient and the excess entropy,
in the pair approximation, S
2
for the smaller atomic species B, exploring a wide range of
both below and above the critical value. S
2
was derived using (6) from two partial radial
distribution functions g
AB
(r) and g
BB
(r). The results of this simulation are presented in Fig.
2. It is clear that the results agree well with relation (6) for which corresponds
to the stable liquid domain. For higher values of the packing fraction corresponding to the
metastable liquid domain, the diffusion coefficient demonstrates a negative deviation from
the prescription of the scaling relation (6) which increases rapidly with increasing
Another simulation demonstrating effects of non-ergodic diffusion that we consider here
explores a simple monatomic liquid [32] with predominantly icosahedral short-range order
induced by a properly constructed pair potential. Because of its structure, this liquid can,
under supercooling, avoid crystallization and remain in a state of metastable equilibrium for
a time sufficiently long for observation of its essential dynamics [33]. It was found that the
117
Figure 3. Reduced diffusion coefficient in a simple monatomic glass-forming liquid [32] as a function of
temperature. Dashed line corresponds to the universal scaling law (6) relating the diffusion coefficient to the
excess entropy, in the pair approximation, S
2
, with A = 0.049. In this liquid T
A
= 0.6
Figure 4. Time evolution of energy-based measure of ergodicity as defined by Eq. (2) calculated for the glass-
forming liquid [32] shown in Fig. 3 for three temperatures indicated in the inset.
118
domain of temperatures where the liquid exhibits the characteristic anomalies of supercooled
dynamics regime is bounded by T
A
= 0.6 [33]. In Fig. 3 we present its diffusion coefficient
plotted as a function of temperature in a manner convenient for testing (6). As the liquid is
cooled below T
A
, the diffusion coefficient displays a pronounced deviation from the scaling
behaviour with respect to the excess entropy as postulated by relation (6).
OTHER EFFECTS OF ERGODICITY BREAKING
The latter model of the two discussed in the previous section can also be exploited to
directly show the separation of the relaxation time scales that we associate with non-ergodic
dynamics. Based on the microscopic information provided by molecular dynamics, the rate
of the ergodicity restoring structural relaxations can be assessed using the ergodicity metric
introduced by Eq. (3). Fig. 4 shows the evolution of this quantity as a function of the average
mean square displacement at different stages of cooling. At sufficiently high temperatures,
the results demonstrate an apparent universality in the relationship between the structural
relaxation and diffusion. In the supercooled dynamics domain below T
A
, this universality
breaks down as the structural relaxation gets strongly retarded as compared with diffusion.
This observation, together with the results shown in Fig. 3, explicitly confirm our analysis
that conjectures non-ergodicity of the liquid diffusion below T
A
.
The long-time decomposition of the region of motion can be viewed as reduction of the
time-dependent effective entropy which measures the volume of the accessible phase-space
region, as compared with the thermodynamic entropy. On the other hand, the entropy, can,
as was pointed out above, be related to the correlation length. Therefore, on the time scale
characteristic of the fast (itracomponent) relaxation, the liquid is expected to exhibit time-
limited correlations the length scale of which would exceed the range of the static structural
correlations which correspond to the ensemble average (or infinite time average). Therefore,
the separation of time-scales in the relaxation dynamics of a supercooled liquid that we regard
as a signature of dynamical non-ergodicity, suggests the existence of large-scale clusters with
the life time exceeding the time-scale of the fast relaxation processes.
These time-limited clusters are conjectured to confine an energy favoured local structure
incompatible with periodicity [9], and, therefore, their growth is limited by geometric frus-
tration. In the case of the molecular dynamics simulation presented in Fig. 3, this structure
corresponds to icosahedra packing. To detect such clusters, we looked for extended structures
comprising connected icosahedra. Two icosahedra were regarded as connected if they shared
at least three atoms. The results of this analysis are presented in Fig. 4 which depicts the
largest clusters of interconnected icosahedra found in the liquid at different stages of cooling.
One can indeed see a rapid increase of the cluster size as the temperature decreases below
T
A
. Development of domain structure in supercooled liquids is consistent with a recent ob-
servation of long-range cooperativity in diffusive dynamics [34]. We also remark that the
development of a network of icosahedral clusters that is observed here resembles the picture
of bonding network that was discussed in the context of chalcogenide glass-formers [35].
Experimentally, the characteristic length of the described cooperative effects can be di-
rectly probed by measuring the rate of liquid dynamics in confined geometries [36], It has to
be emphasized that the time-limited cooperativity is decoupled from the static structure and,
therefore, from the thermodynamic entropy. Attempts to interpret the cooperative dynamics
of supercooled liquids in terms of the Adam-Gibbs theory which refers to the (static) config-
urational entropy, although common, are logically incorrect. In order to detect the growth of
time-limited domain structure in supercooled liquids, one has to compare the cooperativity
range derived from the dynamical measurements in confined geometries with the range of the
static structural correlations which, at least for quasi-simple liquids, can be assessed from the
diffraction measurements. Ergodicity breaking implies that the former considerably exceeds
119
Figure 5. Maximum size icosahedrally structured clusters observed in the glass-forming liquid [32] presented
in Figs. 3,4 at different temperatures: (a) T = 1.0, (b) T = 0.5, (c) T = 0.45
the latter.
DISCUSSION
We considered here the concept of dynamical ergodicity in the context of liquid dy-
namics. According to this concept, ergodic behaviour of a liquid is identified with globally
connected chaotic behaviour in the phase space. Macroscopically, this behaviour is charac-
terised by a single time-scale in the relaxation dynamics. It is shown that the proposed sce-
nario of ergodic phase-space behaviour of a liquid is inherently consistent with the recently
observed universal relationship between the dynamic properties, like transport coefficients or
the KS entropy, and the static properties as quantified by the thermodynamic entropy. On the
other hand, decomposition of the phase space into separate regions of connected stochasticity
results in long-time confinement of a phase trajectory. This leads to the appearance of char-
acteristic anomalies, like separation of the time scales and spatial heterogeneity, which are
commonly associated with the behaviour of a supercooled liquid.
An important conclusion of the discussion presented here is that the long-time decom-
120
position of the phase space into components developing in the supercooled liquid domain
breaks any conceivable relation between the rate of liquid dynamics and the thermodynamic
entropy. In particular, this conclusion concerns the Adam-Gibbs relation which is commonly
used for interpreting the behaviour of supercooled liquids approaching glass transition.
ACKNOWLEDGEMENT
This work was supported by the following Swedish agencies: Natural Science Research
foundations (NFR), Technical Research Foundation (TFR) and National Network for Applied
Mathematics (NTM). We used the graphics software from ref. [37].
REFERENCES
1. Jackle, J. (1986) Models of the glass transition Rep. Prog. Phys., 49, 171-231
2. Ediger, M.D., Angell, C.A. & Nagel, S.R. (1996) Supercooled liquids and glasses J. Phys. Chem. 100,
13200-13212
3. Angell, C. A. (1991) Relaxation in liquids, polymers and plastic crystals - strong/fragile patterns and
problems, Journ. of Non-Cryst. Solids, 131-133, 13 (1991)
4. Angell, C. A. (1997) Entropy and fragility in supercooled liquids Journ. of Res. of the Nat. Inst. of Stand.
and Techn 102 171-185
5. Kirkpatrick, T. R, Thirumalai, D. & Wolynes, P. G. (1989) Scaling concepts for the dynamics of viscous-
liquids near an ideal glassy state. Phys. Rev. A 40, 1045-1054
6. Gtze, W., and Sjgren, L. (1992) Relaxation processes in supercooled liquids, Reports on Progress in
Physics, 55, 241-295
7. Blackburn, F. R., Wang, C.-Y. & Ediger, M. D. (1996) Translational and rotational motion of probes in
supercooled l,3,5-tris(naphthyl)benzene. J. Phys. Chem. 100, 18249-18257
8. Sillescu, H. (1999) Heterogeneity at the glass transition: a review. J. Non-Cryst. Solids 243, 81-108
9. Kivelson, D., Kivelson, S. A., Zhao, X., Nussinov, Z. & Tarjus, G. (1995) A thermodynamic theory of
supercooled liquids. Physica A 219, 27-38
10. Fischer, E. W. (1993) Light scattering and dielectric studies on glass forming liquids. Physica A 201,
183-206
11. Palmer, R. G. (1982) Broken ergodicity Adv. in Phys. 31, 669-735
12. Ma, S.-K., (1996) Statistical mechanics, World Scientific, Singapore (1985)
13. Lichtenberg, A. J. and Lieberman, M. A. (1983) Regular and Stochastic Motion, Springer Verlag, NY
14. Mountain, R. D. and Thirumalai, D. (1989) Measures of effective ergodic convergence in liquids Journ.
of Phys. Chem, 93, 6975
15. Krylov, N. S. Works on the Foundations of Statistical Physics, Princeton Series in Physics, Princeton
1979; see also Sinai, Ya. G. Development of Krylovs Ideas, pp. 239-281 of the same volume.
16. Sinai, Ya. G. (1966) Izv. Akad. Nauk SSSR. Mt 30, 15-32 (in Russian)
17. Dzugutov, M., (1996) Dynamical diagnostics of ergodicity breaking in supercooled liquids J. Phys. Cond.
Matter, 11, 253-259
18. Dzugutov, M., (1996) A universal scaling law for atomic diffusion in condensed matter Nature, 381,
137-139
19. J. P. Boon and S. Yip, (1980) Molecular Hydrodynamics, McGraw-Hill, New York
20. Cohen, E. D. G. (1993) Fifty years of kinetic theory Physica A, 194 , 229-257
21. Stillinger, F. H. & Weber, T. A. (1984) Packing structures and transitions in liquids and solids. Science
225, 983-989
21. Hansen, J. P. and McDonald, I. (1976) Theory of Simple Liquids, Academic Press, London
23. S. Chapman and T. G. Cowling, (1939) The mathematical theory of non-uniform gases, Cambridge Uni-
versity Press
24. Mountain, R. D. & Raveche, H., (1971) Entropy and correlation functions in open systems. II Two- and
three-body correlations Journ. Chem. Phys. 35, 2250-2255
25. Pesin, Ja. B. (1976) Lyapunov characteristic exponents and ergodic properties of smooth dynamical
systems with an invariant measure Sov. Math. Dokl., 17, 196-203 (in Russian)
26. Dzugutov, M., Aurell, E., and Vulpiani, A., (1998) A universal relation between the Kolmogorov-Sinai
entropy and the thermodynamic entropy in simple liquids Phys. Rev. Lett. 81, 1762
121
27. Adam, G. and Gibbs, J.H., (1965) On the temperature dependence of cooperative relaxation properties in
glass-forming liquids Journ. Chem. Phys. 43, 139-146
28. Rosenfeld, Ya., (1977) Relation between the transport coefficients and the internal entropy of simple
systems Phys. Rev. A, 15, 2545-2549
29. Di Marzio, E. A. and Yang, A. J. M. (1997) Configurational entropy approach to kinetics of glasses,
Journ. of Res. of the Nat. Inst. of Stand. and Techn 102, 135-157
30. Sastry, S., Debenedetti, P., and Stillinger F. H. (1998) Signatures of distinct dynamical regimes in the
energy landscape of a glass forming liquid Nature, 393, 554-557; Angell, C. A. Liquid landscape ibid.,
521-524
31. Eldridge, M. D., Madden, P. A., and Frenkel, D. (1993) Entropy driven formation of a superlattice in a
hard sphere binary mixture Nature, 365, 35
32. Dzugutov, M. (1992) Glass formation in a simple monatomic liquid with icosahedral inherent local order.
Phys. Rev. A 46, R2984-R2987
33. Dzugutov, M. (1994) Hopping diffusion as a mechanism of relaxation stretching in a stable simple
monatomic liquid. Europhys. Lett. 26, 533-538
34. Donati, C., Douglas, J. F., Kob, W., Plimpton, S.J., Poole, P.H. & Glotzer, S.C. (1998) Stringlike cooper-
ative motion in a supercooled liquid. Phys. Rev. Lett. 80, 2338-2341
35. Phillips, J.C., and Thorpe, M.F.,. (1985) Constraints theory, vector percolation and glass formation,
Dynamics of glass-forming materials confined in thin films Sol. St. Comm. 53, 699-702
36. Jrme, B. (1999) Dynamics of glass-forming materials confined in thin films Journ. Phys. Cond. Matter
11, 189-199
37. Humphrey, W., Dalke, A., and Schulten, K., (1996) VMD - visual molecular dynamics, Molecular Graph-
ics 14, 33-38
122
NETWORK STIFFENING AND CHEMICAL ORDERING IN CHALCOGENIDE
GLASSES: COMPOSITIONAL TRENDS OF T
g
IN RELATION TO
STRUCTURAL INFORMATION FROM SOLID AND LIQUID STATE NMR
CARSTEN ROSENHAHN, SOPHIA HAYES,
GUNTHERBRUNKLAUS,
and HELLMUT ECKERT
Institut fr Physikalische Chemie;
Westflische Wilhelms-Universitt Mnster,
7, D-48149 Mnster, Germany
INTRODUCTION
During the past few years, there has been a resurgence of the interest in non-
oxide chalcogenide glasses based on the sulfides and selenides of the group 13-15
elements. From a technological point of view, arsenic and germanium sulfide-based
glasses in particular are promising low-phonon host materials for luminescent rare-earth
dopants, with potential applications in the fiber optic laser industry [1-4], From a
scientific point of view, non-oxide chalcogenide glasses represent interesting model
systems for testing simple mean-field concepts used to discuss structure-property
relations. It has been pointed out that certain physical characteristics of chalcogenide
glasses, such as glass transition temperatures, melt viscosities, and molar volumes often
depend non-linearly on their chemical composition [5-9]. An explanation has been
offered on the basis of percolation theory, by considering the overall connectivity of the
network [8,10-12]. For any two- or three-dimensional network one can define an
average coordination number
where n
i
is the number of atoms of species i, r
i
is the number of bonds formed by it, and
the summation extends over all of the different atomic species present in the network.
Any covalent network constrained by bond-stretching and bond bending forces
possesses a critical connectivity threshold at Glasses having average
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 123
coordination numbers higher than this threshold value represent single infinite clusters
with no residual zero-frequency modes [12-16]. Accordingly, abrupt changes in physical
properties are expected at such critical compositions. In this regard, much discussion has
been focused on the dependence of the glass transition temperature on composition [17-
23]. While universal dependences of T
g
on <r> have been found in certain binary and
ternary systems, there are numerous other glass systems in which no such universality
has been observed.
The applicability of the average coordination number concept implies that bond
formation proceeds in a random fashion, i.e. without any chemical bond preferences.
Thus, the concept works at its best when homo- and heteropolar bond energies are of
equal magnitudes. In chalcogenide glasses, however, this situation is rarely encountered.
Indeed, a considerable body of evidence proves that this situation is unrealistic in many
arsenic and germanium-based chalcogenide glasses, which have rather been found
strongly chemically ordered. Because of significant differences in the corresponding
bonding energies, heteropolar As-Se and Ge-Se bonds dominate over homopolar As-As,
Ge-Ge or weakly polar As-Ge bonds in these systems. Tichy and Ticha were the first to
recognize that in such chemically ordered systems the T
g
value is well correlated to the
overall mean bonding energy <E> [24,25]. Thus, maximum T
g
values have been
predicted (and found) at the composition of a chemical threshold, defined by R = 2z/(4x
+ 3y), where x, y, and z refer to the atomic fractions of Ge, As, and Se, respectively
[26]. In systems with a strong dominance of heteropolar over homopolar bonds, the
chemical threshold R = 1.0 marks the minimum selenium content at which a chemically
ordered network is possible without metal-metal bond formation.
Driven by this interest, from both the technological and the scientific point of
view, more detailed fundamental concepts concerning the local structure and
intermediate range order of these materials are being developed. Amongst other
techniques, solid state nuclear magnetic resonance (NMR) has played an important role
in this endeavour. In particular, the favorable NMR properties of the
31
P and
77
Se
isotopes have been exploited to advantage for a structural determination of binary and
ternary phosphorus sulfide, selenide, and telluride materials [27]. In the solid state,
magic angle spinning
31
P NMR spectra have provided information about the
microstructural units present, while dipolar spin echo decay methods have given
information about the extent of P-P bonding. Complementary high-temperature liquid
state NMR studies of the glassforming compositions have provided important insights
into the chemical equilibria and the kinetics leading to glass formation [28]. These
studies have been recently extended to arsenic and germanium selenide glasses [29]. It is
the purpose of this contribution to review the current state of this field, and to discuss
physical and chemical threshold effects on the compositional dependence of T
g
in
relation to the short range order in various binary and ternary chalcogenide glass
systems. Based on this structural information we will show how the concepts of physical
and chemical threshold behaviors can be unified within a more comprehensive model
providing satisfactory correlations between elemental composition and macroscopic
physical properties.
124
BINARY PHOSPHORUS-CHALCOGENIDE GLASS SYSTEMS.
Short Range Order
Both the binary systems P-S and P-Se have a strong tendency for glass formation,
extending until about 50 at.% phosphorus [6,9,30], In spite of this similarity, the
principles governing the structural organization of these glasses are fundamentally
different from each other. This difference becomes immediately obvious when comparing
the compositional dependences of the glass transition temperatures T
g
(see Figure 1).
Figure 1: Compositional dependence of the glass transition temperature T
g
in P-S and P-Se glasses. The
curve is drawn as a guide to the eye.
In P-Se glasses, an increase in the phosphorus concentration is accompanied by a
T
g
increase up to about 40 at.% P, followed by an increase at a greater rate at higher
phosphorus contents. This behavior can be understood on the basis of percolation theory
as applied to a polymeric network: the composition 40 at.% P corresponds to the rigidity
percolation threshold, at which the average coordination number <r> equals 2.4 bonds
per atom. Such a change in slope, dT
g
/d<r>, occurring at <r> = 2.4 has been observed
for various other binary and ternary chalcogenide glass systems, and it appears to be a
general property of polymeric covalent networks.
In contrast, the binary P-S glasses have a decidely different compositional trend
[27]. At smaller phosphorus contents the T
g
values are larger than in the P-Se glasses.
There is a maximum near 25 at.% P, followed by a decrease at larger phosphorus
contents. For glasses containing more than 40 at.% phosphorus the glass transition
temperatures lie below room temperature. For this glass system, the rigidity percolation
concept seems to be inapplicable.
31
P nuclear magnetic resonance studies have given substantial insights into the
structural origins of these differences in behavior. Figure 2 compares the
31
P MAS-NMR
spectra of both systems. On the right, spectra of P-S glasses are displayed. Up to P
contents of 15 at. % the spectra are dominated by a broad resonance centered near +110
ppm, which can be assigned to S=PS
3/2
units. As the P content increases above 15 at.%,
novel sharp resonances appear, signifying the formation of P
4
S
10
, P
4
S
9
and P
4
S
7
molecules embedded in the polymeric matrix [27,31]. The sharpness of these resonances
125
indicates efficient averaging of the anisotropic interactions on the NMR timescale by
virtue of fast molecular reorientation processes. At P concentrations above 30 mole% the
glass can be considered an assembly of molecular (zero dimensional) P
4
S
n
units
As these molecules make no contribution to network connectivity, a dramatic
decrease of T
g
is observed for P-S glasses in this concentration region. Figure 2, left
shows the
31
P MAS-NMR spectra of the P-Se glasses. The rather broad resonances
apparent in these spectra reveal a wide chemical shift distribution as is typical for
polymeric chalcogenide networks. The two peaks at 130 and zero ppm at low P content
have been assigned to three-coordinate PSe
3/2
and four-coordinate Se=PSe
3/2
groups,
respectively [32]. At higher P contents, a gradual shift of the high-frequency peak
signifies the appearance of other types of three-coordinate phosphorus species.
Figure 2:
31
P MAS-NMR spectra of glasses in the systems P-S (left) and P-Se (right)
Complementary spin-echo NMR experiments reveal that P-P bonds make an
increasing contribution to this resonance as the phosphorus content increases above 25
at% [33]. Altogether the structure can be described in terms of four different local units,
namely Se=PSe
3/2
, PSe
3/2
, Se
2/2
P-PSe
2/2
, and (Se-Se)
2/2
fragments. Figure 3 summarizes
the results obtained by this analysis [33]. They indicate that in the local structure of these
glasses there is competition between heteropolar P-Se bonding and the formation of
homopolar P-P and Se-Se bonds, respectively. While a number of P-P bonds are being
formed even at the smaller phosphorus concentrations, nevertheless their fraction
remains sub-statistical within the entire region of glass formation. On the other hand the
preference of heteropolar bond formation is not nearly as large as in arsenic or
germanium selenide glasses. On the basis of this structural information the characteristic
dependence of T
g
on P content in the P-Se glasses can now easily be understood: with
their efficient competition and balance of homo- and heteropolar bonds these glasses are
the most ideal model systems for mean-field theory: the composition 40 at% phosphorus
results in an average coordination number of 2.4, corresponding to the Phillips-Thorpe
rigidity percolation threshold.
126
Figure 3: Site speciation in P-Se glasses as deduced from the combined analysis of
31
P spin echo and
MAS NMR. A: Se-Se bonds, B: Se=PSe
3/2
groups, C: PSe
3/2
groups, D: P-P bonded units.
Medium-Range Order
Important complementary information about how these local units are
interconnected was obtained from
77
Se NMR studies [34,35]. In principle, several
medium-range order scenarios are possible: The P-bearing units could be homogeneously
dispersed (isolation model), they could be clustered together in domains, or all of the
units could be randomly linked. These interlinkage scenarios make different predictions
for the fraction f
Se-Se
of those Se atoms that are only bonded to other selenium atoms.
Experimental values of f
Se-Se
can be estimated from an analysis of the
77
Se chemical shift
measured in the liquid state at temperatures above T
g
[35]. Figure 4a shows a typical
experimental data set of temperature dependent
77
Se NMR spectra for a glass containing
20 at.% P. Above T
g
, the P-bonded and non-P-bonded selenium atoms show initially
distinct resonances, which are subsequently motionally narrowed and affected by
chemical exchange as the temperature is increased. Above 210 C the resonances
collapse into a single peak which narrows continuously as the fast-exchange limit is
approached at higher temperatures. A detailed inspection of the average chemical shift as
a function of composition (Figure 4b) has shown that the data can be analyzed
straightforwardly according to the expression:
where and refer to the chemical shift of Se-only and P-bonded Se atoms, f
Se-Se
estimates are also available for the glassy state, by differentiating between P-bonded and
non-P-bonded selenium species based on their differences in the magnitude of the
31
P-
77
Se magnetic dipole-dipole couplings [34]. As shown in Figure 5, the experimental data
obtained by both methods are consistent with a random linkage scenario. This
consistency is the result actually expected for a polymerized glass structure.
127
Furthermore, the experimental data rule out a previously proposed model [36] based on
phosphorus-rich P
4
Se
n
clusters (n= 3,4,5) in a selenium-rich matrix.
Figure 4a: Temperature dependent
77
Se NMR spectra of a P-Se melt containing 20 at.% Phosphorus.
Figure 4b: (left) Temperature dependence of the 77Se chemical shifts in P-Se melts in the fast exchange
limit. Reproduced from reference 35.
Figure 5. (right) Fraction of Se-only bonded selenium species extracted from liquid state
77
Se NMR and
from
77
Se-
31
P spin echo double resonance NMR (SEDOR). Predictions from various intermediate-range
order scenarios are shown for comparison. Reproduced from reference [35].
128
BINARY ARSENIC SELENIDE GLASSES
Chemical Bond Distribution
Figure 6 juxtaposes the compositional dependence of T
g
for the binary system P-
Se and As-Se, respectively. Again, in spite of the chemical homology, very different
trends are observed: In both systems the composition of 40 at % pnictogen (<r>=2.4,
R=l) appears to be of special significance. The T
g
maximum observed in the As-Se
system indicates the dominance of chemical threshold behaviour at R=l whereas the
steep upturn in the P-Se system, signifies rigidity percolation for <r>=2.4. In view of the
NMR results previously discussed this difference in the behavior is now well understood
in terms of the chemical bond distribution: the P-Se system is characterized by an
efficient competition of homopolar and heteropolar bonds, whereas the As-Se system has
a large degree of chemical ordering. Numerous structural studies have indeed suggested
that these glasses are primarily constructed by pyramidal AsSe
3/2
groups, however, the
extent of As-As bond formation below the corresponding composition of 40 at.% As has
been subject to some discussion. Likewise, such chemical disordering would imply the
existence of Se-Se bonding at larger As contents. Furthermore, the open question
remains as to whether the various structural building blocks are randomly linked or if any
clustering into domains exists. While it has been demonstrated previously that high-
temperature
77
Se NMR is able to differentiate between Se- and As bonded selenium in
arsenic-selenium liquids [37,38], new information on short- and intermediate range order
in this system is also available. Figure 7 shows a typical temperature dependent set of
77
Se NMR spectra of binary As
x
Se
1-x
melts containing 12.5 mole% As [29]. At low
temperature (<200C), we observe three well-resolved resonances in this sample, which
we assign to three different types of selenium species. The good resolution indicates that
narrowing by rapid molecular motion is sufficiently complete to produce isotropic
spectra, while, on the other hand, averaging by chemical exchange is still slow on the
NMR timescale. The peak near 1361 ppm relative to a solid CdSe reference (species 1)
coincides with the spectrum of pure amorphous selenium and is therefore assigned to Se
atoms that are part of selenium chains only. The peak near 1114 ppm (species 2) is found
in all of those samples containing 20 mole% As or less. Since its fractional area increases
with increasing As content (data not shown), this resonance must signify selenium
involved in bonding with arsenic. Based on arguments presented below, we assign it to a
selenium species bonded to one arsenic and one selenium atom, i.e. the selenium atom in
the midst of a As-Se-Se fragment. The assignment of the resonance at 1289 ppm
(species 1`) is the most tentative one. Based on the fact that in samples with different
arsenic contents, the area of this peak remains in a fixed relation to that at 1114 ppm, we
assign this resonance to a selenium with a second-nearest As neighbor, i.e. to the
selenium atom at the end of an As-Se-Se- fragment. A third, smaller feature near 980
ppm (species 3) becomes also evident at temperatures > 180 C. We attribute this feature
to selenium atoms bridging between two As species. Unlike for species 1, 1 and 2, the
temperature windows separating motional narrowing from chemical exchange averaging
are not sufficiently distinct for species 3. Therefore, it is difficult to quantify its spectral
contribution.
129
Figure 6: T
g
vs. <r> correlation in the systems P-Se (crosses), As-Se (circles), and Ge-Se (diamonds).
Figure 7: Experimental temperature dependent
77
Se NMR spectra of As
12.5
Se
87.5
melt
As illustrated in Figure 7, signal coalescence owing to the onset of chemical
exchange is observed above T = 180 C, indicating that bond breaking and re-forming is
occurring on the NMR timescale. At a sufficiently high temperature (>280C) only a
single resonance remains, whose frequency approximately corresponds to the weighted
average of the individual peak positions. In glasses containing more than 20 % arsenic,
the spectra in the slow-exchange limit are not observable at temperatures above T
g
, most
likely because molecular motion is slowed down by increased network rigidity, making
it impossible to identify separate temperature regimes for motional narrowing and
exchange narrowing. At appropriate temperatures >300 C, these spectra show the
lineshapes typical of the fast-exchange limit.
Figure 8 illustrates that the average peak position depends on arsenic content and
temperature. In the sample containing 12.5 mole% As, the chemical exchange process is
revealed in the chemical shift trend shown (see above), while at larger As contents only
the exchange-averaged spectra were studied. The position of the exchange averaged
130
spectra does not coincide with the calculated center of gravity from the resolved spectra
at low temperatures. This result indicates that part of the arsenic-bound Se species are
not affected by motional narrowing and do not contribute to the sharp resonances seen
in Figure 7. This statement applies in particular to species 3 near 980 ppm, attributed to
selenium atoms bridging between two As atoms.
Figure 8: Dependence of
77
Se chemical shift on Composition and Temperature in As-Se melts.
The intrinsic linear temperature dependence observed in Figure 8 which is similar
to the effect seen in Figure 4 for each composition probably arises from a thermal
increase of the unpaired electron spin concentration in the liquid.
The dependence of chemical shift on As content is also evident from Figure 9,
which compiles data measured at a uniform temperature of 400 C. As the As content is
increased, the chemical shift decreases in a linear fashion up to a limiting value of near
1000 ppm at composition 40 at.% As. Within the concentration interval extending from
40 to 50 mole% arsenic, the
77
Se chemical shift remains approximately constant.
The nearly linear dependence of the average chemical shift within the composition
region at % As, and its approximate invariance at larger As concentrations
are consistent with a chemically ordered network structure composed mostly of AsSe
3/2
groups. Thus, the change in chemical shift with As content is due to the progressive
conversion of Se-Se-Se- to As-Se-Se to As-Se-As fragments. As predicted by such a
chemically ordered model, above 40% As, the chemical environment of Se remains
constant, and further introduction of As results in the formation of additional As-As
bonds only.
131
Figure 9:
77
Se NMR chemical shifts of As
x
Se
100-x
glasses at 400C.
Medium-Range Order
In analogy to the discussion in the P-Se system the detailed composition
dependence of the
77
Se chemical shifts enables an assessment of various medium range-
order scenarios, relating to the linking of the As- and Se-bearing units in the network.
Three basic scenarios are considered, from which different predictions with respect to
the fraction of Se-only bonded selenium species are made. First of all, a molecular
clustering model, based on the formation of As
4
Se
4
molecules dispersed in a selenium-
rich matrix can be considered. This model is of certain interest, because a similar
proposal was previously made for P-Se glasses [36]. A second scenario (domain model)
envisions the AsSe
3/2
units congregating into domains, implying that the majority of the
arseni oonded selenium species are bridging between two As atoms. The remaining
scenario is based on a random dispersal of AsSe
3/2
and Se
2/2
groups, simulated either
analytically or by means of a graph theoretical approach [29]. In the latter, AsSe
3/2
groups are placed onto a two-dimensional triangular grid at random (according to the
respective As concentration of the glass considered), and the number of Se-only bonded
Se atoms is determined by a simple counting algorithm. No As-As bonds are allowed.
The construction of the network from these units produces four distinct next-nearest
neighbor (NNN) environments for each arsenic site, being linked via selenium to zero,
one, two or three AsSe
3/2
groups. The populations of these NNN environments are
plotted in Figure 10 as a function of composition. Figure 11 indicates that the three
medium-range order scenarios make different predictions concerning the fraction 'f
1
'
of Se atoms that are exclusively bonded to Se. Experimentally, this number is
available from the chemical shift data measured in the fast exchange region. In
principle, the experimental shift corresponds to the weighted average of the three
basic types of selenium species:
132
where the f
i
are fractional contributions with f
1
+ f
2
+ f
3
= 1.
Each one of the three selenium species has a characteristic shift For species 1,
the chemical shift of molten Se is a good approximation, while for species 3, the
chemical shift of 1000 ppm observed at 400 C in liquid As
2
Se
3
(40 at.% As) can be
taken as a reference. For the chemical shift of species 2 we assume that it lies close to
the average (1200 ppm) of the two other species. The principal difficulty in extracting f
1
from the average chemical shift, then, lies in the lack of knowledge of the relative
contributions f
2
and f
3
of the singly and doubly As - bonded selenium atoms. We are
able to address this question quantitatively only in glasses containing 12.5 mole% As
and 17.5 mole% As, where we know the ratio f
1
/f
2
independently from the low-
temperature data (Figure 7). Figure 11 reveals that the f
1
extracted from this additional
information for the 12.5% As and the 17.5% As glasses are close to the f
1
-s predicted
from a statistical linking scenario.
For the other glasses in our study, only upper and lower limits for f
1
can be
specified, as represented by the vertical bars displayed in Figure 11. Clearly, the
experimental data eliminate both the molecular and the domain scenarios from
consideration, and show best consistency with the random linkage model. As such the
results are in agreement with a continuous network structure of arsenic selenide glasses
and are not in agreement with any scenarios involving clustering and phase separation
processes, at least within the concentration region extending from zero to 40 mole%
arsenic. On the other hand, the situation at larger As concentrations is less evident
from
77
Se NMR data. For such As-rich glasses, the formation of As-As bonds has been
most clearly detected from
75
As NQR spectroscopy [39] carried out on the glasses
themselves. We anticipate that the combination of liquid state
77
Se NMR with solid
state
75
As NQR spectroscopy offers new prospects for discussing chalcogenide glass
chemical order and medium range order in future applications.
Figure 10: Compositional dependence of NNN environments as generated by random linking
133
Figure 11: Comparison of the different scenarios with respect to the fraction f
1
of the Se-only bonded Se
atoms. Possible ranges of f
1
deduced from the analysis of the experimental
77
Se chemical shifts are
indicated (see text).
BINARY GERMANIUM SELENIDE GLASSES
The compositional dependence of T
g
for the binary Ge-Se glasses which has been
included in Figure 6, shows the the clear manifestation of rigidity percolation and
chemical threshold behavior occurring well-separated at the chemical compositions of
23 at% Ge (<r>2.46) and 33.3 at.% Ge (R=l), respectively. The slight upward shift of
the rigidity threshold to values above 2.40 has been noted previously by Boolchand [40]
and will not be discussed here. Figure 12 shows temperature dependent
77
Se NMR
spectra of Ge-Se melts at variable compositions. In this case, no distinct Ge-bonded
selenium species are detectable in the slow-exchange limit, presumably because the
mobility of these species is too low to permit averaging of the anisotropic interactions.
Therefore, in Ge-Se melts the effect of bonding to selenium is only visible in the fast-
exchange regime. Due to limitations in the temperature range available, the data
currently available are restricted to low-melting glasses with Ge contents < 20 at.%.
Since, consequently the limiting chemical shift of Ge-only bonded Se-species is not yet
available, a quantitative analysis in terms of intermediate-range order is not possible on
the basis of these data. Nevertheless, this plot illustrates the great sensitivity of
77
Se
chemical shifts to changes in the short range order in germanium selenide melts.
134
Figure 12: Temperature dependent
77
Se NMR spectra of glassy Ge-Se melts with variable Ge contents.
THE TERNARY SYSTEMS GE-P-X (X=S, SE, TE)
Recently, ternary systems have been studied with Te as the only chalcogen
component. Although no glassforming region exists in any of the binary systems P-Te,
P-Ge, or P-Si, the formation of ternary P-Si-Te and P-Ge-Te glasses has been reported
[5] within compositional limits which have been reproduced in our laboratory.
31
P MAS
and spin echo decay data obtained on such glasses indicate that in the Ge-P-Te glass the
phosphorus environment is entirely dominated by phosphorus-phosphorus bonding.
Heating these glasses above the glass transition temperature produces P
4
molecules [41].
Thus, Ge-P-Te glass can be viewed as being composed of P
n
clusters or polymeric
bands embedded in a Ge-Te matrix. In contrast, the
31
P chemical shifts of the P-Si-Te
glasses occur in a region that is typical for Si-bonded phosphorus. The spin echo decays
of these glasses are substantially slower suggesting weaker
31
P-
31
P dipole-dipole
interactions. From a detailed analysis of these results the extent of P-P bonding in this
glass system has been quantified, and found to be approximately equal to that expected
from statistical probability under explicit exclusion of P-Te bond formation [41].
A comparison of the spectroscopic behavior in the ternary systems P-Ge-S, P-Ge-
Se, and P-Ge-Te reveals a striking influence of the chalcogen ion on the short-range
order of phosphorus. While in the sulfide system, the short-range order of P is
dominated by S=PS
3/2
groups [42], in Ge-P-Se glasses the formation of GeSe
4/2
groups
takes precedence, and controls the balance of P-Se vs. P-P bonding (see below) [43].
Finally, in Ge-P-Te glasses only P-P bonding occurs.
Most recent
31
P MAS NMR results on phosphorus-germanium-sulfur glasses are
shown in Figure 13. These glasses have recently attracted considerable interest as low-
phonon hosts for rare-earth luminescent ions, and the P-atoms are thought to play a
crucial role in increasing rare-earth solubility. Structurally, these glasses can be viewed
as polymeric networks with strong chemical ordering into GeS
4/2
and S=PS
3/2
units.
Compared to binary P-S glasses (see Figure 1), the presence of germanium second
nearest neighbors to the S=PS
3/2
groups produces unusually large resonance
displacements, illustrating the great sensitivity of
31
P MAS NMR to next-nearest
neighbor effects in the structure of these glasses [42].
135
Figure 13:
31
P MAS NMR spectra of two GeS
2
-P
4
S
10
glasses.
Above all, the most interesting results have been obtained for the glass system Ge-
P-Se [43]. Figure 14 shows that no chemical threshold behavior near R = 1 is observable
in this system. This result implies that in glasses with R<1 the selenium deficiency is
balanced by the formation of P-P bonds, whereas Ge remains bonded in the form of
GeSe
4/2
groups. This conclusion has been confirmed by a detailed analysis of the
31
P-
31
P
dipole-dipole interaction in this system. Figure 15 plots the
31
P-
31
P dipolar second
moment M
2
, measured by spin echo decay spectroscopy, as a function of glass
composition. As shown in reference [43] the second moment serves as a measure of the
fraction of P-bonded phosphorus atoms. Addition of Ge to a glass of fixed P/Se ratio
produces a marked increase of M
2
owing to the tendency of Ge to attract selenium to
form GeSe
4/2
units. Thus, the increase of M
2
reflects an increased fraction of P atoms
engaging in P-P bonding as the germanium content of the glasses is increased. Figure 15
demonstrates a universal correlation of M
2
with the compositional parameter
Here [Se
eff
] = [Se] -2[Ge] is the selenium content available for bonding to phosphorus
after the amount needed for the formation of GeSe
4/2
groups has been subtracted. The
first factor in eq. (4) accounts for the competition effect in chemical bond formation and
the second one for the overall dilution effect arising from the presence of Ge as a third
component. Overall the universal correlation, which also includes the data on the binary
P-Se glasses, provides strong evidence for a pronounced hierarchy of homopolar bond
formation in P-Ge-Se glasses: Even in highly Se deficient glasses the formation of GeSe
136
bonds takes precedence, while the Se deficiency is balanced by an increased formation of
P-P bonds.
Figure 14: Plot of T
g
vs. <r> and R for glasses in the system Ge-P-Se. Note the absence of chemical
threshold behavior
Figure 15: Correlation of the
31
P-
31
P dipolar second moment with the compositional parameter X =
P/Se
eff
) (P) (see text) for Ge-P-Se glasses. Open symbols denote data for binary P-Se glasses
THE TERNARY SYSTEMS GE-AS-SE AND GE-SB-SE.
The ternary systems Ge-As-Se and Ge-Sb-Se are homologues of the phosphorus-
based system discussed above. Tichy and Ticha noted that in spite of the strong chemical
ordering observed in the binary As-Se and Ge-Se glass systems (strong T
g
-maxima at R
= 1), a plot of T
g
vs. R does not yield a corresponding maximum in ternary Ge-As-Se
glasses and only a weak effect in the Ge-Sb-Se system. New insight into this behavior
has been recently obtained by
119
Sn spectroscopy [44]. As previously
demonstrated by Boolchand [45], the tin isotope substitutes isomorphically as Sn(IV) for
Ge in glassy GeSe
2
. The formation of homopolar Ge-Ge bonds in Se-deficient glasses
can then be monitored by the appearance of Sn(II) in the spectra. In contrast,
our recent results show that Se-deficient As-Se-Sn glasses can be prepared over a fairly
wide composition range (down to R=0.7) without Sn(II) ever appearing in the
137
spectra[46]. Thus, the Se-deficiency in such glasses is balanced by As-As bond
formation in analogy to the situation in Ge-P-Se glasses. The same balancing
mechanism, then, is likely to occur in the Ge-As-Se ternary, explaining the absence of
chemical threshold behavior. For Ge-Sb-Se glasses a similar explanation may hold for
the attenuation of the chemical threshold effect, however, from
121
Sb Mssbauer
spectroscopy only weak evidence for Sb-Sb bond formation has been found. Overall,
these results suggest that in ternary Ge-X-Se glasses (X = P, As, Sb) there is a strong
secondary hierarchy governing the formation of homopolar bonds, with the tendency
strongly decreasing in the order P-P, As-As, Sb-Sb, Ge-Ge. Over wide compositional
ranges of Se-deficient glasses this hierarchy ensures that the number of Ge-Se bonds is
maximized and no ,,unnecessary Ge-Ge bonds are being formed as long as there are
alternate possibilities of homopolar bond formation. In highly Se-deficient glasses the
presence of Ge-Ge bonds is finally indicated by the appearance of divalent tin species in
the
119
Sn Mssbauer spectra. Figure 14 and 17 illustrate that universal T
g
vs. <r>
correlations are observed for all of the three ternary glass systems. In the system Ge-P-
Se this correlation includes all of the glasses studied. In the systems Ge-As-Se and Ge-
Sb-Se deviations from universality are only observed for those (highly Se-deficient)
glass compositions, for which the
119
Sn Mssbauer NMR spectra indicate a fraction of
tin to be divalent, hence suggesting that part of the germanium constituents are no
longer part of GeSe
4/2
units.
CONCLUSIONS
In summary the results of the present study illustrate the utility of solid- and
molten-state NMR spectroscopy to provide valuable information on short- and medium-
range ordering effects in non-oxide chalcogenide glasses. P-Se glasses with their
efficient competition between homo- and heteropolar bond formation are the most ideal
model systems for mean-field theory. Accordingly a dramatic change in dT
g
/d<r> is
observable at the percolation threshold <r> = 2.40. In contrast, As-Se and Ge-Se
systems display a pronounced preference for heteropolar bond formation, resulting in
chemical threshold effects that are superimposed upon the effects of physical
percolation. These chemical threshold effects produce T
g
maxima at compositions
corresponding to R =1, where heteropolar bond formation is maximized. In ternary Ge-
Se-X systems (X = P,As,Sb) these chemical threshold effects disappear because the
formation of homopolar bonds is controlled by a strong secondary hierarchy. Over wide
compositional regions of Se-deficient glasses this hierarchy serves to minimize the
formation of Ge-Ge bonds and can be held responsible for the observation of universal
T
g
vs. <r> dependences. Based on these results we predict that physical threshold
behavior will in general be observable not only in those glass systems having no
bonding preferences, but also in those ternary glass systems in which chemical ordering
is observed, but where there is a clear secondary hierarchy of heteropolar bond
formation. The most striking example is the Ge-P-Se glass system; other examples can
be envisioned for tellurium-based chalcogenide glass systems Ge-X-Te. In contrast, no
such secondary hierarchy is expected for sulfide-based glasses (such as Ge-X-S), where
we envision heteropolar Ge-S and X-S bonding to dominate the structure to such an
extent that no secondary hierarchy effects are expected. The experimental varification of
these ideas is currently under investigation in our laboratories.
138
Figure 16: Plots of T
g
vs. <r> in the glass systems, Ge(Sn)-As-Se (top), and Ge(Sn)-Sb-Se (bottom).
Open symbols indicate those glasses showing evidence of Sn(II) in the spectra. Note that only
these glasses show deviations from a universal correlation.
ACKNOWLEDGMENTS
Financial support of this research by the U.S. National Science Foundation (DMR
92-21197) and the Deutsche Forschungsgemeinschaft under grant Ecl68/l-2 is most
gratefully acknowledged. Thanks are also due to the Wissenschaftsministerium
Nordrhein-Westfalen for supplemental support. C.R. and G.B. acknowledge support by
personal stipends awarded by the Verband der Chemischen Industrie, Stiftung
Stipendien Fonds. S.H. appreciates support by a University of California Presidential
Dissertation Fellowship.
139
REFERENCES
1. Nishii, J., Morimoto, S:, Inagawa, I., lizuka, R., Yamashita, T., Yamagashi, T. (1992) Recent
advances and trends in chalcogenide glass fiber technology: a review, J. Noncryst. Solids 140, 199-
208.
2. Simons, D. R., Faber, A. J., Waal, H. (1995) GeS
x
glass for Pr
3+
doped fiber amplifiers at
J. Noncryst. Solids 185, 283-288.
3. Aitken, B. G., Quimby, R. S. (1997) Rare-earth doped multicomponent Ge-based sulphide glasses,
J. Noncryst. Solids 213/214, 281-287.
4. Turnbull, D.A., Bishop, S.G. (1998) Rare-earth dopants as probes of localized states in
chalcogenide glasses, J. Noncryst. Solids 223, 105-113.
5. Hilton, A. R., Jones, C. E.,. Brau, M (1966) Non-oxide IVA-VA chalcogenide glasses. Part I.
Glass-forming regions and variations in physical properties, Phys. Chem. Glasses 7, 105-112.
6. Borisova, Z. U. Glassy Semiconductors, Plenum Press New York, 1981.
7. Feltz, A. Amorphous Inorganic Materials and Glasses, Verlag Chemie Weinheim 1993.
8. Zallen, R The Physics of Amorphous Solids, John Wiley New York 1986.
9. Phillips, J. C. (1979) Topology of covalent non-crystalline solids, 1. Short-range order in
chalcogenide alloys and a-Si(Ge), J. Noncryst. Solids 34, 153-181.
10. Phillips, J. C. (1981) Topology of covalent non-crystalline solids. 2. Medium-range order in
chalcogenide alloys and a-Si(Ge), J. Noncryst. Solids 43, 37-77.
11. Thorpe, M. F. (1983) Continuous deformations in random networks, J. Noncryst. Solids 57, 355-
370.
12. Thorpe, M. F. (1985) Rigidity percolation in glassy structures, J. Noncryst. Solids 76, 109-116.
13. Thorpe, M. F. (1995) Dynamics of glassy systems, bulk and surface floppy modes, J. Noncryst.
Solids 182, 135-142.
14. Boolchand, P., Thorpe, M. F. (1994) Glass-forming tendency, percolation of rigidity, and onefold
coordinated atoms in covalent networks, Phys. Rev. B 50, 10366-10368.
15. Phillips, J. C., Thorpe, M. F. (1985) Constraint theory, vector percolation and glass formation,
Solid State Commun. 53, 699-702.
16. He, H., Thorpe, M. F. (1985) Elastic properties of glasses, Phys. Rev. Lett. 54, 2107-2110.
17. Tatsumisago, M., Halfpap, B. L., Green, J. L., Lindsay, S. M., Angell, C.A., (1990) Fragility of
Ge-As-Se glass-forming liquids in relation to rigidity percolation, and the Kauzmann paradox,
Phys. Rev. Lett., 64, 1549-1552.
18. Sreeram, A.N., Varshneya, A.K. and Swiler, D.R. (1991) Molar volume and elastic properties of
multicomponent chalcogenide glasses, J. Non-Cryst. Solids 128, 294-309.
19. Giridhar, A., Mahadevan, S., J. (1992) The T
g
versus Z dependence of glasses of the Ge-In-Se
system, J. Noncryst. Solids, 151,245-252.
20. Mahadevan, S., Giridhar, A., Singh, A. K. (1994) Volumetric effect of topology in chalcogenide
glass systems, J. Noncryst. Solids, 169, 133-142.
21. Senapati, U., Varshneya, A. K., (1995) Configurational arrangements in chalcogenide glasses: a
new perspective on Phillips constraint theory, J. Noncryst. Solids, 185, 289-296.
22. Wang, Zh., Chen, Q, (1995) Structure and some physical properties in relation to average
coordination number, <r>, in TeX and TeXAs glasses, J. Noncryst. Solids, 184, 177-183.
23. Mahadevan, S., Giridhar, A., (1992) Floppy to rigid transition and chemical ordering in Ge-
Sb(As)-Se glasses, J. Noncryst. Solids 143, 52-58.
24. Tichy, L., Ticha, H., (1995) Covalent bond approach to the glass transition temperature of
chalcogenide glasses, J. Noncryst. Solids, 189, 141-146.
25. Tichy, L., Ticha, H. (1999) Is the chemical threshold in certain chalcogenide glasses responsible
for the threshold at the mean coordination number of approximately 2.7?, Phil. Mag., 79, 373-380.
26. Tichy, L., Ticha, H., (1994) On the chemical threshold in chalcogenide glasses, Mater. Lett., 21,
313-319.
27. Mutolo, P, Witschas, M., Regelsky, G., Schmedt auf der Guenne, J., Eckert H. (1999), Nuclear
magnetic resonance (NMR) studies of phosphorus-based chalcogenide glasses: an overview, J.
Noncryst. Solids 256-257, 63-72, and references therein.
28. Maxwell, R.. Eckert, H. (1994) Chemical equilibria in glass-forming melts: high-temperature
3I
P
and
77
Se NMR of the system phosphorus-selenium, J. Am. Chem. Soc. 116 (1994), 682-689.
29. Rosenhahn, C., Hayes, S., Rosenhahn, B., Eckert, H.(2001), Structural organization of arsenic
selenide liquids: new results from liquid state NMR, J. Noncryst. Solids, in press.
140
30. Heyder, F., Linke,D. (1973) Zur Glasbildung in den Systemen Phosphor-Schwefel und Phosphor-
Selen, Z. Chem. 13, 480-481
31. Lyda, C.M., Leone, J.M., Bankert, M.A., Xia, Y., Eckert, H. (1994) Structural studies of
phosphorus-sulfur-tellurium glasses by
31
P MAS-NMR and vibrational spectroscopies, Chem.
Mater. 6, 1934-1939.
32. Lathrop, D., Eckert, H. (1989) Chemical disorder in non-oxide chalcogenide glasses. Site
speciation in the system phosphorus-selenium by magic angle spinning NMR at very high spinning
speeds, J. Phys. Chem. 93, 7895-7902.
33. Lathrop, D., Eckert, H. (1991), Quantitative determination of structural units in phosphorus-
selenium glasses by
31
P dipolar and MAS spectroscopy, Phys. Rev. B 43, 7279-7287.
34. Lathrop, D., Eckert, H.(1990), Dipolar NMR spectroscopy of non-oxidic glasses. Structural
chacterization of the system phosphorus-selenium by
31
P-
77
Se spin echo double resonance NMR, J.
Constraints from
3I
P and
77
Se NMR spectroscopy, J. Noncrysl. Solids 188, 75-86.
36. Price, D.L., Misawa, M., Susman, S., Morrison, T.I., Shenoy, G. K., M. Grimsditch, M., (1984),
The structure of phosphorus-selenium glasses. 1. Concentration dependence of the short-and
intermediate-range order. J. Noncryst. Solids 66, 443-465.
37. Bishop, S.G., Taylor, P.C. (1972) Atomic reorientation rates in liquid chalcogenide glasses: NMR
in Se and As
2
Se
3
, Solid State Commun. 11, 1323-1326
38. Brown, D., Moore, D.S., Seymour, E.F.W. (1972), NMR of
77
Se and
125
Te in liquid and
amorphous semiconductors, J. Noncryst. Solids 8-10, 256-261.
39. Saleh, Z. M., Williams, G. A., Taylor, P.C. (1989) Nuclear quadrupole resonance in the glassy Cu-
As-S and Cu-As-Se systems, Phys. Rev. B 40 10557-10563
40. Feng, X, W., Bresser, W. J., Boolchand, P. (1997) Direct evidence for stiffness threshold in
chalcogenide glasses, Phys. Rev. Lett. 78, 4422-4425.
41. Witschas, M., Regelsky, G., Eckert, H. (1997) NMR studies of phosphorus in Si-P-Te and Ge-P-Te
glasses, J. Noncryst. Solids 215, 226-235.
42. Brunklaus, G. (1999),
31
P-NMR spektroskopische Untersuchungen zur Struktur der Glser des
ternren Systems Ge-P-S, Diploma Thesis Universitt Mnster.
43. Lyda, C., Tepe, T., Tullius, M., Lathrop, D., Eckert, H., (1994) Chemical bond distribution in
chalcogenide glasses. A
31
P MAS and spin-echo NMR study of glasses in the phosphorus-
germanium-selenium system, J. Noncryst. Solids, 171, 271-280.
44. Mllmann, R., Mosel, B.D., Eckert, (1999) H. Physical and chemical threshold behavior in
chalcogenide networks:
119
Sn Mssbauer spectroscopy of Ge(Sn)-As-Se glasses, Phys. Chem.
45. Boolchand, P., Stevens, M., (1984) Evidence for isoelectronic Sn for Ge substitution in crystalline
and glassy GeSe
2
, Phys. Rev. B, 29, 1-7.
46. Rosenhahn, C., Mllmann, R., Mosel, B.D., Eckert, H. Hierarchy of homopolar bond formation in
ternary chalcogenide glasses:
119
Sn Mssbauer spectroscopic results in Ge(Sn)-As-Se and Ge(Sn)-
Sb-Se glasses; to be published.
35. Maxwell, R., Lathrop, D., Eckert, H. (1995) Medium-range order in phosphorus-selenium glasses.
Am. Chem. Soc. 112, 9017-9019
Chem. Phys. 1, 2543-2550.
141
This page intentionally left blank
GLASS TRANSITION TEMPERATURE VARIATION AS A PROBE FOR
NETWORK CONNECTIVITY
M. MICOULAUT
Laboratoire de Physique Thorique des Liquides,
Universit Pierre et Marie Curie, Boite 121
4, place Jussieu 75252 Paris Cedex 05 France
INTRODUCTION
The liquid-glass transition is one of the unsolved challenging problems [1] in actual solid
state and materials sciences, and its basic phenomenology can be described as follows: a liq-
uid which has been cooled sufficiently fast in order to avoid crystallization can be denoted as
supercooled when its temperature lies below the melting temperature T
m
. While decreasing
the temperature, the supercooled liquid becomes so viscous that the structural relaxation time
becomes of the order of the experimental time texp. The temperature at which the glass
transition occurs is denoted T
g
and is best defined at which It corresponds to the
temperature at which structural arrest occurs in the sense that the relaxations are too slow to
observe. As a consequence, the glass transition seems to be primarily dynamic in origin, in
contrast with usual thermodynamic phase transition.
Another definition of Tg, denoted in the literature as the caloric glass transition tem-
perature, is related to the thermodynamic behavior of the supercooled liquid close to the
transition. At Tg, thermodynamic quantities such as enthalpy and volume (the first deriva-
tives of the Gibbs energy) exhibit a first-order behavior while some of the second derivatives
(such as the heat capacity) are continuous, but with an inflexion point at T
g
. A standard mea-
surement of the glass transition temperature consists therefore in following the endothermic
heat flow through the transition in differential scanning calorimetry (DSC technique). We
should note that the value of T
g
depends on the heating rate (typically 10 K.min
1
) in the
DSC experiment, which signals a significant non-equilibrium behavior. However, the con-
ventional technique to access the glass transition temperature has been recently improved by
superposing a sinusoidal variation on the usual linear T ramp (modulated DSC) [2]. This per-
mits deconvolving the total heat flow near the transition into reversing and a non-reversing
contributions. The reversing component tracks the sinusoidal T variation and yields a mea-
sure of the heat capacity during glass transition, while the non-reversing heat flow is ascribed
to the thermal history of the melt (kinetic processes and the enthalpic changes accompanying
structural reorganization close to T
g
). This means that the measure of Tg from the reversing
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 143
heat flow is almost expunged from kinetic effects and represents therefore a true value of
the glass transition temperature [3].
Surprisingly, little attention has been given to the variation of the glass transition tem-
perature with respect to alloying, although it is fundamental for the understanding the nature
of the glass transition, and for technological applications. For instance, it is well known that
a few percent of sodium oxide lowers the glass transition temperature of vitreous silica from
1200C to 600C, thus permitting to obtain a glass with LT furnaces. Several proposals have
been made in trying to relate the value of Tg with other physical or chemical measurable
quantities. Kauzmann suggested for example that T
g
scales with the melting temperature
of the corresponding liquid (the so-called two-third rule). Other authors have proposed
relationships based on the boiling temperature, the Debye temperature of the phonon spec-
trum. The importance of structural factors has been stressed, in particular the valence of the
involved atoms in the glass-forming material, but these T
g
predictions remain mostly quali-
tative [4]. The main quantitative advances in this field are due to Varshneya and co-workers
[5], who applied with success the Adam, Gibbs and Di Marzio theory [6] on glassy poly-
mers to multicomponent chalcogenide network glasses. The chalcogenide glass systems can
indeed be represented as networks of chains made of selenium or sulfur atoms, in which
cross-linking units such as germanium atoms are inserted. The increase of T
g
, identified with
the second-order phase transition temperature in the initial theory, is produced by the growing
presence of these cross-linking agents.
The aim of this article is to show that the variation of the glass transition temperature
upon alloying is mostly related to the connectivity of the glass network. First, we will show
that the increase of the structural relaxation time can be translated in a very simple minded
way, when considering the glass transition in covalent glass-forming liquids. We give the
general ideas concerning the way one should define the glass transition temperature. Then
we will apply it to binary and ternary glass formers for which the theoretical description is en-
tirely solvable. This will give universal relationships that can be compared with experimental
data. Last but not least, we explain why there is a systematic deviation of the prediction with
respect to experiment in all investigated glass systems and how this is related to the constraint
theory [7].
STOCHASTIC AGGLOMERATION THEORY
Stochastic agglomeration theory [8,9] provides a quantitative means to analyze compo-
sitional trends in the glass transition temperature. As we shall see below, this theory seems to
be particularly well adapted for the description of covalent glass-forming systems, where the
increase of the the structural relaxation time with decreasing temperature is Arrhenius-like:
Another quantity of interest behaves similarly, which is the viscos-
ity of the glass-forming liquid, via the relation represents the instantaneous
shear modulus, which is almost constant. The increase of the viscosity with decreasing tem-
perature can be viewed in a very simple way in covalent systems where the viscosity should
be directly proportional to the number of dangling bonds remaining in the liquid structure.
Each time, a new bond is created, the viscosity increases. Therefore, agglomeration between
well-defined atomic or molecular entities (i.e. creating bonds between these entities) should
be one of the most representative physical process taking place during the glass transition.
Also, the activation energy appearing in the Arrhenius behavior of should be intimely
related to the involved bond energy.
In order to describe these agglomeration processes, we have to define more precisely the
structural entities, which are going to be involved. The most simplest way of describing the
short range order (SRO) in the liquid should rely on star-like entities sharing a certain num-
ber of dangling bonds (we shall call them local structural configurations (LSC) or singlets).
144
Figure 1. A schematic representation of the realized experiment for N = 2. a) At time t = 0 (C) has a certain
statistical distribution of LSCs, which is modified if diffusive LSCs are trapped by agglomeration during the
time step (panel b). One can also imagine collective rearrangements where two bonds are created during the
same time step (panel c). Note that the size of (C) can increase in order to avoid densification.
These LSCs have a central atom and have a clear, unambigous experimental evidence. Their
coordination number is defined by e.g. X-ray or inelastic neutron diffraction techniques,
which exhibit sharp and characteristic peaks at the corresponding bond length and give infor-
mation about the number of nearest possible neighbors of the central atom. For instance, we
can mention the tetrahedron which is the lowest possible short-range structure in silicate or
germanate glasses (in this case the LSCs are molecules) or or a four-valenced silicon atom
in network glasses (here the LSCs are atoms). An alternative choice could be the
tetrahedron and a selenium atom. Of course, the results do not depend on the initial
choice of the LSCs. The idea is to start from an initial LSC, corresponding to the short-
range structure of an inital glass (or liquid-like) network and to see how the agglomeration is
affected by the presence of another kind of LSC or several kinds of LSCs.
Now, let us assume that we are able to realize the following experiment. In a supercooled
liquid at T > T
g
, we consider always the same region of space (C), defined by its particular
spatial coordinates. There, we can make at a time t = 0 a statistics of the different LSCs,
yielding a probability distribution {p
i
(0)}, where i = 1...N and N is the total number of
different LSCs. If the liquid is homogeneous and at equilibrium, the value of the probabilities
should be directly related to the macroscopic concentration. As long as the system is in the
supercooled state, the low viscosity allows still intermolecular diffusion of remaining isolated
LSCs [10], passing through the particular region of space (C) we are considering. The
probability distribution {p
i
(0)} at time t = 0 will be modified if these LSCs are trapped after
a certain finite time inside the region, by agglomerating on other existing LSCs, leading
to a distribution and a fluctuation for the LSC i which is We can
repeat this experiment for different times ad decreasing temperature, and still looking at the
quantity
The next issue we have to addresss is the evaluation of a time dependence for the LSC
probability distribution, during the well-defined discrete time-step and derive a master
equation which imposes the minimization of local fluctuations at a certain temperature which
will be identified with the glass transition temperature. We can write the master equation as:
where we have taken into account possible effects due to the cooling rate q = dT/dt.
Why should the LSC fluctations vanish at the temperature T
g
?
If the viscosity is very high (of the order of e.g. 10
13
Poise), reaching the glassy state,
the diffusion of LSCs inside the liquid should be enhanced, and agglomeration should not be
possible anymore. One should therefore expect that after the same time step the distribu-
tion remains constant inside the considered region (C), and equals
145
In other words, at glass transition temperature, the local probabilities reach a stationary value,
which satisfy:
Equation (2) represents a set of (N-l) non-linear equations, which are closed by a normaliza-
tion condition.
What is the form of appearing in equ. (2) ? It represents the probability that
an LSC i can stick onto another LSC j in the considered region during the time step TO,
averaged over all js. is also related to the probability that a bond i j is created. This
probability should be proportional to the products of the probabilities of the LSCs at time
a Boltzmann factor which takes into account the energy of creation of the respective
bond formation at temperature T, and a statistical factor which may be regarded as the
degeneracy of the corresponding stored energy, because there are several equivalent ways to
join together two coordinated LSCs. If one has single bond formation, the statistical factor
is simply the product of the coordination numbers of the involved LSCs. For example, the
number of equivalent ways to connect two LSC i and j with coordination numbers and is
and can be related either to the number of outer shell electrons
or to the coordination number (following in this case the so-called 8 rule [11]).
Here, Z is a normalizing factor.
One obtains the solution of (2) by solving the set of equations in terms of the variable
since the are found as functions of these quantities. It is easy to remark that
such a set of solutions satisfying (2) varies with the glass transition temperature
and depends strongly on (related to the connectivity of the network) and (related to
the bond strength).
The space of solutions of equ. (2) is a (N 1) dimensional symplex whose vertices
correspond to the solutions identified with pure structural phases because all the are
equal to zero, except one. Beside these trivial solutions, which do not depend on the temper-
ature, it is possible to obtain other ones, lying either on the edges of the symplex (partially
pure phases, at least one solution being equal to 0) or inside (mixture of all solutions
Finally, the nature of the agglomeration process can be analyzed by studying the sta-
bility of the master equation in the vicinity of these stationary solutions. The set of solu-
tions can correspond to possible structurally stable (if the solution is an attractor) or
metastable states (if the solution is a saddle point) of the network. In most of the situations,
the pure structural phases are attractive (or repulsive) and the solution found inside the sym-
plex is a saddle point. The existence of the latter depends on the involved bond energies E
ij
,
also on the value of T
g
.
In network glasses (e.g. the
LSCs are the atoms of the network. Thus the concentration x
i
of the atoms is directly re-
lated to the probability p
i
(0) of finding the LSCs B
i
inside the network, and p
i
= x
i
. Their
respective derivatives with respect to the glass transition temperature are equal. Note that
in this notation, A represents the atom of the initial network, when all x
i
are equal to zero
(e.g. the selenium network in selenides). One interesting chemical parameter is the average
coordination number < r >, which is widely used in these systems, since the introduction of
this quantity by Phillips and Thorpe in constraint theory [7,12]. < r > is defined by:
146
where r
i
is the coordination number of the atom B
i
, given eventually by the 8 rule. The
second part of equ. (4) is obtained from the normalisation condition. Then:
The expression (5) can be compared with numerous experimental data [3,13,14], which have
been obtained since 1979 in order to check the validity of the constraint theory [7,12] and its
mechanical threshold when < r > reaches the magic number of 2.4.
In other glass systems, such as binary glasses, (made of a network former and a mod-
ifier), the LSCs are molecules. For example, the silicon tetrahedra with one and two non-
bridging oxygens are used in order to describe systems. Here, the
relationship between concentration x and the solutions p
i
(0) of the LSCs satisfying equation
(2) can be obtained by recalling a charge conservation law [9]. In particular, the concentration
of the modifier cation (such as or must be equal to the anionic contribution located
on each different LSC. In binary glasses with formula (1 x)network former+x modifier,
this conservation law reads:
where is the anionic contribution of the LSC with probability is a stoichiometric
factor. Similarly to (5), we can then compute the variation of with respect to the reduced
concentration R.
The expressions appearing in (5) and (7) are obtained by differentiating the
implicit functions obtained from (2), and by solving the system:
In the following, we will describe the simplest cases. To do this, we neglect the dependence
of the cooling rate q = dT/dt in equation (1) because network chalcogenide glasses or binary
oxide glasses form very easily and have critical cooling rates of the order of
From fig. 2, we can also remark that the cooling rate dependence of equation (1) (and
consequently on the value of T
g
) is rather weak, although is changing from several orders
of magnitude [The figure refers to the construction of the next section with N = 2]. Also, since
we are interested in a variation of the glass transition temperature with respect to structural
modification, if the preparation of the glass systems upon alloying is always realized with the
same procedure, the effect of cooling rate should be negligible (in other words
Then, we can solve the case of N = 2 and N = 3.
The simplest case: N=2
In this section, we consider the application of stochastic agglomeration theory to glasses
having two different kinds of LSCs, which will be denoted by A (the initial or regular LSC,
with probability 1 p(0) and coordination number and B (the modifier LSC with prob-
ability p(0) and coordination number The archetypal systems are the network former
and the low-modified alkali silicate glass. For the latter, the building blocks of the
147
Figure 2. A plot of (equation (1)) for N = 2 as a function of temperature T for different cooling rates
q. The system corresponds to a glass, with (solid line), (dashed line) and
K (dotted line). The glass transition temperature corresponds to the vertical value 0, where
vanishes.
network are the Q
4
(the tetrahedron in NMR notation, which is the short-range struc-
ture of vitreous silica) and the Q
3
units (the tetrahedron), which appear when
a small amount of sodium oxide is added in the network.
Starting again from equation (2), one can see that there is only one single equation to
solve, and for low modifier concentration, only two possible connections can occur during
glass transition (or the time step A-A and A-B connections, neglecting B-B connections.
Z normalizes the whole, and the probability to find the LSC B among this distribution is
We obtain the following equation at T = T
g
, from (2):
Equation (11) has two trivial solutions: p(0) = 1 and p(0) = 0 and a third one:
which defines a relationship between p(0) and T
g
. We stress that only one energy difference
is essential here: The relationship between T
g
and the probability of finding the
LSC B depends therefore only on one parameter. The expression (12) will be physically
acceptable only if which is true when the following condition is satisfied:
148
The condition (13) implies also that the border solutions p(0) = 0 and p(0) = 1 are unstable
stationary solutions, which means that agglomeration out of A-A and B-B bonds only is
prohibited. In this case, the bond statistics is given by the solution (12) and the expressions
of (9) and (10). If the condition (13) is not satisfied, a stable attractive solution is found at
p(0) = 0. There, the agglomeration tends to phase separate at a microscopic level. We note
that the absence of B-B bondings automatically produces a repulsive solution for p(0) = 1.
Equation (12) can be made parameter-free by considering the limit of the pure A network
with glass transition temperature T
g
= T
0
and of course p(0) = 0. It is easy to check from
(12) that:
Last but not least, we can see how the glass transition temperature changes in the very low
concentration limit, by computing the derivative T
g
with respect to p(0):
with the energy difference (14) established, this leads to a complete parameter-free equation:
Following the considered system, we will use either the equivalence p(0) = x or the charge
conservation law in order to relate the probability of the local B configuration to the concen-
tration. For network glasses the formula has been obtained in [8]:
The important and rather surprising point is that (17) depends neither on dynamical, nor on
visco-elastic quantities, although they are used to define T
g
and to track the glass transition.
It becomes obvious that the bond strength between the LSCs does not influence the variation
of T
g
with respect to T
0
. Connectivity plays here the major role. Equation (17) gives also the
mathematical transcription of a well-known empirical rule extracted from numerous experi-
ments in glass science, which states that the glass transition temperature T
g
increases with the
addition of a modifier that increases the network connectivity, i.e. when r
B
> r
A
.
A more sophisticated case: N=3
The next step we can accomplish in agglomeration theory is also entirely solvable if one
neglects again the effects of the cooling rate in equation (1). In solving this case, we will use
the previous results obtained for N = 2.
We consider now systems where there can exist three different kinds of LSCs. This is
of course the situation in ternary chalcogenide glass systems, where three different atomic
species are involved (for instance, the As-Ge-S(Se) network glasses). Nevertheless, the con-
struction will be also valid for binary glasses. Here, three different LSCs can be found,
although there are only two chemically different components (the network former and the
149
modifier). This is for example the situation in the glass in the concen-
tration range [0,0.5], where three different types of tetrahedra can be observed: the afore-
mentioned Q
4
and Q
3
units, but now also the Q
2
unit, sharing two non-bridging oxygens, and
being a part of the metasilicate chain structure at x = 0.5. The difference with the chalco-
genides lies in the fact that both Q
3
and Q
2
contribute to the single concentration R (or x) of
alkali oxide.
We will focus in this article only on chalcogenides. We shall denote the LSC of the two
modified states by B and C, and the basic state by A. A represents then the local structure of
the initial glass network. If we use again the previous examples, the selenium or the sulfur
atom is identified with the A configuration in chalcogenides and the As and Ge atoms with
the B and C LSCs.
We denote their coordination numbers by r
A
, r
B
, and r
C
and their corresponding local
probabilities by 1 p
B
p
C
, p
B
and p
C
(for simplicity, we have removed the bracketts and
the mention of t = 0). Again, we shall consider the situation of a ternary network glass in the
case of low modification. This allows us to take only four possible connections into account,
between the LSCs A, B and C, and neglecting B B and C C bonds. In many situations,
these bonds occur only beyond the stoichiometric glass composition. The corresponding
probabilities of A-A, A-B, A-C and B-C bonds can be written as follows:
where Z is the normalizing factor given by:
and represent the bond energies involved in such types of bonds. As
one can see, the construction remains here very similar to that described in the previous
section, but the main difference lies in the number of equations (2) to solve. Instead of a
single equation with one variable p(0), as shown in (11), we obtain here a set of two non-
linear equations with two variables and These are obtained by defining the local
concentration of B and C:
and writing an equation of type (1) for the B and C LSCs at glass transition temperature,
where the different (with i = B,C) vanish:
150
Figure 3. The solutions of the system (24). The solutions a), b) and c) correspond respectively to pure A, B
and C LSCs networks, the solution d) to a half mixture of B and C, the solution e) and f) to the related binary
glasses (A,C) and (A,B). Finally, the saddle point solution g) corresponds to the ternary glass solution.
where we have set: and
for a clearer presentation. It is worth mentioning that the system (24), which has
to be solved in terms of the variables p
B
and p
C
, is simpler than it looks at the first glance,
especially because of the symetric role of the LSCs B and C. Also, the space of variables p
B
and p
C
is contained in a two-dimensional symplex (i.e. a triangle)
All the meaningful solutions are contained in this triangle (see fig. 3).
There are still some trivial solutions for the system (24), namely: pB = 0, pC = 0 (so-
the vertices of the symplex (fig.3), also (middle of the edge 1 p
B
p
C
= 0,
solution d). The remaing singular solutions correspond either to the related binary systems
(A,C) (solution e) or (A,B) (solution f), i.e.:
and:
Solution a) represents a glass with only A configurations, identified with the network former
(for example v Se). Solution b) and c) correspond to a system made of 100% B or C
configurations and are repulsive when their stability is analyzed. Solution d) is identified
151
lution a), pB = 1, pC = 0 (solution b) and p
B
= 0, p
C
= 1 (solution c) which correspond to
with a half mixture of these LSC. It is interesting to note that the solutions e) and f) have
the same expression as the obtained for the binary glass, with N = 2. Of more interest is the
saddle point solution made of a mixture of A, B and C which is the ternary glass solution
(solution g):
(27)
(28)
The latter solution is located inside the symplex, but its existence depends on the value of the
involved structural and energetical factors and The other useful solutions
are the binary (A,B) and (A,C) solutions e and f. If the system can be considered as
phase separated, and composed of a mixture of the two binary systems. As a consequence,
such kind of networks should display two glass transition temperatures, satisfying the ad-hoc
normalized solutions (25) and (26).
The application of stochastic agglomeration theory with N = 3 involves three indepen-
dent parameters, which are the energy differences: and
However, two bond energy differences can be computed by considering the related A B and
A C binary systems, and performing on solutions (25) and (26) the limit or
in a manner similar to what has been done for the case N = 2. We find again the energy
difference (14) and:
(29)
We have only one free parameter left, the Boltzmann factor involving the bond energy
difference This parameter can be determined by considering the derivatives of
with respect to and in the limit of low modification:.
(30)
and a similar form is obtained for the derivative with respect to y, when y = 0.
(31)
with:
(32)
152
where and and refer to the above mentioned bond energy
differences, but either with glass transition temperature (when z = x) or to
(when z = y). They have been inserted in order to avoid a cumbersome presentation.
If one considers the limit values of the slopes (30) and (31) when x 0 and y 0, one
should recover the results of the slopes for binary (A,B) and (A,C) glasses. This means that
these equations should have the following form, either identical to (17) or symetric in (B,C):
(33)
(34)
where is still the limiting glass transition temperature of the pure A LSC network. In
the limit x = y = 0), both equations are second order linear equations in since
and There is only one
solution satisfying simultaneously both equations (33) and (34), which is:
(35)
The last energy difference to be computed is therefore:
(36)
With all energy differences established, one is able to have access to the value of the glass
transition temperature with respect to for different concentrations of B and C LSCs. Also,
and will depend only on the Boltzmann factors and
(37)
(38)
An interesting point is the prediction of the glass transition temperature with the average
coordination number of the network. The average coordination number of the network is
defined by:
(39)
which leads to the final expression:
(40)
with: and The relationship (38) is, again,
parameter-free and can be plotted for any bonding numbers r and given the initial glass
transition temperature of the chalcogenide network. We compare now the results of the
stochastic agglomeration theory for N = 2 and N = 3 with experimental data on chalcogenide
network formers. This is accomplished by using equations (17) and for network systems, and
equ. (40) for the ternary glasses.
153
Figure 4. Glass transition ratio in IV-VI network chalcogenide systems, as a function of the concentra-
tion of the element of Group IV. Data are taken from [3,15-18]. For a clearer presentation, the telluride systems
have been shifted upwards with an arbitrary value of 2. The solid lines represent the slope equation (17) with
and
COMPARISON WITH EXPERIMENT
Figure 4 gives the glass transition temperature prediction in chalcogenide network for-
mers of the type involving an element of Group IV and a chalcogenide
(Group VI, Consequently, for all the systems displayed in the figure, following equa-
tion (17) the slope is We can observe that this value is in a very accurate agreement
with the experimental data, and for all and However, it is interesting to remark
that only the sulphide and selenide systems exhibit a systematic deviation around the concen-
tration of x = 0.2, to be put in contrast with the and systems for which
the variation remains linear and for which the slope equation (17) agrees over the entire
displayed concentration range.
The reason for this difference is the following. Vitreous selenium or sulphur consist
of polymeric chains beside which some eight-membered rings can exist. The addition of
an atom of Group IV creates cross-linking between these chains in a random fashion and
globally the whole network can be considered as random. The structural modification goes
up to the characteristic concentration x = 0.2 which corresponds to the optimal glass com-
position where mechanical stability reaches its maximum. The latter behavior is very well
understood in terms of the constraint theory developed by Pillips and Thorpe which predicts
a rigidity transition when the average coordination number of the network reaches 2.4 [7,12].
For larger concentrations, the glass structure ceases to be random because it approaches the
chemical stability composition (at x = 0.33). Close to this value, the local structure will be
stoichiometrically balanced and stable crystalline compounds or can be
formed. Therefore, the local structure of these glass systems in the range [0.2,0.33] will be
very close to the crystalline counterpart and each chalcogen will tend to be surrounded by
two germanium atoms and vice-versa. In any case, the description in terms of random A-A
and A-B bondings will become inappropriate (more and more A-B connections are possible),
and the glass transition temperature will vary superlinearly, in systematic deviation with the
154
Figure 5. Glass transition ratio in V-VI network chalcogenide systems, as a function of the concen-
tration of the element of Group V. Data represent the systems P S [20], P Se [21], Bi Se [22] and As Se
[23]. The solid line represents the slope equation (17) with and The dashed and dotted lines
correspond to the slope (41) taking into account four-fold coordinated species with a fraction of 0.3 and 0.5.
predicted slope (17).
In the telluride systems, there is no chemical stability composition at x = 0.33. The
reason is that there are neither crystalline nor compounds, but and GeTe
instead, corresponding to the respective concentrations x = 0.4 and x = 0.5, But there is
still a rigidity transition (mechanical stability) at x = 0.2 which can be detected either by
observing the coalescence of the crystallization temperatures of the floppy and rigid parts of
the network (for [15]) or the reversing heat flow window (for [19]). As
a consequence, the chemical stability composition does not influence the local structure as it
does for the selenide and the sulphide glasses, and therefore the network can remain random
over a larger concentration range, as predicted by the linear increase from the slope (17). We
are not aware of more data on these systems, but we believe that the deviation should occur
close to x = 0.4.
We now turn to the systems involving an element of Group V (a priori, supposed
and one can clearly observe the same kind of agreement, at least for the bismuth selenide and
the phosphorus sulfide glass. For the arsenic and the phosphorus selenides, there can be a
reasonable prediction in the very low concentration limit, when the concentration x is less
than 0.15. Nevertheless, the chosen coordination number is in contradiction with
the observed structure (orpiment-like for As, i.e. three-fold coordinated units) of
these glasses. Thus, one should reasonably expect which does not predict the
trends at all. However, recent experimental results on both glasses (NMR [21] and modulated
DSC [24]) suggest that at low concentration, there should be also four-fold coordinated units
These units share a double As=Se (or P=Se) bond, which has to be taken
into account in the stochastic agglomeration theory. Indeed, the double As = Se bond has a
different statistical contribution than the single bonded P Se. The slope (17) is then
slightly modified [22]:
(41)
155
Figure 6. Glass transition in ternary network selenide systems, as a function of the average coordination
number. Data represent the systems Ge As Se [13,25] and Ge Sb Se [3]. The solid line represents the
inverse of equation (40) with and and
where represents the fraction of the four-fold species. If goes to zero, one obtains the
slope for III-VI systems. The value of yields the correct trend for the glass transition
temperature variation in glasses (dotted line in fig. 5), a result which is confirmed by
NMR and Raman spectroscopy [22]. Similarly, one expects a rate of in the arsenic
system (dashed line in fig. 5) and spectroscopic investigation is currently in progress. In
the systems having really a five-fold coordinated LSC (such as bismuth), the deviation of
the stochastic prediction of (17) is here also supposed to occur at the average ccordination
number of < r > = 2.4, which corresponds to the concentration x = 0.13. One can remark
that the bismuth system exhibits the systematic deviation around this concentration.
In the next figures, the prediction of in ternary glasses displays the same agreement as
for the binary glasses, and the systematic deviation occurs, again, around the value < r > =
2.4. In fig. 6 we have represented data from the ternary IV-V selenides. In these compounds,
the As and Sb atoms are three-fold coordinated, whereas the germanium is still four-fold
coordinated, thus and Fig. 6 shows also that is a universal function of < r >,
described by the inverse of equ. (40) up to and somewhat beyond the stiffness transition, a
feature which was already pointed out by Angell and co-workers some ten years ago [13], The
effects depending specifically on composition occur only below this value of 2.4. Stochastic
description in terms of As-Se, As-Ge, Se-Se and Ge-Se bonds fails beyond this value, because
of the chemical organization of networks in terms of stoichiometric balanced structures (e.g.
or in the As-Ge-Se compound). Surprisingly, ternary Sb-Ge-Se glasses seem
to support the random network picture up to < r > = 2.4 although the binary Sb-Se system is
already chemically ordered at very low Sb concentrations (formation of clusters) [26].
Again, the similar behavior of the As and Sb glasses clearly confirms the major influence of
connectivity, although the bond strength of these atoms is different.
In the analogous IV-V telluride systems, displayed in fig. 7, we can note the same kind of
behavior, although less data are available in the literature. Note that the stochastic description
(and still is accurate in the low modified limit, when the average coordination
number of the network is lower than 2.4. Deviation occurs also around the stiffness transition
156
Figure 7. Glass transition in ternary network telluride systems, as a function of the average coordination
number. Data represent the systems Ge As Te, Ge Sb Te [27], Si As Te [28] and Si P Te [29].
The solid line represents the inverse of equation (40) with and and
and certain data become strongly composition dependent for < r >> 2.4. However, the
solid lines represented here provide a useful information on the absolute magnitude of the
glass transition temperatures, which should be measured in a manner similar to the As-Ge-Se
system. Finally, equation (40) with and permits also to investigate germanium
incorporated chalcohalides [9,26].
GENERATING MEDIUM RANGE ORDER AND SELF-ORGANIZATION
We have already mentioned that the systematic deviation of the stochastic agglomera-
tion prediction in the glass transition temperature trends was due to the occurence close to
the rigidity transition of larger structural correlations, so that a description in terms of only
A-A and A-B bondings was not valid anymore. On a microscopic level, this means that the
modifier LSC B will not be randomly diluted inside the A network. It seems that for a large
class of systems, mechanical rigidity onsets progressively and that stressed-free (isostatic)
regions could emerge inthe network backbone. The signature of this phenomenon has been
detected by experiments on different chalcogenide glasses [16,25] and by computer simula-
tions [30]. The presence of rings may also play an important role and their existence should
modify substantially the nature of the rigidity transition.
How can agglomeration theory take into account such effects ? We can extend the
construction described at the beginning of this paper to larger structures. One can indeed
imagine that during glass transition, multiple bonds (two, as in the panel c of fig.l, but also
three, four, etc.) will be formed during the time step
0
, in the region (C).
The extension has several advantages. On the one hand, it will give a more realistic pic-
ture of what is really happening during the glass transition, where collective rearrangements
(a number of bonds created during a finite time step) occur close to the transition and seem
to be mostly responsible for the huge increase of the relaxation time. On the other hand,
we have a systematic way [31] to construct structures with medium-range order (fig.8), since
each new multiplet created takes into account a corresponding energetical factor similarly to
157
Figure 8. From doublets (A-A and A-B) to quadruplets (A-A-A-A, A-A-A-B, etc.) in a IV-VI network glass.
The numbers represent the statistical weight W of each structure.
equ. (3 ) and a statistical weight W that can be computed from the coordination number of the
involved LSCs. Last but not least, the construction permits to nucleate rings of growing size,
which play the key role in the rigidity transition of IV-VI network glasses, such as
[13].
We have performed the construction of multiplets in this type of glass up to structures
having five atoms, allowing the creation of 4 and 5-membered rings. Both are present in
the celebrated chalcogen-rich ribbon layer structure [11], which has been used in order to
explain the chemical and physical properties of v One can remark that the statistical
weight of rings and large B correlated structures is much larger than simple (A,B) cross-
linked structures (2880 and 1920, against 256 or 768). We can again compute for each size
of multiplets the probability and solve an equation of type (2). What we expect is that
the growing size of the multiplets will enable us to give the prediction of on a increasingly
larger concentration range.
Several results of this analysis are interesting. First, one recovers for all the different
sizes the energy difference (14) in the limit where p(0) = 0. Secondly, in case of dendritic
growth (no rings at all) the prediction diverges even more rapidly than for the A-A, A-B con-
struction. Rings have therefore to be taken into account. The ring energy is the parameter
which is used in order to fit the solution of (2) with the experimental data on
From the determined value of we can then compute the ring statistics of this glass, as a
function of the concentration (fig. 9). Both kinds of rings (4 and 5-membered rings) exhibit a
monotonic increase with concentration. They should therefore be present in large proportions
at the stoichiometric glass composition. We believe that this construction is worth to be devel-
oped to larger structures where rings of larger sizes will be produced. Entropic (through the
weights W) and energetical factors (through equation (14)) will select the dominant struc-
tures and give substantial information about medium range order and self-organization in
these network glasses.
158
Figure 9. The ring statistics in glasses as a function of concentration, obtained from the stochastic
agglomeration of quintuplets. The solid line corresponds to 4-merabered rings, the dashed line to 5-membered
rings.
ACKNOWLEDGMENTS
LPTL is Unit Mixte de Recherche du CNRS n. 7600. It is a pleasure to acknowl-
edge continued discussions with P. Boolchand, R. Kerner, G.G. Naumis, J.C. Phillips and
M.F. Thorpe.
REFERENCES
1. Angell, C.A. (1995) Formation of glasses from liquids and biopolymers Science 267, 1924.
2. Modulated DSC Compendium (1997) Reprint TA-210, TA Instruments, Inc., New Castle, DE
3. Feng, X.W., Bresser, W.J. and Boolchand, P. (1997) Direct evidence for stiffness threshold in chalco-
genide glasses, Phys. Rev. Lett. 78, 4422.
4. Elliott, S.R. (1989) Physics of amorphous materials, J. Wiley & Sons, New York.
5. Sreeram, A.N.,Varshneya, A.K. and Swiler, D.R. (1991) Molar volume and elastic properties of multi-
component chalcogenide glasses, J. Non-Cryst, Solids 128, 294.
6. Gibbs, J.H. and Di Marzio, E.A. (1958) Nature of the glass transition and the glassy state, J. Chem. Phys.
28, 373.
7. Phillips, J.C. (1979) Topology of covalent non-crystalline solids I: short range order in chalcogenide
alloys, J. Non-Cryst. Solids 34, 153.
8. Kerner, R. and Micoulaut, M. (1997) On the glass transition temperature in covalent networks, J. Non-
Cryst. Solids 210, 298.
9. Micoulaut, M. (1998) The slope equations: a universal relationship between the local structure and glass
transition temperature, Eur. Phys. J. B1, 277; Kerner, R. (1998) A theory of glass formation, R.J.
Elliott, M. Balkanski Eds, World Scientific Press, Singapore; Micoulaut, M., Naumis, G.G. (1999) Glass
transition temperature variation, cross-linking and structure in network glasses: a stochastic approach,
Europhys. Lett 47, 568.
10. This is another reason why SRO structures are more adapted than larger structures in this description. On
can indeed reasonably assume that diffusion is only possible for atoms or small molecules.
11. Phillips, J.C. (1982) The physics of glass, Physics Today 35(2), 27.
12. Thorpe, M.F. (1983) Continuous deformations in random networks, J. Non-Cryst. Solids 57, 355.
159
13. Tatsumisago, M., Halfap, B.L., Green, J.L., Lindsay, S.M. and Angell, C.A. (1990) Fragility of Ge-As-Se
glass formi ng liquids in relation to rigidity percolation and the Kauzmann paradox, Phys. Rev. Lett. 64,
1549.
14. Asokan, S., Prasad, M.Y.N., Parthasarathy, G. et al. (1989) Mechanical and chemical thresholds in IV-VI
chalcogenide glasses, Phys. Rev. Lett 62, 808.
15. Norban, B., Pershing, D., Enzweiler, R.N., Boolchand, P., Griffiths, J.E. and Phillips, J.C. (1987)
Coordination-number-induced morphological structural transition in a network glass, Phys. Rev. B36,
8109.
16. Selvanathan, D., Bresser, W.J., Boolchand, P. and Goodman, B. (1999) Thermally rversing window and
stiffness transitions in chalcogenide glasses, Solid State Commun. 111, 619.
17. Saffarini, G. (1994) Glass transition temperature and molar volume versus average coordination number
in bulk glasses Appl. Phys. A59, 385.
18. Sarrach, D.J., deNeufville, J.P. and Haworth, H.L. (1976) Studies of amorphous Ge-Se-Te alloys (1):
preparation and calorimetric observations J. Non-Cryst. Solids 22, 245.
19. Boolchand, P., private communication
20. Feltz, A., Aust, H. and Blayer, A. (1983) Glass formation and properties of chalcogenide systems-
permittivity and the structure of and J. Non-Cryst. Solids 55, 179.
21. Saiter, J.M., Ledru, J., Saffarini, G. and Benazth, S. (1996) About the coordination number of indium in
glasses. Materials Lett. 28, 451.
22. Georgiev, D.G., Mitkova, M., Boolchand, P., Brunklaus, G., Eckert, H. and Micoulaut, M. (2000) Molec-
ular structure, glass transition temperature variation, agglomeration theory and network connectivity of
binary P-Se glasses, Phys. Rev. B, submitted.
23. Myers, M.B.and Felty, E.J (1967) Inorganic polymers, Materials Reser. Bull. 2, 535.
24. Georgiev, D.G., Boolchand, P. and Micoulaut, M. (2000) Rigidity transitions and molecular structure of
glasses, Physical Review B, submitted.
25. Wang, Y., Boolchand, P. and Micoulaut, M. (2000) Glass structure, rigidity transitions and the intermedi-
ate phase in the Ge-As-Se ternary, Europhys. Lett., submitted.
26. Tonchev, D., private communication.
27. Lebaudy, P., Sailer, J.M., Grenet, J., Belhadji, M. and Vaulier, C. (1991) Idenlification of amorphous
zones ine the GeTeSb system Mater. Sci. Eng. A132, 273.
28. Dai, Y., Chen, S.W., Chen, C.H. and Wang, J.L. (1999) Glass formation, density, microhardness and
optical properlies of As-Si-Te alloys J. Non-Cryst. Solids 255, 217.
29. Wilschas, M., Regelsky, G. and Eckert, H., (1997) NMR studies of phosphorus in Si-P-Te and Ge-P-Te
glasses J. Non-Cryst. Solids 215, 226.
30. Thorpe, M.F., Jacobs, D.J., Chubynsky, M.V. and Phillips, J.C. (2000) Self-organizalion in nelwork
glasses, J. Non-Cryst. Solids 266-269, 859.
31. Micoulaut, M., Kerner, R., dos Sanlos-Loff, D.M. (1995) Statistical modelling of structural and thermo-
dynamical properties of vitreous J. Phys: Cond. Matt. 7, 8035.
160
FLOPPY MODES EFFECTS IN THE THERMODYNAMICAL
PROPERTIES OF CHALCOGENIDE GLASSES
GERARDO G. NAUMIS
1
1
Instituto de Fisica, Universidad Nacional Autonoma de Mexico,
Apdo. Postal 20-364, 01000, Mexico D.F., Mexico.
INTRODUCTION
One of the most fascinating problems in solid state theory is the glass transition (GT),
where a non-crystalline amorphous solid (a glass) is formed by supercooling a melt. Not
all materials are able to form glasses, and many different semi-empirical criteria have been
proposed in order to explain the ability of a material to reach the glassy state [ 1 ]. This means
that there are many factors involved in the process. Of these, an important one is the time
available for crystallization of the material with respect to the cooling rate, since the crystal
has a lower free energy with respect to the glass. Thus, the melt needs to be cooled fast
enough to avoid thermodynamical equilibrium. Due to this reason, the GT is usually not
considered as a true phase transition, although there are discontinuities in the specific heat
(C
p
(T)) and the thermal expansion coefficient at constant pressure that lead to the
idea of an underlying phase transition. Another criterion for the GT is the development of
shear elasticity and a cessation of viscous flow [1]. As we will see, this increase in shear
viscosity is often exponential with respect to the temperature (Arrhenius law) or follows the
Vogel-Fulcher-Tammann law [2], The type of the viscosity behavior is known as fragility:
glasses that follow the Arrhenius law are known as strong glasses, while the others are called
fragile [3].
At a microscopic level, the development of shear viscosity is due to a slowing down
of the atomic diffusion which is responsible of atomic rearrangements and structural relax-
ation. Below GT, the atomic structure is fixed so the contributions of structural relaxation to
thermodynamical quantities is nearly absent [1]. One can say that the liquid stops exploring
ergodically the available phase space (determined by the energy landscape) and stays in a
region for a long-time. The temperature where this phenomena happen is known as the glass
transition temperature (T
g
).
A lot of attention has been devoted to correlate T
g
with physical and chemical properties
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 161
in different glasses. Among the most studied systems, we can cite the chalcogenide glasses
(formed with elements from the column VI of the periodic table, like Se or S), which are a
bench-mark test for the understanding of the GT. For example, T
g
and (the jump in the
specific heat during the GT) of a chalcogenide glasses can be raised or lowered by adding
impurities like Ge or As, and the fragility of the glass can be changed from fragile to strong
[4].
For the change of T
g
with the composition, a method based on the statistics of agglom-
eration processes [5-9] succeeded in obtaining the empirical modified Gibbs-DiMarzio law,
that allows to calculate the change of T
g
as a function of the concentration of modifiers [9].
This stochastic method predicts the value of the characteristic constant that appears in the
Gibbs-DiMarzio law for almost any chalcogenide glass, and gives a topological explanation
[5] for the origin of T
g
. The changes of and fragility as a function of the composition has
been less studied, but there are some experimental works about the subject [4,11,12]. A care-
ful study of these experiments can reveal many interesting features of the GT as a function of
the chemical composition.
The constraint theory introduced by Phillips [13] and further refined by Thorpe [14-15]
has a very fundamental role in these problems. Phillips considered the covalent bonding as
a mechanical constraint experienced by the atoms in order to explain the ability for making
a glass. As we will see in the next section, when the number of constraints is less than the
degrees of freedom given by the dimensionality of the space, the lattice has zero frequency
modes that are called floppy. However, altougth constraint theory has been very sucessfull
in describing qualitative features of GT, not so much effort has been done in order to test the
theory in a quantitative way. In this work, we will discuss the implications of the constraint
theory approach in terms of the thermodynamical properties of the glass. The main idea is to
connect the entropy of the liquid phase, with the range of possible disordered structures, since
rigidity percolation operates more effectively in the liquid state of the chalcogenide systems
[ 11] . This idea also allows to explain the change in fragility of the glass, which in some sense
gives an idea of the di ffi cul t y degree for making the glass. The layout of this work is the
following. In the next section we present a brief introduction to the constraint theory and
floppy mode counting. In section 3 we present a discussion of the thermodynamical implica-
tions of the floppy mode approach and a comparison with the experimental data. Finally, in
section 4 we give the conclusions.
CONSTRAINT THEORY AND FLOPPY MODES
Covalent bonded atoms, like those found in chalcogenide glasses, follow the 8 N rule,
where N is the number of valence electrons and the coordination number of the atom r is given
by 8 N. For example, a Se atom has r = 2, since it is in the group VI of the periodic table.
The coordination number of the atoms that are present in the glass, allows the construction of
an average coordination number < r >, which in some sense is the coordination of a pseudo-
atom forming a structure whose topology is similar to that of the system [2]. For a ternary
alloy of the type A
x
B
y
C
1xy
(as Ge
x
As
y
Se
1xy
), < r > is,
while in general, if the total number of atoms is N and there are n
r
atoms with coordination
r, the average coordination number is defined as,
In 1979, Phillips proposed a connection between glass-forming ability and the average
coordination number [13]. The idea is that the tendency for glass formation is maximized
162
when the number of mechanical constraints N
c
in each atom is equal to the number of degrees
of freedom N
d
given by the dimensionality of the space. As we will see, N
c
is related with
the average coordination number. A system with more constraints than N
d
is overconstrained
and cannot easily form a glass.
In a covalent network, the atoms are constrained due to the fact that the interatomic
lengths and angles are well defined, altougth there are small departures from the equilib-
rium position. The strain potential energy of these deviations can be expressed as sum of
contributions from bond-stretching and bond bending forces,
where and are the bon-stretching and bond-bending force constants, and and are
small deviations from equilibrium in bond-length and bond-angles respectively.
Using this potential, one can solve the movement equations by finding the eigenmodes
and eigenvalues of the dynamical matrix. Since the matrix is of size 3Nx3N, it has 3N
eigenvalues which corresponds to the frequencies of oscillation of each eigenmode Of
these modes, 6 are zero, since correspond to a symmetry in the translation of the center
of mass and an arbitrary rotation in space of the whole system. However, for low average
coordination the lattice can be deformed without cost in energy since there is no term in the
potential that couples to the dihedral angles [15], and thus the rank of the dynamical matrix
is reduced. These eigenmodes with zero frequency are called floppy. They give the number
of possible deformations of the network without a cost in energy.
Following the work of Phillips, Thorpe estimated the number of constraints by using
the Maxwell counting [14], which is a procedure originally devised to determine the stability
of a network of rods connected with pivot joints. In this counting, each r coordinated atom
is associated with r/2 bond-stretching constraints, since there are r bonds, each of them are
shared by two atoms. In each atom, there are 2r 3 constraints that come from the angular
forces. Using this counting, the fraction of floppy modes with respect to the degrees of
freedom is,
Of special importance is the point < r >= 2.4, where the number of floppy modes is zero.
This point was recognized by Thorpe to be a rigidity transition [14], between a floppy net-
work and a rigid one; it corresponds to a strong tendency for making a glass. In some sense,
this transition is akin to the percolation transition, but rigidity percolation is more difficult
to study, since it is a vectorial and non-local problem. However, the counting presented here
works remarkably well except near the transition point, where more sophisticated methods,
like the pebble game are needed [16].
THERMODYNAMICAL EFFECTS OF FLOPPY MODES
In this section we will study of the effects of floppy modes in the thermodynamical
properties during GT. We start by considering the effect of the constraints or floppy modes in
the internal energy of the glass if they were really at zero frequency. To make this calculation,
we write the Hamiltonian of the glass in terms of the normal modes of vibration (Q
i
) that are
deduced from the dynamical matrix given by the potential,
where P
i
and m
i
are the momentum and mass of the particle i. The most important point
in the last equation, is that the sum in the potential energy is not carried over all the 3N
163
modes. In this case, each floppy mode does not contribute because in principle it has zero
frequency. Using this Hamiltonian, we can obtain the internal energy of the glass by using
the equipartition theorem of statistical mechanics, where at high temperature each degree of
freedom contributes with 3kT/ 2 to the internal energy,
( i t is worthwhile mentioning that the degrees of freedom in statistical mechanics correspond
to constraints in rigidity theory). The first term is the contribution of the kinetic energy, and
the second is the contribution of the potential energy. The specific heat is,
As is known, the solids at high temperatures follow the Dulong-Petit law, which shows that
the specific heat is cal/mol k. In the previous result, we have obtained a reduction
of the specific heat given by 3Nkf /2. However, since the glass is a solid, it is difficult to
imagine that it does not follows the Dulong-Petit law. Furthermore, different experiments for
glasses show that C
p
is 6cal/mol k independent of < r > [11]. Thus, we are lead to the
conclusion that floppy modes do not have zero frequency, instead they are shifted to small
values close to zero due to the non-linearity of the potential or the residual Van-der Waals
forces. This conclusion that comes from a thermodynamical argument, is supported by an
inelastic neutron scattering experiment, where the density of states was measured
in a Ge
x
Se
1x
, and Ge
x
AS
y
Se
1xy
glass, and it was found a blue-shift of the floppy modes,
peaked around 5meV [17]. Note that although at high temperatures the floppy modes does
not have any effect in the specific heat, at very low temperature we can expect changes in the
thermodynamical properties with the composition, since for a given temperature, the number
of accessible states is bigger in the floppy glass.
In the l i qui d melt, floppy modes are expected to be more important in the thermodynam-
ical properties, since the system can explore a wider energy landscape due to its extra thermal
energy. Floppy modes produce regions of less rigidity, and these regions are more suitable to
produce a richer landscape. The idea that we are going to develop, is to relate floppy modes
wi t h the presence of local regions with less stress in the system, i.e., we relate low-frequency
vibrational modes with disorder, since configurational modes have been shown to explain the
additional specific heat of the melt [18], and these configurational modes are identified with
the degree of stress. The rigidity of the network is related to how amenable is the glass to
continuous deformations [15], and thus floppy modes correspond to motions in the system
that require only very small energies.
In principle, floppy modes are eigenvectors of the dynamical matrix with zero frequency
in the glass, but in the supercooled and normal liquid phase, this matrix is not well defined,
except if we consider that the eigenrnodes can be identified with instantaneous normal modes
[18]. However, rigidity is a static concept, involving virtual displacements, so while it is
useful to use a dynamical matrix for a given potential, any potential would give the same
results for geometric aspects of rigidity [16].
To test these ideas, we start by evaluating the strain energy of the liquid near T
g
as a sum
of three contributions,
U(melt) = U
kin
+ U
harm
+ U
conf
, (8)
where U
kin
is the contribution from the kinetic energy of the atoms, U
harm
comes from the
harmonic vibrations and U
anharm
is the contribution from anharmonic configurational inter-
action. The contribution from the kinetic energy is given by the equipartition theorem which
gives kT/2 for each degree of freedom. Since we have 3N particles, U
kin
= 3Nk T/ 2. For
the harmonic contribution, we must remember that liquids cannot withstand shear stress,
164
and thus they cannot sustain transverse modes of vibration, therefore, they have only N vi-
brational modes, corresponding to longitudinal phonons. The contribution from this term
is U
harm
= NkT/2. Observe that we are counting all the 3N modes, instead of writing
U
harm
= (N 3 Nf ) k T/ 2 , since we already noticed that floppy modes have finite frequen-
cies. The third term represents the interaction which is responsible for nucleation and crys-
tallization. One can think that near T
g
, it is reasonable to suppose that U
config
must be of the
same order of U
kin
= 3NkT/ 2, since the energetic barrier for configurational motion needs to
overcome the thermal barrier. Then, one obtains for the internal energy,
and for the j ump in the specific heat during glass transition (the difference between the spe-
cific heat of the glass compared to that of the melt),
Surprisingly, this result coincides with the value reported for Ge
x
Se
1x
by Boolchand et. al.
[12] and Ge
x
As
y
Se
1xy
by Tatsumisago et. al.[4] when < r >= 2.4, as can be seen in Fig.
1. In fact, when < r >= 2.4, is minimum, and the configurational entropy S
c
, calculated
from the following expression,
where T
k
is the Kauzmann liquid-crystal isoentropy temperature, is also a minimum. Thus,
for < r >< 2.4 or < r >> 2.4, we are underestimating the contribution from relaxation or
configurational rearrangements. At this point, we can introduce the idea that constraints in
the lattice produce regions with different degrees of stress. For the underconstrained melt,
it is reasonable to think that underconstrained regions have a bigger configurational entropy,
since the existence of floppy modes means that there are many equivalent structures with the
same energy. Here we propose to consider a hole activation model to describe the relaxation
modes, with the number of holes determined by the number of floppy modes. As we will see,
this model gives us a good fit of the experimental data. Before continue with this approach,
we must comment the reason of separating anharmonic and floppy modes . As was said
previously, the shape of the pair potential does not affect the existence of floppy modes, it
only changes the small finite frequency of these modes. In fact, these modes we expect to
be more related with terms in the potential that produce stress, as for example happen in
the modified soft-model, where a linear term is added in order to account for stress [18].
Anharmonic terms are related to melting, while we suppose that floppy modes are related
with relaxation of strain.
Here we will assume that U
flop
is a function of the number of floppy modes. We suppose
that each mode acts as a effective extra degree of freedom for relaxation in the region where
< r >< 2.4, since the number of floppy modes gives the different possible deformations of
the network. Thus, U
flop
is proportional to f3NkT (where f = 2 (5 < r > /6) is the fraction
of floppy modes). Observe that this idea is connected with the work of Duxbury et. al., who
showed that the number of floppy modes behaves as a free energy for both connectivity and
rigidity percolation [19]. In fact, a specific heat can be defined using the second derivative of
the free energy [19]. By adding this contribution to the previous calculation of we find
the specific heat to be,
165
Figure 1. as a function of <r >. The line corresponds to Eq. (12).
Experi ment al data from Ge
x
As
y
Se
1xy
(taken from Ref. [4]) are shown
wi t h squares. The star corresponds to As
2
Se
3
. A best fit line is shown
with a dashed line. The circles correspond to the data of Ge
x
Se
1x
from
the work of Feng et. al. [12]. The dotted line corresponds to the best fit
of their data.
This equation does not contain free parameters, and can be compared with the experi-
mental data available. In Figure 1 we plot Eq. (12), and from the experimental data of
As
x
Ge
y
Se
1xy
and Ge
x
Se
1xy;
taken from the work of Tatsumisago et. al. [4] (squares, cir-
cles and diamond), and Feng et. al [12] (circles). As
x
Ge
y
Se
1xy
and Ge
x
Se
1x
systems are
used because they are a bench-wood test for the constraint theory; specially the first system
since a given average coordination number can be reached with many different compounds.
We observe that although Eq.(12) does not have free parameters, it gives a good fit of the
experimental data. One special exception is the binary As
2
Se
3
(shown as a star), which is
atypical in the sense that it does not follow the isocoordination rule in its properties. This is
related to its raft-like structures of two dimensional aspect [4]. If we exclude this point,
then a linear regression of the experimental data from the work of Tatsumisago et. al. gives a
slope of 5.07 cal/(mol K), with a correlation of 0.993, which is very similar to the value
of 4.96 cal/(mol K) predicted by the floppy mode approach. This best fit is shown as a
dashed line in Figure 1. The data from the work of Feng et. al. [12], give a slope for the best
fit of 4.76, with a correlation of 0.985 (shown as a dotted line).
More recently, the evolution of the modulated differential scanning calorimetry (MDSC)
has made possible to obtain more information about the thermodynamical processes that
ocurr due to rigidity [20]. In the MDSC technique, a sinusoidal variation of the heat is
imprinted over the linear T-ramp that is used for decreasing the temperature in the usual
differential scanning calorimetry. The response of the system can be deconvoluted in two
contributions, one is the reversible heat flow and the other the non-reversing heat flow, which
is the one that does not follow the variations in T. The importance of this technique is that
the non-reversing heat flow gives an idea of the non-reversing processes that ocurr
in the glass-liquid transformation [21]. An important experimental observation, is that
is nearly zero for optimally coordinated glasses (< r >= 2.4), furthermore, the minimum
in is due to the non-reversing component [20]. The fact that means that the
glass stops exploring the energy landscape, and rest in a deep minimum [20]. The glass
166
transition is thermally reversing. However, as we change the composition begins
to rise, as the glass explore the landscape. Observe that these results from MDSC are in
agreement with the approach used in this work, since we assumed that rigidity is related with
the number of accessible structures of the network via relaxation. In fact, this explains why
the first estimation of cal/mol k gave the correct estimation only for < r >= 2.4,
since for other values of < r > there is an extra contribution as the system explores the
energy-landscape. It is interesting that in some systems , the rigidity transition is not very
sharp, since there are compositional windows [21,22] where for a certain region
of < r >. Another interesting feature of the thermodynamical behavior with respect to the
composition is that the volar volume of the glass mimics the shape of i.e., the network
packing is optimized at the rigidity transition [20],
For the overconstrained region (< r >> 2.4), Phillips proposed the following formula
for the total strain energy [13],
(13)
where N
d
is the dimension of the space, N
co
the number of constraints. According to Phillips,
a,b and c are constants that are estimated in the following way. a = 1/2 since it corresponds
to the greatest relaxation allowed by atom movements during the configuration freeze at T
g
.
b = 1 since it corresponds to the harmonic contribution. The third term is the anharmonic
configurational interaction among the excess coordinates, and must be of the same order of
magnitude than the first term. The expression for the strain energy can be written in terms of
the average coordination number,
(14)
However, from the data of Figure 1 we can see that seems to be linear in the overcon-
strained region. A liner regression of the data gives a slope of 4 cal/(mol K) and correlation
0.991.
The magnitude of the j ump in C
p
at the glass transition is also related with the fragility
of the glass. Strong glass forming liquids are resistant to changes in the medium range order
[2] because the amount of configurational entropy in the liquid is relatively small. Fragile
glass forming liquids have a high entropy.
In the present approach, from Eq. ( 11) is clear that S
c
has a linear dependence on < r >;
i.e., fragility is related with the number of floppy modes. Observe that this is the natural
link between the floppy mode theory and a quantitative property that measures the ability for
making the glass [23]. A key quantity that allows to classify the fragility of the glass is the
behavior of the viscosity. Fragile glasses forming liquids follow the Volger-Fulcher law [2],
(15)
where D and T
0
are constants. Strong glasses follows an Arrhenius law. However, both
behaviors are to be expected from the Adam-Gibbs equation [2],
(16)
since if is small, from Eq.(11), S
c
is almost T-independent and Eq.(16) follows an Ar-
rhenius form. The Vogel-Fulcher law is recovered from Eq. (12) when is bigger, with a
functional form of the type B/T [2], where B is a constant that must be adjusted in order to
account for the total value of during the transition. In this case, is known, and we
get,
(17)
167
Figure 2. D as a function of < r >. The squares are from reference [4].
The circles are the values obtained from Eq.( 18), and the line is a visual
guide.
T
m
is the temperature limit where C
p
begins to descent towards T
g
. Using this expression and
Eq.(16), the constant D of the Volger-Fulcher law is given by:
(18)
which is a measure of the strength of the liquid. Higher values of D corresponds to strong
glasses. The relation defined by Eq.(18) between D and < r > can be tested with the ex-
perimental data, if the constant C is fixed from one of the experimental points. For pure Se,
the experimental data [4] shows that D = 10, T
K
= 240K and to give
C = 38400 cal/mol. Using this constant and the values of T
K
and T
m
from the experiment
[4,11], from Eq.(18) we obtain the points that are shown with circles in Fig. 2. The experi-
mental data of Tatsumisago et. al. [4] are shown with squares in Fig. 2. As it can be seen,
there is a good correspondence between the prediction of Eq. (18) and the experimental data.
Another quantity of interest is the excess expansion coefficient The present ap-
proach allows to obtain its functional form, although we cannot obtain the values of the
constants. According to the free volume theory, we expect that constant, thus
is of the form:
(19)
where C
1
and C
2
are two constants. This can be corroborated in Fig. 3, where we show a plot
of the experimental data fo r in the As
x
Ge
y
Se
1xy
system [4] and the corresponding linear
regression, which has the following form: with a correlation
coefficient of 0.936.
SUMMARY
In this work, we made a discussion about the thermodynamical effects of the floppy
mode theory. For the glass, we obtained that floppy modes have a finite frequency shift,
since they follows the Dulong-Petit law. For the liquid melt, rigidity has a contribution to the
168
Figure 3. as a function of < r > from Ref. [4]. The line corresponds
to the best fit.
configurational entropy, that depends on the number of floppy modes. Using this, the jump in
C
p
during glass transition can be estimated, and the change of fragility can be also obtained.
This allows to find a quantitative measurement of the ability for making the glass in terms of
the floppy mode theory.
Acknowledgments
I would like to thank P. Boolchand for his comments and for bringing to my attention
the MDSC measurements and the windows of rigidity. Thanks also to R. Kerner and M.
Micoulaut for enlightening discussions. This work was supported by the projects DGAPA-
UNAM IN-108199 and CONACyT 25237-E .
REFERENCES
1. Jckl e J. (1986) Models of the glass transition. Rep. Prog. Phys.49, 171-229.
2. Elliot S.R.. (1989), Physics of Amorphous Materials, Wiley, New York.
3. Zallen, R. (1998) The Physics of Amorphous Solids, John Wiley & Sons, New York.
4. Tatsumisago M., Halfpap B.L, Green J.L., Lindsay S.M., Agnell C.A., (1990) Fragility of Ge-As-Se
glass-forming l i qui ds in relation to ri gi di t y percolation and the Kauzmann paradox Phys Rev. Lett. 64
1549-1553.
5. Kerner R. (1995) Two simple rules for covalent binary glasses, Physica B 215, 267-272.
6. Barrio R.A.. Duruisseau J.P., Kerner R. (1995) Structural properties of alkali-borate glasses derived from
a theoretical model. Philosophical Magazine B72, 535-550.
7. Naumis G.G., Kerner R. (1998) Stochastic matrix description of glass transition in ternary chalcogenide
systems. J. of Non-Cryst. Solids 231, 111-117.
8. Naumi s G. G. (1998) Modelling of growth and agglomeration processes leading to various non-crystalline
materials, J. of Non-Cryst. Solids 232-234, 600-606.
9. Mi coul aut M., Naumi s G.G. (1999) Glass transition temperature variation, cross-linking and structure in
network glasses: a stochastic approach, Europhys. Lett. 47 (5), 568-574.
10. Sreeram A.N., Swiler D.R., Varshneya A.K. (1991) Gibbs-DiMarzio equation to describe the glass tran-
sition temperature trends in multicomponent chalcogenide glasses, J. Non-Cryst. Solids 127, 287-297.
11. Senapati U., Varshneya (1995) Configurational arrangements in chalcogenide glasses: a new perspective
on Phillips constraint theory, J. Non-Cryst. Solids 185, 289-296.
169
12. Feng X., Bresser W.J., Boolchand P.( 1997) Direct evidence for stiffness threshold in chalcogenide glasses.
Phys. Rev. Lett. 78,4422-4425.
13. Phi l l i ps J.C. (1979) Topology of covalent non-crystalline solids I: short-range order in chalcogenide al-
loys, J. Non-Cryst. Solids. 34, 153-181.
14. Thorpe M. (1983) Continuous deformations in random networks J. Non-Cryst. Solids 57, 355.
15 He H., Thorpe M. (1985) Elastic properties of glasses Phys. Rev. Lett., 54 2107-2110.
16. Jacobs D.J., Thorpe M.( 1995) Generic percolation: the pebble game Phys. Rev. Lett, bf 75, 4051- 4054.
17. Kamitakahara W.A., Capelleti R.L., Boolchand P., Halfpap B., Gompf F, Neumann D.A., Mutka H.
( 1991) Vibrational density of states and network rigidity in chalcogenide glasses Phys. Rev. B44 94-100.
18. Zrcher U., Keyes T., (1997) Soft-modes in glass-forming liquids: the role of local stress, in Fourkas
J.T., Kivelson D., Mohany U., Nelson K. (eds.). Supercooled liquids: advances and novel applications,
American Chemical Society, Washington D.C., pp. 82-94.
19. Duxbury P.M., Jacobs D.J., Thorpe M., MouzkarelC., (1999) Floppy modes and the free energy: rigidity
and connectivity percolation on Bethe lattices Phys. Rev. B59, 2084-2092
20. Boolchand P., Feng X., Selvanathan D., Bresser W.J., (1999) Rigidity transition in chalcogenide glasses,
in Thorpe M.F., Duxbur y P.M., Rigidity Theory and Applications, Kluwer Academic/Plenum Publishers,
p.p 279-295.
21. Selvanathan D., Bresser W.J., Boolchand P., Goodman B., (1999) Thermally reversing window and stiff-
ness transitions in chalcogenide glasses. Solid State Communications 111 619-624.
22. Selvanathan D., Bresser W.J., Boolchand P., (2000) Stiffness transitions in Si
x
Se
1x
glasses from Raman
scattering and temperature-modulated differential scanning calorimetry, Phys. Rev. B61 15061-15076.
23. Naumis G.G., (2000) Contribution of floppy modes to the heat capacity jump and fragility in chalcogenide
glasses Phys. Rev. B61 R9205-R9208.
170
THE DALTON-MAXWELL-PAULING RECIPE FOR WINDOW GLASS
RICHARD KERNER
Laboratoire de Gravitation et Cosmologie Relativistes,
Tour 22-12, 4-me tage, Bote 142
Universit Pierre-et-Marie-Curie, 4 Place Jussieu
75252 - Paris Cedex 05, France
INTRODUCTION
The aim of this article is to merge together two methods of understanding and model-
ing the structural and thermodynamical properties of covalent glasses: the constraint theory
describing glasses in terms of rigidity transition [1,2], and the more recently created local
agglomeration model [3,4,10], on the particular example of silica-based glasses with alkaline
oxydes as modifiers. Although these ternary glasses, with the molecular weight composition
are the oldest and most widely used, [5,6] their local medium-
range structure is very complicated, and their modeling is much more difficult than what is to
be achieved in the case of simple network glasses such as [7-9].
Nevertheless, the knowledge of local configurations which are stable enough at glass
transition temperature, makes the task much easier. We shall also argue that the two ap-
proaches are complementary because they serve to probe the glass structure on different
scales. The constraint theory gives its best predictions when applied to the atomic level,
when the network is treated as a ball-and-stick structure, and taking into account all atoms
and bonds present in the network, whereas the agglomeration theory, whose equations ex-
press the principle of minimal local fluctuations (i.e. maximal homogeneity), takes into ac-
count the characteristics averaged over clusters with a well defined medium-range structure
(with chains and rings involving up to 25 to 30 atoms).
We shall show that many predictions of the two models often coincide, their combination
offering a better insight into the process of glass formation. New predictions can be given, in
particular, concerning the glass transition temperature as a function of chemical composition
(and local connectivity of the random network), the variation of thermal capacity during
the glass transition, and many others.
It seems very clear by now that in spite of high degree of randomness, glasses are char-
acterized by well-defined local structures, such as stiff -tetrahedra and four-, five- or
six-membered rings in silicate glasses, or stiff tripods and three-membered boroxol
rings in borate glasses, etc.
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 171
The shape and energetics of these highly organized structures depend on the chemistry
of the melt about to undergo the glass transition at the temperature
In silicate glasses, the alkaline or lime and alumina modifiers , must
occupy certain topologically well-defined positions in the oxygen matrix pre-existing in the
pure network, because no dangling oxygens can be observed later on in the resulting
glass structure. This leads to the formation of stable building blocks containing these addi-
tives, which coexist with the silicon-centered tetrahedra.
For example, the molecules split and form two truncated Si-centered tetrahedra,
whereas the CaO molecules form stable bridges linking two Si-centered tetra-
hedra together, giving rise to stable six-coordinate blocks
With these elementary blocks, whose abundance is controlled by molar concentrations
according to Daltons principles, we can perform computations based on two quite different
and independent approaches, which are the constraint theory introduced by J.C. Phillips and
M.F. Thorpe, and the stochastic agglomeration model (R. Kerner, M. Micoulaut, R.A. Barrio,
G.G. Naumis). The use of the constraint theory based on Maxwells principles requires a very
careful analysis of the character of bonds formed in the network during glass transition. This
is achieved with the help of Paulings theory of resonant chemical bonds, which attributes the
(1 /n)-th part of chemical valence to each of the closest neighbors of a given covalent atom.
The agglomeration model is able to predict the dependence of on the modifier con-
centrations not only close to i.e. pure but also at higher concentrations if one takes
into account the ring-forming ability. The homogeneity of ring distribution in silica glasses
become then an important factor for determining the optimal composition of the melt, com-
bining the highest degree of homogeneity with minimal stress and possibly the lowest glass
transition temperature.
With the help of these principles, we are also able to predict optimal chemical composi-
tions of other silica-based glasses, which combine the best proximity to the stiffness transition
with maximal homogeneity.
The article is organized as follows. First, we expose a simple version of the agglomer-
ation model applied to binary and ternary
glasses, and we show how this model is able to predict the correct behavior of the glass tran-
sition temperature as a function of modifiers concentration.
Next, we recall the definition of spinodal and binodal curves that define the immisci-
bility region for binary liquids, and compare it with the immiscibility dome observed in the
mixtures. We then draw certain conclusions concerning the average size of
clusters agglomerating during glass transition.
Finally, we show how the combination of these ideas with the application of the con-
straint theory to silicates makes it possible to derive the best compositions of silica-based
window glasses.
THE AGGLOMERATION MODEL AND IN SILICATE GLASSES
Let us first recall the simplest version of the agglomeration model set forth in Refs.
[11,12,15,16]. We assume that two different stable building blocks are present in the liquid
melt about to undergo glass transition. These elementary entities can be just atoms (e.g. Se
and Ge, or Se and As in the chalcogenide glasses, or more complicated molecules or clusters,
like e.g. the tripods and 4-coordinate boron atoms with a new
oxygen bond appended by a -ion.
The simple mathematical model of a network forming process during the glass transition
consists in building up the probabilities of creation of new bonds between given stable build-
172
Figure 1. in and glasses.
ing blocks, or microclusters, as the case may be. These probabilities contain the products of
respective concentrations of these entities in the melt, the statistical factors acknowlidging for
the number of distinct ways in which a given new bond can be created (which in the simplest
case is just the product of free valencies displayed by the building blocks under considera-
tion), and finally, the factor which takes into account the energy barrier, or chemical potential
relative to this bond creation, entering the corresponding Boltzmann factor, and perhaps an
extra factor related to the kinetics of the bond creation process (e.g. when the cross-sections
for the encounters of various pairs of building blocks are very different).
Let us analyze the first step of agglomeration process with two different elementary
building blocks present, and at very low concentration of molecules. This should de-
scribe the decrease in the glass transition temperature via addition of some amount of
to pure quartz This decrease is quite spectacular, because it is sufficient to add
no more than 4 to 5% (molar) of in order to make the glass transition temperature
fall down from 1480 K to about 1000 K (see Fig. 1).
It turns out that there is no point in adding more pure because one enters the
immiscibility domain, which extends up to the concentrations of about 15% of molar
Even if the glass transitions occur in this range of concentrations, the result is an opaque
material in which a very high degree of inhomogeneity is observed, due to the spontaneously
formed micro-domains of two disctinct types: sodium-rich and sodium-poor. This problem
will be analyzed in detail in the next section.
After leaving the immiscibility region we arrive at the concentrations of which are
higher than 15 16%. The resulting glass is again homogeneous, but it becomes chemically
unstable, i.e. is very prone to water attack.. This is why some amount of CaO must be added
to the melt, making increase again, but not very much; the positive slope of versus the
CaO concentration y at x ~ 15% is much smaller than the initial negative slope of versus
the concentration x at low values of x.
The idea of maximal homogeneity of the network suggests that during the formation of
the random glassy network from liquid melt all local fluctuations are close to the minimum.
This enables us to put forth the equations expressing this principle and relate to local con-
centrations x and y.
Let us start with binary mixture containing only two elementary
173
Figure 2. The elementary building blocks A and B and the doublets created by pairing AA, AB and BB.
building blocks : the four-coordinate A and the three-coordinate B with one dangling bond
saturated with a Na atom, as represented in Fig. 1.
If we denote by and the probability of finding one of these local configurations in
the melt about to freeze and undergo the glass transition, at low concentration we can
use the relations
Let us evaluate now the probabilities of creating the doublets, i.e. the new bonds be-
tween the elementary blocks. Supposing that the process is not very far from thermal equi-
librium (slow quenching rate), we can use Boltzmann factors with respective energy barriers
and The differences between these energies acknowledge the change in the
rigidity of the respective tetrahedrons and the repulsion between the ions. The statis-
tical weights attributed to each agglomeration are computed in an obvious way, taking into
account the valency of each block. The common denominator Q normalizes the sum of the
probabilities to 1. We get then:
where the normalizer Q is given by
Neglecting the BB-pairs at very low concentrations of the local concentration
evaluated on the doublets (denoted by reads as follows :
The agglomeration theory [3,4] provides us with a simple criterion for defining the glass
transition, identifying it with minimal local fluctuations during the bonding of elementary
building blocks creating amorphous randomnetwork. Close to x = 0 this leads to the follow-
ing equation:
174
Explicitly, this gives the following equation:
Reducing this expression to the common denominator, in which we also neglect the term
containing we obtain the following approximation for (equivalent to x ~ 0) :
Besides the obvious solution there is another one, which we can identify with glassy
state, given by
In the limit when and the glass transition temperature of pure amorphous
we must have
which leads to the relation
which fixes the difference between the energy barriers corresponding to the creation of an AA
or an AB bond. It is equal to
We may also evaluate the slope of the function describing the dependence of the glass
transition temperature on the concentration x. We should keep in mind that in the linear
approximation, when the variables and x are proportional up to a multiplicative constant,
the slope at the origin is given by the same universal expression, because for any
continuous function one has
According to the formula (its derivation can be found in the refs. [3,4,10]), we obtain
when deriving with respect to T, inversing it and going to the limit
which has a negative value, as it should be, in agreement with the experiment. Inserting the
value of we find that in the linear approximation,
the glass transition temperature goes down to 760 C = 1040 K when the concentra-
tion is only 5% high, which means that the glass transition temperature is reduced by as much
as 105 K with the addition of one molar percent of which is in good agreement with
the experimental data [19,20].
Further extrapolation of this linearized formula is of no use anyway, because when the
concentration of gets close to 5%, the resulting glass is of no use, presenting strong
opaqueness due to local phase separation - on the scale of many microns - into sodium-rich
175
and sodium-poor domains.
We can now make an estimation of another slope, with respect to the variable y, the
molar concentration of CaO (equal to in the limit y ~ 0,) at a given constant (but quite
low, i.e. 0.05) value of x.
In order to do this, we shall use again the magic formula from [3], which tells us that
the initial slope of the function versus the concentration of a modifier in covalent glasses
is given by
where is the concentration of the building blocks of the modifier with coordination number
m, and m is the coordination number of the dominant glass-former.
In our case, m = 6, because as we know, each molecule of CaO creates a strong and
rigid bond keeping together two tetrahedra, resulting in a six-fold C-unit, whereas
m should be replaced by the average coordination number of a mixture of A and B blocks,
defined (at and 1 as
Therefore, recalling that we can write
Leaving only terms linear in x in Taylors expansion of this expression yields
so that after integration we get
Even with this very rough approximation, whose validity does not go beyond 10% con-
centration of the total of both modifiers, the basic behavior comes out right. In the very
beginning, at x = 0 and y = 0, the glass transition temperature decreases with the addition
of and increases even more considerably while a small amount of CaO is added. This
is why the glass makers start always with adding more than 10% of first, and only then
add some amount of CaO.
Moreover, the positive slope with respect to y, which is very high at x = 0, y = 0, because
after inserting the numerical values, we have
becomes much smaller after adding a relatively small amount of
which means that at x = 0.05 the glass transition temperature has become 1060 K instead
of 1480 K at x = 0, and the increase in temperature due to the addition of certain amount of
CaO is now very gentle:
176
However, the above formulae are no more valid on the other side of the immiscibility
region, where the concentration of is above 15%, and the glass transition temperature
is well below 800 K. The slope of the curve (x) is now much lower than before, i.e. about
10 K per 1% of variation of x.
It is not difficult to spot the source of eventual discrepancies between the linear approxi-
mation and the experimental curve. At the glass transition temperature close to 700 800
K, which is the case at higher concentrations on the sodium-rich side of the immisci-
bility gap, the approximation that local agglomeration in melt can be described by means of
freely floating tetrahedra can not be maintained. The real situation is more complex
then, i.e. the melt certainly contains a lot of clusters with various sizes.
As soon as bigger stable local configurations do appear, the most important growth mode
is through the creation of rings, most of which are four-, five and six-coordinate, (this is dif-
ferent from what is observed in most of the crystalline phases of where the six-fold
rings dominate) The creation of all these rings can be modeled quite successfully if doublets
and triplets of tetrahedra become elementary building blocks, instead of simple Si-centered
tetrahedra or NaO-modified tripods.
The creation of rings has a much higher statistical weight than a single bond creation
producing local chains of corner-sharing tetrahedra, although this mode of agglomeration
should also be taken into account. Also the energies associated with the creation of various
rings are now different, and should be adapted in order to give a good fit with the experimen-
tal curve; the nature and value of potential energies stored in rings have been discussed in the
papers by F.L. Galeener [13] and others [14].
A model taking into account the ring-forming tendency has been developed recently,
(see [17]), and it gives reasonable predictions for the glass transition temperature in ternary
silica-soda-lime glass. Because of the complexity, no analytical formulae are available. We
shall give here only the example how the influence of the CaO-additive can be evaluated with
new statistical weights.
If the ring-forming process is dominant, with three elementary building blocks A, B
and C, the last one representing the O Ca O bridge with two Si-centered tetrahedra on both
sides, we should replace the valencies m and m by statistical factors defining the ring-forming
ability of each of the above building blocks. It is given by the corresponding binomial coef-
ficients, and respectively.
Therefore, the ring-forming ability of tetrahedron is equal to 6, the ring-
forming ability of a truncated (3-fold) unit with one NaO group is equal to 3, whereas the
6-forld unit C can give rise to 15 rings.
For the relatively low concentrations of CaO we can still maintain the approximate
equality Considering the binary mixture as a uniform substrate,
and the CaO additive as modifier, we can introduce the average ring-forming ability of the
binary as
Then, inserting this value and the value 15 (the ring-forming ability of calcium-centered six-
fold unit C) into the same universal formula for the slope of but with replaced now by
we obtain the following result:
which leads to the following linear approximation fr the sodium-rich domain:
177
which is fairly well confirmed by the experimental data.
IMMISCIBILITY, SPINODAL AND THE AVERAGE CLUSTER SIZE
The existence of the immiscibility dome for the binary is well
known by all glass-makers. It represents a convex parabola-shaped curve in the (x, T) -plane
cutting the x-axis at x = 0 and at 0.2, attaining its maximum at about 1200K at x
between 0.06 and 0.08.
When the glass transition temperature goes below this curve (as it happens close to the
low concentrations of at x as low as 4 5%), it becomes extremely difficult to obtain
a homogeneous glassy network; one observes an opaque material in which partial demixtion
took place, producing microscopic inhomogeneous regions of two kinds, sodium-rich and
sodium-poor ones.
Let us demonstrate how the shape of this curve, called a spinodal, can be derived by
means of simple thermodynamical arguments, and how its parameters (i.e. its width and the
position of the maximum) give a clue about the average cluster size in the liquid melt close
to the glass transition temperature.
From our previous analysis it is clear that in a binary melt, at low
concentrations of only two main kinds of clusters are important, whatever their aver-
age size might be: the agglomerates containing a NaO unit attached to one of the Si-centered
tetrahedra, and alternatively, the agglomerates composed of pure this because at low
values of x, the probability of two NaO groups coming together is extremely low. This is why
we shall pursue our analysis with this simplifying assumption.
Assuming that the potential energy of interaction between two units (denoted symboli-
cally by A and B) coming close enough to each other, is also proportional to the number of
dangling bonds, we can evaluate its contribution to the internal energy of the hot liquid as
follows. Let the average number of free bonds available in an A-type unit be and re-
spectively, for a B-type unit, we shall suppose that the potential energies of interacting
elementary agglomerates (but without a stable bond forming) are and These
are not the same as the previously introduced activation energies and corre-
sponding to bond formation during the glass transition; they might be different, and they are
counted with the opposite sign.
Then, if we denote by (1 c) the concentration of the A-type clusters, and by c the con-
centration of the B-type clusters, the potential energy stored in the bonds created around and
average elementary cluster is:
so that the total potential energy per building block in the mix is:
where the factor is put in front in order to count each bond only once.
Now, in the case of pure A or pure B configurations present, we would have found
Therefore, the extra potential energy resulting from mixing is readily evaluated to be
178
The extra entropy resulting from mixing is given by the usual expression
so that the Gibbs free energy variation becomes
with
According to the laws of thermodynamics, which we believe to be still applicable here, the
system is stable against the separation into two different phases if the following inequality is
satisfied:
where denotes the usual chemical potential function. More explicitly, we have
If and if T is above the critical value, there are two solutions to the equation
defining two inflexion points of the curve (c), called spinodal points, which we shall de-
note by and These points are found an equal distance from the central point due
to the obvious symmetry of the above equation.
The equation which gives the condition of vanishing first derivative of is also sym-
metric with respect to the substitution
When the temperature T is above the critical temperature it has also two solu-
tions, called binodal points, which we shall denote by and With this in mind it is easy
to draw the curves for a given temperature T (Fig. 3).
The pairs of binodal and spinodal points, and respectively, enable us
to draw two curves which give the implicit relation between these solutions and the corre-
sponding values of temperature T. These curves are called the respectively the binodal and
the spinodal; they meet at the critical temperature attained when
(Fig. 3).
According to the first principles of thermodynamics, in the range between the spinodal
and binodal curves the system is stable against the phase separation, whereas inside the spin-
odal curve it becomes unstable, and local phase separation is preferred.
The immiscibility dome for the binary as measured by different
authors, is displayed in Fig. 4. below. It is easy to see that it is not perfectly symmetric
around the maximum which is attained for ythe value 7 8%; upon extrapolation to-
wards higher values of the molar concentration x we see that the extreme value of the
spinodal curve (corresponding to c = 1 is attained at x slightly more than 20%.
This provides us with another clue concerning the average size of entities which should
serve as a basis for the calculus of the spinodal, and at the same time, as the elementary
179
Figure 3. a) The shapes of b) The binodal and spinodal curves.
and meta-stable building blocks which to be used in the agglomeration theory calculations.
Quite obviously, the hot melt can not be arbitrarily divided into just two types of clusters,
with and without sodium ions attached, although at very high temperatures and with a few
molar percent of the use of just two small entities in the agglomeration model, the Si-
centered tetrahedra (building block A) and the truncated tetrahedra with one ion and one
non-bonding oxygen (building block B) gave satisfactory predictions for the glass transition
temperature behavior.
However, using the same entities in the reconstruction of the spinodal leads to the ob-
vious discrepancy with the experiment; indeed, when we set and the
spinodal curve would attain its maximum at which corresponds to the molar concen-
tration of whereas the maximum possible value c = 1 would correspond to
x = 1/3 (which represents the molar rate of in the B-type building blocks, one half of
an molecule versus one molecule).
But if we assume that most of the elementary building blocks are represented by dou-
blets (of AA or AB type), then the 100% concentration of AB doublets (one half of an
molecule molecules will correspond to x = 0.2, which is much closer to the observed
immiscibility dome; moreover, its maximum will occur at c = 0.5, which corresdponds now
to x = 11.11%, which is also close to the reality.
We can go furher and postulate the presence of bigger elementary blocks, which will
move the maximum even closer to the observed value of about 7 8%. The real curve is
certainly the result of a superposition of several spinodal curves corresponding to the clusters
of various size (see Fig. 4).
The best fit can be obtained if we assume that the only two species of clusters present
are just triplets of the type AAA and AAB (together with ABA, with proper statistical weights),
in which case, setting (1 c) equal to the concentration of the AAA-triplets, the maximum of
the immiscibility curve at c = 0.5 will correspond to x =1/13 7.7% of molar
It is worth noticing that the presence of doublets and triplets of Si-centers tetrahedra is
enough to acknowledge the formation of rings of all sizes up to the 6-fold ones (6 Si atoms
connected via oxygen bonds).
180
Figure 4. The immiscibility dome, after [22].
RIGIDITY, CONNECTIVITY AND THE RING DISTRIBUTION
As argued in previous section, the knowledge of shapes and chemical properties of the
elementary configurations that can be considered as the most stable ones at a given tempera-
ture is crucial for the subsequent construction of our model. Quite fortunately, the based
glass, with the addition of certain amount (from 12 to 16%) of and about 7 to 14% of
CaO, displays two well established tendencies:
* The addition of results in creation of saturated bonds, which amounts to local
breaking of oxygen bonds in the random network. It is to be stressed that each molecule
of creates two such bonds, which can be no longer connected to other bonds in the net-
work, thus decreasing the average coordination number . Each tetrahedron
is transformed into a tripod (or truncated tetrahedron) with coordination number = 3 :
** The addition of CaO results in creation of stable and practically undestructible units
containing two tetrahedra, connected by a O Ca O bridge. Each of these new
elementary building blocks can be considered as a stable unit,
with coordination number 6.
It should be stressed here that in both modified blocks the molecular concentration of
the modifier (be it or CaO) is exactly the same, and equal to one molecule
creating two modified blocks of and one molecule of CaO also creating
a stable bridge between two such entities. Therefore, if we denote the elementary pure
tetrahedron by A, the tripod with one saturated bond as B, and the six-coordinate doublet
linked together by the O Ca O bridge by C, we can easily define the relation between x, y
and : where and denote the respective probabilities to pick up an A, B or C
building block in the network, with
181
The inverse relations read:
Obviously, if no other local configurations can be found in the network, the maximal molar
concentration of each modifier, and CaO is equal to : which
is more than enough in order to describe real silicate glasses in which the part of is
always higher than 70%.
To be more precise, consider a very huge sample composed of N building blocks. It
contains, by definition,
A blocks, B blocks and C blocks
Each B-cluster contains one half of an molecule, while each C-cluster contains one
CaO molecule.
On the other hand, an A-block contains the equivalent of only one molecule a
B-block contains the equivalent of one molecule and a half (one and half of
finally, a C-block contains the equivalent of three molecules (two and one CaO).
Therefore, the molar concentrations of (denoted by x) and of CaO (denoted by y)
are computed respectively as:
Inserting simplifying the above expressions by the
common factor N and taking into account that we obtain
It is worthwile to notice that in the limit of very low concentrations, when both and
tend to zero, x behaves as and y as whereas when is close to 1, one has
and when is close to 1, one has
We need two independent equations in order to fix a particular value for the molar con-
centrations x, y that is observed in the most currently used silicate glasses.
We derive the first equation from the constraint theory, which sets the rigidity threshold
at the average coordination number equal to 2.4 [1,2]. We shall evaluate this parameter
on the atomic scale, counting all atoms,including the one-valenced ions.
Although one may find in current litterature the discussions of radial distribution func-
tions obtained with various techniques such as the Mssbauer spectroscopy or Nuclear Mag-
netic Resonance, which tend to acknowledge the idea that sodium atoms are often surrounded
by four or more oxygen atoms, we shall follow Paulings idea of resonating bonds, and base
our counting of coordination number on chemical valence only. If we consider our random
network as a conglomerate of single Si, O, Na and Ca atoms, then the computation of the
average coordination number should be done in the following way. In terms of two indepen-
dent probabilities and we can evaluate the average atomic coordination after observing
what follows:
- an A block, found with the probability contains three atoms, one 4-
coordinate atom (Si) and two (four halves !) 2-coordinate (O) atoms;
- a B block, appearing with the probability accounts for four atoms and a half, of
which one (Si) four-coordinate, one (Na) one-coordinate, and five halves of two-coordinate
O atoms;
182
- finally, a C block (with the probability p
C
) contains two four-coordinate atoms (2
Si), and six two-coordinate atoms (five O and one Ca).
This enables us to write the following expression for the average atomic coordination
number
More precisely, we acknowledge that is the total number of valencies divided by the
total number of atoms in a randomly chosen huge sample, containing the total of N building
blocks, out of which A-blocks, B-blocks and
C-blocks. The total number of atoms in the sample is thus
The total number of valencies (bonds that each atom is able to create with its immediate
neighbors) is:
1 4 + 2 2 = 8 for an A-block;
1 4 + 2.5 2 + l 1 = 10 for a B-block;
2 4 + 6 2 = 20 for a C-block.
(therefore the total number of valencies (2 the total number of bonds) is
leading to the above expression for The
constraint theory predicts that the best glass-forming conditions occur for ; there-
fore, we must have
which in terms of molar concentrations x, y reads
Now we need a second equation in order to fix the values of concentrations x and y
corresponding to the ideal glass forming conditions. As we argued before, the principle of
maximal homogeneity of the network should povide us with an independent relation. The
most important point now is the choice of the right scale to which the homogeneity criterion
whould be applied.
The fact that with the new building blocks we are able to produce (mostly 4, 5 and 6-
fold) rings, which is certainly one of the most important characteristics of silicate glasses,
suggest a criterion of homogeneity, which is the potential ability to form rings, evaluated on
a given distribution and of elementary clusters.
What is important here is the number of bonds coming out of the central unit, on which
the rings will be constructed. We have the following situation:
- the A - configuration can give rise to 6 different rings (because we can pick up 6
different couples out of 4 free bonds;
- the B - configuration can give rise to only 3 rings;
- the C - configuration considered as a 6-coordinate unit, can participate in as many as
15 different rings (Fig. 5).
If we want to keep the ring-forming ability constant on the average, and equal to 6, like
in the case of pure then we get readily the equation:
183
Figure 5. The ring-forming ability of elementary clusters A,B and C.
leading to
Combining this extra relation with the previous result of fixing the we obtain the
following equation:
after inserting the relation takes on the form:
resulting in
which corresponds to molar concentrations x = 15.79%, y = 10.52%,
i.e. 73.79% of 15.79% of and 10.52% of CaO.
It results from the above formulae that pure is not a very good glass former ac-
cording to the constraint criterion; this can be seen from the fact that x = 0, y = 0 does not
satisfy the above equation, which is quite obvious if we remember that for pure one
has However, there is an evidence that in this case the bond-bending con-
straints are broken, so that the counting that leads to the rigidity threshold definition should
be done in a slightly different way, suggested by P. Boolchand [7], The most important, and
as a matter of fact, primary feature of a given network defining its rigidity threshold at which
the glass forming tendency should be at its maximum, is the number of degrees of freedom
per atom, which should be equal to 3 in a 3-dimensional network. Let us consider the ternary
glass
From the ring homogeneity condition we derive the relation that x and y should obey, i.e.
2x = 3y, or which corresponds to the cation ratio Na : Ca = 3 : 1, because each
molecule destroys 6 rings, whereas each CaO molecule creates 9 new rings as compared with
the average number of rings created around each Si-centered tetrahedron in a pure amorphous
Therefore, we can write now :
The counting of constraints has to be performed carefully, taking into account the exis-
tence of non-bonding oxygen atoms, connected to the ions:
184
* Each molecule creates two non-bonding oxygens, which have only one
constraint on them, and each ion has only constraint on it.
** Each silicon atom, with the coordination number (and valence) equal to 4, has
of bond-stretching constraints (called constraints) and of bond-bending con-
straints (called constraints), the total
*** Each Ca atom, as well as each bonding oxygen, have 2 constraints.
Keeping this in mind, we can summarize the constraint counting per atom as follows:
In the above formula, we have substracted one non-bonding oxygen from the molecule,
and added it to the molecule, because one of the non-bonding oxygens created is bor-
rowed from the surrounding structure.
Then we can add up all the constraints, and compare the result with the average number
of degrees of freedom per atom. Let us remind that the average number of atoms per mole is
here:
The two formulae combined yield the following equation:
Substituting i.e. the average number of constraints per atom equal to 3, so that the
average number of remaining degrees of freedom is also equal to 3, we obtain the relation
leading to
yielding the same solution again, i.e. and consequently,
and 1 x y = 73.77%, the same solution as the previous one. This
result seems satisfactory when compared with the composition of commonly used glasses :
here are a few examples of most widely used chemical compositions :
Type
Window glass: 71-73% 12-15% 8-10% ~ 0.5-1.5%
Bottle glass: 71-73% 12-14% 10-12% 1-4% 0.5-1.5%
Glass for bulbs: 73.6% 16% 5.8% 3.6% 1%
In our computations we have neglected the presence of which is often added as
a disorder-increasing agent. Some amount of is also added in order to enhance local
tendencies of crystallization of with CaO; besides, its behavior is the same as that of
CaO.
Supposing that each molecule produces two new elementary blocks named D,
whose relative abundance shall be denoted by whose ring-forming ability is as high as
36, because the three pending Si-centered tripods display now 9 oxygen bridges which must
be inserted into the network, and there are 36 ways of picking up two bridges out of nine in
order to start building a ring.
Let us denote the molar concentration of by z, so that we have now the following
constitutive relations:
185
obtain easily the following two equations:
acknowledging the fact that the average number of rings produced by building block of any
kind remains equal to six, as in the pure amorphous Si network, and
expressing the constraint theory requirement that
Of course, these two equations are not enough in order to fix the values of all the three
variables x, y and z. We can use them to check if they give right answers when one of the
above variables is fixed at a given value. For example, if we consider the glass in which the
amount of has been fixed as equal to 75%, we get :
If we consider ternary mixture with sodium and aluminium modifiers only, without any cal-
cium, i.e. fixing and y = 0, then we get
showing that the maximal amount of is about 2%, which also does correspond to the
current compositions. It is often argued that the Al atoms are four-coordinate in the silica-
based glasses with alkali modifiers, because the ions stick to the Al atoms creating an
extra oxygen bond. It is not difficult to check that even with such an assumption the above
count will lead to a similar result, because it reinforces the ring-forming tendency provided
by Al-centered groups, but at the same time it raises the number of constraints on them.
ACKNOWLEDGEMENTS
It is a pleasure to express my thanks to P. Boolchand, J.C. Phillips and M.F. Thorpe
for invaluable help during the preparation of this article. The innumerable and enlightening
discussions with R. Aldrovandi, R.A. Barrio, Ph. Jarry and M. Micoulaut are gratefully
acknowledged.
REFERENCES
1. J.C.Phillips (1979) Journal of Non-Crystalline Solids, 34, 153.
2. M.F.Thorpe (1983) Journal of Non-Crystalline Solids, 57, 355.
3. R.Kerner(1991) Journal of Non-Crystalline Solids, 135, 155.
4. R.Kerner (1995) Physica B 215, 267.
5. J.Zarzycki (1979) Verres et Etat Vitreux, Ed. CNRS, Montpellier.
6. S.R.Elliott (1990) Physics of Amorphous Materials, Longman, London.
7. P.Boolchand etal. (1995) Journal of Non-Crystalline Solids, 182, 143; also: P.Boolchand and M.F.Thorpe
(1994) Phys. Rev. B 50, 10366; also: W.Bresser and P.Boolchand (1986) Phys.Rev.Lett, 56, 2493
8. R.Kerner, G.G.Naumis (2000) Journal of Physics: Cond.Matter, 12 1641
9. P.Boolchand (2000), private communication; also: P.Boolchand and W.Bresser, Phil.Mag. B, (to appear
in 2000)
10. M. Micoulaut (1998) European Journal of Physics, B 1, 277.
11. R.Kerner, D.-M. dos Santos (1988) Phys.Rev. B , 37, 3881.
12. R.Kerner (1991) Journal of Non-Crystalline Solids, 135, 155.
186
13. F.L.Galeener (1990) Journal of Non-Crystalline Solids, 123, 182.
14. R.Kerner (1995) Journal of Non-Crystalline Solids, 182, 9.
15. R.A.Barrio, R.Kerner, M.Micoulaut and G.G.Naumis (1997) Journal of Physics : Cond. Matter, 9, 9219.
16. R.Kerner, M.Micoulaut (1997) Journal of Non-Crystalline Solids, 210, 298.
17. R.Aldrovandi, R.Barrio, Ph.Jarry and R.Kerner (2000) submitted to Phys.Rev.B.
18. S.J.Gurman (1991) Journal of Non-Crystalline Solids, 125,191.
19. V.K.Leko et al (1977) Fizika and Khimiya Stekla, 3, 204.
20. Ph.Jarry, private communication.
21. Z.Strnad, P.W.McMillan (1983) Physics and Chemistry of Glass, 24, 57.
22. Yu.S.Tveryanovich et al (1998) Khimiya Stekol i Rasplavov, Ed. St.-Petersbourg University, (in Russian).
187
This page intentionally left blank
LOCAL BONDING, PHASE STABILITY AND INTERFACE PROPERTIES OF
REPLACEMENT GATE DIELECTRICS, INCLUDING SILICON OXYNITRIDE ALLOYS
AND NITRIDES, AND FILMAMPHOTERIC ELEMENTAL OXIDES AND SILICATES
G. LUCOVSKY
Departments of Physics, Electrical and Computer Engineering, and Materials
Science and Engineering, North Carolina State University
Raleigh, NC 27695-8202, USA
INTRODUCTION
The scaling of silicon integrated circuits to smaller in-plane lateral dimensions of
individual devices to increase speed and reduce cost through higher packing densities
requires that insulating oxides and nitrides other than non-crystalline or amorphous SiO
2
be
incorporated into the gate stacks of field effect transistors, FETs. These replacement
insulators must have dielectric constants, k, significantly larger than that of SiO
2
, e.g., in
the range of 10 to 30, in order to increase the gate dielectric capacitance and thereby
provide a sufficiently high density of channel charge under operating bias voltages to meet
current drive requirements. The increased physical thickness associated with the higher
values of k must reduce tunneling currents for the same effective capacitance as referenced
to an SiO
2
film. Finally, the alternative dielectrics must also have electronically-acrtive
defect densities in the bulk, and at interfaces with crystalline Si that are respectively the
same or less than those of thermally-grown SiO
2
and Si-SiO
2
interfaces.
This paper provides a science base that underpins the technology challenges identified
above. The approach is predicated on an empirical observation that any direct replacement
for SiO
2
must be a deposited, non-crystalline dielectric thin film. For example, the leakage
current of films of non-crystalline deposited Al
2
O
3
thin films increases by more than five
orders of magnitude upon crystallization which occurs after a 900C anneal. Since there are
a large number of potential elemental and binary oxides to consider as replacement
dielectrics, it is essential to develop a classification scheme that helps to narrow the process
of identifying viable replacements to SiO
2
. This separates the potential replacement
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kl uwer Academic/Plenum Publishers, 2001 189
dielectrics into two broad categories, i) silicon oxynitride and nitride films that retain a
continuous random network structure and ii) the so-called high-k dielectrics that include
elemental oxides, such as Al
2
O
3
and ZrO
2
, as well as binary oxides, including transition
metal silicates and aluminates.
THERMALLY-GROWN SiO
2
THE STANDARD THAT ALL REPLACEMENT
DIELECTRICS MUST MEET
Crystalline silicon is unique among semiconductor materials since device-quality SiO
2
layers can be prepared by thermal oxidation of silicon surfaces. The resulting oxides are
stoichiometric to better than one part in a million, have bulk defect densities < 10
16
cm
3
(or
equivalently < 10
11
cm
2
), and after annealing at ~900C, they have interfaces with
crystalline silicon with densities of interface traps and fixed charge, ~l-3l0
10
cm
2
, that
are less than one part in 10
4
. These densities of electrically-active defects define a standard
that must be achieved by any deposited replacement gate dielectric.
SiO
2
films grown at temperatures between 800 and 1000C are non-crystalline and
remain so after annealing or other thermal exposures at 900-1100C. Stoichiometric,
device-quality SiO
2
thin films have also been prepared by rapid thermal chemical vapor
deposition, RTCVD, and by direct and remote plasma-enhanced CVD, RPECVD [1].
Deposited dielectrics require annealing at temperatures between about 900 and 1000C to
have electrical properties similarly to thermally-grown SiO
2
films grown and/or annealed at
the same temperatures. The infrared absorption spectra of thermally-grown and deposited
oxides annealed at temperature of 900 to 1050C are essentially the same as bulk SiO
2
bulk
glasses. Based on these comparisons with bulk glasses, the thermally-grown oxides are also
chemically-ordered, continuous random networks with 4-fold coordinated Si-atoms bonded
to bridging or 2-fold coordinated O-atoms. This includes a very weak bonding force at the
2-fold coordinated oxygen atoms and a large spread in the Si-O-Si bond angle from about
120 to 180 degrees [2].
There are several important aspects of the atomic scale bonding structure of SiO
2
that
provide a quantitative basis for a proposed classification scheme that can be applied to
potential replacement dielectrics. These are i) the bond-ionicity, I
b
[3], and ii) the average
bonding coordination or number of bonds per atom, N
av
, and iii) the average number of
bonding constraints per atom, C
av
[4,5].
The simplest operational definition of bond-ionicity is the one originally introduced by
Pauling [3], and based on atomic electronegativity. If X(O) is the Pauling electronegativity
of oxygen, 3.44, and X(Si) is the Pauling electronegativity of silicon, 1.90, then the
electronegativity difference, is 1.54. Using the definition of bond-ionicity proposed by
Pauling, I
b
for Si-O bonds is ~ 45 %. In the next section it is
shown that the bond-ionicities for other good glass-forming oxides and chalcogenides are
generally less than Ib for SiO
2
, i.e., the bonds are generally less ionic.
The network structure of SiO
2
has been classified as a continuous random network, or
crn. The randomness, which supplies the configurational entropy necessary for good glass
formation, derives from two sources: i) a large spread in bond angle at the O-atom sites,
and ii) a random distribution of dihedral angles.
Phillips has shown that another important metric for characterizing a continuous
random network is the average number of bonds/atom, N
av
. using a valence force field
model, this in turn can be related to an average number of bonding constraints/atom, C
av
[4]. In a series of seminal papers, Phillips has demonstrated that a simple criterion for good
190
bulk glass formation is that C
av
~3. Based on bond-stretching and bond-bending valence
forces, C
av
is directly proportional to N
av
, and is given by C
av
= 2.5 N
av
3. If m is the
coordination of a network atom, then the number of stretching constraints per atom is m/2,
and the number of bending constraints per atom is 2m-3 for a large number of different
bonding arrangements. Using this relationship for C
av
, N
av
for an ideal glass is given by
2.4, as for example in As
2
S(Se)
3
. One exception to the constraint counting is that if the
atomic coordination is three or more, and the atom is in a planar bonding arrangement, e.g.,
B-atoms in B
2
O
3
, then the number of bending constraints per atom is reduced to m-1.
For SiO
2
, N
av
= 2.67, so that C
av
= 3.67; however, since the bending force constant at
the O-atom site is extraordinarily weak, it can be neglected and C
av
is then reduced to three,
explaining the ease of glass formation for SiO
2
. Phillips has recently extended the bond-
counting and bond-constraint counting relationships to silicate glasses which have network-
modifier ions [5]. This extension of constraint theory provides a basis for defining C
av
for
silicates with bonding structures that represent departures from the crns of oxide and
chalcogenide glasses. For example, at a metal atom ionic bonding site within a silicate
network, such as a sodium atom in a sodium silicate glass, the effective coordination for
bond counting and bond constraint counting is taken as the number of resonating bonds as
defined by Pauling, or equivalently is the product of the number of actual bonding
neighbors and their effective bond-order.
The success of this approach to silicates is illustrated by application to ordinary
window glass which is a ternary silicate alloy with a nominal composition of SiO
2
(74%),
Na
2
O (10 %) and CaO (16%). Based on the resonating bond model of Phillips, N
av
=
0.74(2.67) + 0.1(1.33) + 0.16(2) = 2.43, so that C
av
is very nearly equal to ideal value of
three [7].
CLASSIFICATION SCHEME OF DIELECTRICS
The classification scheme is empirical and is based on several observations First,
dielectrics with such as SiO
2
, B
2
O
3
, As
2
O
3
, P
2
O
5
, As
2
S(Se)
3
, are generally good
glass formers (see Table I which includes values of and I
b
). For example, they meet
the Phillips criteria for N
av
and C
av
, with values C
av
less than or equal to 3. There is a
second glass of good glass formers that are fluorides rather than oxide or chalcogenides,
and are considerably more ionic (see Table I). This class includes and
Bond ionicities are greater than 70 %. Bond coordinations are
essentially the same as SiO
2
with the formal number of bonds/atom being 2.67. However,
since the bonding is not covalent, a valence force field model is not appropriate, and the
constraint theory of Refs. 3 and 4 cannot be applied. A better way to describe the BeF
2
(and
ZnF
2
) glasses is to consider them as random closed packed ionic structures [8].
A third class of glasses with partial ionic character are the silicates or alternatively
pseudo-binary alloys of SiO
2
and metal oxides such as Na
2
O, CaO, Al
2
O
3
, ZrO
2
, PbO, etc.
In these silicates, the metal atoms are network modifying ions whose charge is
compensating by terminal negatively oxygen atoms that are covalently bonded to the Si-O
network. The bonding in these silicates can then be characterized as being amphoteric with
the Si-O network being the acidic or negatively charged component, and the metal ion the
basic or positively charged component. Since these metals are less electronegative than
silicon, the compositionally-averaged electronegativty difference will be greater than that
of SiO
2
, but less than that of the metal oxide. Since homogeneous silicate glasses and as-
deposited thin films can be prepared with compositions up to 20-40 atomic % of the metal
191
oxide, this means that values for these silicates can be as large as 2. For example
for alloys of 70 atomic % SiO
2
and 30 atomic % or ZrO
2
and La
2
O
3
are 1.74 and 1.81,
respectively.
Based on i) the large differences in and bond ionicity between the crn glasses,
exemplified by SiO
2
, and the rcp glasses, exemplified by BeF
2
, and ii) intermediate values
of average bond ionicity in silicate glasses, we propose that non-crystalline oxides can then
be grouped into three glasses according to their average bond ionicities: i) continuous
random networks, crns, with and I
b
< 48 %, ii) modified random covalent
networks, mrcns, with between 1.6 and 2.0m and I
b
between about 48 and 65 %, and
iii) rcp ionic structures with 2.0 and I
b
> 65 % (see Table I). The boundaries between
the classes are not to be construed as exact, but are representative of ionicities at which
there are significant changes in bonding structure and microstructure. Additionally, and
based on the bonding in the silicate alloys, there are a small number of elemental oxides,
defined by electronegativities between approximately 1.6 and 2.1, that can also be
characterized amphoteric or auto-compensating. For example, Table I includes several
elemental oxides with for which is > 1.6, but less than 2: Al
2
O
3
(1.83), Ga
2
O
3
(1.63),
Ta
2
O
5
, (1.94), and Nb
2
O
5
(1.84). The local atomic structure of at least two of these oxides,
Al
2
O
3
and Ga
2
O
3
, supports the validity of the classification scheme.
REPLACEMENT GATE DILECTRICS
The transition from thermally-grown SiO
2
gate dielectrics to deposited dielectrics with
higher static dielectric constants will likely proceed in two steps that are defined by so-
called technology nodes associated with targeted values for scaling of feature sizes, see for
192
example the National Technology Roadmap for Semiconductors, 1999 [9]. The first
generation of deposited gate dielectrics will be silicon oxynitride and nitride dielectrics,
and the second will be metal oxides, silicates or aluminates. The boundaries between these
generations of devices are marked by an equivalent oxide thickness, EOT, defined by the
gate capacitance of a metal-oxide-semiconductor device structure, and a limiting value of
direct tunneling current at an operating bias of about 1 volt above the threshold for field
effect transistor operation. Based on a nominal criterion that the direct tunneling current be
less than 1-5 A-cm
2
for high power, desk-top devices, these three generations are: i)
thermally-grown SiO
2
and lightly-nitrided SiO
2
can be used for EOT down to ~1.6 nm, ii)
silicon nitride and oxynitrides in stacked structures with nitrided Si-SiO
2
interfaces can be
used for EOT down to about 1.1 nm, and iii) metal oxides, silicates and aluminates, etc.,
will be required for EOT of 1 nm and below. It is likely that this scaling will not extend
below about EOT of ~0.6 to 0.8 nm, even though NTRS targets are set at about 0.5 nm.
The limiting direct tunneling currents for devices incorporated into mobile devices, e.g.,
cellular phones, must be about three orders of magnitude less than for the desk top devices,
and the three generations of devices are defined by proportionally larger values of EOT. In
the discussion that follows the metrics for the high power devices will serve as the
reference.
Based on Table I, the most important boundary between these three generations of
devices is the one between the crns, which included SiO
2
, and the silicon nitrides and
oxynitride alloys, and the mcrns, which include amphoteric oxides, and silicate and
aluminate alloys. The microscopic physics that is important in the technology applications
is qualitatively different for the crns and the mcrns, and these differences will be
addressed in more detail the remainder of this paper.
LIMITATIONS IMPOSED BY NETWORK CONSTRAINTS IN CRNs
The limitations on the performance and reliability of field effect transistors with SiO
2
,
and Si nitride and oxynitride gate dielectrics with EOT < 2.5 nm have been shown to derive
primarily from bonding defects at the Si-dielectric interfaces. These defects have their
origin in mechanical bonding constraints. A recently published paper has addressed this
limitation, and the results of that paper are summarized below [10], Several studies have
addressed chemical bonding arrangements at thermally annealed Si-SiO
2
interfaces and
have shown that optimized interfaces display transition regions ~0.3 nm thick with excess
sub-oxide bonding arrangements different from those expected at abrupt Si-SiO
2
metallurgical interfaces. These studies include: i) X-ray photoelectron spectroscopy, XPS,
on ultra-thin Si-SiO
2
interfaces using monochromatic synchrotron radiation [11], ii) in-situ
Auger electron spectroscopy, AES [12] and iii) in-situ Fourier transform infra red, FTIR
[13]. From the perspective of constraint theory, these transition regions are at an interface
between a rigid crystalline Si material and an effectively floppy amorphous material,
SiO
2
, with a continuous random network structure [14].
Applications of constraint theory to non-crystalline solids have focused primarily on
bulk glasses [4], thin films [15] and Si-SiO
2
/Si
3
N
4
interfaces [16]. Recently, both theory
[14] and experiment [17] have identified a new aspect of constraint theory by
demonstrating that rigid' or over-constrained and floppy' or under-strained
microstructural regions in an alloy glass are separated by monolayer scale, self-organized
interface layers that are over-constrained with respect to bonding coordination, yet not
mechanically-strained. Experiments cited in Refs. 11-13, combined with the theory of
Refs. 14 and 18, suggest that the interfacial transition regions in advanced Si FET gate
193
stacks are intrinsic, and result in part from entropy effects. These transition regions
constitute a basic limitation for the aggressive-scaling of CMOS silicon devices, as well as
other semiconductor devices with similar steps in interfacial bonding coordination.
For example, the data in Figs. l(a) and (b) have been explained in terms of model based
on XPS data of Ref. 11 (see Figs. 2(a) and (b)). Analysis of the interfacial features in XPS
data of Figs. 2(a) and (b) labeled Si
l+
, Si
2+
and Si
3+
, yields concentrations of Si-atoms in
Figure 1. Current density as function of gate voltage in (a) the Fowler Nordheim, and (b) the direct tunneling
regimes for devices with plasma-grown interface layers, and plasma-deposited SiO
2
bulk dielectric films.
These devices have been subjected to a 30 s 900C rapid thermal anneal in a He ambient.
suboxide bonding environments in excess of those that are required for an ideal and abrupt
interface. The excess silicon atom concentrations have been equated to an average
composition of SiO, thereby converting thye Si areal densities into equivalent transition
region widths of 0.27 nm for the non-nitrided interface and 0.35 nm for the nitrided
interface [19]. Combining these XPS data with differences in optical SHG for interfaces
with and without interfacial nitridation [20], band models have been generated for the
interfacial transition
Fig. 8. Synchrotron XPS data for remote plasma processed interfaces without nitridation in (a) and with
monolayer interface nitridation in (b). These interfaces have been subjected to a 30 s 900C rapid thermal
anneal in a He ambient.
194
regions (see Fig. 3(a)). After completing the stacked dielectrics with films of SiO
2
subject
to the constraint of maintaining the same values of EOT, calculations of direct tunneling
for substrate accumulation in Fig 3(b) have yielded the reductions in the tunneling current
density for devices with nitrided interfaces essentially the same as shown in Fig. 3(b) [19].
It is important to note that the direct tunneling current is higher for an optimized interface
with minimal suboxide bonding than for an abrupt interface, and that insertion of one
monolayer of nitrogen atoms at the Si(100) interface by remote plasma processing reduces
Figure 3. (a) Interfacial band structure for tunneling calculations for plasma processed and nitrided plasma
processed interfaces as derived from XPS data of Fig. 2. (b) Calculated direct tunneling for devices with EOT
= 2 nm, for the interfaces of Fig. (a) and ideal Si-SiO
2
interfaces with no suboxide bonding.
the direct tunneling current to approximately the same value as for an ideal abrupt Si-SiO
2
interface.
Constraint theory has also been invoked at semiconductor dielectric interfaces to
explain differences in the defect densities between Si-Si
3
N
4
and Si-SiO
2
interfaces [16,21].
In this application, mechanical bonding constraints at the interface have been characterized
in terms of the average number of bonds per atom in the interfacial region. Following Ref.
15, the interfacial bonding structure is defined by 0.5 molecular layers of Si (0.5 atoms and
two bonds), and 1.5 molecular layers of the dielectric film (SiO
2
or Si
3
N
4
. Interface
nitridation has been taken into account by inserting one atomic layer of nitrogen between
the Si substrate and SiO
2
layer. The Si-SiO
2
interface is used as a reference interface and is
characterized by an average number of interfacial bonds/atom, N
av
* = 2.86. Based on the
model interface bonding model described above, N
av
equals 3.47 for a Si-Si
3
N
4
interface,
and 2.89 for a monolayer nitrided Si-SiO
2
interface. As shown in Fig. 4, the concentration
Figure 4. Normalized defect density at interfaces as a function of over-coordination.
195
of defects relative to the Si- SiO
2
interface scales as (N
av
- N
av
*)
2
. This scaling is based on
an empirical observation that bonding stretching force constants are significantly stronger
than bonding bending force constants. It is therefore assumed that increased bonding
coordination contributes predominantly to bond angle strain, so that
(1)
Strain energy, E
s
, is proportional to and if it assumed that the bond defect
concentration [D] is proportional to strain energy, then the following scaling relationships
apply:
Based this scaling, the defect concentration at an Si-Si3N4 interface is expected to be
about three orders of magnitude higher than at a monolayer nitrided Si-SiO
2
interface,
consistent with experimental results [16,21], Constraint theory and the associated scaling
neither predict defect concentrations, nor identify the defect bonding arrangements.
Instead, they yield scaling relationships that provide a useful guideline for comparisons of
the type discussed above. In the spirit of the scaling arguments, a value of interfacial N
av
~
3 has been proposed in Ref. 16 as a demarcation between device-quality and increasing
defective interfaces, paralleling a similar criterion applied to defects in homogeneous
amorphous thin films [15].
Scaling theory also provides an explanation for differences in fixed charge at internal
dielectric interfaces. In this application, the appropriate scaling variable is the difference in
the average number of bonds per atom, on either side of the internal dielectric
interface. This model will now be applied to the C-V data of Figs. 5(a) and 5(a) for
interfaces between interfacial SiO
2
and Si nitride and oxynitride. The negative shifts of the
Figure 5. Capacitance-voltage, C-V, curves for (a) PMOS and (b) NMOS devices (EOT ~ 2 nm) with remote
plasma oxidized interfaces (~0.6 nm) and remote plasma deposited oxide, oxynitride and nitride bulk
dielectrics.
shifts of the C-V curves from the oxide, to the oxynitride alloy and nitride, correspond
respectively to fixed charge levels. Q
f
, of ~2l0
11
cm
2
and 7.5x10
11
cm
2
, respectively.
The relative shifts in the C-V traces are essentially the same for NMOS and PMOS
capacitors indicating that they derive from fixed charge. Since there is no detectable fixed
charge associated with the remote plasma oxidized Si-SiO
2
interface, the charge must
reside at the internal dielectric interface. These data are consistent with the scaling
196
(2)
from constraint theory. The ratio of the defect concentrations, ~ 3.8, is approximately equal
to the ratio in the values of at these two internal dielectric interfaces, [4,10].
LIMITATIONS IMPOSED BY BOND-IONICITY IN MCRN DIELECTRICS
The silicon oxynitride and nitride continuous random network insulators of the last
section have dielectric constants that are at most two times that of SiO
2
. Coupled with the
necessity for nitrided SiO
2
interface layers, the minimum EOT that can be obtained using
silicon oyxnitride or nitride stacked gate dielectrics is about 1.1-1.2 nm [1]. Extension to
smaller EOT requires dielectrics with larger values of k, preferably in the range of 15 to 25.
This has created interest in both elemental oxides such as TiO
2
, Ta
2
O
5
, Al
2
O
3
, ZrO
2
and
HfO
2
, and silicate alloys, initial of Zr and Hf, and more recently extending to Y and La.
This section of the paper addresses three issues relative to these potential alternative gate
dielectric materials.
Stability Against Chemical Phase Separation
An important issue in the replacement of thermally-grown SiO
2
by an alternative
deposited gate dielectric material is process integration. This aspect of manufacturing
science and technology addresses the entire process sequence by which an integrated
circuit is fabricated. The thermal stability of deposited silicon oxynitride and nitride
dielectrics is sufficient for direct substitutions to be made. In particular, post deposition
processing temperatures up to about 1050-1100C can be utilized, and this is sufficient for
dopant activation in both polycrystalline Si gate electrodes and the source and drain
contacts of the PMOS and NMOS field effect transistors. This is not generally the case for
deposited dielectrics. Many elemental oxides such as TiO
2
, ZrO
2
and HfO
2
are either
polycrystalline upon deposition, or crystallize and relative low post-deposition
temperatures. This places serious limitations on device structure and process integration,
e.g., processing sequences may require all high temperature steps done prior to deposition
of the gate dielectric, and then restrict the gate electrode to deposited thin film metals rather
than doped polycrystalline Si.
Our group has studied post deposition stability of remote plasma deposited dielectrics
by three different techniques, Fourier transformation infrared spectroscopy, FTIR, X-ray
diffraction, XRD, and high resolution transmission electron microscopy, HRTEM, lattice
Figure 6. FTIR spectra of (a) A1
2
O
3
and (b) Ta
2
O
5
films as-deposited at 300C and after 30 s anneals in Ar at
the temperatures indicated. The feature at 1100 cm
1
in (a) is a substrate artifact.
197
imaging. Figures 6(a) and (b) illustrate FTIR results for A1
2
O
3
and Ta
2
O
5
and as a function
annealing temperature. The Al
2
O
3
films show evidence for crystallization at an annealing
temperature of 900C, and the Ta
2
O
5
films show evidence for the onset of of crystallization
at about 800C.
Figure 7(a) displays absorbance for Zr silicate films ([SiO
2
]
x
[ZrO
2
]
1-x
) prepared by
remote plasma enhanced chemical vapor deposition [22,23]. Spectra of as-deposited films
and those annealed at temperatures up to 800C are essentially the same, whereas spectra
of films annealed at 900C are markedly different. After a 900C anneal, changes in these
spectra indicate a chemical phase separation into i) a non-crystalline low Zr-content silicate
alloy with ~1-2 atomic % Zr, or simply SiO
2
, and ii) non-crystalline (x~0.1 and ~0.23) or
crystalline ZrO
2
(x~0.5) [23]. Figures 7(b) and (c) compare the XRD and FTIR results for a
film with x ~ 0.5, the compound silicate composition, ZrSiO
4
.
Figure 7. (a) Absorbance versus wavenumber for
three Zr silicate alloys with different ratios of of
Zr:Si, as-deposited and after a 30 second, 900C
anneal in Ar. (b) XRD and (c) FTIR spectra of an
alloy with x~0.5, for as-deposited and annealed
films. The changes in XRD and FTIR at 900C
marking the onset of crystallization are evident.
Interface Properties
There are two aspects of interface properties that are important in high-k replacement
gate dielectrics: i) conduction band offset energies, E
cb
, that determine the barrier for direct
tunneling out of the Si substrate, and ii) interfacial defects in the form of traps, D
it
, or fixed
charge, Q
f
. The band offset energies for high-k oxides have estimated by Robertson [24],
and are generally reduced by about 1 volt or more with respect to the 3.15 eV value of E
Cb
the for the Si-SiO
2
interface. The calculated values in Ref. 24 for Si
3
N
4
and Al
2
O
3
, are
198
larger by out 0.3 to 0.8 eV than those obtained from XPS studies [25]. The very low values
of Ebc calculated for TiO
2
and Ta
2
O
5
, ~0.0 and 0.36 eV, respectively, and confirmed by
experiment, make these dielectrics unsuitable for replacement gate dielectrics. The low
band offset energies of all of the transition metal elementary and binary oxides result from
two effects, increased ionicity with lowers the valence band energies, and reduced energy
gaps, which are associated with the d-state parentage of their conduction bands. The values
of E
cb
for the Zr and Hf silicates are about 1.5 eV, and it remains to be determined if these
are sufficient for tunneling current reductions. The tunneling transmission probability is
proportional to ~ exp ( -2 a t
phys
[E
bc
m
e
* ]
0.5
), where a is a constant, t
phys
is the physical
thickness of the high-k replacement dielectric, ~[EOT]x[k/k(SiO
2
)], and m
e
* is the tunneling
mass of electrons. Reductions in E
bc
will clearly offset gains in t
phys
, and this remains an
important issue in gate dielectric replacement.
A second issue relates to fixed positive charge. This interfacial charge which can have
two different origins. For the stacked gate dielectrics with silicon oxynitride or nitride gate
Figure. 8 C-V characteristics for PMOS and NMOS capacitors with Al
2
O
3
gate dielectrics.
dielectrics this charge results from mechanical bonding constraints at the dielectric
interfaces. For the dielectrics with mrcn structures, the charge exists because of increased
ionic bonding. C-V studies of NMOS and PMOS capacitors with plasma deposited Ta
2
O
5
,
A1
2
O
3
and ZrO
2
-SiO
2
silicate alloy high-k gate dielectrics demonstrated fixed charge at the
Si-high-k gate dielectric interface. These dielectric films were formed by remote plasma
processing. Post-deposition analyses indicated that detectable interfacial SiO
2
layers were
not formed during the processing. Fixed charge levels were determined from comparisons
of flat band voltages for devices with different values of EOT, generally from 1 to 4 nm.
For example, the C-V traces in Fig. 8 show positive flat band voltage shifts which increase
with decreasing capacitance, or equivalently with increasing dielectric thickness. These
charge levels are of order 1-510
12
cm
2
. The charge was positive for the test capacitors
with Ta
2
O
5
and ZrO
2
-SiO
2
silicate alloy dielectrics, but negative for those with A1
2
O
3
.
Table II bond ionicity, ionic fraction, and fixed charge levels for devices with A1
2
O
3
, Ta
2
O
5
and Zr-silicate alloy (x ~ 0.3) gate dielectrics.
The three dielectrics in Table 1 can be characterized as modified covalent random
networks including a charged network and discrete ions. For example, a structural formula
for A1
2
O
3
is given by 2 A1
2
O
3
=3 (AlO
4/2
)
1
+ A1
3+
[26]. The term (A1O
4/2
)
1
represents a
network structure similar to SiO
2
, wherein Al
1
is isoelectronic with Si. The A1
3+
ions
occupy octahedral interstitial sites of the network, and form three resonating or dative
199
donor-acceptor pair bonds with bridging oxygen atoms of that network. The average
number of bonds/atom is 3. Similarly based on FTIR, Raman and EXAFS, the Ta atoms in
Ta
2
O
5
, assumed to have a formal charge of +1, are in octahedral bonding sites. The
octahedra are corner- and edge-connected, respectively in a network through 2-fold and 3-
fold coordinated O atoms. These O-atoms have a net negative charge that balances the
positive charge of the Ta-atoms. Finally, the (ZrO
2
)

(SiO
2
)
1-x
alloys for values of x < 0.5
are in a prototypical silicate structure in which Zr
4+
ions are nearest-neighbors to four or
more terminal and negatively-charged oxygen atoms of the network; hence the designation
of these ions as network modifiers. Combining the results presented in Table 2, with these
structural models, and with a propensity for covalent bonding of O to Si at the Si-dielectric
interface, the respective bonding arrangements at the Si-Al
2
O
3
, Si-Ta
2
O
5
, and Si-Zr silicate
(x=0.3) interfaces are, Si-O-Al
1
, Si-O-Ta
1+
and Si-O
1
--Zr
4+
, suggesting a correlation
between the sign of the fixed charge and the metal ion of the network component of the
oxide in Al
2
O
3
and, Ta
2
O
5
, and of the metal ion in the silicates. Studies of group IIIB
silicates in which the bond-ionicity is higher, e.g., Y and La silicates, indicate fixed
positive charge consistent with the incorporation of Y and La as 3+ ions rather than as
network constituents. Finally, the magnitude of the fixed charge is more difficult to
account for. The convention has been to calculate an areal density of fixed charge defects
of unit charge based on the macroscopic charge density obtained from the analysis of C-V
data. This is not the only way to interpret the charge density. It can be described in terms of
partial charges on atoms, or differences between charge distributions of both signs as in
interfacial dipoles. Since the way the fixed charge is distributed influences the channel
mobilities of charged carriers, analysis of mobility data from PMOS and NMOS FETs
using different microscopic models for the fixed charge bonding arrangements may help to
resolve some of these issues.
Dielectric Constant Enhancement
MOS capacitors with SiO
2
-rich Zr and Hf silicates with 3 to 6 atomic percent Zr(Hf)
have been reported to display increased dielectric constants and reduced tunneling currents
[27-30]. Reported values of k, extracted from capacitance-voltage curves are ~8 to 11, and
more than 50 % larger than values estimated from a linear extrapolation of k between SiO
2
,
~3.9, and the compound silicates, ~12 (Fig. 9). These enhanced values of k can not be
reconciled with macroscopic dielectric theory that predicts a downward bowing of k
between end-members in a mixed materials system [31]. Since macroscopic theory applies
to mixtures in which chemical bonding of the constituents does not change with
composition, it is important to determine if these SiO
2
-rich Zr(Hf) silicate alloys satisfy this
condition.
200
Chemical bonding has been studied in Zr silicates ( [SiO
2
]
x
[ZrO
2
]
1x
) by Fourier
transform infrared spectroscopy, FTIR [23]. Similar bonding is expected in Hf silicates as
well. These results have already been discussed earlier in this paper. Several features in as-
deposited films with x ~0.1 and 0.23 have been assigned to Si-O-Si groups in corner-
connected arrangements, as stretching modes at ~1150 cm
1
, 1065 cm
1
, and 810 cm
1
, and
a rocking mode at ~450 cm
1
[2,23]. Two other features are assigned to Si-O-Zr bond-
stretching vibrations, a terminal Si-O mode at ~950 cm
1
that is a shoulder on the 1065 cm
1
Figure 9. The dotted curve indicates the dielectric constant calculated from Eqn. 2 as a function of the percent
Zr(Hf)O
2
relative to the stoichiometric silicate compound composition, Zr(Hf)SiO
4
. Experimental points are
from Refs. 27-30. The dashed line is a linear extrapolation between the dielectric constant of SiO
2
, and a
nominal dielectric constant of 12 for Zr(Hf)SiO
4
.
band, and a broader Zr-O feature at ~450 cm
1
that is accidentally degenerate with a
rocking mode of the Si-O-Si group. Since both stretching modes involve predominantly O-
atom motion, their frequencies reflect a significantly smaller force constant for the Zr-O
vibration. This results from the increased Zr-O bond-length of 0.22 nm relative to 0.16 run
for Si-O [9]. Far-IR spectra extending to 50 cm
1
show no additional discrete features.
Broader spectral features of an as-deposited x ~0.5 film have been attributed to a random
close packing of Zr
4+
and SiO
4
4
ions [8,23]. The 800 to 1200 cm
1
band includes internal
SiO
4
4
vibrations, and the ~450 cm
1
band is assigned to Zr-O vibrations. After a 900C
anneal, changes in these spectra indicate a chemical phase separation into i) a non-
crystalline low Zr-content silicate alloy with ~1-2 atomic % Zr, or SiO
2
, and ii) non-
crystalline (x~0.1 and ~0.23) or crystalline ZrO
2
(x~0.5) [23]. As in bulk silicate glasses,
introduction of oxides of electropositive Zr(Hf)-atoms into the SiO
2
network results in a
break-up or modification of that network [7,8]. Homogeniety in bulk silicate glasses
quenched from high temperatures is limited by chemical phase separation, and in many
instances homogeneous glasses are obtained only at relatively low metal oxide
compositions, <5-10 mol percent. This is not a limitation in thin film silicates deposited at
temperatures <500C [22,23,27-30], The analysis presented below applies to these films, as
well as films prepared at low temperatures, and subsequently processed at temperatures
<800C. The capacitors with Zr(Hf) silicate dielectrics in Refs. 27-30 have not been
subjected to processing temperatures >600C and meet these temperature constraints. The
introduction of a Zr(Hf)O
2
molecule into SiO
2
is assumed to break two Si-O bonds of that
network [8], so that the concentration of terminal Si-O terminal bonds is linear in alloy
201
composition. The coordination of silicon to oxygen remains at four, and there are then five
tetrahedral bonding groups with different distributions of O-atoms that are either i)
connected to the network through bridging Si-O-Si bonds, or are ii) in terminal Si-O
groups. Figure 10 gives the fractional concentrations of these tetrahedral groups as a
function of alloy composition. Their concentrations are obtained from the appropriate terms
of a binomial expansion,
(3)
in which w is the fraction of bridging O-atoms, and z = 1 - w is the fraction of terminal O-
atoms. This bonding model applies to both Zr and Hf silicates.
Figure 10. Relative fractions of five tetrahedral silicon-oxygen bonding groups with different numbers of
bridging and terminal oxygen atoms plotted as a function of the percent of Zr(Hf)O
2
relative to the
stoichiometric silicate compound composition, Zr(Hf)SiO
4
.
Based on Fig. 10, the majority of Zr atoms (or Zr
4+
ions) in alloys with x <0.1 are
incorporated as network modifiers with four terminal negatively-charged O-atom neighbors
in corner-connected arrangements [8] (Fig. 4). As the mol fraction of ZrO2 is increased, an
increasing fraction of these groups must contain two or more terminal Si-O bonds. This
causes the coordination number of the Zr-atoms to increase above four, including edge- as
well as corner-connected bonding arrangements. In crystalline silicates, the Zr
4+
ions have
a coordination of 8, with each Zr-atom making edge-connections to four different
tetrahedral SiO
4
4
groups [22,28]. The respective Zr-O and Si-O bond lengths are ~0.22 nm
and 0.16 nm. This same bonding coordinations, four for Si and eight for Zr (or Hf) are
assumed to apply in amorphous compound silicates.
The contribution of a vibrational mode to the dielectric constant is proportional to the
square of its transverse infrared effective charge, e
T
*, and is different for different bonding
coordinations of the same atom pair [32,33]. The FTIR spectra indicate a broadening of the
950 cm
1
feature with increasing x. Based on Fig. 10, this broadening occurs at alloy
concentrations where the ratio of edge- to corner-connected arrangements and Zr-atom
coordination have increased. If e
T
* scaled directly with increases in the number of terminal
Si-O bonds, this broadening would also be accompanied by a marked increase in
202
absorbance of the 950 cm
1
feature relative to the spectral peak from Si-O-Si network
bonds at 1065 cm
1
. To the contrary, the FTIR results indicate that relative absorbance of
terminal Si-O groups does not scale in this way, and therefore decreases with increasing
Zr-atom coordination. The bond-order a Zr-O bond is defined as the ratio of number of
valence electrons available from each Zr-atom (four) to the number of nearest O-atom
neighbors. Following this definition, four-fold coordinated Zr-atoms have the largest bond
order, one. They also have the highest degree of covalency [34], so that dynamic
contributions to e
T
* are expected to be larger than for higher bonding coordinations in
which the bond order is reduced and the bonding becomes more ionic [32,33]. Scaling of
local bond properties such as bond energies and stretching force constants with bond-order
is well-established [34]. Consistent with the relative absorption strengths of single, double
and triple carbon-oxygen bonds, this scaling can also be extended to e
T
*. Since infrared
active mode contributions to the dielectric constant are proportional to (e
T
*)
2
, the
appropriate scaling variable is the square of the bond order.
Figure 1 1 . Transition in local bonding arrangements of Zr-atoms in Zr silicate alloys from low to high ZrO
2
concentrations. At low concentrations the dominant bonding arrangements are between the Zr-atoms and four
terminal O-atoms in a corner-connected geometry. At higher concentrations (fraction of ZrO
2
>30-35
percent), there is a transition to bonding including more than one O-atom in edge-connected geometries.
Based on Eqn. (3) and an assumption that contributions of Si-O-Zr arrangements to the
dielectric constant scale quadratically with the Zr-O bond order, the variation of k with
alloy composition is approximated by Eqn. (2),
(4)
in which the a
ij
are the product of i) the number of terminal Si-O bonds per group and ii)
the square of an effective average bond order. The a
ij
s are approximated by a
1,3
~lx(4/4)
2
,
a
2,2
~2x(4/(5-6))
2
~1.05, a
3,1
~3x(4/(6-7))
2
~1.15, and a
4,4
~4x(4/8)
2
= 1, and are each of
order 1. The curve is Fig. 1 is for a
i,j
= 1; values of a
2,2
and a
3,1
>1 increase k for alloys with
more than 45 % ZrO
2
, or x > 0.25. Since values of k for Zr(Hf) content > 7 atomic percent
have not been reported, the application of the model is restricted to alloys with lower
concentrations. The value of 3.9 fixes k for SiO
2
, and the value of the prefactor, 8.1, fixes
203
k at 12 for compound silicates [1,4]. The experimental data in Refs. 27-30, with alloy
content <7 atomic % (or <50% ZrO
2
relative to ZrSiO
4
) fall close to calculated model
indicating that the empirical relationship of Eqn. (4) provides a quantitative description of
the enhanced dielectric constant behavior for alloys in this composition range.
There is also an additional, and smaller contribution to the dielectric constant from
electronic transitions. A similar enhancement in the index of refraction, or equivalently the
optical frequency dielectric constant, is expected to occur in Zr(Hf) silicate alloys. Our
initial experiments have indicated that this enhancement is present, and that it scales with
bond order as well. In the spirit of the analysis presented above, this contribution is
contained in the a
i,j
terms of Eqn. (4).
It has been shown that the enhanced dielectric constants of the group IVB Zr and Hf
SiO
2
-rich silicates are due primarily to the four-fold coordination of the Zr(Hf)-atoms in
the alloy composition range below about 8 atomic %. Similar enhancements in k at low
metal-atom concentrations are also anticipated for SiO
2
-rich silicate alloys with i) TiO
2
, ii)
Y(La)
2
O
3
, but only for compositions up to the first silicate phase, La(Y)
2
SiO
5
[35], as well
as iii) and oxides of with other polarizable atoms such as Pb, Bi, Tl, etc. Interfacial
properties of these SiO
2
-rich silicates are anticipated not to depart significantly form those
of Si-SiO
2
interfaces. Therefore these interfaces of Si with SiO
2
-rich silicate alloys are
expected to have properties similar to Si-SiO
2
interfaces, as well as providing significantly
increased capacitance and reduced direct tunneling.
INTERFACIAL LIMITATIONS FOR ALTERNATIVE GATE DIELECTRICS
Figure 12 provides a graphic description of the effect of introducing a nitrided SiO
2
mechanical or chemical, interfacial buffer layer between an alternative gate dielectric and
the Si substrate. Under these conditions EOT is given by:
Figure 12. Plot of EOT versus physical thickness of the high-k dielectric for k = 25 (e.g., Ta
2
O) and 15 (e.g.,
Zr or Hf silicate alloy) without and without a 0.35 nm EOT)
int
layer.
204
where EOT)
int
is the equivalent oxide thickness of the interfacial layer, ~0.35nm for a
remote plasma oxide film with a physical thickness ~ 0.5-0.6 nm and monolayer interface
plasma nitridation, t
phys
is physical thickness of the high-k dielectric [1]. In the absence of
an interfacial layer EOT = t
phys
[k (SiO
2
) / k], so that the reduction in the physical thickness
of the high-k dielectric that is associated with the interfacial layer is given by
(6)
Since EOT)int ~ 0.35 nm, and k (SiO
2
) ~ 3.9, the reduction in physical thickness in nm, is
approximately 0.9 k. For values of k ~ 10-20, this decreased physical thickness is
equivalent to an increase in direct tunneling current of three to four orders of magnitude.
This estimate is based on a dielectric constant of k~15, a 1.5 to 2 eV conduction band
offset energy, E
bc
, an electron tunneling mass of 0.5 m
0
, and an applied bias of one volt
across the dielectric film. Therefore it is critically important to determine i) if interfacial
nitride oxides are required for the high-k dielectrics, and ii) if there are any other interface
options that would lead to a smaller vlaue of EOT)
int
. The nitrided interfaces required with
silicon oxynitride and nitride dielectrics limit the application of these layers to EOT ~ 1 . 1
nm, for a direct tunneling current limit of 1 -5 A-cm
2
.
SUMMARY
A classification scheme for gate dielectric materials based on bond ionicity as
characterized by and I
b
has been introduced. Based on this scheme there are three
classes of deposited dielectrics that can be implemented into aggressively scaled silicon
MOS devices. These are continuous random networks such as silicon nitride and silicon
oxynitride alloys. The major limitations for the implementation of these dielectrics have
been shown to be mechanical bonding constraints at Si-dielectric interfaces and internal
dielectric interfaces in stacked structures. These limitations derive from increases in N
av
which in turn promote values of C
av
in excess of three and lead to defect concentrations
increasing well into the 10
11
cm
2
regime. The second class of dielectrics are based on
modified continuous random networks. These dielectrics have a strong ionic contribution to
their bonding, but retain aspects of network connectivity. Since the constraint theory of
Refs. 4 and 5 is based on valence forces, it can only be applied to some members of this
class that include a strong network component of bonding, e.g., Al
2
O
3
and SiO
2
-rich
silicate alloys. The major limitations in this class of materials come from their ionic
character, e.g., fixed charge at Si-dielectric interfaces. The third class of materials are
elemental metal oxides such as ZrO
2
, HfO
2
, La
2
O
3
where the bonding is ionic, i.e., the
bonding in the non-crystalline state is best described by a random close packed array of
spherical ions. The major limitation in the application of these materials is their instability
against crystallization in a polycrystalline thin film, either as deposited or a relatively low
post-deposition processing conditions. The discussion below is restricted to the crn and
mcrn dielectrics.
Continuous Random networks
The paper has addressed three different aspects of interface bonding and defect
structure that limit device performance and reliability of gate dielectrics that fall into the
category of continuous random networks.. These are i) interfacial transition regions
between the Si substrate and a dielectric film, ii) defect concentrations at over-constrained
205
interfaces between Si and different dielectric films, and i i i ) defect concentrations at internal
interfaces between dielectrics with different bonding constraints.
It has been shown that the interfacial transition regions with suboxide bonding in
excess of what is required at an abrupt interface between Si and SiO
2
are typically about
0.3 nm thick. These regions are formed during oxidation of the silicon substrate and
minimization of their spatial extent requires a post-oxidation anneal, e.g., 30 s to 1 minute
at 900C. It is particularly noteworthy that interfaces prepared by thermal oxidation at
900C, also require an anneal at the same temperature to optimize interfacial smoothness or
equivalently minimize suboxide bonding. The existence of an interfacial transition region
between crystalline silicon and SiO
2
is anticipated on the basis of constraint theory.
Constraint theory predicts that a transition region of the order of one molecular layer must
be present at the interface between two materials in which the number of bonds/atom is
different, or equivalently at which the number of bonding constraints per atom is different.
The resulting interfacial region is over-constrained with respect to the lower constraint
partner of the interface, and is self-organized in a way that minimizes the development of
mechanical strain. Interface layers of this type have been reported within glassy alloys
with microstructure composed of a floppy-or under constrained constituent, and an over-
constrained bonding partner. Floppy is defined in terms of a maximum average bonding
coordination of 2.4, except for SiO
2
where, 2.67 is more appropriate due unusually small
bonding forces at the oxygen atom sites. Values of Nav <2.4 (or 2.67) define a regime in
which the average number of bond constraints per atom is lower than the network
dimensionality. These concepts, originally applied to bulk glasses, have been extended to
thin film amorphous materials and more recently to internal interfaces between crystalline
and amorphous materials. It has been shown that control of interfacial bonding structure
by interfacial nitridation can modify interfacial transition regions between Si and SiO
2
and
produce significant reductions in tunneling currents, and thereby improve device
performance. Experiments, combined with theory of suggest that the interfacial transition
regions in advanced Si FET gate stacks are intrinsic, and result from entropy effects.
Entropy is not a factor in the crystalline substrate, but configurational entropy is one of the
most important factors in allowing for the formation of amorphous continuous random
networks. Therefore it is not surprising that it plays a role at the interface between a
crystalline solid and an ideal amorphous covalent random network solid.
Constraint theory has accounted for differences in the fixed charge at internal interfaces
of stacked gate dielectrics. In particular, the quadratic scaling with has accounted for
differences in fixed charge at SiO2-Si
3
N
4
and SiO
2
-(SiO
2
)
0.5
(Si
3
N)
0.5
internal interfaces.
This interfacial charge shifts flat band voltages and therefore will also change threshold
voltages in FETs; however, it does not appear to degrade reliability. If the charge is high
enough, and if the interfacial layer is sufficiently thin, it can also reduce effective channel
mobilities as in p-channel FETs with bulk Si
3
N
4
dielectrics.
Modified Continuous Random Networks
These materials are qualitatively different than the cnr oxides, nitrides and oxynitride
alloys. The increased ionic bonding of elemental oxides such as Ta
2
O
5
and Al
2
O
3
, and the
introduction of ionic metal oxides into SiO
2
, introduces local bonding groups were ionic in
character. This markedly changes the nature of mechanical bonding constraints, since a
description based on valence forces is no longer applies, accept for very dilute metal
silicate alloys, for up to 10 to 15 atomic percent metal oxide content, or for group IVB Zr
and Hf oxides, 3-5 atomic percent Zr or Hf. This regime of metal oxide doping provides
206
increases in k, exceeding those of silicon oxynitride alloys and nitrides and may eventually
find its way into the technology.
The paper has identified two significant issues relative to incorporation of these
materials in device technology. These are thermal stability and fixed interfacial charge.
Virtually all of the elemental oxides and binary silicate alloys, respectively either
crystallize or chemically phase separate with or without crystallization at temperatures of at
most 900C, placing thermal budget restrictions on post deposition processing. For
example, this limitation may require that high temperature processing steps such as source
drain formation is done prior to dielectric depositions, and that dopant activation in
polycrystalline silicon gate electrodes be replaced by in-situ doping during a low
temperature deposition, e.g., at 800C.
The interface limitation is equally challenging. There are several approaches that can
introduce neutral interfacial bonding on a macroscopic scale, e.g., three-five analogs of
SiO
2
such as AlTaO
4
. However, if the there are fluctuations in interfacial charge neutrality
on a scale of 0.1 to 1 percent, corresponding the densities of scattering centers of the order
of 10
11
to 10
12
cm
2
, these may reduce the mobilities of holes and/or electrons in the
channel regions of field effect transistors.
On the bright side, the identification of a mechanism for dielectric enhancement
discussed above opens up new opportunities to explore low concentration silicate alloys
where both thermal stability and interface properties may be closer to SiO
2
.
Acknowledgements
Research support from the ONR, AFOSR, NSF and SEMATECH/SRC Front End
Processing Center is acknowledged. The contributions from current and former graduate
students and postdoctoral fellows in my research group at NC State University is gratefully
acknowledged. These include Sunil Hattangady, Hiro Niimi, Hanyang Yang, Yider Wu,
Bruce Rayner, Bob Johnson and Bob Therrien. The author also acknowledges many
important discussions and collaborations with Jim Phillips of Lucent Bell Laboratories and
Mike Thorpe of Michigan State University.
REFERENCES
[1] Lucovsky, G. (1999), IBM J. Res. and Develop. 43, 301.
[2] R. Zallen, R. (1983) The Physics of Amorphous Solids, John Wiley and Sons, New York, pp. 49-72.
[3] Pauling, L. (1948) The Nature of the Chemical Bond, Cornel University Press, Ithaca, pp. 58-73.
[4] Phillips, J.C. (1979) J. Non-Cryst Solids 34, 153 ; (1981) J. Non-Cryst Solids 43, 37.
[5] Phillips, J.C. (2000) J. Vac. Sci. Technol. B 18, 1749.
[7] Kerner, R., and Phillips, J.C., (unpublished).
[8] Zallen, R. (1983) The Physics of Amorphous Solids, John Wiley and Sons, New York, pp. 100-101.
[9] National Technology Roadmap for Semiconductors, (SIA, San Jose, 1999)
[10] Lucovsky, G., Yang, H., Niimi, H., Keister, J.W., Rowe, J.E. Phillips, J.C., and Thorpe, M.F. (2000)
J. Vac. Sci. Technol. B 18, 1742.
[11] Keister, J.W., Rowe, J.E., Kolodzie, J.J., Niimi, H., Tao, N.-S., Madey, T.E., and G. Lucovsky G.
(1999), J. Vac. Sci. Technol. A 17, 1340.
[12] Weldon, M., Queeny, K.T., Chabal, Y.J., Stefanov, B.B., and K. Raghavachai, K.. (1999) J. Vac. Sci.
Technol. B 17, 1795.
[13] Lucovsky, G., A. Banerjee, A. Hinds, B., Claflin, B., Koh, K.., and H. Yang, H. (1997) J. Vac. Sci.
Technol. B 15, 1074.
[14] Thorpe, M.F., Jacobs, D.J. and Chubynsky, M.V. (2000) J. Non-Cryst Solids 266, 859.
[15] Lucovsky, G. and Phillips, J.C. (1998) J. Non-Cryst Solids 227, 1221.
[16] Lucovsky, G., Wu, Y., Niimi, H., Misra, V., and Phillips, J.C. (1999) Appl. Phys. Lett. 74, 2005.
[17] Selvanathan, D., Bresser, W.J., Boolchand, P., Phys. Rev. B 61, 15061 (2000).
[18] Tu, Y and Tersoff, J. (2000) Phys. Rev. Lett. 84, 4693.
207
[19] Yang, H., Ni i mi , H., Keister, J.W., Lucovsky, G., and Rowe, J.E. (2000) IEEE Electron Device
Letters 21, 76.
[20] Lucovsky, G., Ni i mi , H., Koh, K., Lee, D.R., and Jing, Z (1996) H.Z. Massoud, E.H. Poindexter and
C.R. Helms (eds). The Physics of SiO
2
and Si-SiO
2
Interfaces-3, The Electrochemical Society,
Pennington, N. J. , p. 441.
[21] Misra, V., Wang, Z., Wu, Y., Niimi, H., Lucovsky, G., Wortman, J.J., and Hauser, J.R. (1999) J.
Vac. Sci. Technol. B 17, 1836.
[22] Wolfe, D.,Flock, K., Therrien, R., Rayner, L. Gnther, L., Brown, N., Claflin, B., and Lucovsky, G.
(1999) Mat. Res. Soc. Symp. Proc. 567, 343.
[23] Rayner. B., Therrien, R., and Lucovsky, G. (2000) Mat. Res. Soc. Symp. Proc. in press.
[24] Robertson, J. (2000) J. Vac. Sci. Technol. B 18, 1785.
[25] Hirose, M. and S. Miyazaki, S. unpublished.
[26] Lucovsky, G., Rozaj-Brvar, A., and R.F. Davis, R.F. (1983) P.H. Gaskell, J.M. Parker and E.A.
Davis (eds) The Structure of Non-Crystalline Materials, Taylor and Francis, London, p. 193.
[27] Wilk, G.D. and R.M. Wallace, R.M. (1999) Appl. Phys. Lett. 74, 2854.
[28] Wilk, G.D., Wallace, R.M., and Anthony, J.M. (2000) J. Appl. Phys. 87, 484.
[29] Wilk, G.D. and R.M. Wallace, R.M. (2000) Appl. Phys. Lett. 76, 112.
[30] Misra, V., unpublished.
[31] Jayasundere, N and Smith, B.V. (1993) J. Appl. Phys. 73, 2462.
[32] W.A. Harrison, W.A. (1999) Elementary Electronic Structure,World Science, Signapore, Chap. 11.
[33] Burstein, E., Brodsky, M.H., and Lucovsky, G. (1967) International J. Quantum Chem. 1S, 759.
[34] Cotton, F.A. and G. Wilkenson, G. (1972) Advanced Inorganic Chemistry, 3rd Edition, John Wiley
and Son, New York, pp. 122-124.
[35] Torpov, N.A. and I.A. Bondar, I.A. (1961) Izv. Akad. Nauk. SSSR, Ofd. Khim. Nauk 4, 547.
208
EXPERIMENTAL METHODS FOR LOCAL STRUCTURE
DETERMINATION ON THE ATOMIC SCALE
E.A. STERN
Department of Physics, University of Washington,
Box 351560, Seattle, WA 98195-1560
I. INTRODUCTION
It is generally agreed that for a fundamental understanding of the properties of the con-
densed state it is necessary to know the composition and arrangement of the atoms in the
material. In this article I will review the main experimental techniques for directly deter-
mining the local structure of materials in the condensed state with emphasis on the type of
information each technique determines and the strengths and limitations in obtaining such in-
formation. The theory of the structure information obtained by the experimental techniques
is already well known and only that aspect of it related to the focus of this review will be
presented here.
The most generally employed method is coherent scattering of either x-rays or neutrons.
The most common application of the technique is on crystals where the scattering is peaked
about discrete directions corresponding to Bragg peaks. By an appropriate analysis of the
intensity and line shapes of only the Bragg peaks one can obtain the average periodic structure
of the materials, e.g., by the Rietveld refinement method [1].
In recent years the interest in condensed matter physics has focused on phenomena such
as high T
c
superconductivity and colossal magnetoresistivity where the materials are rela-
tively complex, and there are suggestions and experimental evidence that local deviations
from the average periodicity for such materials may be important for understanding their
properties. Even for the well-studied phenomenon of ferroelectricity, it has recently been
shown that local deviations from periodicity occur which were previously not suspected [2].
This has resulted in a reassessment of the previous theories of ferroelectricity leading to a
new theory which includes the local deviations [3]. This theory based on a new physical
mechanism of ferroelectricity gives greatly improved quantitative and qualitative agreement
with experiment.
To obtain the local deviations, the standard Bragg diffraction method needs to be sup-
plemented. If any such local disorder is present in the crystal, the standard Bragg diffraction
method is able to detect this only as a decrease in intensity in the Bragg peaks, since such
disorder diverts some scattering into angles between the Bragg peaks, so-called diffuse scat-
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 209
tering. Such a decrease in intensity would be interpreted only as a larger thermal factor of
the vibration of atoms about their lattice sites. Detecting the local disorder requires including
in the analysis the diffuse scattering between the Bragg peaks. An analysis of both the diffuse
scattering and the Bragg peaks leads directly to the local structure [4] in the form of the pair
distribution function (PDF), as defined below.
Disorder in the distribution of atoms can be classified as (a) local disorder which does
not destroy the long-range order of the crystal or as (b) in amorphous solids and liquids where
the disorder destroys long-range order. For the crystal with local disorder, only some of the
Bragg peak intensity is transferred to the diffuse scattering, while in the sample without long-
range order all of the scattering is diffuse. By neglecting the diffuse scattering the refinement
methods of Bragg reflection do not detect the local disorder. In this review I wi l l focus on
methods that detect the local disorder. For scattering techniques this means including the
diffuse scattered intensity in addition to any Bragg peaks that may be present when analyzing
the data.
As discussed in more detail below, the local structure determines the actual distance
between atoms while, when disorder is present, the diffraction structure may not correctly
display the actual distances between atoms. For any fundamental understanding of the prop-
erties of materials, a knowledge of the actual distances between atoms is important, since the
atom-atom interaction is a strong function of this distance. An example was mentioned above
in relation to ferroelectricity.
When an x-ray photon or neutron particle is scattered from atoms in materials there
is transfer of momentum and energy to the material. The and are given by
the momentum and energy change, respectively, of the scattered particles. The intensity
distribution of the scattering particles as a function of and can be analyzed to obtain the
structure factor As discussed in Sect. II, the pair distribution function, also called
the pair density or density correlation function, all being denoted by PDF, can be obtained
from the structure factor. An alternate method to obtain the local structure including any local
disorder is the x-ray absorption fine structure (XAFS) technique as discussed in Sect. IV.
The techniques of both XAFS and diffuse scattering directly measure the local structure.
Other techniques such Raman scattering and magnetic resonance are also employed to obtain
information on the local structure, but their information is more indirect and of less general
applicability. In this review I will focus on the two techniques that directly measure the local
structure, namely diffuse scattering and XAFS.
The outline of the review is as follows. Sect. II defines local structure in terms of
the PDF, the partial PDF (PPDF) and the structure factor The relationship between
the local structure and the average periodic structure will also be discussed in this section.
Sect. Ill presents the aspects of the local structure determined by diffuse coherent scattering,
discussing the strengths and limitations of using x-rays and neutrons. Sect. IV discusses the
XAFS technique for determining the local structure, presenting its strengths and limitations.
Sect. V presents a discussion, and a summary and conclusion are given in Sect. VI.
II. AVERAGE PERIODIC AND LOCAL STRUCTURE
From measurements of the coherent scattering of x-rays or neutrons it is, in principle,
possible to obtain directly the time-dependent pair distribution function of atoms,
which gives the joint probability of finding an atom at point at time t = 0 and an atom at
point at time t later. Because the typical experiment measures the intensity of the scattered
radiation instead of the amplitude and phase of the scattered waves, absolute phase is lost
and only relative phase between two scatterers is retained. This gives a correlation between
two atoms leading to In particular, diffuse coherent scattering gives directly the structure
210
factor which is given by [5]
(1)
Equation (1) shows that is the space-time fourier transform of
The single particle density operator is given by:
(2)
where is the position of the jth particle at time t and the angle brackets in Eq. (1) indicate
a thermal average at temperature T of the probability that a particle is at at and any
particle is at at later time t. The average periodic structure determined by using only
Bragg peaks is given by while the PDF
(3)
The most common samples studied by diffuse elastic scattering are either liquids, amor-
phous solids or polycrystals. In this case, the PDF is averaged over angle and is a function
of where and is called the radial distribution, corre-
lation or density function.
To obtain the single particle distribution function requires determining the absolute
phase of the scattered wave. For large unit cells as in biological macromolecules, additional
measurements are required, such as heavy atom substitution or multiwavelength anomalous
diffraction (MAD) techniques [6]. For small unit cells the absolute phase problem can be
solved by refinement methods, such as the Rietveld method [1]. These methods of solving
for are only applicable to crystals with long-range periodicity and determine the average
periodic structure, since these techniques use only the sharp Bragg peaks in their analysis.
An important difference between and should be noted. In a spatially homogeneous
amorphous solid or liquid (as expected for the condensed state in a single phase) is a
constant. In a single crystal has a periodic array of atomic lattice sites determined by
the space group symmetry of the crystal. The typical refinement of the structure accounts for
the disorder due to vibrations (i.e., the Debye-Waller factor) by assuming a gaussian disor-
der about the lattice sites and then determines the gaussian-width about each lattice site. On
the other hand, determines the relative distances between pairs of atoms and the disorder
or distribution about this relative distance. The most striking difference between these two
distribution functions is illustrated for the amorphous and fluid state. Whereas is a con-
stant at the average density value, displays the short-range order of peaks at the first few
nearest neighbor distances and then becomes blended into the constant average density as
schematically shown in Fig. 1. In crystals, if there is local disorder present, then both and
show peaks, but their relative spacings and distribution may not coincide. Coincidence
occurs only if there is no local disorder. For example, PbTiO
3
, which is ferroelectric and has
a tetragonal structure at low temperatures, displays a cubic perovskite structure by Bragg re-
flection measurements in the paraelectric phase above 760 K. However XAFS measurements
[7] determine that the local structure in the paraelectric phase has a displacement of the Ti
atoms from their cubic symmetry sites relative to the oxygen and Pb atoms, which are almost
as large as occurs in the ferroelectric phase. The paraelectric phase occurs because the local
displacements lose their long-range correlations above T
c
, the ferroelectric to paraelectric
phase transition temperature, i.e., they become disordered so that their average displacement
is zero.
When structure is the goal of measurements, one is not interested in the general
but only the instantaneous snapshot of the structure given by the PDF. By referring
to Eqs. (1) and (3) and using the relation that
(3a)
211
Figure 1. A schematic of the radial pair distribution function as a function of the radial distance R for a
liquid. The short-range order of peaks of the first few neighbors are discerned and then the distribution blends
into the constant average density of the liquid.
it is found that
(4)
Thus, if the diffuse scattering measurement does not energy discriminate but detects all
of the scatterings, both elastic and inelastic, then the spatial fourier transform of gives
the PDF.
In the above discussion it is assumed that the atoms cause all inelastic scatterings, which
neglects electronic excitations. In practice, the inelastic scattering cross-section from elec-
trons is much smaller for neutrons than the inelastic scattering from the nucleus of atoms,
while the scattering of x-rays, though dominated by the intact atom without electronic excita-
tions, still has a significant inelastic signal from the inelastic electron excitations of Compton
scattering and the Raman effect. The inelastic energies for both neutron and x-ray scattering
from intact atoms is typically of the order of phonon energies, namely the order of 15-100
meV, while the electron excitation energies are in the eV to many eV range. The experi-
mental measurement of x-ray scattering is performed in a manner as to correct for both the
Compton scattering and all other electronic inelastic loss contributions, but integrates over
all of corresponding to the atomic motion contibutions. In short, if the experimental tech-
nique determines where then it accurately determines
As we discuss next, in some cases neutron scattering measurements do not perform a correct
integration over to accurately determine
First, to understand the problem introduced by not integrating over a large
enough frequency range we evaluate Eq. (1) when integrating over
(5)
where
Note that the weighting factor of (sinu/u) in the integral over du gives its largest con-
tribution over the range of and thus the term in brackets is integrated over the range
212
from If atoms do not move appreciably in the time then
and the integral in Eq. (5) over u gives the PDF, and is accurately determined.
On the other hand, if atoms move a large distance in the time then the integral in Eq.
(5) may give an inaccurate estimate of and the PDF. This is certainly the case for fluids
where atoms can diffuse large distances from an initial starting point.
In crystalline and amorphous solids where diffusion is negligible, and at low temper-
atures so that the thermal energy is small compared to the zero point energy and quantum
effects dominate, the PDF is a gaussian independent of time and the integration over time or
u gives a correct PDF even when However, when the thermal energy is much larger
than zero point energy, classical physics dominate and the atoms typically vibrate harmoni-
cally about a fixed point with an average angular frequency that is denoted here by The
net displacement is small when averaging over the time interval even when
so that atoms classically vibrate many periods over the time interval . However, the
PDF is still greatly distorted when It can be shown that what is determined for a
one-dimensional oscillator of amplitude A about the origin is a distribution function that is
a constant in space between while the correct distribution is
Note that since typically for oxygen bonds where T
R
is room temperature and k
B
is the Boltzmann constant, then for temperatures below T
R
the quantum regime applies and
the correct PDF would be determined for all values of
Moreover, there is evidence that in some cases the motion of atoms may not be har-
monic, e.g., a bifurcation of the planar Cu2 to apical oxygen (O4) relative distance has been
detected [8] in with some indication that a snapshot of the distribution is not
being measured. This shows up by the anomalous q-dependence of the bifurcation as shown
in Fig. 2. If the snapshot distribution is being determined, then decreasing q
max
(Q
max
in
Fig. 2) should simply broaden the two peaks proportionally to The change between
q
max
= 23
1
. Such is not the case. At q
max
= 23
1
the distribution is not simply broad-
ened but the splitting and positions of the peaks have greatly changed, placing into doubt any
quantitative measure of splitting, or even if the splitting exists. In the presence of a dynamic
non-harmonic motion, in order to determine the snapshot PDF, requires where
is the characteristic time of significantly changing the distribution. For the neutron scatter-
ing measurements of Fig. 2, which corresponds to For
the anharmonic dynamics of atoms it is quite possible that for this value of that the re-
lation is not satisfied, even for oxygen atoms, and the PDF will not be accurately
determined in this region, as indicated by the anomalous dependence of Fig. 2.
III. DIFFUSE SCATTERING BY NEUTRONS AND X-RAYS
The coherent scattering of x-rays from a single atom is predominantly caused by the
interaction of the electric field of x-rays with the average charge density of the atom when the
x-ray photons have energy much greater than the most tightly bound electrons. A measure
of the scattering of the incident x-rays electric field by the j
th
-type atom in the sample is
given by its atomic form factor which is the fourier transform of the electronic charge
density of the j
th
atom with respect to The value of f
j
(0) is Z, the number of electrons
in the atom and the intensity of the scattered radiation from that atom is proportional to Z
2
.
The PDF determined by diffuse scattering of x-rays is a sum of the distribution of the relative
distance between all pairs of atoms with the contribution from an atom proportional to its
Z
2
. Thus, the scattering power of low Z atoms like oxygen compared to that of a heavy atom
like Cu are less by a factor of (8/28)
2
= 0.08, making it difficult to detect oxygens in the
background of the scattering from heavy atoms. For that reason, diffuse x-ray scattering is
213
q
max
= 31
1
and q
max
= 27
1
should be about the same as between q
max
= 27
1
and
Figure 2. The dependence of the positions of the Cu2-O4 peaks of as presented in ref. 8. The
maximum q (Q
max
in the figure) is varied which should just decrease the resolution between the two peaks.
Instead the plot shifts the peak positions and increases the splitting, indicating that the instanta-
neous PDF is not being measured in this region as discussed in the text,
not the method of choice for materials where oxygen plays an important role, such as the
Cu-O high T
c
materials. As discussed below, the scattering of neutrons is more favorable for
detecting oxygen atoms.
When x-ray photons have energies near absorption edges, so-called anomalous or reso-
nance scattering occurs where the scattering strength of the atom is energy dependent. Uti-
lizing this anomalous scattering allows the determination of the partial PDF (PPDF), i.e., the
PDF about the atom with anomalous scattering [9,10]. The PDF is the sum of the PPDF
about each type of atom of the material so the PPDF gives more detailed information about
the structure and is preferred when available.
X-rays used for diffuse scattering are in the hard x-ray region, greater than 3KeV, be-
cause soft x-rays have too long wavelengths to resolve atomic dimensions. Unless heroic
efforts are used to obtain very high energy resolution, the inelastic losses due to atomic mo-
tion, typically 20 meV, are not resolved and the detected x-rays then are integrated over the
full w range of the atomic motion and determine In order to be able to utilize anoma-
lous scattering to obtain the PPDF with atomic resolution, the absorption edge energy of the
atom of interest must be in the hard x-ray range. This precludes the use of light atoms as
anomalous scatterers. In addition, the need to use x-rays near absorption edge energies limits
the q
max
that is obtainable, except for the very heaviest atoms. More details are given in this
Section below, in Sect. V and in the extensive reviews of Waseda [9,10].
Neutrons scatter from the nucleus of the atoms and the scattering strength from atoms
depend only on their nuclear properties and not on the atomic Z. The scattering of neutrons
from a nucleus of charge Z is strongly spin- and isotope-dependent. Even for a sample com-
posed of a single isotope, the scattering depends strongly on the projection of the nuclear spin
on the neutron spin and, unless the nuclear spin is zero, the scattering from these atoms will
differ dependent on this projection. This introduces a disorder in the scattering that produces
both coherent and incoherent scattering, where the coherent scattering amplitude is propor-
tional to the average of the scattering factor of the nuclei at each site, and the incoherent
scattering by the rms fluctuation of the scattering factor from the average value. Only the
coherent scattering defines the PDF. When it is intense enough the coherent scattering can be
214
separated from the incoherent scattering. For some nuclei, e.g., vanadium and hydrogen, the
coherent component is weak compared to the incoherent one and it is difficult to obtain their
contribution to the PDF. However, for most nuclei such as Cu and O their coherent signal
can be used to obtain the PDF. Neutrons have a great advantage over x-rays for determining
the O contribution to the PDF, since its coherent scattering is comparable to that of heavy
atoms such as Cu. For that reason neutrons are the particles of choice for determining both
the average periodic structures and the PDF of the Cu-O high T
c
materials.
Neutrons can also be used to obtain the PPDF by isotropic substitution. For instance, a
recent diffuse scattering investigation of YBa
2
Cu
3
O
6.92
determined the PPDF about Cu atoms
by utilizing the contrast in the neutron scattering cross-section between
63
Cu and
65
Cu in two
isotropically-pure samples with identical composition [8]. However, isotope substitution is
limited by the availability of suitable isotopes and by how identical the two isotope samples
can be made in composition and perfection.
Neutrons have another advantage over x-rays in that the scattering amplitude from a
nucleus is essentially constant as a function of permitting the determination of the scattering
signal to high values of up to 35
1
and beyond. This feature is a consequence of the
small size of the nucleus which for practical q-values appears as a delta-function in space.
In contrast, the scattering amplitude f
j
of x-rays decreases significantly with q for values of
where the radius of the atomic mean charge, causing a practical limit
for measuring scattering of In addition, for atoms lighter than Tc, anomalous
scattering requires lower q x-rays, lowering below 20
1
. Thus, neutrons can determine
up to significantly larger values of q which translates into a higher spatial resolution in
the PDF. To illustrate this point, consider a solid that has a structure containing an atom pair
whose relative distance is split. To detect the split requires having enough spatial resolution
in the measurement. Similar to the Rayleigh criterion for distinguishing the separation
between two points, the criterion for distinguishing the splitting from another distribution,
such as a gaussian broadening, is that
(6)
This criterion guarantees that the measurement detects the unique beat in the interference
between the scattering from the two sites involved in the splitting. At q-values below the
beat, the signal for a gaussian and split is quite similar and only near the beat occurring at
is the distinction clear. Thus, to resolve a splitting of requires including
q-values up to at least q
max
= 30
1
To obtain such large q-values in neutron scattering requires a source of higher energies
than obtainable from a continuously emitting thermal reactor. Pulsed sources of neutrons
are used to obtain such energetic neutrons, e.g., the Intense Pulsed Neutron Source (IPNS)
of the Argonne National Laboratory. The scattered neutrons are analyzed by a time of flight
measurement determining both their speed and angle of scattering, allowing the determination
of by these two quantities. A detector at the IPNS used to determine PDFs is the Glass,
Liquids, and Amorphous Materials Diffractometer (GLAD), The detected by GLAD
is integrated over energy with an energy cutoff at about 20 meV. However, this spread in
causes the q not only to depend on the angle of scatter, but also on since the final magnitude
of the scattered neutron momentum depends on Thus and the integral over
is not the correct one to determine S(q), since that requires that remains constant as
is integrated. This problem does not occur for x-rays because is orders of magnitude
smaller than the photon energy while it is not negligible compared to the neutron energy. This
problem is additional to the issue discussed in the previous section which required that the
cut-off energy be large compared to phonon energies to guarantee a correct value of A
procedure has been developed to correct the dependence only when extended phonons
are involved [11], but not for localized modes. In practice, it appears that errors introduced
215
in the PDF by these effects are limited to temperature ranges around phase transitions where
dynamics in splittings of metal-oxygen distances apparently occur with time scales that are
sensitive to both the -dependent effects, causing the anomalous behavior shown in Fig. 2
and discussed above and in ref [8]. The result is that some possibly very interesting behavior
of the PDF for high T
c
and colossal magnetoresistance phenomenon detected by neutron
scattering may be occurring near the transition, but unfortunately one cannot be confident
that diffuse coherent neutron scattering is correctly determining this behavior.
In any experimental determination of the PDF it is very important to assess the uncer-
tainties in the result. Both x-ray and neutron scattering can obtain the PDF with reasonable
accuracy. However, the PPDF accuracy is not as favorable as it is derived from a smaller
difference between larger measured spectra, which introduces greater uncertainties in the fi-
nal result compared to that in each measured spectrum. This is especially severe for x-ray
scattering where the anomalous differences for high concentration atoms are typically 10%
or less, thus increasing the uncertainties by more than an order of magnitude. On the other
hand, anomalous x-ray scattering determines its various spectra on exactly the same sample,
whereas neutron scattering requires fabricating two different samples with different isotopes
introducing systematic uncertainties which are hard to assess. This important issue of uncer-
tainties is discussed further in Sect. V.
In summary, neutrons have significant advantages over x-rays for determining the PDF
(better spatial resolution and better sensitivity to low Z atoms) but, because of experimental
limitations, neutrons measurements sometimes do not have a rapid enough time response
nor the correct dependence to obtain an accurate snapshot determination of the PDF of
anharmonic motions in temperature ranges near some phase transitions, which, unfortunately,
are physically very interesting regions.
IV. X-RAY ABSORPTION FINE STRUCTURE (XAFS)
XAFS is another technique that determines directly the PPDF using different physics
than diffuse coherent scattering. XAFS is the fine structure that occurs in the x-ray absorption
coefficient on the high-energy side of absorption edges [12]. The absorption edges occur as
the x-ray photons attain enough energy to just dislodge electrons bound in atoms. The fine
structure occurs only when the atoms are in the condensed state and the dislodged electron
can backscatter from surrounding atoms and interfere with the outgoing portion of its wave
function. By appropriately analyzing this XAFS, the arrangement of atoms about the x-ray-
absorbing atom, the probe atom, can be determined [12]. Recent advances in theory [13]
and analysis [14] have extended the range of reliable detailed structural information from
the first neighbor to four or more neighboring shells of atoms about the probe atom. Among
other information, XAFS directly obtains the PPDF of the relative distance between the probe
atom and its neighbors with a high spatial resolution (typically the former
for light atoms and the latter for heavy atoms) and with high sensitivity. Since the XAFS
signal is produced by the scattering of the photoelectron from its neighboring atoms, this
scattering does not have the decrease in signal from light atoms compared to heavy atoms
that x-ray scattering suffers from. The electron scattering from light atoms has a comparable
signal to heavy atoms but differs in the energy dependence of this signal. This dependence
allows the identification of the surrounding atoms, while the x-ray absorption edge energy
identifies the probe atom. XAFS measures the PPDF for the most dilute concentrations of
impurities by far than the coherent scattering techniques because it measures this feature
directly. As mentioned above, since the other techniques of x-ray and neutron diffuse elastic
scattering measure the PPDF by subtracting two large numbers from one another to obtain a
very small result for dilute impurities, the answer has a much larger relative uncertainty than
either measurement. XAFS, by using the atom specific experimental technique of detecting
216
its characteristic signal by x-ray fluorescence or electron emission, obtains the spectra from
only that atom, giving the PPDF about that atom directly without adding noise from the signal
coming from the PPDF of the other atoms, leading to its much greater sensitivity for detecting
PPDF about dilute impurities, down to ppm.
In addition to determining the PPDF, which is a two-body correlation function, a unique
feature of XAFS is its capability to determine three-body correlations with high accuracy in
special cases, namely, when three atoms are collinearor nearly so, due to a focusing effect of
the intermediate atom on the propagation of the photoelectron [12,15]. Compared to neutron
scattering, XAFS has the same advantage of x-ray scattering of a stronger interaction with
matter allowing the use of much smaller samples. The XAFS measurements are much quicker
and less tedious to complete than the diffuse scattering techniques. Orientation information
on the PPDF is obtained in the special cases of anisotropic single crystals by varying the
angle between the x-ray polarization (which is horizontal for synchrotron radiation sources)
and the crystal axes.
To observe the instantaneous PPDF, the characteristic time scale of the measurement
must be much shorter than the characteristic time in which the arrangement of atoms can
change. Since its characteristic time is typically less than 10
15
seconds, XAFS easily sat-
isfies this criterion. For comparison, the indirect experimental techniques of magnetic res-
onance and Raman scattering have characteristic times of 10
10
seconds or longer and are
limited in obtaining a snapshot of dynamic distortions to ones that are relatively slow. For
example, in second-order structural transitions the dynamics of local distortions become fast
somewhat above T
c
, and these techniques typically lose their ability there to determine any
disordered local distortions, e.g., as occur in ferroelectricity [2].
Two other aspects of the information content of XAFS should be mentioned. The XAFS
analysis software, e.g., UWXAFS [14], contains reliable objective estimates of the uncertain-
ties of the PPDF determination. This is important in assessing the reliability of the results.
The present analysis procedures for the diffuse elastic scattering techniques do not present
objective estimates of the uncertainties of their PDFs [8,9] as discussed in more detail in the
next section. In addition to the PPDF information, the XAFS spectrum within 15 eV of the
edge (the XANES) contains electronic structure information, e.g., the valence state of the
probe atom, and in some cases its coordination geometry [12]. The XAFS spectrum that
extends far beyond the edge (typically from 15-1000 eV), the EXAFS, contains the PPDF
information.
XAFS has two limitations. The region around the probe atom where XAFS data can
be analyzed is limited by the mean free path of the photoelectron and the complications of
multiple scattering effects to typically four nearest neighbors. Also, because of both the
physics of the photoelectron excitation process and difficulties in distinguishing the XAFS
fine structure from the rapidly increasing total absorption near the edge, XAFS presently
does not obtain the values below q = 6
1
. This latter deficit translates into XAFS not
being able to measure distribution functions with widths greater than about 0.4. In practice
this typically limits XAFS to determining the PPDF of only the first neighboring atoms in
disordered liquids, glasses and amorphous solids as illustrated in the next Section.
Fortunately many interesting physical phenomena occur with local disorder in crystals
with long-range order where XAFS can probe the PPDF to four nearest neighbors. Examples
are structural phase transitions, including ferroelectricity and strongly interacting materials
such as colossal magnetoresistance and high temperature superconductors. For these materi-
als XAFS is the premier technique for determining the local structure up to 5 6. Even in
the highly disordered materials XAFS has decided advantages for the PPDF of the first neigh-
bors, namely, more accurate determination of ligand distributions and greater sensitivity to
dilute components.
217
V. DISCUSSION
By determining the relative distribution of local displacements of atoms from the average
structure, the pair distribution function (PDF) supplements diffraction measurements, which
determine only the average periodic structure. In cases where structural disorder is induced
by compositional disorder, this information allows the determination of the actual structure,
including visualizing the displacements of atoms from the periodic structure [16]. In cases
of crystal structures with no local disorder and the location, of atoms within the unit cell is
not completely defined by the space group crystal symmetry, refinement techniques on the
diffraction measurements, such as the Rietveld [1] method, are required to solve the structure.
When, for experimental limitations, not enough diffraction data are available for refinement,
e.g., measurements under high pressure, adding PPDF using XAFS, allowed the solving of
the structure of the intermediate pressure phases of AgCl [17].
Information contained in the PPDF is important for understanding the atomic interac-
tions that are present in solids because they determine the actual distances between atoms
which may differ from the average periodic structure distances when local disorder is present
as mentioned in Sect. I in regard to perovskite ferroelectrics [2,3].
In a sense coherent scattering and XAFS are complementary in determining the PDF.
The much stronger interaction of the photoelectron with its surroundings compared to x-rays
and neutrons limits the spatial range of XAFS for probing the PDF to the much smaller range
of about 4-5 . In addition, the inability of XAFS to determine low q information limits the
detection of spatially slowly varying distributions. On the other hand, XAFS is able to more
accurately determine the PPDF in the range it covers.
Figure 3 shows the difficulty that the lack of low q data introduces for determining the
PPDF beyond the first coordination shell about Cu ions in water solutions under hydrother-
mal conditions, and also illustrates the clean signal XAFS gives for determining the first shell
to high accuracy in spite of its relative dilution. There are just two discernable peaks above
the noise for all four plots which have been measured by Fulton et al. [18] at the three tem-
peratures indicated in Fig. 3. The highest temperature 350C is at supercritical conditions.
Besides varying the temperature and its concomitant pressure the solutions had varying con-
centrations of Cl ions added by dissolving various amounts of NaCl and HCl besides CuCl.
Without going into the details of how the Cl concentration relative to the Cu concentration
was varied, the issue to focus on is what can XAFS detect in such a disordered medium. The
first and largest peak is the first neighbor ligands which for the T = 25C sample is six wa-
ter molecules. The second and third plots have two Cl atoms as first neighbors without any
detectable water molecules in that same shell. The last plot contains one Cl atom and one
oxygen atom without any other atoms detected in the first shell. Although the Cu atoms have
only a fraction of a molarity concentration (0.2-0.4), there is a very clean signal allowing the
determination of distances to better than 0.01 and coordination number to about 10%. Note
that the second smaller peak is roughly twice the distance of the first peak. The locations
of the peaks are not at the actual distances, but the theory can interpret the data to give the
necessary corrections to determine the true distances to the stated accuracy.
It turns out that the small second peak is not due to an actual position of a peak in the
PPDF but is due to multiple scatterings of the photoelectron due to a pair of first neighbors
being collinear with the Cu probe atom, as shown in the drawings in the upper right-hand
corner of Fig. 3. The relevant multiple scattering consists of the photoelectron scattering
backwards from one neighbor, scattering forward through the Cu atom to the opposite neigh-
bor and finally backwards to the Cu atom. The total distance the photoelectron travels is twice
the single scattering path from the Cu to its neighbor and back again to the Cu, causing the
small peak at the larger distance. This multiple scattering signal is greatly enhanced by the
forward scattering and would not be visible if the atoms were not collinear. Thus the pres-
ence of the multiple scattering peak and the first peak are evidence of both the collinearity of
218
Figure 3. The magnitude of the transform of the EXAFS signal from Cu ions in water solution under various
temperatures and Cl ion concentrations as presented by ref. 18. This plot is a pseudo-PPDF which, after
appropriate analysis, determines the correct PPDF relative to the Cu ions, leading to the three ligand structures
shown in the diagrams in the upper right-hand corner. The text explains the origin of the two peaks that are
significantly above the noise level in each plot as being caused by the first neighbors of the Cu ion. No other
neighbors are detected because of their broad distribution in the PPDF.
pairs of first shell atoms, and the strength of their bonding as verfied by its rms disorder of
0.02, sharp enough to be fully measurable by XAFS. However, no other shells are visible
indicating that all the rest of the surrounding atoms are not tightly bound to the Cu atoms and
have a distribution too broad to be detected by XAFS, rms > 0.2.
Thus, because of the high sensitivity of XAFS to determining the PPDF it is able to
determine the near portion of the PPDF to higher accuracy and for far more dilute atoms than
can the coherent scattering techniques. In addition, in the special cases when three atoms
are near collinearity, XAFS has good sensitivity for measuring three-body correlations, in
particular, it determines directly, where is the angular deviation from collinearity.
Finally, the XANES portion of the XAFS (not shown) determines the valence of the Cu ion
and independently verifies the coordination number about the Cu atoms.
The complementarity between XAFS and diffuse coherent scattering was first exploited
by Raoux and Flank [19] who combined the results of the two techniques to obtain a better
model of the PPDF of a series of amorphous Cu-Y alloys than could be obtained from either
separately. More recently di Cicco et al. [20] combined the two techniques to obtain an
improved PPDF of silver-halides melts. The addition of XAFS to the coherent scattering
results improved the short-range atom-atom correlations in both cases.
It is striking that based on the anomalous x-ray scattering measurements of Saito and
Waseda [10] a direct numerical solution for the three partial structure factors of molten CuBr,
namely, the q-dependent interference pattern due to scattering between the Cu-Br, Cu-Cu,
and Br-Br pairs of atoms which contain the information leading to the three PPDFs for these
219
pairs of atoms, contain large errors as shown in Fig. 4. The cause is apparently due to the
required solution of three simultaneous linear equations being ill-conditioned at various q
values because of the quite small difference of the anomalous scattering and the magnitude
of the inversion determinant in the denominator becoming small which greatly magnify un-
avoidable experimental errors. Worse yet is that apparently the errors are not random, but
are systematic, since a direct fourier transform to the corresponding PPDFs gives physically
unreasonable results!
Figure 4. The three partial structure factors of molten CuBr at 810K as presented in ref. 10. The solid plot
corresponds to the values calculated by the reverse Monte Carlo method while the vertical lines denote the
uncertainties estimated after a direct determination from the experimental data.
To overcome this fatal problem the reverse Monte Carlo simulation technique [21] was
employed to obtain a best fit to the original experimental data. This technique is a method to
vary a distribution of a cluster of atoms with the experimental composition and determine a
best calculated fit to the experimental data. The result of this best fit is shown in Fig. 5 and
the resulting fit to the three partial structure factors are shown by the solid lines in Fig. 4.
Unfortunately, there is no objective way to estimate the uncertainties in the PPDF determined
in this way. Some uncertainties are estimated in the publication though no description is
given of how this is determined.
In the recent neutron scattering paper [8] discussed in Sect. III there is no quantitative
estimate of uncertainties given at all. This is a serious defect since one cannot judge whether
small changes in the modelled PPDF are real, i.e., whether they are outside of experimental
errors. For example, it is claimed that two types of Cu1-O4 (Cu chain and O apical) bonds
are likely to be present locally which differ in their bond distances by 0.027 . Using the
criterion of Eq.(6) for resolving a splitting of from another distribution with the same rms
220
Figure 5. The three experimentally measured interference functions as presented in ref. 10 are plotted with
the solid lines. The top is determined from the anomalous scattering from the Br atoms, the middle from the
anomalous scattering from the Cu atoms while the bottom plot is determined far from any anomalous scattering
region. The fit of the reverse Monte Carlo method is shown by the dotted lines. From these three seta of data
the partial structure factors of Fig. 4 are obtained by direct solution giving the large uncertainties and then by
the reverse Monte Carlo method giving the solid plot there.
variation such as a gaussian distribution, leads to the requirement of a Since
the measurement had a q
max
= 36
1
the existence of the splitting is not credible. In addition,
the errors introduced by the limited range of the integration of for neutrons are not
understood when dynamic effects are present for anharmonic motion, as discussed in Sect.
III in relation to Fig. 2.
It should be emphasized that the above discussion on uncertainties introduced in deter-
mining the PPDF by diffuse coherent scattering applies to samples where each component is
a substantial fraction of the total composition. When a component is dilute, as per impurities
or dopants, then it is not possible to obtain any reliable PPDF for that component.
The fact that XAFS can measure the PPDF directly without the signal and the concomi-
tant noise from the rest of the atoms in the sample is a decided advantage, not only because
of the greater sensitivity for measuring more dilute atoms, but also for obtaining objective
estimates of uncertainties. The introduction of uncertain sources of systematic errors is sig-
nificantly less of a problem for XAFS. It is possible to separate random and systematic errors
since the signal and noise comes from only the atoms of interest as opposed to the diffuse
scattering measurements where the major source of noise comes from the other atoms while
the signal comes predominantly from the atoms of interest. This latter situation leads in most
cases to the need to smooth the result by special techniques such as maximum entropy [22] or
221
the reverse Monte Carlo method [21] which introduce uncontrolled uncertainties. For exam-
ple, in XAFS it is possible to define the total information content of the data and to control the
fitting procedure to guarantee that the number of variable parameters is less than the relevant
number of independent points, N
I
, the data contain (the information content), which allows
a meaningful estimate of the errors. Diffuse scattering also has inherently a limited amount
of information due to the limited q-range of the data, as per the Nyquist sampling rate
criterion [23], namely, where N
I
is the maximum number of variable paramters
allowed and is the range of real space being fit. When the smoothing procedure is intro-
duced, what corresponds to N
I
is obscured and how objectively to assess uncertainties is a
problem not generally addressed in the diffuse coherent scattering literature.
VI. SUMMARY AND CONCLUSIONS
Refinement of diffraction data allows the determination of the average periodic structure
of crystals, while diffraction gives no information on materials without long-range order
since no diffraction peaks occur for these materials. However, the PDF and PPDF contain
local structure information on both of these classes of materials. The information from PPDF
or PDF and diffraction are complementary. The local structure of PDF and PPDF give the
actual distances between atoms while diffraction gives the average position on crystal sites.
When local disorder is present, the actual distances between atoms need not agree with that
determined by diffraction. Since the interaction of atoms, neglecting coulomb interactions,
is typically of the order of interatomic distances, the PPDF is most important for accurately
determining this local interaction which is strongly distance-dependent.
Two types of experimental techniques are presently utilized to determine the PDF and
the PPDF. They are diffuse coherent scattering and XAFS. For diffuse scattering, use of neu-
trons instead of x-rays is the technique of choice for determining the PDF of solids containing
light elements, such as the Cu-oxide high temperature superconductors, because of the ability
to better detect the distribution of the light elements. Neutrons have the added advantage of
obtaining data to higher q, resulting in higher spatial resolution of the PDF. However, neutron
measurements do not integrate over a large enough frequency range to be able to freeze the
motion of rapidly moving atoms as required for an accurate determination of the PDF, which
sometimes results in spurious results. Also, pulsed neutron sources, as needed for high spa-
tial resolution, do not allow the clean separation between and as required for obtaining
reliable PDF for some anharmonic atom motions.
XAFS, similar to neutrons, has the advantage of a favorable signal from light atoms
comparable to heavy atoms. This occurs because the particle being scattered in XAFS is the
photoelectron. XAFS also has similar spatial resolution to neutron scattering. It has decided
advantages compared to diffuse coherent scattering in ease and speed of measurement, and in
accuracy and sensitivity of determining the partial PDF, especially for dilute atoms, because
it can measure the partial PDF directly. However, this advantage is typically limited to the
first neighbor atoms in non-crystalline materials and up to about the fourth neighbor atoms in
crystalline materials. Beyond those distances diffuse coherent scattering has the advantage.
Another advantage of measuring the partial PDF directly is that XAFS can estimate the er-
rors objectively as discussed in the previous section. Unfortunately, recent diffuse coherent
scattering papers on the partial PDF do not present methods to estimate objectively the uncer-
tainties in their results, a serious defect. An important future goal should be the development
of such a method for error estimation.
When unit cell dimensions are less than the range of XAFS, combining XAFS results
with diffraction allowed the solving of the actual structure of the crystal, including local
disorder, if present [16,17]. Knowing this structure then allows the calculation of coulomb
interactions in addition to the local interatomic interactions.
222
XAFS contains more information than the PPDF. The near edge fine structure contains
information on the electronic state of the probe atom and, in some cases, geometry of the near
neighbor angular distribution. In the special case of three atoms which are within 20deg of
being collinear, XAFS can measure the three-body correlations corresponding to the
average of where is the angle deviation from collinearity [12,15].
Thus, the PDF, and particularly the PPDF, is an important complement to the average
periodic structure determined by diffraction. When local disorder is present it is important
to determine the PPDF in addition to the average structure, in order to have the complete
information necessary for a correct theory, as illustrated for the case of ferroelectricity. For
distances up to about the first four neighbors, XAFS is usually the premier technique to
determine the PPDF in crystals. For non-crystalline matter, XAFS typically can only obtain
first-neighbor information. Only XAFS can determine the PPDF for dilute atoms while only
diffuse coherent scattering can determine the PDF or PPDF for intermediate distances beyond
the range of XAFS.
ACKNOWLEDGEMENTS
I am grateful to Dr. Daniel Haskel for many discussions, that have helped to crystallize
my thoughts on the topics presented here. Research involved in the writing of this article was
partially supported by DOE grant no. DE-FG03-98ER45681.
REFERENCES
1. Rietveld, H.M., (1969) A profile refinement method for nuclear and magnetic structures. J. Appl. Cryst
2, 65-71; von Dreele, R.B., Jorgensen, J.D. and Windsor, C.J. (1982) Rietveld refinement with spallation
neutron powder diffraction data, J. Appl. Cryst. 15, 581-89.
2. Yacoby, Y. and Stern, E.A., (1996) Structural disorder in crystals undergoing displacive type structural
phase transitions as revealed by XAFS. Comments of Cond. Matter Phys 18, 1-19. Stern, E.A. and
Yacoby, Y. (1996) Structural disorder in perovskite ferroelectric crystals as revealed by XAFS, J. Phys
and Chem. Solids 57, 1449-55.
3. Girshberg, Y. and Yacoby, Y. (1999) Ferroelectric phase transition and off-centre displacements in sys-
tems with strong electron-phonon interaction. J. Phys: Condensed Matter 11, 9807-22.
4. Warren, B.E., (1990) X-ray diffraction, Dover, Mineola, New York, Ch. 10.
5. Kittel, C. (1967)Quantum Theory of Solids, John Wiley & Sons, New York, Ch. 19.
6. Hendrickson, W.A., (1991) Determination of macromolecular structure from anomalous diffraction of
synchrotron radiation. Science 254, 51-58.
7. Sicron, N., Ravel, B., Yacoby, Y., Stern, E.A., Dogan, F. and Rehr, J.J., (1994) Nature of the ferroelectric
phase transition in PbTi03. Phys. Rev. B 50, 13168.
8. Louca, D., Kwei, G.H., Dabrowski, B. and Bukowski, Z. (1999) Lattice effects observed by the isotope-
difference pair density function of the super conductor. Phys. Rev. B 60, 7558-64.
9. Waseda, Y.(1984) Novel application of anomalous x-ray scattering for structural characterization of
disordered materials, Springer-Werlag, New York.
10. Saito, M., Waseda, Y. (2000) Anomalous x-ray scattering for determining the partial structural functions
of binary liquids, J. Synchrotron Rad. 7, 152-159.
11. Placzek, G., (1952)Phys. Rev. 86, 277.
12. Stern, E.A., and Heald, S.M. (1983) Basic principles and applications of EXAFS, in E.E. Koch (ed.),
Handbook of Synchrotron Radiation, North-Holland, New York, vol. 1, ch. 10, pp 955-1014.
13. Ankudinov, A.L., Ravel, B., Rehr, J.J. and Conradson, S.D. (1998) Real-space multiple-scattering calcu-
lation and interpretation of x-ray absorption near-edge structure. Phys. Rev. B. 58, 7565-7576.
14. Stern, E.A., Newville, M., Ravel, B., Yacoby, Y., and Haskel, D. (1995) The UWXAFS analysis
package: philosophy and details, Physica B 208 & 209, 117-120. The UWXAFS is licensed by
the University of Washington and information about obtaining it can be obtained from the web site
http://depts.washington.edu/uwxafs.
223
15. Haskel, D., Stern, E.A., Dogan, F. and Moodenbaugh, A.R. (2000) XAFS study of the low temperature
tetragonal phase of La
2x
Ba
x
Cu 0
4
: disorder, stripes, and T
c
suppression at x=0.125, Phys. Rev. B. 61,
7055-7076.
16. Frenkel, A.I., Stern, E.A., Voronel, A., Qian, M. and Newville, M. (1994) Buckled crystalline structure
of mixed ionic salts, Phys. Rev. Lett. 71, 3485-88; (1994) Solving the structure of disordered mixed salts,
Phys Rev. B49, 11 662.
17. Kelly, S., Ingalls, R., Stern, E.A., Voronel, A. (2000) (unpublished).
18. Fulton, J.L, Hoffman, M.M., and Darab, J.G. (2000) Copper (I) chloride coordination structure under
hydrothermal conditions up to 325, submitted to Chem Physics Letters.
19. Raoux, D., Flank, A.M. (1984) Local Ordering of Metallic Glasses by EXAFS and X-ray Scattering:
Cu-Y alloys, in EXAFS and Near Edge Structure III K.O. Hodgson, B. Hedman and J.E. Penner-Hahn
(eds), Springer-Verlag, pp. 321-26.
20. Di Cicco, A., Taglienti, M., Minicacci, M., and Filipponi, A. (2000) Local structure of liquid and solid
silver halides probed by XAFS, The 11th International Conference on XAFS, to be published.
21. McGreevy, R.L. and Pusztai, L. (1988) Mol. Simul. 1, 359-367.
22. Skilling, J. and Bryan, R.K. (1984) Maximum entropy image reconstruction: general algorithm Mon. R.
Astron. Soc. 211, 111-125.
23. e.g.,Brigham, E.O., (1974) The fast fourier transform Prentice-Hall, NY; Also see Stern, E.A., (1993)
Number of relevent independent points in XAFS data. Phys. Rev. B48, 9825.
224
ZEOLITE INSTABILITY AND COLLAPSE
G.N. GREAVES
Department of Physics, University of Wales,
Aberystwyth, Ceredigion, SY23 3BZ, UK
INTRODUCTION
It is well known that zeolites are metastable materials and that retaining their low-
density friable structures is dependent upon maintaining the majority of the connections
within the expanded alumino-silicate network. However, it is the accessibility of the
massive internal structure of zeolites for heterogeneous chemistry that is responsible for
their huge industrial versatility [1,2,3]. Zeolites and their microporous analogues such as
ALPOs, find diverse application as ion exchange materials in agriculture, water softeners
in detergents, the primary catalysts for cracking heavy oils, hosts for nanoparticle
synthesis, and as the most commonly used molecular sieves and drying agents in the
laboratory. Many of the processes involved in these wide-ranging applications require the
chemistry to be altered locally, for instance through ion exchange, hydroxylation and
calcination, but these modifications can in turn result in silicon and aluminium being
removed from the network. Even bound oxygen in the network can exchange with gas-
phase oxygen molecules. The removal of any of the essential linkages in the network
depletes its rigidity. As-prepared zeolites generally contain molecules within the structure
which must be removed by modest heating (300C - 400C) in order to generate porosity.
Water is the most common species but larger molecules are often also engaged as
templates in synthesising the open network. Porosity is accomplished at modest
temperatures. The removal of intra network molecules, however, puts the mechanical
integrity of the whole atomic scaffolding under threat. The microporous structure can be
destablisied by heating in the vicinity of the glass transition temperature, T
g
, of the
corresponding alumino-silicate glass [2,4]. This can also be accomplished at ambient
temperature by the application of pressure [5]. When sufficient linkages are broken the
low density microporous structure collapses to an amorphous alumino-silicate of higher
density a crystalline-to-amorphous transition.
STRUCTURE OF ZEOLITES
All zeolites share the same general formula: M
x/m
m+
Al
x
Si
2-x
O
4
.nH
2
O, where M
m+
is
the charge compensating cation for tetrahedral AlO
4
+
units in the framework and
AlO
4
+
and SiO
4
tetrahedra are linked through common bridging oxygens (BOs) to form a
completely compensated alumino-silicate network. The different zeolite structures, rather
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 225
like the analogous feldpsars which share the same stoichiometry, depend primarily on the
Si:Al ratio but are also influenced by type of charge compensating cation, M
m+
, present.
This paper concentrates on Na zeolite-A (Na
12
Al
12
Si
12
O
48
) which is one of the most well-
studied zeolites. For Na zeolite-A Si:Al = 1 and the structure is shown in Figure 1. The
length of the cubic unit cell is 24.61 [6,7] and contains eight Na
12
Al
12
Si
12
O
48
units. In
common with other zeolites, the low density results from the stacking together of large
polyhedral units [2]. For zeolite-A these comprise sodalite or cages and double four ring
cubes (D4R). Together cages D4R cubes generate the large 26-hedra or cages which
can be clearly seen in Figure 1. The windows into this open atomic framework are also
shown: 4.2 for cages and 2.2 for cages. Water molecules, whose maximum
molecular dimension is < 2, can easily pass through the channels in zeolite A whereas
methanol (CH
3
OH) molecules, with an effective diameter of ~ 4, are adsorbed only
slowly. Ethanol (C
2
H
5
OH) and glycerol (C
3
O
3
H
6
) are excluded, even though they could be
readily accommodated within the girth of the cage. This is the action of a molecular
sieve. The diameters of the and cages are in fact 11.4 and 6.6 respectively and these
voids are largely responsible for the low density of the structure shown in Figure 1. For Na
zeolite-A this is 1.52 gm.cm
-3
, compared to 2.62 gm.cm
-3
and 2.50gm.cm
-3
-the densities of
nepheline (NaAlSiO
4
) and nepheline glass. When Na zeolite-A amorphises with
temperature or pressure, the density increases by more than 60%.
Figure 1. The open framework structure of zeolite A. The comers of the various polyhedra consist of Si or
Al atoms whilst bridging oxygens lie along the edges linking adjacent tetrahedral SiO
4
or AlO
4
-
units. The
sizes of the channel openings are also shown [2].
AMORPHISATION
Amorphisation is the transformation of a crystalline phase to an amorphous phase
without melting and vitrification. It is a progressive process that results from chemical,
thermal or pressure-induced disruption of the crystalline structure [8,9]. There are
numerous examples ranging from bombardment of minerals with -particles, neutron and
electron beams [10] to continuous grinding [11]. Ion implantation processes used in
fabricating very-large-scale integrated circuits can also result in sufficient radiation damage
so that the surface layers amorphise, which must then be annealed to regenerate electronic
activity [12].
226
Temperature-induced amorphisation of Zeolites
The temperature-induced amorphisation of zeolites is well known
[2, 4,13,14, 15, 16]. Collapse occurs over a wide range of temperatures, starting with
phillipsite (K
2
Ca
1.5
NaAl
6
Si
10
O
32
) which amorphises at ~ 250C, extending to faujusite
(Na
12
Ca
12
Mg
11
Al
58
Si
134
O
384
) whose collapse occurs at ~500C. The most resilient are
zeolite A and ultrastable zeolite Y which amorphise at ~900C and ~1000C respectively
[2]. It is useful to define T
amorph
as the temperature at which half of the zeolite has
amorphised. This is because, as will be shown later, the onset of amorphisation and its
completion are both protracted and difficult to judge. The precise value of T
amorph
, even for
the same zeolite structure, can vary widely and is influenced by a variety of factors.
Figure 2. DSC scans at 10 deg.min
-1
tracing the calcination, amorphisation and devitrification of Na zeolite-
A (a), Zn exchanged zeolite-A (b) and Cd exchanged zeolite-A [4,19]. There are large endotherms between
130C and 220C which relate to the removal of water from the zeolite structures. Calcination is complete by
~400C. Amorphisation can be identified by the first of the sharp exotherms, the remainder relating to
recrystallisation. Note the large differences in the amorphisation temperature, T
amorph
, from 900C in Na
zeolite-A to 825C in Zn exchanged zeolite-A.
227
The rate at which water is removed during calcination can lead to hydrolytic
splitting of the Si-O-Al bonds (dealuminisation) and therefore a weakening of the network.
This effect can be exaggerated with steam treatment which can lower T
amorph
for zeolite-A
to ~350C [15,17]. The type and size of the charge compensating cation, M
m+
will also
influence the temperature of collapse [18]. Figure 2 compares the DSC scans for Na
zeolite-A (a), Zn (b) and Cd (c) exchanged zeolite-A. Not included are broad endotherms
from 130C to 220C which vary with composition and which coincide with where the
major water loss occurs. These are followed by sharp exotherms which mark temperatures
of collapse, the peak approximating to T
amorph
. For Na zeolite-A this occurs at 900C but
collapse is lowered to 825C for Zn exchanged zeolite-A [4]. Zn
2+
clearly destabilises the
zeolite-A framework. T
amorph
is lowered to 849C for Cd exchanged zeolite-A, suggesting
that the destabilisation caused by Cd
2+
is less severe than Zn
2+
[19].
Beyond the narrow amorphisation exotherm in each of the DSC scans in Figure 2
are broad endotherms which are decorated at higher temperatures with exotherms
identifying devitrification processes. The amorphisation and recrystallisation process is
well-illustrated by the in situ X-ray Powder Diffraction patterns illustrated in Figure 3(a)
which follow the course of events for Zn exchanged zeolite-A between 595C and 811C
Figure 3. In situ combined XAFS-XRD experiments following the heat treatment of Zn exchanged
zeolite-A [4]: XRD powder patterns (a) and Fourier transforms of zinc K-edge XAFS (b). The coordination
numbers of oxygen around zinc versus the variance in Zn-O distances analysed from XAFS (c). For a heating
rate of 1.7 deg.min
-1
, the temperature of amorphisation, T
amorph
, for Zn exchanged zeolite-A is 692C.
228
[4]. Amorphisation at 692C can be clearly identified. Whilst amorphised Na zeolite-A
recrystallises totally to the feldspar nepheline (NaAlSiO
4
) [14,15], Zn exchanged zeolite-A
converts to the spinel gahnite (ZnAl
2
O
4
) in a glass matrix [4]. By contrast, Cd exchanged
zeolite-A devitrifies to an anorthite (CdAl
2
Si
2
O
8
) glass ceramic [19]. These differences in
recrystallisation are illustrated in Figure 4 where the X-ray powder diffraction pattern for
the starting zeolite-A at room temperature is contrasted with patterns for Na zeolite-A, Zn
and Cd exchanged zeolite-A after heat treatment at 1100C [19].
Figure 4. XRD powder diffraction profiles for Na zeolite-A and exchanged with Zn and Cd with heat
treatment [19]: zinc exchanged zeolite-A (a) and after treatment at 1200C showing the formation of the
gahnite (ZnAl
2
O
4
) phase (b).
229
Figure 4 (cont). The formation of nepheline (NaAlSiO
4
) from Na zeolite-A at 1100C (c); the formation of
an anorthite phase (CdAl
2
Si
2
O
8
) from Cd exchanged zeolite-A at 1000C (d).
In situ XAFS enables the environment of exchanged cations to be followed from
the starting zeolite, through amorphisation to recrystallisation [4,18,19] as illustrated in
Figure 3(b) and (c). These measurements reveal, for example, how Zn switches from
octahedral coordination in the zeolite to tetrahedral coordination as the collapse
commences [4], identifying the diffusion pathway of zinc from charge compensating sites
in the vicinity of the cage to network sites displacing host tetrahedral cations and hence
230
disrupting the network. The eventual nucleation of ZnAl
2
O
4
(Figure 3(a) and Figure 4(b))
suggests that zinc initially destabilises the zeolite-A network by preferentially removing
aluminium and therefore distorting the Si:Al ratio in the surrounding crystal. In situ Cd
XAFS also reveals cadmium transforming with rising temperature from 6-fold extra-
framework sites to tetrahedral network sites [19]. The subsequent nucleation of an
anorthite composition, however, means that cadmium destabilises the network by removing
silicon and aluminium in equal numbers, leaving the local zeolite Si:Al ratio in tact.
Cadmium therefore should have less effect compared to zinc, which explains the marked
differences in T
amorph
illustrated by the DSC scans in Figure 2.
Figure 5. The decline in the strength of zeolite-A diffraction peaks with risng temperature (a) and increasing
time (b) showing the reduction in the crystalline fraction through the process of collapse for Zn and Cd
exchanged material. The solid curves show the result of fitting to the Avrami-Erofeev expression (Eq. (1)).
231
The fact that zeolites are cheaply made in terms of energy, collapse and
recrystallisation has been exploited in devising energy-saving processes for synthesising
ceramics [20,21,22]. Devitrification from these amorphous alumino-silicate sources avoids
the much higher temperatures required for fabricating ceramics by annealing quenched
glasses above the glass transition temperature, T
g
, or by sintering crystalline powders [23].
As we have seen the capacity for altering the composition of the starting zeolite by ion
exchange, can facilitate the formation of specific ceramics. Refractory dielectrics of
commercial importance include anorthite itself (CaAl
2
Si
2
O
8
) [21] a refractory dielectric
for electronics applications and cordierite (Mg
2
Al
4
Si
5
O
18
) [22] the preferred support for
exhaust catalysts.
SOLID STATE ASPECTS OF COLLAPSE
The zeolite crystalline fraction, x, given in Figure 5(a) as a function of temperature,
can also be plotted as a function of time (Figure 5(b)). Although the in situ XRD
measurements are non-isothermal, the temperature rise through which the major collapse
occurs is <5% of T
amorph
. The overall shape of the decline of the crystalline phase and the
growth of the amorphised phase (1-x) with time is characteristic of solid state nucleation
and growth processes [24]. These commence with an induction or nucleation phase
followed by an acceleration in growth. An inflection point at x~0.5 heralds a deceleration
in growth and the process terminates on completion at x=l. In this well-developed area of
solid state chemistry there are proven rate equations for each of the above regions of which
the Avrami relation
(1)
models the sigmoidal region. In Eq. (1) n equals the number of steps involved in nucleus
formation plus the dimension of growing nucleating sites. Typically n lies between 2 and
4. The solid curves in Figure 5(b) are the result of fitting XRD amorphisation data for Zn
and Cd exchanged zeolite-A to eq.(l). The values of n for Zn (3.4) and Cd (3.2)
exchanged zeolite-A are consistent with approximately three dimensional growth following
an initial nucleating period. We can equate k
-1
with the characteristic
amorphisation time. is approximately the time for half the zeolite to amorphise.
This is shorter for Zn (2635s) than for Cd (3361s) and reflects the greater efficiency of Zn
in disrupting the zeolite-A network, implicit in XRD and XAFS experiments.
In terms of solid state nucleation and growth kinetics we might expect to
decrease as the collapse temperature, T
amorph
, increases, possibly in an Arrhenius fashion.
For a given composition T
amorph
is indeed strongly affected by the thermal conditions. This
can be clearly seen by comparing the rates of collapse for Zn and Cd exchanged zeolite-A
in Figure 2 with those in Figure 5(a). With the heating rate of ~ 1.5 deg.min
-1
used for in
situ XRD, collapse is lowered to 692C for Zn and 772C for exchanged zeolite-A,
compared to 825C and 849C respectively for the DSC heating rate of 10 deg.min
-1
(Figure 2(b) and (c)). For Na zeolite-A Thomas, Lutz and others have shown that T
amorph
can be reduced to 700C after isothermal heating for 80 hours [16]. These values of T
amorph
should be compared with a collapse temperature of 900C when Na zeolite-A is heated at
10 deg.min
-1
(Figure 2(a)).
As T
amorph
decreases with heating rate the collapse time, rises sharply. This
can extend to many hours when T
amorph
<T
g
[16], reducing to a few minutes when T
amorph
>T
g
(Figure 2). These different aspects of the dynamics of temperature-induced amorphisation
are drwan together in Figure 5, which takes results for and T
amorph
for Na zeolite-A
and Cd and Zn exchanged zeolite from the literature and from unpublished data, plotting
232
them as log
10
versus 1/T. The differentials in the values between the three
zeolite-A compositions are clearly maintained over a wide temperature range. The
approximately Arrhenious behaviour yields an activation energy for collapse of ~2.5eV, a
value intermediate between the magnitude expected for bond breaking processes compared
to diffusion processes.
Figure 6. Comparison of amorphisation time, and the temperature of collapse, T
amorph
, for Na zeolite-
A [13,15,4], Cd exchanged zeolite-A [Greaves and Sankar, unpublished results] and Zn exchanged zeolite-
A [4]. Error bars indicate the extent of non-isothermal conditions. Log
10
versus T
amorph
-1
plots yield
activation energies of ~2.5eV.
AMORPHISATION AND MELTING
The huge depressions in the temperature of zeolite collapse, T
amorph
, as the heating
rate is reduced, however, are reminiscent of the behaviour of the glass transition
temperature, T
g
, with thermal history. In this context it is worth noting that T
g
(at 10
12
Pa.s) for nepheline glass (NaAlSiO
4
) is 804C [25], approximately 100 deg less than the
T
amorph
of Na zeolite-A (Na
12
Al
12
Si
12
O
48
) measured with DSC at 10 deg.min
-1
. Conversely
the viscosity of molten nepheline at T
amorph
= 900C is ~ 10
9
Pa.s, representative of a
viscous, albeit fragile fluid. The dependence of for different compositions of
zeolite-A with temperature is contrasted in Figure 6 with Toplis careful measurements of
the viscosity, of nepheline melts [25]. It is tempting to consider the collapsing zeolite as
a conventional fluid and to equate the viscosity of nepheline glass, with the viscosity of
the collapsing zeolite during amorphisation. In this case the rigidity or shear modulus, G
s
,
can be related to the relaxation time, by Maxwells relation:
(2)
Taking parameters from Figure 6 for the different zeolite-A compositions where
T
amorph
>T
g
, we obtain G
s
increasing from ~2MPa for Na zeolite-A to ~500MPa for Zn and
Cd exchanged zeolite-A, which together with the viscosity values are indeed consistent
with the rheological properties of viscous fluids.
233
Figure 7. Scanning electron micrographs of cubic zeolite-A crystals at room temperature (top) and after
amorphising at 750C for 2 hours (bottom) [19].
The allusion of amorphisation to melting can be clearly seen in electron
micrographs of zeolite-A before and after amorphisation illustrated in Figure 7. The cubic
morphology of the initial zeolite prior to heat treatment is transformed to the globular
shapes following annealing at 750C, well below T
g
but for 2 hours [19]. For isotropic
solids G
s
becomes identical with c
44
, the shear elastic constant. However, the much
stronger temperature dependence of compared to means that for conditions where
T
amorph
>T
g
, Eq.(2) leads to unrealistically high values of C44. Accordingly, whilst
amorphisation of zeolites is clearly a Theological phenomenon, it is complicated by the
very large difference in density between the zeolitic and the amorphised phases of >60%
234
which results in a massive negative expansion coefficient in the vicinity of T
amorph
. This is
in contrast to the viscosity properties of a conventional glass where the density changes
only modestly and continuously through T
g
.
In common with melting, though, amorphisation can be viewed as a heterogeneous
process promoted by instabilities which start at surface nucleating sites and advance
inwards [8]. The micrographs shown in Figure 7 are a graphic illustration of this. In
particular, even the initial stages of amorphisation in zeolites are expected to have serious
consequences for the porosity of the remaining crystalline phase. Indeed Thomas and co-
workers detected the onset of amorphisation in Na zeolite-A by a reduction in water
absorption [13] at temperatures as low as 550C. With successive heating this falls rapidly
to zero by 670C. However, subsequently grinding the partially amorphised material
results in much of the original porosity returning. It would appear that from the start the
amorphised phase forms a film around the intact crystal, as is clearly the case when
amorphisation is complete (Figure 7). Whilst this amorphised shell blocks the sorption
capacity, Thomas et al report that this can be recovered if the shell is then broken by
grinding. A further increase in porosity can also be achieved with subsequent sintering,
suggesting an additional closed macoporosity. These interesting results are reproduced
in Figure 8, where new product phases include devitrified as well as amorphised
material.
Figure 8, Changing macroscopic sample composition during heat treatment of Na zeolite-A [13].
Proportions are determined pro rata from water absorption capacity for specimens heated for 2 hours at each
of the above temperatures and subsequently ground and sintered. See text and ref. 13 for details.
COMPRESSIBILITY
Compressibility, K, is the inverse of the bulk modulus, B, and is defined to be
(3)
Inorganic solids respond to the application of pressure by the contraction of interatomic
distances, by changing the configuration of cation polyhedra and by tilting polyhedral
linkages. The latter two processes coincide with first order phase transitions, whereas
polyhedral compression can often be accommodated without a change of symmetry.
235
Indeed the compressibility of polyhedral has been shown to be independent of structure
type [26], with
(4)
where <R
ca
> is the mean action-anion distance and z
c
and z
a
are the cation and anion
charges, respectively. This is essentially a relationship between compressibility, K, and
volume and is different for different families of solids. By using scaling factors
S for each anion system (S=0.5 for silicates, for instance), polyhedral compressibilities for
a huge range of mineral families can be projected onto the common universal relationship
(eq.(4)) plotted in Figure 9.
Figure 9. Polyhedral Compressibility (K
p
-1
) - Volume relationship for 4-fold 6-fold (x) and 8-fold (o)
sites in a wide range of minerals [26]. The scaling factor S=0.5 for silicates, 0.75 for halides, 0.4 for
chalcogenides, 0.25 for pnictides and 0.2 for carbides. Note: 1Mbar
-1
= 0.0lGPa
-1
.
Compressibility of zeolites
At ambient temperature the application of external pressure to zeolites can lead to
new phase transitions with high compressibility [27,28] and eventually to collapse [5].
The changes in cubic phase are accompanied by the discontinuities in unit cell volume
illustrated in Figure 10 [28] but also by dramatic decreases in diffraction intensity
indicative of amorphisation. For example, in this classic series of high-pressure
experiments on zeolite A, Hazan and Finger demonstrated that compressibility was
critically dependent on the molecular size of the hydrostatic pressure medium [27]. If
water was used, they found that the cubic structure of zeolite A was retained up to 4 GPa
with a compressibility of 0.007 GPa
-1
, comparable to that of -quartz found in the bottom
corner of Figure 9. Water molecules are sufficiently small and rigid to fill the zeolite
cages. However, if an alcohol mixture was used instead, penetration was not optimum. A
series of cubic transitions occured, starting around 2GPa, and the zeolite exhibited
compressibilities in the range 0.014 0.028 GPa
-1
a third of the way up Figure 9. By
comparsion, 0.03GPa
-1
is the compressibility of silica and silicate glasses. Finally when
236
glycerol, which is impervious to zeolite A, was employed as the hydrostatic fluid much
damage was done to the starting crystal and the compressibility reached 0.046 GPa
-1
half
way up Figure 9 and in the zone of open-framework silicates and octahedral oxides.
Accordingly as the sizes of the hydrostatic liquid molecules increase they find it less easy
to enter the zeolite micropores through the channel openings (Figure 1), internal support
for the open network reduces and, importantly, amorphisation becomes less reversible.
Figure 10. Compression of Na zeolite-A in different hydrostatic fluids [27]. The unit cell volume, V, is
plotted as a function of pressure and the compressibility, K, given by eq. (3). K increases for water
(0.007GPa
-1
), which is completely absorbed by the microporous structure, to glycerol (0.046GPa
-1
) which is
virtually impervious.
PRESSURE-INDUCED AMORPHISATION
Amorphisation induced by pressure pressure melting was first discovered by
Mishma in 1984 when ice I was observed to vitrify at 77K and 1 GPa [29]. By 1990 the
room temperature amorphisation of -quartz at 25 GPa [30] and of berlinite (A1PO
4
) at 15
GPa [31] had been reported. Since then the number of examples has grown and reviews of
this new field can be found in refs. 8 and 9. There is a broad relationship between the
pressure of amorphisation, P
amorph
, and the density of the crystalline precursor at ambient
pressure. Figure 11 illustrates this for a range of minerals [8] where P
amorph
can be seen to
scale from ~5 GPa for the natural zeolite, scolecite (Ca
8
Al
16
Si
24
O
80
.24H
2
O), to ~55 GPa
for the dense perovskite, forsterite (Mg
2
SiO4). The upshot of this empirical relationship is
that amorphisation demands the exisitence of higher density more stable phases.
Conversely a dense and stable ambient phase kinetically hinders the route to
amorphisation. It will be already be clear that, with their intrinsically low densities and
with their manifold of higher density crystalline and amorphous phases, zeolites, like
scolcerite, fall in the bottom corner of Figure 11. They are clearly metastable structures.
The gradual destruction of the periodic structure during the amorphisation of dense
materials like ice and quartz is presumed to be initiated by mechanical instabilities
237
[33,34,35,36,37]. For microporous materials these can be identified with the chemical and
thermal instabilities described above in connection with temperature-induced collapse.
Figure 11. The range of pressure-induced amorphisation in silicates as a function of their ambient densities
(SCO:Ca
8
Al
16
Si
24
O
80
.24H
2
O; SP:Mg
3
Si
2
O
5
(OH)
4
; QZ:SiO
2
; AN:CaAl
2
Si
2
O
8
; MS: KAl
2
(Si
3
Al)O
10
(OH,F)
2
;
WO: CaSiO
3
; FO: Mg
2
SiO
4
; EN: Mg
2
Si
2
O
6
) [8].
Reversibility of Amorphisation
Unlike temperature-induced amorphisation, when amorphisation takes place under
pressure at ambient temperature it may be reversible. The crystalline-to-amorphous
transition, however, is only one-way for ice I and for -quartz [29,30]. Indeed non-
hydrostatic processes during compression are believed to play an important role in
engendering this irreversibility, -quartz being particularly studied in this respect
[38,39,40]. Never the less, routes back to the starting crystalline state have been developed
for some systems. In the case of ion-beam damaged silicon, for example, thermal epitaxial
processes have been perfected to reactivate implanted dopants in vlsics [12]. Reversibility
of amorphisation has also been reported for A1PO4 [31], giving rise to the notion of a
Memory Glass. Molecular Dynamics calculations of AlPO
4
point to the retention of
rigid PO
4
tetrahedra within the pressurised amorphous alumino-phosphate structure which
then provide a key for reestablishing the crystalline phase when the pressure is released
[32,41]. For zeolites [27,28] and other low density materials [42] it is also claimed that
amorphisation can be reversed or even prevented. In these cases the pre-requisite for
recovering or retaining the starting crystalline phase is that the microporous structure is
internally supported by non-deformable molecular units. Figure 12 illustrates the benefits
of this for a model of the microporous clathrasil dodecasil-3C or D3C [42]. In the
calculations the open silicate structure is filled with spherical guests to mimic non-
deformable molecules (a). As the pressure is increased amorphisation is fully established
by 10 GPa (b). Further compression to 15 GPa densifies the amorphous structure (c), but,
as the pressure is gradually released, the original crystalline structure is recovered.
Irreversible Amorphisation in Zeolites
If the hydrostatic fluid in the pressure cell cannot penetrate the microporous
structure of the zeolite, irreversible amorphorphisation will occur. The diffraction
experiments relating to the location of scolecite (Ca
8
Al
16
Si
24
O
80
.24H
2
O) on the P
amorph
versus
Density relationship (Figure 11) are reproduced in Figure 13. These are high pressure
238
energy dispersive XRD measurements and used non-penetrating solid KBr as the pressure
transimitting medium. Irreversible amorphisation takes place at 6.9 GPa and this is corob-
Figure 12. Molecular Dynamics simulation of reversible amorphisation in the clathrasil D3C [42]. The
microporous silicate structure is supported by 8 spherical guests replicating internally located molecules:
ambient conditions (a); 10 GPa (b); 15 GPa (c); and after gradual depressurisation to 5 GPa.
orated by in situ Raman measurements, where spectra for amorphised material bear strong
similarities to spectra for aluminosilicate glasses [5]. The amorphous phase of scolecite is
retained when the pressure is removed. Similar behaviour is reported for mesolite
(Na
16
Ca
16
Al
48
Si
72
O
240
.64H
2
O), another of the naturalite series.
The bars in Figure 11 for scolecite and each of the other minerals recorded depict
the range of P
amorph
observed, when amorphisation takes place. Typically
values are around 30% of the amorphisation pressure, P
amorph
reaching 100% for
the microporous scolecite. is attributed to pressure inhomogeneities that develop
within the pressure cell as the crystalline-amorphous transition takes place [5,8]. These are
also a problem during decompression and are very likely to be a major cause of the
hysterisis reported in reversible systems like berlinite [32].
239
Figure 13. Energy dispersive diffraction patterns for scolecite at the pressures shown [5]. Amorphisation of
this natural zeolite is irreversible in the non-penetrating hydrostatic medium, solid KBr.
Being lower in desnity than scolecite, zeolite-A should collapse at even lower
pressures than 5-10 GPa. Preliminary results for Zn exchanged zeolite-A are plotted in
Figure 14 [43]. Unlike Gillet and co-workers experiments [5], these XRD patterns are of
higher resolution and were obtained in angle-dispersed geometry using image plate
detection. The reduction in the intensity of the diffraction lines with increasing pressure
indicates the onset of amorphisation. This is matched by the increase in the diffuse
background resulting form the accumulating amorphous phase. By 4.8 GPa the Zn
exchanged zeolite-A is fully collapsed and remains so on decompresion to ambient
pressure. It is clear from Figure 15 that the diffraction lines shift to higher angles between
ambient and 2 GPa, indicative of crystalline compressibility. Indeed K ~ 0.07 GPa
-1
for Zn
exchanged zeolite-A pressurised in silicone fluid, similar to the value reported by Hazan
and Finger for Na zeolite-A pressurised in glycerol [28].
The decline in XRD diffraction intensity for pressure-induced amorphisation is
comaparable in profile to that for temperature-induced amorphisation with time (Figure
5(b)). This points to a similar sequence under compression, viz: nucleation, accelerated
growth, an inflection point, decelerated growth and termination when all of the zeolite has
collapsed. Likewise an Avramiesque expression can be employed to fit the process of
pressure-induced amorphisation. Defining P
amorph
by analogy with T
amorph
, as the pressure
when half the crystal is amorphised (x=0.5), P
amorph
can be easily interpolted from the
experimental points. A value of 2.5 GPa is obtained for Zn exchanged zeolite-A and,
interestingly, a higher value of 3.9 GPa for Cd exchanged zeolite-A [43]. These
differences in P
amorph
mirror differences in T
amorph
discussed earlier, suggesting that the
different instabilities created by the two charge compensating cations at high temepratures
may well be present as different mechanical instabilities under compression at ambient
temperature. The massive difference in density between the starting crystal and
amorphous feldspar mean that pressure-induced amorphisation is accompanied by a huge
rise in the compressibility of the combined system from the crystalline value as P
amorph
is
approached. This rises from the unsupported zeolite-A value of ~0.05 GPa
-1
and returns
to ~0.02 GPa
-1
, typical of a aluminosilicate glass [43].
240
From the picture of solid-state melting developed earlier for temperature-induced
amorphisation, the effect of pressure is expected to impinge at the surface of zeolite
crystals and work inwards, as the electron micrographs in Figure 7 suggest for
amorphisation with temperature. Accordingly, considerable pressure inhomogeneities are
expected from the huge differences in density between the crystalline and amorphous state.
In particular, if the amorphised shell is rigid it will serve to reduce the pressure in the
crystalline core. The initial positive compressibility, evident from the XRD patterns in
Figure 14, switches sign as the pressure advances [43], which serves to explain the pressure
imhomogeneities observed by Gillet and others [5,8]. The large uncertainties,
evident in Figure 11 [8] will of course be compounded if P
amorph
is associted with complete
amorphisation, the sigmoidal tail-off as the growth of the amorphous phase is completed is
difficult to pin-point.
Figure 14. Image plate XRD patterns for Zn exchanged zeolite-A at the pressures shown [Greaves, Sapelkin
and Sankar, unpublished results]. The hydrostatic fluid silicone fluid does not enter the microporous structure
so amorphisation is irreversible.
Finally it is worth looking again at the reversible compressibility results for Na
zeolite-A (Figure 10) [28], where discontinuities in crystalline compressibility occur at
pressures in the range 1 to 3 GPa, pressures over which Zn and Cd exchanged zeolite-A are
collapsing [43]. Hazen reports considerable time variations in the high pressure
crystallography of up to 100h in traversing between the different phases. These could well
241
reflect the gradual removal of pressure inhomogeneities between the external amorphised
shell and the internal crystal.
MECHANICAL RIGIDITY AND ZEOLITE COLLAPSE
Zeolite collapse exhibits all the hallmarks of the ridity percolation threshold [44].
This coincides with the elastic constant c\\ of a solid falling to zero and with an abrupt rise
its compressibility, K. As we have seen, when zeolites collapse by temperature-induced
amorphisation, c
44
for the crystalline-amorphous composite falls to fluid-like values.
When collapse occurs under pressure at ambient temperature, K rises sharply with
pressure. Since c
11
= 1/K + 4c
44
/3 for isotropic materials, then, when K approaches infinity
and c
44
zero , c
11
must also fall to zero.
In the context of an atomic framework, a material becomes a fluid when the number
of constraints per atom, n
c
, falls below the number of degrees of freedom that it enjoys, n
d
.
In a solid n
c
generally exceeds n
d
and the rigidity percolation threshold occurs when the n
c
= n
d
[45]. When n
c
< n
d
, zero frequency or floppy modes are present, the fraction being
given by (n
d
- n
c
). As n
c
can be re-expressed in terms of the average coordination number
<r>, the fraction of zero frequency modes, f, can be written [44]
(5)
where the average coordination number is defined to be
(6)
At the rigidity percolation threshold f=0, so the average coordination number
(7)
For the 3D aluminosilicates we are considering zeolites, fully amorphised zeolites
or recrystallised feldspars like nepheline etc. the average coordination number <r> is
equal to that of -quartz, which equals 8/3 [45]. However, for networks which contain
large numbers of 2-fold coordinated anions (oxygen and the chalcogenides, S, Se, Te)
and/or cations whose coordination number is less than 4 (like the pnictides, P, As, Sb), <r>
can fall to 2.4 or less. For incomplete networks which include terminal bonds or singly
coordinated atoms, the average coordination of the remaining atoms in the network can
also reduce to 2.4 or less. Hydrogenated amorphous semiconductors like a-Si:H or a-C:H
are a case in point. Here, covalent linkages are broken by singly coordinated hydrogens.
In this event Thorpe has introduced the concept of the skeleton network, where only
complete linkages are included in counting constaints [44]. By ignoring non-bridging
linkages, this approach enables the above criterion for defining the rigidity percolation
threshold ((eq. (5,7)) to be extended to incomplete networks.
Applying the skeleton approach to an incomplete silicate or aluminosilicate
network, i.e. one that incorporates terminal or non-bridging oxygens, we get from eq.(6)
where x
4
= fraction of Si/Al sites, x
2
= fraction of bridging oxygens and x
1
= fraction of
non-brdging or terminal oxygens. Recognising that x
4
+ x
2
+ x
1
= 1, the average
coordiantion number for network atoms is given by
(8)
242
In most minerals and stable aluminosilicate glasses, oxygen bridges are complete and x
1
=
0 and <r> = 2.67 which is also the value for -quartz. (When bon-bending constraints are
removed, as has been proposed for silica glass, this will fall to 2.4 [46].) For modified
crystalline silicates, however, where <r> < 8/3. In particular for disilicates, of
which -Na
2
Si
2
O
5
is an example, x
4
= 2/7, x
4
= 3/7 and x
1
= 2/7. From eq.(8) <r> = 2.4,
which marks the rigidity percolation threshold (eq.(7). For these compositions, there is one
non-bridging oxygen per tetrahedral site on average, i.e. all silicons occupy Q
3
configurations. Crystalline disilicates are 2D layered structures and the melting points of
disilicates are massively reduced compared to molten quartz or indeed silica [46], i.e. n
c
<3.
As oxygen linkages are gradually removed from a zeolite structure by nucleation,
thermal or mechanical instabilities, we expect that the 3D rigidity of the starting structure
will be irreversibly lost when the equivalent proportions for a disilicate structure
composition are achieved locally. Taking the collapse of Na zeolite-A as an example, for
some intermediate stage of amorphisation we can write:
(9)
where (1 x) is the fraction of low density zeolite that has collapsed to a high density
amorphous nepheline. During the nucleation process local oxygens from each component
will be shared. It can be shown that the ruptured zeolite reaches the point where each
tetrahedral site has a Q
3
configuration when x ~ 0.2. For temperature-induced
amorphisation, it is clear from Figure 5 that amorphisation accelerates around this point.
The same is true for pressure-induced amorphisation [43]. It is also clear that
amorphisation induced either way decelerates when x ~ 0.8.
There are, then, two sides to the amorphisation process - collapse of the low
density crystal and re-assembly of the higher density amorphous phase. In between, the
average coordination number <r> will be less than 2.4, n
c
will fall below n
d
and zero
frequency modes will emerge. This provides a natural explaination of the sharp exotherms
that mark amorphisation (Figure 2) and which are so conter-intuitive with respect to
conventional melting, which is strongly endothermic. In particular, if is the
enthalpy of zeolite collapse, we can equate the fraction of floppy modes created, f, with
H
collapse
/3RT
amorph
. From eq.(5) the minimum value of average coordination number,
<r>
min
, will approximately be given by
(10)
If <r> falls to 2.2, H
collpase
/3RT
amorph
= 0.17 and taking T
amorph
= 900C gives
kJ.mol
-1
. This value is close to the enthalpies associated with the exotherms shown in
Figure 2 for Na, Zn and Cd zeolite-A. Each is followed by a broad endotherm which is
expected as f returns to zero and <r> rises back to 2.4, eventually reaching 2.67 as the fully
connected alumino-silicate is formed. Interpretting this endothermic re-assembly process
is complicated, however, by the entropic nature of the crystalline-to-amorphous transition
and by the exothermic devitrification processes that are beginning to start.
EXPERIMENTAL POSTSCRIPT
Many of the results described in this review have come about very much in the
wake of the development of new light sources and new experimental techniques. The
brilliance of synchrotron radiation enables diamond anvil cells and multi-anvil presses to
be penetrated so that X-ray diffraction can be measured routinely at high pressures and also
at high temperatures [8,9,47]. High temperature X-ray measurements have necessitated
the development of furnaces with sufficient window area for angle dispersed XRD and also
for accessing other X-ray probes [48]. With tunability XAFS measurements can be made
243
at the high temperatures encountered in temperature-induced amorphisation and these can
be combined with XRD [4,18,22,49]. As we have seen, XAFS enables dopant
environements in zeolites to be followed through the course of collapse and
recrystallisation, the crystallography being monitored with XRD. In situ Small Angle X-
ray Scattering (SAXS) combined with XRD [50] facilitates the observation of the
microstructural changes that accompany zeolite amorphisation [51,52]. Because
amorphisation is a complex phenomenon and often embroiled with the uncertainties of
hysteresis, multiple experimental probes offer the most consistent approach. Laboratory
techniques like Raman and IR [8] are especially complementary to X-ray technqiues as is
ionic conductivity [53]. Bringing some or all these together is technically demanding but
the returns would be immense.
ACKNOWLEDGEMENTS
Thanks are particularly extended to G. Sankar and A. Sapelkin for their skill in
applying combined X-ray techniques and high pressure techniques to the study of zeolite
collapse. L. Colyer and C. Altru are also acknowledged for their willingness to try new
things and to analyse the consequences.
REFERENCES
1. Barrer, R.M. (1978) Zeolites and Clay Minerals as Sorbants and Moelcular Sieves, Academic Press,
New York
2. Dyer, A. (1988) An Introduction to Zeolite Molecular Sieves, John Wiley & Sons, New Work
3. Cusumano, J.A. (1992) New technology and the environment, Chemtech, 22, 482-489
4. Colyer, L.M., Greaves, G.N., Carr, S.W. and Fox, K.K. (1997) Collpase and recrystallisation processes
in zinc-exchanged zeolite-A: A combined x-ray diffraction, XAFS, and NMR study, J. Phys. Chem.,
101, 10105-10114
5. Gillet, P., Malzieux, J-P and Iti, J-P (1996) Phase changes and amorphisation of zeolites at high
pressures: the case of scolecite and mesolite, Am. Mineral, 81, 651-657
6. Smith, J.V. and Dowell, L.G. (1968), Zeit. fr Kristall , 126, 135
7. Gramlich, V. and Meier, W.M. (1971), Zeit fr Kristall., 133, 134
8. Richet, P. and Gillet, P. (1997) Pressure-induced amorphisation of minerals: a review, Eur. J. Mineral.,
9, 907-933
9. Sharma, S.M. and Sikka, S.K. (1996) Pressure-induced amorphisation of materials, Progr. Materials
Sci., 40, 1-77
10. Schwartz, R.B. and Johnson, W.L. (1988) Remarks on solid state amorphising transformations, J. Less
Common Metals, 140, 1-6
11. Lin, I.J., Navid, S., Grodzian, D.J.M. (1975) Changes in the state of solids and mechano-chemical
reactions in prolonged comminution processes, Miner. Sci. Eng., 7, 313-336
12. Mller, G., (1998) Amorphisation processes in silicon, Current Opinion in Solid State & Materials
Science, 3, 364-370
13. Thomas, J.L., Mange, M., and Eyaud, C. (1971) in Molecular Sieve Zeolites-I, Advances in Chemistry
Series 101, Americal Chemical Society, Washington D.C., pp. 443.
14. Scmitz, W., Siegel, H. and Schllner (1981) , Cryst. Res. Technol., 16, 385
15. Lutz, W., Engelhardt, H., Frichner-Schmittler, H., Peuker, Ch., Lffler, E. and Siegel, H. (1985) The
influence of water steam on the direct phase transformation of zeolite NaA to nepheline by thermal-
treatment, Cryst. Res. Technol., 20, 1217-1223
16. Lutz, W., Frichner-Schmittler, Richter-Mendau, J., Becker, G., and Blow, M. (1986) Formation of a
special intermediate during phase-transformation of zeolite NaA to nepheline, Cryst. Res. Technol., 21,
1339-1344
17. Lutz, W., Fahlke, B., Lohse, U. and Seidel, R. (1983) Investigation of the hydrotheraml stabilities of
NaA, NaCaA and NaMgA zeolites, Chem. Technol., 35, 250-253
18. Colyer, L.M., Greaves, G.N., Dent, A.J., Fox, K.K., Carr, S.W. and Jones, R.H. (1995) In situ study of
ceramic formation from Co
2+
and Zn
2+
exchanged zeolite-A using combined XRD/XAFS techniques,
Nucl. Instr. and Meth. in Phys. Res. B, 97, 107-110
19. Colyer, L.M (1996) Recrystallisation processes in transition metal exchanged zeolite-A, PhD Thesis,
University of Keele.
244
20. Corbin, D.R., Parise, J.B., Chowdhry, U. and Subramanian, M.A. (1991), Mater. Res. Soc. Symp.
Proc., 233, 213
21. Subramanian, M.A., Corbin, D.R. and Chowdhry, U. (1993), Bull. Mater. Sci., 16, 665
22. Sankar, G., Wright, P.A., Natarajan, S., Thomas, J.M., Greaves, G.N., Dent, A.J., Dobson, B.R.,
Ramsdale, C.A. and Jones, R.H. (1993) QuEXAFS-XRD: A new technique in high temperature
materials chemistry. An illustrative in situ study of the zinc oxide enhanced solid-state production of
cordierite from a precursor zeolite, J. Phys. Chem., 97, 9550-9554
23. MacMillan, P.M. (1979) Glass-ceramics, Academic Press, London
24. Barnford, H., and Tipper, C.F.H. (1980) Theory of Solid State Reaction Kinetics in Comprehensive
Chemical Kinetics, Vol 22, Academic Press, New York.
25. Toplis, M.J., Dingwell, D.B, Hess, K-U. and Lenci, T. (1997) Viscosity, fragility, and configurational
entropy of melts along the join SiO
2
-NaAlSiO
4
, Am. Mineral.82, 979-990
26. Hazen, R.M. and Finger, L.W. (1979), J. Geophys. Res., 84, 6723
27. Hazan, R.M. (1983) Zeolite molecular sieve 4A: anomalous compressibility and volume discontinuities
at high pressure, Science, 219, 1065-1067
28. Hazan, R.M. and Finger L.W. (1984) Compressibility of zeolite 4A is dependent on the molecular size
of the hydrostatic pressure medium, J. Appl. Phys., 56, 1838-1840
29. Mishma, O., Calvert, L.D. and Whalley, E. (1984) Melting iceI at 77K and 10kbar: a new method of
making amorphous solids, Nature, 310, 393-395
30. Hemley, R.J., Jephcote, A.P., Mao, H.K., Ming, L.C. and Manghnani, M.H. (1988) Pressure-induced
amorphisation of crystalline silica, Nature, 334, 52-54
31. Kruger, M.B. and Jeanioz, R. (1990) Memory glass: an amorphous material formed from A1PO
4
,
Science, 249, 647-649
32. Gillet, P., Badro, J., Varel, B. and MacMillan, P.F. (1995), High-pressure behaviour of A1PO
4
:
Amorphisation and the memory glass effect revisited, Phys. Rev. B, 51, 11262-11269
33. Handa, Y.P., Tse, J.S., Klug, D.D and Walley, E. (1991) Pressure-induced phase transitions in
Clathrate Hydrates, J. Chem. Phys., 94, 623-627
34. Williams, Q. and Jeanloz, R. (1989) Static amorphisation of anorthite at 300K and comparison with a
diaplectic glass, Nature, 338, 413-415
35. Tse, J.S. and Klug, D.D. (1991), Phys. Rev. Lett. Mechanical instability of alpha-quartz a Molecular
Dynamics Study, 67, 3559-3562
36. Tse, J.S. (1992), J. Chem. Phys., 96, 5482-5487
37. Binggeli, N., Keskar, N.R. and Chelikowsky (1994) Pressure-induced amorphisation, elastic instability,
and soft modes in alpha-quartz, Phys. Rev. B, 49, 3075-3081
38. Kingma, K.J., Hemley, R.J., Mao, H.K., Veblen, D.R. (1993) New high-pressure transformation in -
quartz, Phys. Rev. Lett., 70, 3927-3930 and 72, 1301-1302
39. Watson, G.W., and Parker, S.C. (1995) Quartz amorphisation A dynamic instability, Phil. Mag. Lett.,
71, 59-64
40. Badro, J., Barrat, J-L and Gillet, P. (1996) Numerical simulation of -quartz under nonhydrostatic
compression: Memory glass and five-coordinated crystalline phases, Phys. Rev. Lett., 76, 772-775
41. Tse, J.S. and Klug, D.D. (1992) Structural memory in pressure-amorphised A1PO
4
, Science, 255, 1559-
1561
42. Tse, J.S., Klug, D.D., Ripmeester, J.A., Desgreniers, S. and Lagarec, K. (1994) The role of non-
deformable units in pressure-induced amorphization of clathrasils, Nature, 369, 724-726
43. Greaves, G.N., Sapelkin, A. and Sankar, G., unpublished results
44. Thorpe, M.F. (1995) Bulk and surface floppy modes, J. Non-Cryst. Solids, 182, 135-142
45. Phillips, J.C. (1979) Topology of covalent non-crystalline solids I: short range order in chalcogenide
alloys, J. Non-Cryst. Solids, 34, 153-181; Phillips, J.C. (1979) Topology of covalent non-crystalline
solids II: medium range order in chalcogenide alloys and a-Si(Ge), J. Non-Cryst. Solids, 43, 37-77
46 Zhang, M. and Boochland, P. (1994) The Central Role of BRoken Bond-Bending Constraints in
Promoting Glass Formation in the Oxides, Science, 266, 1355-1357
47. Parise, J.B., Weidner, D.J., Chen, J., Liebermann, R.C. and Chen, G. (1998) In situ studies of the
properties of materials under high-pressure and temperature conditions using multi-anvil apparatus and
synchrotron X-rays, Annu. Rev. Mater. Sci., 28, 349-374
48. Dent A.J., Dobson B.R., Greaves G.N., Sankar G., Roberts M., Catlow C.R.A., Thomas J.M. and
Rayment T.A. (1995) New Furnace Design for Use in Combined X-ray Absorption and Diffraction for
Ceramics and Catalyst Studies: A Study of the Formation of the CuCoMn(CO
3
)
3
Catalyst for CO
Oxidation, Nucl. Instr. & Methods B, 97, 20-22
49. Thomas J.M., Greaves G.N., Catlow C.R.A. (1995) Solid Catalysts studied under Operating
Conditions, Nucl. Instr. & Methods B, 97, 1-10
50. Bras W., Derbyshire G.E., Ryan A.J., Mant G. R., Felton A., Lewis R. A., Hall C. J. and Greaves G.N.
(1993) Simultaneous Time Resolved SAXS and WAXS Experiments Using Synchrotron Radiation,
Nucl. Instr. and Methods A, 326, 587-591
51. Aletru C., Greaves G.N., and Sankar G. (1999) Tracking in Detail the Synthesis of Cadmium Oxide
245
from a Hydroxyl Gel Using Combinations of in situ X-ray Absorption Fine structure Spectroscopy,
X- ray Diffraction, and Small Angle X-ray Scattering, J. Phys. Chem.,103,4147-4152
52. Greaves G.N., Aletru C., Sankar G., Kempson V. and Colyer L. (1999) In Situ Characterisation of
Semiconducting Nano-particles in Zeolites with XAFS, XRD and SAXS, Jap. J. Appl. Phys.,38, 202-
205
53. Secco, R.A. and Huang, Y. (1999), Pressure-nduced disorder in hydrated Na-A zeolite, J. Phys. Chem.
Solids, 60, 999-1002
246
THERMODYNAMICS AND TRANSPORT PROPERTIES OF INTERACTING
SYSTEMS WITH LOCALIZED ELECTRONS
A.L. EFROS
Department of Physics, University of Utah,
Salt Lake City, UT 84112 USA
INTRODUCTION
The problem of strongly correlated electrons is a focus of modern condensed matter
physics. Nobody doubts that the fractional quantum Hall effect results from electron-electron
interaction. Many people thing that this interaction is also the origin of the high temperature
superconductivity and the two-dimensional Insulator-Metal transition.
In this paper we concentrate on a problem of localized in space interacting electrons in a
disordered system. The impurity bands of lightly doped semiconductors, the two-dimensional
electron gas in different structures in an insulating regime are common examples of such
systems.
The study of electron-electron interaction in localized regime has been initiated by M.
Pollak [1] and G. Srinivasan [2], Efros and Shklovskii [3],[4] have argued that the single
particle density of states (DS) tends to zero at the Fermi level due to the long range
part of the Coulomb interaction which, in a sense, remains non-screened. They proposed
the following universal soft gap of the DS near the Fermi level at T = 0 which is called the
Coulomb gap.
(1)
(2)
The reference point for the energy is the Fermi level. The universality means that the above
DS is constructed from the energy and from the electron charge e. Note that this is the only
combination of and e with a proper dimensionality both in 2D and 3D cases. We assume
here and below that the dielectric constant of a lattice is included into electron charge.
The lack of screening is connected with vanishing of the DS at the Fermi level. In
this case the screening radius should depend on the magnitude of the electric field which is
subjected to screening. Both the Coulomb gap and the non-linear screening result from the
Coulomb law and from the discrete nature of electron charge. The question of screening in a
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 247
system of localized interacting electrons has been considered in details by Baranovskii et al
[5].
The main tool for a quantitative study of the Coulomb gap is the computer simulation
[6-14], which mostly confirm the above results on the DS. However, some deviations from the
universal behavior have been reported [9,13] and even the main concept has been equivocal
[15]. We think that in 2D case these deviations can be understood in terms of the work
[14], which shows that the universal physics of the Coulomb gap in the 2D case is valid at
large disorder only. The apparent proximity to the universal result has been observed earlier
because the simulations were restricted by a relatively high values of a disorder.
A renormalization group calculations, performed by Johnson and Khmelnitskii [16],
confirmed Eqs.(1-2) for strong disorder and the results of Ref. [14] on the crossover from
strong to weak disorder.
The experimental manifestations of the Coulomb gap are mostly variable range hopping
conduct i vi t y (VRH) and tunneling experiments. Without interaction the VRH conductivity
obeys the Mott law [17]
(3)
Here s = 1/4, 1/3 for three-dimensional (3D) and two-dimensional (2D) systems respec-
t i vel y, where G
0
is the DS at the Fermi level, is localization length, is
a numerical coefficient which depends on the space dimensionality D.
It is obvious that Eq. (3) can not be valid if the Coulomb interaction makes G
0
= 0. In
the next section we present the derivation by Efros and Shklovskii [3], which gives the law
(4)
for D = 2, 3. Here where is another numerical coefficient which depends on
the space dimensionality.
In some cases, like in neutron-transmutation-doped germanium [18], one can be sure
that experimental data closely follow the Efros-Shklovskii (ES) law while in some other
cases, where conductivity changes only by a few orders of magnitude, it is difficult to dis-
tinguish between the two laws. We think that the most important results have been obtained
recently by groups of Jiang and Dahm [19] and Adkins [20]. They have proved that the trans-
port they observed in 2D case reflects the crucial feature of the Coulomb gap, the lack of
screening in a system of localized electrons. To do that they used a metallic electrode (gate)
parallel to the 2D electron gas at a distance d from the gas. This electrode provides an extra-
screening of the Coulomb interaction. Namely, at a distance, larger than d, the interaction
between electrons becomes dipole-dipole due to the image charges in the metal. Then, the
Coulomb gap becomes smeared and the DS is energy independent at Therefore,
the conductivity obeys the ES law at high temperatures and the Mott law at low temperatures.
Note, that wi t hout an external screening the Mott law may be valid at high temperatures
where the relevant electronic states are outside the Coulomb gap while the ES law is valid at
low temperatures. The experimental observation of the completely different and very unusual
temperature dependence in the gated structures confirms that the the long-range interaction
plays an important role in the VRH. The study of the d- dependence of the VRH gives even
stronger arguments in favor of the Coulomb gap.
The direct observation of the Coulomb gap in tunneling has been claimed by Lee and
Massey [21-22].
The glassy properties are another important manifestation of the electron-electron inter-
action in a system with localized electrons. Davies, Lee and Rice [8] where the first to rise
this problem theoretically. They have coined a term electron glass which have survived
unt i l now, sometimes with a transformation to Coulomb glass. These terms have to stress
248
the relation of an electron system and a spin glass system. Davies et al. introduced an ana-
log of the Edwards-Anderson order parameter, but their calculations already showed that this
parameter is, probably, non-zero at any temperature. The absence of such parameter in the
system with an external (or built-in) disorder has been pointed out by many researches (See
[23] and references therein). However, the glass transition may exist even without Edwards-
Anderson order parameter.
The new experiments, started by the group of Ovadyahu in 1993 [24,25], definitely show
the relation of this system and the ordinary glasses. In these experiments one could control
the electron density in the films of indium oxide by the gate voltage V
g
. In some density range
the mechanism of a conductivity is the VRH, which means that electron states are localized.
The most important result is the discovery of a memory in this range. The sample, slowly
cooled down at some V
g
= V
0
, memorizes the value of V
0
. Conductance at low temperature,
measured as a funct i on of V
g
, has a small minimum at V
0
. The mi ni mum disappears with
time very slowly. The characteristic time is up to 15 hours and increases in a magnetic field.
Similar phenomena have been observed by the group of A. Goldman [26] on ul t rat hi n films of
metals near the superconductor-insulator transition. Slow relaxation has been demonstrated
by Don Monroe et al. [27] in compensated GaAs.
The analogy with the ordinary dielectric glasses can be seen from the data of Osheroff
group [28], where similar effect in the capacitance of silicon oxide has been reported.
As the first step toward the understanding of the glassy effects we have studied the
thermodynamic properties of the Coulomb glass[29].
The paper is organized as follows. In the second section we give a simple derivation of
the Coulomb gap and of the VRH of interacting electrons. In the third section we concentrate
on the glassy properties of an electron system.
COULOMB GAP AND VRH CONDUCTIVITY
Hamiltonian and Some Exact Properties
We consider here a standard classical Coulomb glass Hamiltonian [4]
(5)
The electrons occupy sites on a lattice, n
i
= 0, 1 are the occupation numbers of these sites and
r
ij
is the distance between sites i and j. The quenched random site energies are distributed
uniformly within the interval [ A, A] . To make the system neutral each site has a positive
background charge ve, where v is the average occupation number, i.e. the filling factor of the
lattice. If v = 1/2, the chemical potential is zero at all temperatures due to electron-hole
symmetry.
Hereafter we take the lattice constant l as our unit length and e
2
/l as our unit energy.
Using these units, the single particle energy at site i is given by
(6)
An important result for the Coulomb glass Hamiltonian is that the average occupation
number of a site with energy is given by the Fermi function. This result can be obtained from
a self-consistent equation derived in Ref. [30]. It has been also mentioned as an exact result in
Ref. [31]. Due to the strong electron-electron interaction this is not obvious, and we provide
a short proof here, which has never been published before. The average occupation number
249
of sites with energy can be calculated by considering a single site i = 1 and calculating its
average occupation number < n
1
> subject to the constraint that it has the required energy
By definition, this is given by
(7)
The Hamiltonian Eq. (5) can be written in the form where does not depend
on n
1
. This enables us to separate out Tr
1
which is the trace over the variable n
1
= 0, 1, thus
obtaining
(8)
Here Tr and stand for the trace and sum over all n
i
except n
1
. From Eq. (8) we readily
obtain the Fermi function
(9)
Coulomb Gap in a Single-Particle Approximation
The Coulomb gap in the DS around the Fermi level can be derived in a simple manner
[32].
At zero temperature the distribution of electrons over the lattice sites is determined by
the condition of mi ni mum H at a given total number of electrons. Equivalently, one can look
for the unconditional mi ni mum of the functional
(10)
The single-particle energies have the reference point at the Fermi level
(11)
The functional should be minimized with respect to simultaneous changes of any amount
of occupation numbers n
i
. It is easy to see that the change of one occupation number gives
the condition and which is equivalent to the regular definition
of the Fermi level in a non-interacting Fermi gas.
In the next approximation we consider the transfer of one electron from the site i, oc-
cupied in the ground state, to the site j, which is vacant in the ground state. The energy
increment of is positive if for any pair of such sites
(12)
where R
ij
= r
i
r
j
. The last term in Eq. (12) reflects a simple fact that the ground state
energy includes the potential of electron, which is initially at site i.
One can see the origin of the Coulomb gap from Eq. (12). Since and
the first two terms give a positive contribution, while the third term is negative. Thus, the
sites with energy close to the Fermi level should be separated in space. Consider sites whose
energies fall in a narrow band around the Fermi level. According to Eq. (12),
any two sites in this band with energies on the opposite sides of the Fermi level must be
250
separated by a distance R
ij
not less than Therefore the concentration of such sites
cannot exceed where space dimensionality D = 2, 3. Then, the DS
must vanish at at least as fast as
It is clear that in the approximation based on Eq. (12), no faster law can arise than
Indeed, if it were the case, the interaction energy between the sites would be much less than
their energies Such a week interaction could not be responsible for lowering of the DS.
It is obvious from the above derivation that the long range Coulomb interaction is a
crucial condition for the Coulomb gap. The DS at the energy is determined by interaction
at a distance It has been show by computations that if the localized electrons are em-
bedded into some conducting medium which is able to screen interaction, the Coulomb gap
disappears [10].
This consideration leads to Eqs. (1,2), where numerical coefficients are obtained from
the mean-field equation [4,10]. The Eqs. (1,2) are valid if where G
0
is a bare DS
without interaction. In our model it is 1/ Al
D
, where l is the lattice constant. Thus, at large A
the width of the Coulomb gap is E
g
= ( e
2
/ l )
D
( 1 / A)
D 1
. At the DS is close to
G
0
.
The applicability of the above consideration is limited by the condition or
in dimensionless units The crossover between large and small disorder has been
considered in Ref. [14] for D = 2. It has been shown that A = 1 can be considered as a large
A, since substantial deviation from the law Eq. (1) appears at very low energy only.
The dimensionality D = 1 is a marginal for the Coulomb gap. In this case one gets [33]
(13)
Thus, DS tends to zero at as This result is a starting point for an expansion
used in Ref. [16].
At finite temperature, it has been shown [8,10] that the gap is smeared at energies smaller
than the temperature. Roughly speaking,
Interaction of Excitations
Until now we have taken into account only conditions of the minimum of with respect
to change of one and two occupation numbers. What about many electron transitions? Do
they present an extra restrictions on the DS or everything is already taken into account? The
analysis of this questions can be done in terms of interaction of dipole and charge excitations.
It shows that in the 2D case the single-electron approach, used above, is, probably, good,
while in the 3D case physics is more difficult.
All single-electron excitations are electron transitions from the sites, which are occupied
in the the ground state, to the empty sites. The energy of the transition of one electron is given
by Eq. (12).
For any transition from the ground state the energy should be positive. The result
of each transition is an electron-hole pair. If the pair is just two independent
single-particle excitations: an extra electron on a site j and an extra hole on a site i. They
participate in the VRH as the carriers and they are described by the single-particle DS.
If the pair is rather an exciton with a large binding energy. From electrostatic
point of view this is a small dipole. A dipole can be also formed by transitions of many
electrons in a small region. Such a soft excitation has many-electron nature. At small these
dipoles are spatially separated from each other and they do not participate in dc. However,
they contribute to the ac and they are responsible for the low temperature thermodynamics of
the system. The interaction of these dipoles has been carefully studied (See review [10]) and
has been found not very important even in the 3D case. Assuming that the DS of the dipoles
251
is a weak function of energy, one gets that the specific heat provided by these excitations
should be approximately linear in temperature. The typical size of a pair is r
0
= e
2
/E
g
.
The crucial question now is the interaction of the single particle excitations (carriers)
with this dipoles. Suppose, an extra electron enters the system or a pair with a very large R
is excited. In both cases an extra electron appears at some point which we take as the origin.
This electron creates an electric field E = e/r
2
. The field causes electron transitions leading
to a new ground state. At a large distance the field is small and transitions occur within the
pairs with a small excitation energy. The transition occurs if Here eR is the dipole
moment of the pair with the excitation energy As a result, a pair becomes polarized by the
electric field. It is easy to show that the total number of dipole pairs polarized by this field at
a distance r is of the order of (r/r
0
)
(D 2)
where D is space dimensionality. In these estimates
we use for the DS of pairs g = G
0
and we ignore the logarithmic factor in the DS, which may
be due to the interaction of pairs in 3D case [10].
At D = 2 the interaction between carriers and dipoles is not very important. We think,
however, that it creates a persistent drift between the pseudoground states observed recently
[29] (See the next section).
However, at D = 3 the number of polarized pairs is large and they create a polaronic
shift of the order of the width of the Coulomb gap E
g
[10]. That is why in 1980 we came to
conclusion [34] that carriers have very different nature in 3D- and 2D-cases and that in 3D-
case they are rather electronic polarons than individual electrons. It follows straightforward
[10] that for D = 3 the single-particle DS has a form instead of Eq. (2).
It is obvious now that the concept of electron polarons does not correspond to reality.
The most accurate computational results [9] are inconsistent with exponential behavior of the
DS in 3D-case. Experimental data also do not show any signature of the polaronic effect.
Thus, we have a contradiction in our understanding of the 3D-case.
We would like to propose a new picture of the low energy excitations in the 3D-case
[35]. The basis of this approach is a generalized universality principle. Assume that at large
disorder the DS of pairs with given energy and length is universal. We are considering now
the DS of all low energy excitations at a given length R, in the interval of which are either
carriers or dipoles. In some way this description takes into account interaction between pairs
of all sizes, and the proposed DS is a result of this interaction. In 3D-case only the expression
(14)
is universal and obeys the proper dimensionality(1/energy.volume). Universality means that
Eq. (14) contains only the charge of electron. Here and q are some numerical constants,
and is a positive energy of an excitation. Eq. (14) is valid if both and e
2
/R are less than
E
g
. At large R we get parabolic DS for carriers, which coincide with Eq. (2). At small R one
gets for the DS of short pairs Then, specific heat C
V
/T ~ T
q
.
To choose q we should assume that the polaronic effect is absent. Otherwise Eq. (14)
would be inconsistent because of the polaronic shift in a single-particle DS.
It follows from Eq.(14) that the total number of pairs N
p
with the length R polarized by
one extra electron at a distance smaller than r is
(15)
The concept of constant DS for small dipole pairs and of a strong polaronic effect corresponds
to q = 0. For the new scenario we choose q = 1/2. Then the dipoles are not interacting with
electrons in all scales or, to be precise, q = 1/2 is a marginal exponent for the interaction. In
this sense the picture is the same as in 2D case. One should mention at this point that we try
to avoid the linear divergence in N
p
at large distances and, on this stage, we do not care about
possible logarithmic factors.
252
Finally, the new scenario for 3D-case gives that the DS of short pairs
The pairs with R = r
0
= e
2
/E
g
give the major contribution to thermodynamics. The DS of
carriers obeys Eq. (2).
An appealing feature of the proposed scenario is the unification of all strong interactions.
Consider all dipoles and carriers wi t h the energy less than The distance between the dipoles
is of order of At small this distance is much less than the distance between
carriers However, the interaction between the dipole with length R and the nearest
carrier is of the order of which is of the same order as interaction between the
nearest carriers. Thus, all important interactions become of the same order.
If this scenario is correct, one should expect that specific heat C
V
~ T
3/2
. It might be
possible to check it experimentally in lightly doped semiconductors. It is possible also to
check it by a standard computer modeling at finite T. Such a modeling has been done be-
fore by different methods [23,36-37]. The results show superlinear temperature dependence.
Mobius and Pollak [37] got C
V
~ T
p
with p 1.8 0.2. This is close to the result we expect.
The main problem for the modeling is a large size effect due to the long-range interac-
tion. As far as we know, the size effect in the specific heat has never been studied systemat-
ically. It can be described in the framework of the above scenario. The macroscopic regime
appears when the size of the system is much larger than the average distance between the
pairs We predict that in a sample L L L
(16)
where f(x) is some constant at and f(x) ~ x at The scaling law, given by Eq.
(16), is the best check of the proposed scenario. For this purpose one should neither go very
low in temperature no increase the size very strongly. One should just show that C
V
/T is a
universal function of T
1/2
L. In the same finite temperature computer modeling one can find
directly the form of the function
An experimental evidence for this scenario may come from the measurements of the ac
conductivity of disordered systems. The theory of ac conductivity goes back to the famous
papers by Pollak and Geballe [38] and Austin and Mott [39]. It has been reconstructed
by Efros and Shklovskii [10] taking into account interaction. However, the constant DS
of short pairs has been assumed. If the q = 1/2-scenario will be proved, all the theory should
be reconstructed again. This reconstruction will substantially change the temperature and
frequency dependencies of the conductivity.
We understand that the Nature might be more sophisticated than any scenario. Say,
another possibility for the 3D case would be a glass transition at finite temperature. Finally,
we think that the problem of the DS in 3D case is very important and still unsolved.
Variable Range Hopping Conductivity (VRH)
In 1968 N. F. Mott [17] found out that at sufficiently low temperatures hopping conduc-
tion results from the states, whose energies are in a narrow band around the Fermi level. With
decreasing temperature the width of the band decreases and the hopping length increases.
That is why this mechanism is called VRH.
Mott assumed that the DS in the relevant band near the Fermi level is energy indepen-
dent. To repeat his calculations assume that the hops are allowed within the band (+ w,
w) only. The concentration of sites within the band is N
w
= 2wG
0
and the average distance
between them is Hopping rate has a form
(17)
253
where R is the distance between the two sites and is the localization length. The first term
in the exponent is responsible for the tunneling, while the second one describes activation.
We do not care now about numerical factors which cannot be obtained by this way.
To estimate the logarithm of effective conductivity due to the hops inside the chosen
band one may substitute an average distance between the sites instead of R. Then
(18)
The tunneling term is small at large w, while activation prefers small w. The exponent has
a maximum at This is the width of the VRH band. At this value of
w one gets the Mott law Eq. (3). To calculate numerical coefficient one should solve the
corresponding percolation problem [40].
To take into account the Coulomb gap near the Fermi level one should use
(19)
Repeating similar calculations, one gets Eq. (4). Note that in this case
Therefore we may not to take into account the smearing of the Coulomb gap due to the finite
temperature.
The first principal computer simulation of the VRH in the interacting system is an ex-
tremely difficult problem because of a strong size effect and a huge dispersion of transition
rates given by Eq. (17). We think that all attempts to solve this problem, which we know,
[7,41,42] are not satisfactory due to different reasons. In Ref. [7] we artificially slow down
the transition rates of small pairs, while in Ref.[41,42] the size of a system is too small. The
experimental situation is described in the Introduction.
THERMODYNAMIC FLUCTUATIONS OF SITE ENERGIES AND OCCUPATION
NUMBERS
In this section we are trying to understand the nature of a glassy properties observed
experimentally in Ref. [24-27]. This part is based upon our works [29,43].
Another important effect, coming from the interaction, is that the phase space of the
many particle system has the so-called pseudoground states (PS) which were first described
by Baranovskii et al. [6] and then studied by many authors [44-46] in connection with the
long range relaxation.
The PS are the states with the total energies very close to one another but with very
different sets of the occupation numbers. Thus, many electrons have to be transferred to
go from one state to another. Each state presents a local minimum of the total energy and
transition from one state to another requires passing huge barriers, if electrons are moving by
small groups one after another.
We understand the slow relaxation of a glassy system as traveling around different PS
which is hindered by barriers between them. We study the system of interacting electrons in
thermal equilibrium, where it has a possibility (long enough time) to wander around many PS.
Such a wandering is accompanied by fluctuation of site occupation numbers. The main man-
ifestation of these fluctuations is that the configuration of occupied sites within the Coulomb
gap changes with time, even though the shape of the gap itself is time independent. This
persistent change of the configuration of occupied sites occurs even at temperatures which
are much lower then the Coulomb gap width. A related effect is the fluctuations in the site
energies, the magnitude of which is much larger than the temperature.
A crucial ingredient of the above picture is the assumption that there is no finite tem-
perature thermodynamic glass transition in the system. In the thermodynamic limit, such a
254
transition would prevent the system from reaching thermal equilibrium below the transition
temperature, thus limiting the validity of our results to finite sized samples. In fact, no such
transition has been observed either experimentally or numerically in the two dimensional
(2D) Coulomb glass. Furthermore, our system has much in common with various 2D spin
glass models, where there is a strong numerical evidence that no finite temperature thermo-
dynamic transition occurs [47]. In the work [43] we present results which support such a
conclusion for the Coulomb 2D glass as well,
While we believe that the 2D Coulomb glass can always reach thermal equilibrium,
the experimentally observed glassy dynamics [24-25] indicate that at low temperatures the
equilibration time of the system becomes very long. This results from the large energy bar-
riers that need to be traversed in order for the system to drift amongst the different PSs. As
the aim of the current work is to study thermodynamic fluctuations, it is necessary to de-
sign the simulations so that the system equilibrates within a computationally feasible time
scale. Since the standard Coulomb glass Hamiltonian itself does not contain any dynamics,
we employ dynamics which differ from the physical dynamics of a typical system, but are
significantly faster. Namely, we assume that the transition rate is independent of a distance
between sites. Therefore, any non-equilibrium phenomena, and particularly transport, cannot
be studied directly by the methods discussed here.
Spectral Diffusion
Now we turn to the central topic of this Section, which is the study of the thermodynamic
fluctuations within the Coulomb gap. These fluctuations can be seen in a few ways. One is
the time dependence of the single particle energies, which we call spectral diffusion.
Our computer simulations use the standard Metropolis algorithm, where the rate of a
hopping transition depends only on the energy difference between initial and final config-
urations. The simulations were performed on a square lattice of L L sites with periodic
boundary conditions. In this torus geometry, the distance between any two sites is taken as
the length of the shortest path between them. The filling factor v = 1/2. All results obtained
were averaged over P different sets of the quenched random energies Unless stated
otherwise the value P = 100 was used throughout. Furthermore, it was verified that all our
results saturate as a function of the system size.
To study the spectral diffusion, we first equilibrate the system for t
w
MC sweeps. ( We
define a single MC sweep as a series of N = L
2
/2 consecutive MC attempts). Then we mark
all the sites whose single particle energies are in a narrow interval [E
c
W, E
c
+ W] within
the Coulomb gap as test sites. We follow the evolution of the distribution of these energies
as the simulation proceeds. We observe that after some number of MC steps, this distribution
becomes time-independent. We have also checked that the final form of the distribution does
not depend on the initial time t
w
.
This asymptotic form is shown in Fig. 1 for E
c
= 0, W = 0.1, at two different temper-
atures. Note that the final energy distribution covers most of the Coulomb gap, although the
initial distribution (shown by arrows in Fig. 1) is centered in a small region at the center of
the gap. This is in spite of the fact that the temperature is much smaller than the gap width.
We have also studied the dependence of the final distribution on W, and have found that the
results are independent of W for W < 0.1. Also shown in Fig. 1 is an example where the initial
distribution is asymmetric, namely In this case, the asymptotic distribution is also
asymmetric, however it is nearly as broad as the symmetric distributions. Moreover, while
the initial asymmetric distribution consists of only sites with positive energies (unoccupied
sites), the final distribution also includes sites with negative energies (occupied sites).
Another way to observe spectral diffusion is to measure the time average of the single-
particle energy at site i, and the standard deviation at the same site,
We perform this calculation for all sites and create a function This function is shown
255
Figure 1. Final energy di st ri but i on of sites i ni t i al l y in the energy range [E
c
W, E
c
+ W]. Diamonds are for
E
c
= 0, W = 0. 1, T = 0.05. Squares are for E
c
= 0.3, W = 0.1, T = 0.05. Triangles are for T = 0.1, E
c
= 0
and W = 0. 1. The arrows mark the positions of the two i ni t i al distributions of test sites which were used. All
results are for A = 1.
in Fig. 2 for A = 1 and several temperatures. It is found that for all sites, the standard
deviation of the single-particle energies is much larger than the temperature. Moreover, for
sites wi t h energies near the Fermi level the standard deviation is up to two times larger than
for other sites.
We understand this picture in the following way. Sites with large are expected to
be active sites which change their occupation frequently, as their energies cross the Fermi
level. The changes of the occupation numbers of these sites are accompanied by a reorgani-
zation of the local configuration of occupied sites, which in turn is responsible for the larger
value of in a polaron manner. On the other hand, the sites with smaller are passive
sites, and they change their energy only in response to the random time dependent potential
created by the active sites. In the context of passive and active sites, it is instructive to define
the quantity E
w
as the energy at which The meaning of this is that sites which
satisfy have energy fluctuations larger then their average energy, and therefore are
active. From Fig. 2 it is also apparent that these sites have larger value of thus supporting
our understanding that these are indeed the active sites of the system.
The width of the maximum in Fig. 2 may indicate that the active sites are predominantly
within the Coulomb gap. This is reasonable, since the occupation number of sites within the
gap is strongly affected by interactions. However, at A = 1 all characteristic energies, includ-
ing the gap width, are of the same order. To check whether the active sites are indeed within
the gap, we estimate the dependence of on A for A > 1 and compare it with simulations.
The width of the gap E
g
decreases with A as E
g
~ 1/ A. The electron density within the gap
is If active sites make up a finite portion of all sites in the gap, they
create time dependent potential wi t h the mean square value We
cut off the logarithmic integral at the screening radius which is proportional to the reciprocal
density of states A. We make simulations for and plot the results of simulations in
the scale against A. The temperature for each value of A is T = 0.05/A,
keeping it constant in units of the gap width. If our hypothesis is correct we can chose pa-
rameter b in such a way that all curves collapse in one at least for passive sites. One can see
256
Figure 2. Site energy standard deviation as a funct i on of site average energy for various values of A
and T. For A > 1 the temperature is given by T = 0 . 0 5 / A. Onl y positive energies are shown due to part i cl e hole
symmetry.
from Fig.2 that indeed all curves collapse in one for b = 1.5.
The temperature dependence of the results of Fig. 2 is presented in Fig. 3. The curves
marked and show the maximal and minimal values of the standard deviation
as a function of temperature. The curve marked E
w
shows the width of the function
which was previously defined. All the quantities shown in Fig. 3 appear to be much larger
than the temperature for the entire temperature range shown.
A large number of active sites we understand as a result of wandering of the system
around different PS. Each PS is characterized by a unique set of sites forming the Coulomb
gap. So when the system passes from one PS to another the number of sites that change
their occupancy is a finite fraction of the number of sites in the Coulomb gap. The energy
separation between different PS decreases with increase of the sample size [43]. If the ther-
modynamic fluctuations of the total energy (C is the heat capacitance of the system)
is larger than this energy separation, then the described mechanism of the spectral diffusion
gives a temperature independent contribution to estimated above. For small samples, that
we use for simulations, goes to zero with temperature when becomes smaller than the
energy separation between PS, and the above estimate does not work at too low temperatures.
To get a non-zero result at the l i mi t one should increase the size of the sample while
decreasing temperature. We are not able to perform this procedure in full scale but we check
that there is no size effect in the temperature range presented in Fig. 3.
We think that the temperature dependence in Fig. 3 results from soft dipole excitations
(See previous section). The density of states of the dipole excitations is 1/A, so that the con-
centration of active excitations is ~ T/A. These excitations not only contribute themselves
to the spectral diffusion but also induce an energy change of passive sites leading to linear
temperature dependence of in Fig. 3. Indeed, each dipole creates the potential r
0
/r
2
at a
passive site, where r is the distance between the site and the nearest dipole and r
0
~ 1 /E
g
~ A
is the size of the dipole [10]. This estimate leads to a slope of dependence close to
that in Fig. 3.
257
Figure 3. Temperature dependence of quantities and E
w
for A = 1. See text for a description of these
quant i t i es.
Correlation Function
Thus, the spectral diffusion shows that the configuration of occupied sites within the
Coulomb gap persistently changes in thermodynamic equilibrium. To obtain more informa-
tion about this motion, one can study the correlation function of occupation numbers. We do
this by constructing a vector D(t
w
) after t
w
MC steps have been performed. The components
of D(t
w
) are the occupation numbers n
i
of all sites within a given energy range [W, W]. The
vector is normalized so that D(t
w
)

D(t
w
) = 1. As the simulation proceeds, we check the
occupation number of these same sites, construct the vector D(t
w
+ t), and calculate the cor-
relation function C(t
w
, t) = D(t
w
)D(t
w
+ t). Correlation functions analogous to C(t
w
, t) are
commonly used to measure the similarity between different configurations in systems such
as spin glasses [47]. For two identical configurations C(t
w
, t) = 1, while if there is no corre-
lation C(t
w
, t) = 0.5. Basically, we are interested in which is a
measure of the similarity of two arbitrary states of the system at thermal equilibrium. For a
non-interacting system,
(20)
where is the Fermi function. Thus, for the non-interacting system at
and at
In order to evaluate from the simulation, we measure C(t
w
, t) as a function of t for
a given t
w
, and wait long enough so that C(t
w
, t) becomes independent of t. We denote this
saturated value as In the inset of Fig. 4 we show an example of such a saturation.
We then increase t
w
unt i l becomes independent of t
w
, and thus obtain our estimate
of
The results for for the interacting system are shown in the main part of Fig. 4 as a
function of temperature, for A = 1, W = 0.3 and W = 0.6. The corresponding functions for
the non-interacting system, calculated directly from Eq. (20), are also shown. We observe
that the correlation for the interacting system is much weaker than for the corresponding
258
Figure 4. The correlation function as a function of temperature for A = 1 and W = 0.3 (solid circles)
and W = 0.6 (open circles). The solid line and dashed l i ne show the correlation function for the non-interacting
system, as given by Eq. (20), for W = 0.3 and W = 0.6 respectively . The inset demonstrates the saturation of
the correlation function wi t h time.
non-interacting system at the same temperature.
Note that by increasing W we include more sites in the correlation function How-
ever, our results for the spectral diffusion indicate that most of these additional sites remain
passive as Thus, should increase as W increases, as we indeed observe comparing
the results for W = 0.6 and W = 0.3 in Fig. 4.
The most interesting result coming from the study of the correlation function is the
possibility of a finite value for Clearly, such an extrapolation cannot be
considered conclusive, however, one has to keep in mind the following points: First, while
we have included in the definition of the correlation function only sites that were in the initial
energy range [0.3, 0.3], it is clear that many passive sites are still included in this definition.
These passive sites mask the behavior of the active sites and tend to increase the correlation.
Second, a finite value of means that thermal motion continues down to
zero temperature. This conclusion is consistent with the results obtained from the spectral
diffusion in the previous section, and we understand it in the same way: Namely, we expect
that a nonzero value of may be obtained only if the size of the sample grows with
the decrease of temperature so that remains larger than the energy separation between
different PS.
It is important to point out that our results cannot be explained by assuming that the
excitations of the system are separated pairs of sites, with electrons hopping back and forth
between the sites of each pair. This assumption would mean that electrons are effectively
localized in space. Since the energy density of such excitations is constant at low energies,
meaning the number of available excitations decreases linearly with temperature, one imme-
diately obtains that like in the non-interacting system. The same
temperature dependence is obtained even if excitations involve a few electrons that change
their positions simultaneously (so called many electron excitation [49-50]). In fact, any pic-
ture based upon confined separated excitations which do not interact with each other would
mean that Since our data definitely contradicts this temperature
259
dependence, we conclude that such excitations cannot explain our results.
Thus, a nonzero value of can be explained only by a thermal motion
of such electron configurations in the whole system that cannot be separated into thermal
motion of independent clusters. This means that there is a finite portion electrons that are not
localized in some region of space but move around the whole system. We view this motion
as another manifestation of wandering of the system around different PS.
The conclusions of this section are as follows. We have presented strong computational
evidence that in a disordered 2D system of localized interacting electrons, the configuration
of occupied sites within the Coulomb gap persistently changes with time. This effect persists
down to temperatures well below the Coulomb gap width, and it causes a large time dependent
random potential. This result is an exclusive property of the interacting system. Without
interaction only electrons in a small interval of the order of T change their occupation.
We argue that this effect may exist at zero temperature, if the size of the sample increases
wi t h decreasing temperature so that the separation between local minima of the total energy
is much smaller than the thermal fluctuations of the total energy. However, our computational
abilities are not enough to prove this point. Nevertheless, our simulations have been done at
low enough temperature to compare them with the experimental data.
We have obtained these results for equilibrium state only. Our program does not limit
the length of the electron transition in order to reach equilibrium by the easiest way. In a real
systems hopping distance is limited by tunneling so that applicability of the above results to
a real system arises questions.
We have shown [43] that equilibration time saturates at some value L = R
c
, which is
approximately proportional 1/ T. One can draw a conclusion that equilibration time is size-
independent if localization length is large. Such a condition can be satisfied at very low T
near metal-insulator transition. However we are not able to discuss transport in this situation
because our computations still do not take into account such important factors as narrowing of
the Coulomb gap with increasing (See [22]) and decrease of the random potential because
of the quantum mechanical averaging at large
CONCLUSIONS
We have given a critical review of the theory and experiment on the interaction of local-
ized electrons and hopping conductivity. There is another very interesting field of quantum
mechanical description of the same system which includes such important problems as the
metal-insulator transition. The simplest Hamiltonian of this problem can be obtained by
adding the non-diagonal term to the classical Hamiltonian Eq. (5). Many efforts
have been made in this field, but we do not review them in this paper. We would mention only
that quantum theory should include all the problems of classical theory mentioned above
and something else.
Coming back to the classical theory of the Coulomb gap, we would make the following
conclusions:
1. We believe that the theory of the 2D systems is basically complete and correct. We are
not so optimistic about the 3D systems. We believe that the physics of long-range interaction
is valid in this case as well, we believe that there is a gap near the Fermi level, we even think
that the VRH law Eq. (4) is also valid, but we are not sure about the structure of elementary
excitations in this case and about the possibility of thermodynamic glass transition. However,
we propose in this paper a reasonable scenario which leads to a pretty simple picture, similar
to the 2D case.
2. We think that the pseudoground states are responsible for the glassy effects which has
been observed recently. We claim that in 2D case there is no thermodynamic glass transition,
but the equilibration time strongly increases at low temperatures. We argue that the VRH
260
in the interacting system might be a result of non-ergodic behavior due to the long-time
relaxation at low temperatures.
AKNOWLEDGEMENTS
I am grateful to my long-term friend and coauthor Boris Shklovskii for many long dis-
cussions of the problems, considered here. I am grateful to David Menashe, Boris Laikhtman,
and Ofer Biham, with whom I was working on the problem of thermodynamic fluctuations.
I appreciate helpful discussions with Zvi Ovadyahu and A. Vakhin. The work was funded by
BSF Grant 9800097.
REFERENCES
1. Pollak, M.(1970) Effect of Carrier-Carrier Interactions on Some Transport Properties in Disordered Semi-
conductors, Disc. Faraday Soc. 50, 13-17.
2. Srinivasan, G.(1971) Statistical mechanics of charged traps in amorphous semiconductors, Phys. Rev.
B4, 2581-2590.
3. Efros, A.L., Shkl ovski i , B.I.(1975) Coulomb Gap and Low-Temperature Conduct i vi t y of Disordered Sys-
tems, J.Phys. C8, L49-51.
4. Efros, A.L.( 1976) Coulomb gap in disordered systems, J. Phys.C9, 2021-2031.
5. Baranovskii, S.D., Shklovskii, B.I., Efros, A.L. (1984) Screening in a system of localized electrons, Sov.
Phys. -JETP 60, 1031-1042.
6. Baranovskii, S.D., Efros, A.L., Gelmont, B.L., and Shklovskii, B.I. (1979) Coulomb Gap in Disordered
Systems, J. Phys. C12, 1023-1032.
7. Levin, E.I., Nguen, V.L., Shklovskii, B.I., and Efros, A.L. (1987) Coulomb gap and hopping electron
conduction. Computer simulation, Zh. Eksp. Teor. Fiz. 92, 1499-1510 [Sov. Phys. JETP 65, 842-856.]
8. Davies, J.H., Lee, P.A., Rice, T.M. (1984) Properties of the electron glass, Phys. Rev. B29, 4260-4271.
9. Mbius, A., Richter, H., and Drittler, B. (1992) Coulomb gap in two- and three-dimensional systems:
Simulation results for large samples, Phys. Rev. B45, 11568-11579.
10. Efros, A.L., Shklovskii, B.I. (1985) Coulomb Interaction in Disordered Systems wi t h Localized Elec-
tronics States, in A. L. Efros and M. Pollak (eds.), Electron-Electron Interaction in Disordered Systems,
Elsevier, Amsterdam, pp.409-482.
11. Cuevas, E., and Ortuno, M. (1992) Coulomb gap and tunneling conductance in 1d and 2d systems, Philos.
Mag. B 65, 681-692.
12. Ruiz, J., Ortuno, M., Cuevas, E. (1993) Correlation in two-dimensional Coulomb gaps, Phys. Rev. B48,
10777-10789.
13. Li, Q and Phillips, P. (1994) Unexpected activated temperature dependence of conductance in the presence
of a soft Coulomb gap in three dimensions, Phys. Rev. B49, 10269-10278.
14. Pikus, EG., Efros, A.L. (1994) Critical behavior of density of states in disordered system with localized
electrons, Phys. Rev. Lett. 73, 3014-3017.
15. Pollak, M. (1992) The Coulomb gap: a review and new developments, Philos. Mag. B 65, 657-668.
16. Johnson, S.R., Khmelnitskii, D.E. (1996) A Renormalization Group Approach to the Coulomb Gap, J.
Phys., Cond. Matter, 8, 3363-3372.
17. Mott, N.F. (1968) Conduction in glasses containing transition metal ions, J. Non-Crystal. Solids 1, 1-10.
18. Wang, N., Wellstood, F.C., and Sadoulet, B. (1990) Electrical and thermal properties of neutron-
transmutation-doped Ge at 20 mK, Phys. Rev. B41, 3761-3770.
19. Van Keuls, F.W.,Mathur, H, Jiang, H.W.,Dahm, A.M. (1997) Localization scaling relation in two dimen-
sions:Comparison with experiment, Phys. Rev. B56, 13263-13275.
20. Yakimov, A.I., Dvurechenskii, A.V., Kirienko, V.V.,Yakovlev, Yu.I., Nikiforov, A.I., and Adkins, C.J.
(2000) Hopping transport through an Ensemble of Ge Self-Assembled Quantum Dots, phys.stat.sol. b
218, 99-105.
21. Massey, G., Lee, M. (1995) Direct Observation of the Coulomb Correlation gap in a Nonmetallic Semi-
conductor, Si:B, Phys. Rev. Lett. 75, 4266-4269.
22. Lee, Mark,. Massey, G., Nguyen, V.L., Shklovskii, B.I. (1999) Coulomb gap in a doped semiconductor
near the metal-insulator t ransi t i on: Tunnel i ng experiment and scaling ansatz, Phys. Rev. B60, 1582-1591.
261
23. Tenelsen, K., Schreber, M. (1994) Many localized electrons in disordered systems wi t h Coulomb inter-
action: A si mul at i on of the Coulomb glass Phys. Rev. B49, 12662-1591.
24. Ben-Chorin, M., Ovadyahu, Zvi, and Pollak, M. (1993) Nonequi l i br i um transport and slow relaxation i n
hopping conduct i vi t y, Phys. Rev., B48, 15025-15034.
25. Vakhi n, A.D., Ovadyahu, Zvi, and Pollak, M. (2000) Aging Effects in an Anderson Insulator, Phys. Rev.
Lett. 84, 3402-3405.
26. Martinez-Arizala, G., Grupp, D.E., Christyiansen, C.,Mack, A.M., Markovic, N, Seguchi, Y., and Gold-
man, A.M. (1997) Anomalous Field Effect in Ul t r at hi n Films of Metals near the Superconductor-Insulator
Transition, Phys. Rev. Lett. 78, 1130-1133.
27. Don Monroe, Gossard, A.C., English, J.H., Golding, B., Haemmerle, W.H., Kastner, M.A. (1987) Long-
lived Coulomb gap in a compensated semiconductor-the electron glass, Phys. Rev. Lett. 59, 1148- 1151.
28. Rogge, Sven, Natelson, Douglas, and Osheroff, D.D. (1996) Evidence for the importance of interactions
between active defects i n glasses, Phys. Rev. Lett 76, 3136-3139.
29 Efros, A.L., Menashe, D., Bi ham, O. and Lai kht man, B.D. (2000) Classical delocalization of Int eract i ng
Electrons, phys. stat. sol(b) 218, 17-24.
30 Mogi l yanski i , A.A. and Rai kh, M.E. (1989) Self-Consistent description of the Coulomb gap at fi ni t e
temperatures, Sov. Phys. - JETP 68, 1081-1090.
31 Efros, A.L., Pikus, F.G. (1993) Classical approach to the gap in the t unnel i ng density of states of a two-
di mensi onal electron l i qui d in a strong magnetic field, Phys. Rev B48, 14694-14697.
32. Shkl ovski i , B.I., Efros, A.L. (1984) Electronic Properties of Doped Semiconductors, Springer-Verlag,
Berl i n.
33. Rai kh, M.E., Efros, A.E. (1987) Density of states in one-dimensional system wi t h localized electrons,
Sov. Phys. -JETP Lett. 45, 280-289.
34. Baranovski i , S.D., Shkl ovski i , B.I., Efros, A.L. (1980) El ement ary excitations in disordered systems with
localized electrons, Sov. Phys. JETP 51, 199-215.
35. In a very brief form this idea was published in the last sentence of the paper Efros, A.L., Shklovskii, B.I.
(1987) Influence of electron-electron interaction on hopping conduction of disordered systems, J. of Non-
Cryst. Solids 97&98, 31-36.
36 Baranovski i , S.D., Uzakov, A.A., Efros, A.L. (1982) Thermodynami c Properties of I mpur i t y Band Elec-
trons, Sov. Phys. JETP 56, 422-434.
37. Mbius, A., Pollak, M. (1996) Low Temperature Specific Heat of the Coulomb Glass: Role of Correla-
tions, Phys. Rev B53, 16197-16204.
38. Pollak, M., Geballe, T.M. ( 1961) Low Frequency Conductivity due to Hopping Processes in Silicon, Phys.
Rev. 122, 1742-1751.
39. Aust i n, I.G., Mott, N.F. (1969) Polarons in cryst al l i ne and Non-crystalline Materials, Adv. Phys. 18,
41-50.
40. Ambegaokar, V., Halperin, B.I., Langer, J.S. ( 1 9 7 1 ) Hopping conduct i vi t y in disordered systems, Phys.
Rev B4, 2612-2630.
41. Tenelsen, K., Schreber, M. (1995) Low-temperature many-electron hopping conductivity in the Coulomb
glass, Phys. Rev B52, 13287-13293.
42. Perez-Garrido, A., Ortuno, M., Cuevas, E.,Ruiz, J., Pollak, M. (1997) Conductivity of the two-
dimensional Coulomb glass, Phys. Rev B55, R8630-R8633.
43. Menashe, D., Biham, O, Laikhtman, B.D., Efros, A.L. (2000) Thermodynamic Fluctuations of Site Ener-
gies and Occupation Numbers in the two-dimensional Coulomb Glass, unpublished
44. Kogan, Sh. (1998) Electron glass: Inter-valley transitions and the hopping conduction noise, Phys. Rev
B57, 9736-9744.
45. Diaz-Sanches, A., Mbius, A., Ortuno, M., Perez-Garrido, A., and Schreiber, M. (1998) Coulomb Glass
Simulations: Creation of a set of Low-Energy Many-Particle States, phys. stat.sol (b) 205, 17-19.
46. Yu, C.C. (1999) Time-Dependent Development of the Coulomb Gap, Phys. Rev. Lett. 82, 4074-4077.
47. Marinari, E, Parisi, G., Ruiz-Lorenzo, J.J. (1998) Numerical simulations of spin glass systems, in P.
Young (ed), Spin Glasses and Random Fields, World Scientific, Singapore.
48. Knotek, M.L., Pollak, M. (1972) Correlation effects in hopping conduction: Hopping as multielectron
transition, J. Non-Crystal. Solids 8-10, 505-515.
49. Knotek, M.L., Pollak, M. (1974) Correlation effect in hopping conduction. A treatment in terms of
multielectron transitions, Phys. Rev B9, 644-655.
262
THE METAL-INSULATOR TRANSITION IN DOPED SEMICONDUCTORS:
TRANSPORT PROPERTIES AND CRITICAL BEHAVIOR
THEODORE G. CASTNER, emeritus
Department of Physics and Astronomy, University of Rochester,
Rochester, NY 14627
INTRODUCTION
Most graduate level texts in condensed matter physics devote very little attention to
metal-insulator transitions (MIT) in solids even though magnetism (ferromagnetism and anti-
ferromagnetism) are covered thoroughly. The subject is usually treated in special volumes,
review articles and monographs. Sir Nevill Mott, the individual most responsible for the
development of the field, provided a monograph [1] in 1974 (second edition 1990) that pro-
vided a comprehensive discussion of this field. Mott notes there are many different types of
MIT ranging from the vanadium oxides, nickel sulfide, metal-ammonia systems, alkali-rich
alkali halides, compressed solid rare gases, hydrogen, and iodine, liquid mercury, and the
doped semiconductors. Since 1986 one would add to that list the doped cuprate high tempera-
ture superconductors (HTS), the superconductor-insulator transition, the apparent two-
dimensional (2D) MIT in Si MOSFETS and other 2D systems such as GaAs/AlGaAs hetero-
structures. In this paper the emphasis is on the doped semiconductors. Doped Ge and Si have
been more carefully studied than other doped systems and therefore this review will focus on
these two materials, although there has also been work on 3-5 materials. The feature of the
MIT in the doped systems distinguishing them from the vanadium oxides, etc., is the disorder.
However, many of the features of MIT for ordered systems need also to be considered for
disordered systems. Mott has termed the MIT for these disordered systems an Anderson
transition. Anderson [2] described the lack of diffusion of carriers in a random potential of
sufficient strength. The Anderson transition has become known as the MIT of non-interacting
electrons. However, Belitz and Kirkpatrick (BK) [3] in their 1994 review have noted that
interactions are always present, although not necessarily relevant. They have noted that neither
Andersons nor Motts picture based on screening from interactions, is separately sufficient
to understand the MIT in the disordered doped semiconductors such as Si:P. In this review the
focus will be on the transport in the critical regime (near the critical density n
c
) and what the
transport data tells us about the relative importance of interactions and disorder. Theories
taking account of both interactions and disorder are inherently difficult. BK [3] have provided
the most comprehensive review of the theoretical progress and outstanding problems. How-
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 263
ever, the experimental data provides some very important clues on the relative importance of
disorder and interactions. Some concepts from classical semiconductor physics will be shown
to be crucial in explaining the critical behavior of weakly compensated doped Si and Ge.
Doped Si(Ge) systems consist of random distributions donors (n-type) or acceptors (p-
type). Since the hosts are insulators as with a filled valence band and empty conduction
band any itinerant carriers must arise from the dopants. In the early period of semiconductor
physics there was substantial interest in the scattering mechanism of the itinerant carriers and
what scattering mechanisms determined the mobility. Shockleys text [4] provided an early
discussion of both ionized impurity scattering and phonon (lattice) scattering and provided a
discussion of the early experimental results. The experimental results and theory demonstrated
that ionized impurity scattering (iis) [Rutherford scattering] was the dominant scattering
mechanism at low T and doping densities comparable to n
c
. This result is of vital importance
in understanding the critical behavior of the conductivity of The random
distribution of dopants accounts for the random potential necessary for an Anderson MIT. The
magnitude of the disorder potential can be enhanced by compensation that leads to ionized
acceptor-donor pairs (n-type). Compensation has an important effect on not only the nature of
the random potential, but also on the critical behavior of Motts classic 1949 paper
[5] showed that Thomas-Fermi screening (the result of interactions) provided a remarkably
accurate criterion for the critical dopant density n
c
for the onset of metallic behavior at T=0.
The nature of the screening for the dopant density n just above n
c
, presumed to be Thomas-
Fermi by many theorists, has not yet been documented experimentally. However, it is the
temperature-dependent that provides definitive evidence for the dominant role
of carrier-carrier interactions at finite T. The dominance of interactions in describing the T-
dependent demonstrates that the Anderson transition without interactions is insuffi-
cient to describe the behavior of Si:P, etc.. One of the central themes of the present review is
an effort to reconcile the T-dependent transport that is dominated by interactions
with the critical behavior of where the role of interactions is less well established.
Since the MIT phase transition is viewed as a T=0 transition, or a quantum phase transition
(QPT) it is of particular interest to consider the conductivity at the critical point, namely
There are new theoretical and experimental results on the critical conductivity
that will be discussed.
One of the dominant features on the insulating side is variable range hopping (VRH)
conduction, both Mott VRH [6]and Efros-Shklovskii (ES)VRH [7]. Since the latter features
the Coulomb gap resulting from long range Coulomb interactions, while Mott VRH apparently
ignores interactions one wants to understand the role of interactions for n<n
c
. A new theory
will be discussed that is valid in the critical regime when the Mott hopping energy This
new theory has an impact on the n- and T-dependence of the Coulomb gap and provides a
means of explaining the new experimental results [8,9] for neutron transmutation doped (NTD)
Ge:Ga which are dominated by ES VRH in the regime where Si:P and Si:As exhibit Mott
VRH. A new features of this theory arises from spatial dispersion of the dielectric response.
The outline of the paper is as follows. Section II will discuss important developments in
three periods, namely before 1976, a second period 1976-1986 that was dominated by the new
scaling theories, the theory of carrier-carrier interactions, and major advances in the experi-
mental results near the critical point at low temperatures, in many cases well below 1K. This
second period ended with the discovery of HTS. That discovery diverted attention away from
doped semiconductors, but introduced a new type of MIT with similarities to that in doped Si
and Ge. The post 1986 period introduced new experimental results, particularly on the Hall
effect, and a substantial amount of controversy that lingers today. The mid 90's produced the
apparent 2D MIT in Si MOSFETS and the most important new results in NTD Ge:Ga that will
be reviewed by Watanabe. Section III will provide a new theory and interpretation of the T=0
scaling results and address the problems with the Mott minimum metallic conductivity. Section
264
IV will discuss the T-dependent conductivity for n>n
c
, n=n
c
, and n<n
c
. Particular emphasis will
be paid to a careful comparison of the theory with the best experimental data. Section V will
discuss the effects of inhomogeneity arising from both doping inhomogeneity and, in the case
of uniaxial stress experiments, stress inhomogeneity. The inhomogeneity question is of par-
ticular importance because of the new results for NTD Ge:Ga [8,9 ]. Section VI will close the
re-view with the most significant conclusions and acknowledgements. This review must, of
necessity, omit many important topics such as magnetoresistance, microwave properties,
thermal properties, infrared studies, and much important theory. Even in the transport area
discussed much work could not be included. The hope is that some of the most interesting and
controversial issues are addressed and partially resolved.
II. BACKGROUND
Pre 1976
The earliest model that was useful for predicting the occurence of the MIT was the Her-
zfeld model [10] which was based on the Clausius-Mossotti equation for a dielectric. It can be
written for a doped semiconductor in the form
(1)
where is static dielectric constant, is the host value, N is the density of polarizable impur-
ities of polarizability diverges as The right hand side is usually written as
R/V where R is the molar refractivity and V the molar volume. Materials with R/V<1 are
insulators and those with R/V>1 are metals. The Herzfeld criterion works well for the halo-
gens, the solid rare gases and the alkaline earth and transition metals. Although it has been
applied to doped semiconductors where gets very large as there are serious
doubts as to whether the Clausius-Mossotti result is applicable for a random set of donors,
because the Lorentz local field may not be valid as the localization length di-
verges as The Herzfeld criterion has been reviewed by Ross and Barbee [12].
The Mott criterion (1949), namely where n
c
is the critical density (of don-
ors) and a* is the Bohr radius of the donor atom in the dilute limit has been remarkably suc-
cessful in predicting the n
c
for doped semiconductors and other systems over more than 7
orders of magnitude in n
c
, as demonstrated by Edwards and Sienko [13]. The Mott derivation,
based on a Yukawa potential of the form where k
s
is the screening wave
vector given by for Thomas-Fermi screening. This potential no longer binds
an electron as k
s
a* approaches 1. Despite the success of the Mott criterion, the situation is con-
siderably more complicated because donor electrons becomes itinerant in the impurity band
at energies well below the conduction band edge. Other theories and experimental results will
suggest smaller values of k
s
of order 1/d=n
l/3
. However, the empirical success of the Mott
criterion is perfectly clear. A more sophisticated Mott criterion calculation needs to take ac-
count of the screening, which isnt going to be Thomas-Fermi just above n
c
and the much
smaller energy required to excite an electron to the mobility edge of the donor impurity band.
The Anderson model and Anderson localization are described by the Hamiltonian
(2)
where the a
i
+
and a
i
are creation and annihilation operators and the are the site energies which
are randomly distributed and the distribution function is characterized by a width W. The first
term is the tight-binding term for transfer from site j to site i and is frequently simplified to
265
Figure 1. The DOS versus energy E (where E
cb
=0) for the 1s-bands for Si:As for n near n
c
based on the valley-
orbit splittings from the dilute limit. The 1s-bands are Gaussians with widths of 6, 8, 8, 12 meV for the 1s-A
1
, 1s-
T
2
, 1s-E, and UHB respectiively. E
F
and E
c
are near a minimum (or inflection point) in the DOS . Interaction
contributions are not included.
nearest neighbors with t
ij
= t for i and j neighbors. For large enough W/t all states are localized.
For 3D there is a critical W
c
below which there is a mobility edge separating localized from
ex-tended states within a single band as shown in Fig. 1. The extended states are in the band
center and the localized states are in the band tails. For the multivalley conduction bands of Si
and Ge the situation is complicated by overlapping bands associated with the 1s donor states.
Fig. 1 shows a typical density-of-states (DOS) for Si:As for 8.6x10
18
donors/cc. Including the
overlapping 1s-A
1
, 1s-T
2
, 1s-E, and upper Hubbard bands. One should stress the DOS magni-
tude of the impurity bands is much larger than that for the lower part conduction band. The
other important feature is the large valley-orbit splitting between the 1s-A
1
and the 1s-T
2
(1s-E
bands). The spin-orbit split 1s-T
2
band is responsible for the strongly donor-dependent Orbach
spin-lattice relaxation process in the dilute limit and for the strong donor dependence of the
metallic conduction electron spin resonance (CESR) linewidth for Si:P, Si:As, and Si:Sb,
Shown on Fig. 1 is the Fermi energy E
F
(n) and the mobility edge E
c
(n). For this density E
F
(n)
is near E
c
(n). A characteristic feature of an Anderson transition is the smooth passage of E
F
(n)
thru the mobility edge E
c
(n) as n is increased thru n
c
. A second crucial feature of the Anderson
transition is the two-component model. For a given n and T there are both localized and
itinerant electrons of density and n
a
(T) respectively, subject to the condition + n
a
= n
266
where n is the net doping. At T = 0 these densities are related to the DOS in Fig. 1 by
(3)
Of some importance for the transport at higher temperatures well above 1K is the T-
dependence of the chemical potential (T) [13]]. It is straight forward to calculate E
F
at T=0
for the overlapping bands. This overlap is important and it yields empty states in the upper part
of the 1s-A
1
band and filled states in the lower part of the 1s-T
2
, 1s-E, and upper Hubbard band.
In the past there has been controversy [14,15] about the existence of localized and extended
states at the same energy E. Auxiliary minima in the random potential above E
c
can produce
localized states in the itinerant regime E>E
c
, but their number will be very small and will be
neglected here.
One of the important ideas in the MIT field was the notion of a minimum metallic con-
ductivity
min
developed by Mott [16 ]. This idea is based on the Ioffe-Regel (IR) criterion that
where k
F
is the Fermi wavevector and is the mean free path and that the Boltzmann
conductivity was only meaningful for where d is the atomic spacing (the donor spacing
in n-type Si). Mott employed the Boltzmann result
to obtain Invoking the IR criterion and Mott obtained
(4)
where the coefficient C depends on the number of valley v of a multivalley semiconductor. The
Mott notion was that would drop discontinuously to zero for n<n
c
and would increase above

min
as n increased above n
c
. Section III will give an alternative derivation using classical semi-
conductor results and Anderson localization.
Conduction on the insulating side of n
c
consisted of a variety of processes characterized
by different activation energies leading to a conductivity of the form
where is the energy required to excite a donor electron to the conduction band,
while is the energy to excite an electron to a neutral donor (i.e. to the upper Hubbard band,
UHB). is the activation energy to a nearby neighbor. This last process depends on compen-
sation and is successfully explained by the Miller and Abrahams (MA) theory [17] of hopping.
Of particular interest for the MIT is which scales to zero as Pioneering studies of the
density dependence of these activation energies in Ge were made by Fritzsche [18] and Davis
and Compton [19]. Experimental results in amorphous semiconductors led Mott to develop
a new mechanism of hopping namely VRH with a T-dependent hopping energy and mean
hopping distance. This was first observed in doped Ge or Si in Ge:Sb by Alien and Atkins [20].
A second type of VRH in the presence of long range Coulomb interactions was calculated by
Efros and Shklovskii [7] and observed in many semiconductors. There were also substantial
studies of metallic samples at LT and a particularly comprehensive study was that of Yaman-
ouchi et al. [21] for Si:P of both and the Hall coefficient R
H
(n, T=4.2K) over more
than two decades in n. Fritzsche [22] gave an excellent review of the early transport properties.
This period also featured some controversy. Motts
min
was criticized by Cohen and Jort-
ner [23], who argued there was an inhomogeneous regime near n
c
. They suggested class-ical
percolation theory would describe conduction along metallic channels. would continu-
ously drop to zero as the width of the metallic channels decreased as Mott [24] re-
sponded that one could not divide the sample into insulating (opaque regions) and metallic
267
Figure 2. versus n=N
D
-N
A
for Si:Sb, Si:P, and Si:As from [27]. The increase in with n is more rapid
than predicted by Eq. (1). The scaling exponent for Si:P is 1.15 from [70-f] and is 1.20 for Si:As from [66-
g].The results give (Copyright by the American Physical Society)
regions and gave theoretical reasons for this. However, at this time there was no data that
clearly demonstrated that As a result the Cohen-Jortner idea didnt gain
much support. Nevertheless, the issue of inhomogeneity has always been an important concern
close ton
c
and is addressed in section V.
One of the important features of the MIT is the screening, as reflected in the dielectric re-
sponse. The long range Coulomb interaction between two electrons is directly affected
by Mott frequently mentioned the importance of behavior for MIT systems. There
were far fewer measurements of than for however early measurements of at 10 Ghz
for Na-ammonia solutions [25] and doped Ge [26] demonstrated an enhancement of as
268
n approached a critical density. A comprehensive study of Si:P, Si:As and Si:Sb at low fre-
quencies and low temperatures by Castner et al. [27] demonstrated the onset of a polarization
catastrophe as [see Fig.2)]. (n) increased more rapidly than predicted by Eq. (1).
1976 -1986 - The golden era of scaling and interactions
Using renormalization group methods Wegner [28] obtained an expression for the dy-
namical conductivity where t is a measure of distance from the critical point and for
many cases t=(E
F
- E
c
)/E
c
. Wegners result leads to a DC conductivity
(6)
where v is a correlation length exponent and d is the dimensionality. For d = 3 the conductivity
scaling exponent s = v. From the definition of the correlation length the con-
ductivity can be written as An influential paper by Abrahams et al. (AALR, Gang-
of-Four) [29] used the -function approach with (g) = dlng/dlnL. Their approach, a one-
parameter scaling theory for noninteracting electrons, assumes is only a function of the
dimensionless conductance g [g=G(L)/(e
2
/2h)]. Using for = s ln(g/g
c
) and integrating AALR
obtain
(7)
where the prefactor was identified with Motts
min
and with 1/s = v their result resembles
Wegners result. However, a different interpretation of the AALR result is given below. AALR
noted that which for d=3 yields Ohms law. For d = 2 AALR concluded that
for all g and that all states were localized and there was no MIT in 2D. In the last decade
there has been substantial experimental evidence for a MIT in 2D [30]. There are numerous
more sophisticated scaling theories. The scaling equations and scaling exponents depend on
the symmetry (universality classes) and these are discussed in BK. Of relevance for this
discussion are the orthogonal class (no symmetry breaking) and the unitary class (breaking of
time-reversal symmetry by a magnetic field).The scaling theory for interacting electrons was
initially discussed by McMillan [31], but was followed by by Finkelshtein [32] and others.
These more complex scaling theories may be of interest theoretically, but apparently are not
required to explain the experimental data for Si:P, etc.
An independent theoretical approach developed by Gorkov et al.[33] and by Bergmann
[34] known as weak localization based on coherent back scattering gives a DC conductivity
(8)
where the const is of order unity. Weak localization has been reviewed by Lee and Rama-
krishnan [35] and is also discussed by BK. For doped Si and Ge the experimental results can
be explained without weak localization contributions.
The most important theory (n>n
c
) for finite T is the e-e interaction theory of Altshuler and
Aronov (A-A) [36] for diffusing electrons in strongly disordered metals. A very large variety
of disordered metals exhibit the correction This correction, plus a second
T
1/2
correction from ionized impurity scattering, will provide an excellent description of the
doped Si and Ge data for n>n
c
. The A-A correction for the d=3 case takes the form
(9a)
269
Figure 3. Extrapolated T=0 values of versus the uniaxial stress S from Ref.[42]. The solid line is the
region which is reproducible in 3 samples and yields s= 0.49. The inset shows data from Ref. [41] from individual
samples. (Copyright by the American Physical Society)
where
(9b)
where x = (2k
F
/)
2
where this is the screening wave vector. The factor is known
as the exchange-Hartree factor, or the singlet and triplet (spin=1) scattering amplitudes. The
form of F= x
-1
ln(1+x) is for a Fermi liquid. Because the A-A theory is only first order in the
disorder many believe the theory is not valid very close to n
c
. An early perturbative form of the
theory used (4/3-2F) for the exchange-Hartree factor and this was used by Rosenbaum et al.
[37] to explain their data using x = 0.2x(n/10
18
)
1/3
based on free electron-like expressions for
k
F
and k. Had they used for x in the range 0.3 to 1.0 they would not have been able
to explain their negative m(n) values for n>1.05n
c
unless they reversed the sign of the ex-
change-Hartree factor. An examination of the A-A integral and final result in Eqs 5.3 and 5.4
of their review indicates the sign is negative for d=3 and opposite that for d=1. However, the
integral diverges for all and analytical continuation cannot be used from d=1 to d=3.
Choosing an appropriate cutoff frequency for the integral of (10<p<40) yields a
reasonable result.
Some have considered a localization correction based on the work of
Gorkov et al. [33] and many papers have analyzed the data with both the A-A T
1/2
term and the
BT term. A particularly good example is the analysis of the Ge:Sb data by Thomas et al. [38].
270
Figure 4. Combined data showing versus S-S
c
(upper scale n/n
c
-1) from Fig.3. The
open circles are from Ref. [41], the solid circles are from Ref. [42]. The combined results give s=0.48 0.07.
(Copyright by the American Physical Society)
It will be demonstrated below that an alternative interpretation based on a T-dependent
diffusivity which arises from an overlooked Einstein relation
[39,40] between D and the mobility , can explain the features of
using the A-A result and a new AT
1/2
term from iis.
The most significant experimental results of this period were for Si:P in a series of papers
[41,42] that established that the scaling exponent s for was s~0.510.05. This work
featured dilution refrigerators (DR), and particularly the uniaxial stress experiments reached
the lowest T (~3mK) and probed closer to n
c
(n/n
c
-1<10
-4
) than any previous or subsequent
measurements. Uniaxial stress tuning allowed a detailed study of with 10
-4
<n/n
c
-1<10
-2
with a single sample. These results provided the most detailed, accurate picture of the critical
behavior and were the standard to judge all subsequent data. Only the more recent NTD Ge:Ga
results [8,9] come close to approaching the Si:P results [41,42]. Nevertheless, the seeds for
future controversy had been planted because: 1) the exponent s~ was apparently not explain-
able by the most important scaling theories; 2) the transport results for were inter-
preted with a single-component model and free-electron expressions inconsistent with Ander-
son localization; and 3) there was no proper interpretation of the very strong deviations from
m(n)T
1/2
behavior observed in [42].
On the insulating side Shafarman et al. [43] reported data (see Fig.7 below) for 10 Si:As
samples with 0.73n
c
<n<0.99n
c
in the range 0.5K<T<10K. This data for n>0.88n
c
was a good
fit to the Mott VRH law and the prefactor appears to be nearly independent of T. The
271
Figure 5. Normalized versus n/n
c
-1 for Si:B [49], Si:P [49), Si:As [48], and Ge:Sb [47]. In each
case g<s and is given by g=0.70.07 s. The inset shows normalized Hall number A/n
c
eR
H
versus n/n
c
-1. (Copy-
right by the American Physical Society)
Mott characteristic temperature T
o
scaled rapidly toward zero as The
data for 1-n/n
c
<0.05 suggested 2.3<3v<2.9 and that the localization length exponent v was
closer to 1 than to and should have suggested that s was not equal to v. A second feature of
this data is the very rapid increase in T
o
(n) for 1-n/n
c
>0.06 that is much faster than This
is not consistent with the Mott VRH theory. This data also pinpointed n
c
for Si:As to the inter-
val 8.55<n
c
<8.60x10
18
/cm
3
without DR data.
Phillips (1983, 1988) [44] introduced a novel approach to obtain the scaling exponents
s= employing set theory, topological constraints, and the quantum theory of measurement.
A basic ingredient of Phillips approach is the two-component model embodied in (3) requiring
that This model has many, but not all, of the features required to explain the
data.
Another explanation was proposed by Kaveh [45]. He utilized a weak localization correc-
tion, in addition to an interaction correction His approach utilized a DOS N(n,0) that
was noncritical and required the Hartree term to dominate in Kavehs approach led
to s~1.0 when exchange dominated and s~0.5 when the Hartree term dominated. This explana-
tion was expanded by Kaveh and Mott [46] who wrote for n>n
c
as
272
Figure 6. Scaling behavior of the CESR excess linewidth versus n/n
c
-1 for Si:As and
Si:P from [50J. is a minimum at n~n
c
at low enough T. (Copyright by the American Physical Society)
(10)
where is the second term in (8), with L
i
the inelastic
diffusion length and where L
T
is the thermal interaction length L
T
= (hD/kT)
1/2
seen in (9a). The last term in (10) is just the
A-A term.
Eq. (10) illustrates the common trend for the 3D MIT of employing
where L
cj
is the jth characteristic length. In sections III and IV it will be argued the data
can be explained with only the first and last terms on the RHS of (10). Associated with is
a length that has previously been ignored.
In the late 80s the 3D semiconductor MIT effort was eclipsed by HTS. However Hall co-
efficient results [47,48,49] began to appear for these systems. These results showed
results scaling to zero as where g is in the vicinity of
0.7s. The fact that g<s is important because is confirms the mobility is well behaved and also
scales to zero as Koon and Castner [48] attempted to claim their results were flattening
in the vicinity of n
c
, but their Si:As results yielded g~0.35 further above n
c
. The Hall results,
as summarized in Fig. 5, were particularly important in my own thinking because with the
simplest Kittel-level interpretation determines the density of itinerant electrons,
implying as This notion, however, was at odds with the interpretations given in
the early 80s. This provided the motivation for many of the new developments described
herein.
Another quantity exhibiting scaling is the excess conduction electron spin resonance
(CESR) line width as discussed by Zarifis et al. [50] (see Fig. 6). The
theory for is given by an Elliot-Yafet [51,52] expression
273
(11)
where is the spin-orbit splitting associated with the 1s-T
2
excited state, is the valley-
orbit splitting [the 1s-A
1
- 1s-T
2
splitting], and is the scattering rate associated with the
valley-orbit process. It must be stressed that this is a different quantity than the conductivity

c
in Eq. (12) below. It is believed the ratio is very weakly dependent on n suggesting
that the scaling of originates for For the scaling of can
be approximately explained by the scaling of and a constant cross
section. However, there is no rigorous derivation of a constant cross section for this case. Note
that the scaling exponent for is slightly smaller than that for D(n,0).
Recent developments and controversy
There was substantial frustration that the Wegner and AALR results yielded s~1.0 while
the Si:P results gave s~0.5 and there were many efforts to explain the reasons for this expo-
nent puzzle, however none of the different theoretical approaches to explain s~0.5 produced
any concensus. This problem led to new experimental results and to a controversy that still
remains today. In 1993 Stupp et al.. [53] presented new Si:P results that purported to show a
crossover from s-0.55 to s~1.3 when n/n
c
-1<0.1 in the region where m(n)>0. They defined the
critical regime as that for which m(n)>0. Their approach lowered n
c
from 3.74x10
18
to
3.52x10
18
, namely a 6% drop in n
c
, which is large for this type of phase transition. This inter-
pretation was challenged by Rosenbaum et al. [54] because of homogeneity considerations
when and by myself [55] based on the fact that 7 samples were claimed to be metal-
Figure 7. versus T
-1/2
for 10 insulating Si:As samples [43] showing Mott VRH for T<10K for
Strong deviations from Mott VRH are found for below a certain T
c
(n). (Copyright by
the American Physical Society)
274
lic were actually a much better fit to Mott VRH for 0.1<T<1K than they were to the A-A
m(n)T
1/2
term. Their 3.52 sample yielded a Mott T
o
~2.3K, but T
o
is supposed to scale to zero
at n
c
. The Karlsruhe group also supported their choice of n
c
from their thermopower results [56]
that showed an unusual sign change.as Fig. 7 shows the Mott VRH results for Si:As
while Fig. 8 shows the Stupp et al. data from their Fig. 1 replotted with a Mott VRH plot of
ln (n,T) versus T
1/2
. Above T = 0.09K the data is an excellent fit to Mott VRH. For T<0.09K
there are sample dependent upward deviations that can be attributed to thermal heating of the
electrons. There is a strong similarity between the two sets of data. The Stupp et al. fit to the
A-A m(n)T
1/2
term isnt that good for 3.54<n<3.69xl0
18
since the slope increases with lower
T.
This was followed by some n-type Ge results from Shlimak et al. [57] claiming one could
get the right scaling exponent without extrapolating the data to T=0. Shlimak et al. claimed an
exponent s~1.0. Their interpretation has been criticized by Sarachik and Bogdanovich [58]. It
should be noted their metallic samples showed an m(n)T
1/2
behavior very different than other
n-type Si and Ge samples in that the m(n) dependence on n was very weak.
In 1996 Itoh et al.[8] reported the NTD Ge:Ga results showing s=0.50 for weakly com-
pensated samples. The particular importance of these new Ge:Ga results is that the samples
have a compensation K<0.001, which is much better than in earlier Ge studies. One of the
important claims for the NTD approach is the more homogeneous dopant distribution obtained
as compared to thermal doping.
Figure 8. Data from [53] replotted in [80] to compare with Mott VRH. The results for T>0.09K are a good fit
to the Mott law (vertical dashed line T= 0.09K). For T<0.09K there are sample dependent upward deviations (see
3.63) from thermal decoupling. (Copyright by the American Physical Society)
275
III. A NEW APPROACH FROM TRADITIONAL SEMICONDUCTOR PHYSICS
Although iis was the focus of enormous attention in the 50s, 60s and 70s it never played
a significant role in MIT theory and the effort to explain the scaling results for
1)
s
[s=0.5 for Si:P, Ge:Ga, etc.], nor was it used in efforts to explain the results. It was
believed that Boltzmann transport expressions could not explain the scaling behavior. This is
typified by the approach of Bhatt and Ramakrishnan [59] who attempted to explain the scaling
of with localization and interaction corrections, but employed the Boltzmann result to
explain the prefactor
o
. The pioneers in early Ge and Si transport studies knew that iis was the
dominant scattering mechanism at LT and high doping densities. An excellent review of iis is
given by Chattopadhyay and Queisser [60]. A general expression for (n,, T)
iis
, valid for
arbitrary degeneracy, employed by Mansfield [61]for scattering from a random distribution of
impurities, namely
(12)
where N(E) is the DOS and f is the Fermi function. Mansfield intended his expression for
electrons in the conduction band (E>E
CB
) but it applies equally well for itinerant electrons in
the impurity band with E>E
c
. For E
F
-E
c
/kT1 Eq. (12) leads to the standard Boltzmann result
where n
a
= N(E)f dE. In the present case n
a
is the density of itinerant
electrons above E
c
. The collision rate for iis is where N
i
is the density of ionized
impurities, and the angle-averaged cross section
The phase shifts are restricted by the Friedel sum rule [62 ] where Z is
the valence difference [Z = 1 for Si:P,etc.]. For an arbitrary N(E) = C[(E-E
c
)/E
o
]
p
it can be
shown [63] for where the phase shift sum can be removed from the integral,
that
(13)
where and is the Fermi integral
The charge neutrality condition (see [4] p.238) as is n
a
= N
i
for no compensation,
because the itinerant electrons can only originate from the donors. For a compensation K =
N
A
/N
D
(n-type) one has N
i
= n
a
+ 2KN
D
. Evaluating the Fermi integrals in (13) keeping the
leading terms for and using the charge neutrality condition for K1 one obtains
There are various points to be made about this result: 1) the result is the square root of the
Wegner result; 2) the second [ ] is simply k
F
(measured with respect to E
c
) and is where
is the well known de Broglie wavelength; 3) this result lacks the extra in Motts
min
,
thus demonstrating the Ioffe-Regel criterion is irrelevant for
B,iis
; 4) this result arises from
incoherent scattering; 5) for E
F
- E
c
= E
o
(n/n
c
-1) this result explains the scaling exponent s =
; 6) the form of the DOS has no effect on the exponent s, but does affect the magnitude of
the prefactor
o
which can be calculated using k
F
(2n
c
); 7) this result is consistent with an
Anderson transition with E
F
crossing E
c
smoothly; while screening effects are buried in the
phase shifts and screening affects the prefactor but not the scaling exponent s as long
as the system is sufficiently degenerate; 8) the compensation dependence of
B,iis
is the result
in (14) multiplied by n
a
/(n
a
+2KN
D
). It is worth emphasizing (14) doesnt depend on the itiner-
276
(14)
ant electron density n
a
, but depends only on
i,h
and the E-dependence of
Basically, the derivation of (13) has used Friedel scattering developed for impurities in
metals [64], The MIT in doped Si and Ge is a special case of alloy theory. The difference is
that the host is an insulator as but also k
F
in an alloy changes only a small amount
because N
i
n. In the MIT at A A second difference is that for an alloy
while for the MIT The same formalism, starting with (11),
has been shown [65] in the nondegenerate limit to lead to a new contribution
AT
1/2
, where A is proportional to (m*/m)
1/2
and the ratio of a scattering integral
In this case the phase shift sum term cannot be removed from the integral and one has E- and
T-dependent phase shifts that still satisfy a generalized Friedel sum rule for arbitrary
degeneracy valid for doped semiconductors. For finite T if one considers k(E,T) [or the veloc-
ity as
(15)
This emphasizes the importance of low temperatures and sufficient degeneracy to obtain
the scaling behavior with s = . Paalanen et al. [42] achieved the lowest T (~3mK) corre-
sponding to an energy difference of 0.26eV. For a characteristic energy E, = 12.5meV for
Si:P the results in [42] extended to n/n
c
-1<10
-4
and still maintained
The above derivation should be compared with that of Phillips [15,44,66]. The common
features are the use of the two-component model, and the notion that weak local-
ization is not applicable to these MIT systems. It seems highly likely that both models feature
conducting filaments for n barely above n
c
. However, the crucial difference in
B,iis
is the use
of the charge neutrality condition n
a
=N
i
, so that the k
F
dependence comes from whereas in
the Phillips case it comes from the density of itinerant electrons n
a
. The Phillips approach
yields while the Hall data yields with g~0.7s. Regret-
tably, physicists are more familiar with iis (Rutherford scattering) and Boltzmann transport
than they are with set theory, topological constraints, and Voronoi polyhedra.
Besides the scaling of and [and the Hall mobility the
scaling of the diffusivity also needs to be considered as well as some of the charac-
teristic lengths like the mean-free-path Some of these transport quantities are related by
Einstein relations. Two of these in the strongly degenerate limit are
(16)
The first one between and D is due to Kubo [67], while the second relation is the original
Einstein relation between D and which has been overlooked for many years. For many years
it was thought that and D scaled with the same exponent because of Lees identification [68]
of with the thermodynamic potential Lee argued that since the specific heat
c
v
(n,T) for Si:P varied smoothly thru n
c
then also varied smoothly thru n
c
. However, as
it will be shown that the in (16) is unrelated to c
v
and the new intepretation shows
This quantity diverges as as Classically which using
becomes in the degenerate limit. This result is important because D(n,0) gives
a direct measure of Ioffe-Regel factor even though the treatment of yields a result
independent of From (15) one finds so that The
scaling exponents for these transport parameters are summarized in Table 1 where the experi-
mental values are given for Si:P, Si: As, Si:B, and Ge:Ga, namely for the most weakly compen-
sated samples.
The major focus in experiments has been on the scaling of the conductivity, but there are
277
other quantities as that show scaling with . The scaling of R
H
has been discussed
above and the mobility [ and
H
are expected to scale with the same exponent at low fields
as H approaches a small value of order 0.1 tesla] is given by the product The diffusivity
D has not been measured directly but is inferred from from the A-A theory and the
experimental results as discussed below. The exponent for D from the Einstein relation scale
with is 1+s-g. For a constant or slowly varying effective mass m*(n) D and the Ioffe-Regl
factor the same exponent. The mean-free-path scaling can be inferred from that of D and k
F
and
scales with a positive exponent +s-g. The scaling of to zero as requires more discus-
sion since the traditional viewpoint has been that could not be less than the donor spacing.
On the insulating side the DC conductivity The interesting scaling quantities
on the insulating side besides the dielectric constant are the Mott and ES characteristic
temperatures T
o
and associated with VRH conduction. Several authors [69] have discussed
the fact the exponent and Kawabata has considered the product to be a
universal quantity for noninteracting electrons. The data in Table 1 for Si:P and Si:As show
to 2.4 times s. For the amorphous S-M alloy Al
0.3
Ga
0.7
As:Si (a persistent photo-
conductor) Katsumoto [71] obtained the scaling exponents s~1.0 and hence the ratio is
similar to the n-type Si results. A slightly different explanation for the scaling behavior of the
quantities in Table 1 was given in [40]. However, that explanation depended on the scaling of
m* and gave a different scaling relation of k
F
(n). The scenario given above with k
F
(E
F
-
Ec)
1/2
(n/n
c
-1)
1/2
is a far more convincing solution.
The new theory for k
F
, the Einstein relations, and
H
= appear to give a
self consistent set of scaling relations describing the transport behavior just above n
c
. A prob-
lem not yet resolved is what form of DOS N(E) in the vicinity of E
c
will explain the scaling
exponent g~0.33 to 0.4 using Eq. (3). This value of g is in reasonable agreement with the 3D
percolation prediction by Kirkpatrick [72] for 1/R
H
, although this might be fortuitous since the
exponent s= for is less than 1/3 the percolation prediction of 1.6. The experimental
result g<s appears to be firmly established and is in agreement with the Cohen et al. [73] hypo-
thesis that the mobility must be well behaved (non diverging as This requires All
of the known data, including that for compensated cases like Ge:Sb [47] satisfies g ~ 0.70
0.07 s. One issue still to be resolved is the apparent conflict between the 3D AALR result
(s~1) and the result s~. This requires a second look at the AALR calculation.
The AALR = s ln(g/g
c
) is only valid very close to the critical point and becomes
278
much larger than 1 for gg
c
. However as one must have for 3D and a better
choice for is = sx/(1 + sx) where x=(g-g
c
)/g
c
. This has the same form as the AALR choice
for small x, but approaches 1 as This choice permits an exact result for the integral and
yields, using g(L) = where L is a macroscopic length (sample dimension)
(17)
where the last term results from sx term in the denominator of the new The significance of
this extra term might seem unimportant for s~1 but all the data for the 3d MIT is in the
regime and the result in (17) is basically independent of the magnitude of
s. For a fixed L this is basically Ohms law The 3D data, from the standpoint of the
AALR result is very far from the critical point, even though it shows scaling. In this regime
gg
c
one should not compare (17) with the Wegner result which contains the correlation length
exponent v. The result is independent of v and isnt obviously related to the correlation
length Lee and Ramakrishnan [35] employed a Taylor series expansion of in
terms of powers of E
F
-E
c
which then yielded a result analogous to the Wegner result for 3D
and v=1. This result, using E
F
-E
C
(n/n
c
-1) yielding s~1 explains the scaling exponent for the
a-S-M alloys. However, a different Taylor series expansion in the quantity k
F
-k
c
yields a result
(18)
Since k is measured with respect to the mobility edge at E
c
this implies k
c
= 0 and because k
F
(E
F
-E
c
)
1/2
the first term in (18) yields the exponent s= .Because for 3D e
2
/hL
c
where L
c
is a characteristic scaling length this suggests (18) becomes In this case
the first derivative in (18) is a constant and all higher derivatives d
n
g/dk
n
are zero for This
is a compelling result because it demonstrates that in agreement with Eq. (13).
This form of the AALR result not only explains the exponent s=, but also explains the large
width of the scaling regime. This provides a satisfactory resolution of the exponent puzzle.
What is the physical meaning of in the Einstein relation in Eq. (16)? If one were
to use the Mott result for one obtains the result This free
electron-like expression is familiar for good metals. On the other hand when Eq. (14) [for p=]
is used one finds which in this form is an apparently unfamiliar result.
However, this can be rewritten, using as
(19)
(19) is a free electron-like expression, since n
a
is the density of itinerant electrons and E
F
-E
c
is
the Fermi energy measured with respect to the mobility edge. However, unlike normal, good
metals both n
a
and = E
F
- E
c
scale to zero as The fact that n
a
scales to zero more
slowly than E
F
- E
c
explains the divergence of The simple dependence of in a free
electron manner also strongly suggests the MIT system in Si:P is that of a Fermi liquid.
IV. THE T-DEPENDENCE OF
Metallic samples
The most important new data in the last decade is the NTD Ge:Ga data [8,9] which will
be reviewed by Watanabe, includes measurements to 20mK. This is not as low as the 3mK
279
achieved by Paalanen et al., but is sufficient for most samples, but may not be low enough for
n/n
c
-1 < 0.002 to maintain the strong degeneracy thruout the T-range of the measure-
ments. Watanabe et al. [9] have used the T
1/3
prediction of Altshuler and Aronov [74] to ana-
lyze the data for the samples closest to n
c
. This T
1/3
result, purportedly valid very close to n
c
where is based on the assumption (see Eq. (16)) is independent of T,
since the A-A derivation involves the removal of D from Eq. (9a) using As
will be demonstrated below and D(n,T) have different n- and T-dependences. This can
be seen from the general Einstein relation [75] between D and , namely
where and the is the Fermi integral. Series expansions for and
yield.
(20a)
and
(20b)
At T=0 D scales with an exponent nearly twice that that for The T-dependence of D(n,T)
is linear in T(n,T) in the nondegenerate regime whereas in the degenerate regime
D(n,T) consists of a constant D(n,0) plus a quadratic correction in T
2
. It will be shown below
that the observed deviations from T
1/2
behavior can be interpreted with the A-A term in Eq. (9a)
with a T-dependent D(n,T). The features in the Watanabe et al. Ge:Ga data are similar to the
earlier data in Ge:Sb [38], Si:P, and Si:As. The most important features of the data
are:
1) for large values of n/n
c
-1 with m(n)<0 the deviation from T
1/2
behavior is upward (a
positive change in
2) for very small values of n/n
c
-1 [<0.01 for Si:As] with m(n)>0 the deviation from T
1/2
behavior is downward (a negative change in in
3) there is a special concentration n* where no deviations from T
1/2
behavior are observed.
This can only be explained if m(n)=0. This occurs for n*=1.6x10
17
for Ge:Sb and near
1.912x10
17
for Ge:Ga, but has not been as accurately identified for Si:P and Si:As.
However, the T
1/2
magnitude is not zero at n* but is positive. This suggests there is a
second AT
1/2
contribution which has been found [61] to arise from iis. The fact that at
n* there is no deviation from T
1/2
behavior rules out the localization correction
The features 1) and 2) are illustrated in Fig. 9 showing results versus T
1/2
for Ge:Sb, Si:P
[41] and Si:As [43,76] samples. The two samples with positive slopes show downward devia-
tions from T
1/2
behavior while the four samples with negative slopes show upward deviations.
For Si:P and Si:As these deviations are much smaller than for Ge:Sb and Ge:Ga. The Newman
and Holcomb results give the most accurate picture of these deviations for n-type Si because
the data shown extends to T~20K. It should be stressed that the deviations are in conflict with
a localization correction BT, but are well described by inserting (20a) into the A-A expression
in Eq. (9a). For Ge:Sb and Ge:Ga the deviations from T
1/2
behavior are easily observed below
1K reflecting the much smaller values of E
F
-E
c
for Ge systems. This second contribution AT
1/2
provides an additional reason why the A-A derivation of the T
1/3
term is not correct.
A new analysis [77] of the of the data for Si:P and Si:As based on
= (A +m(n))T
l/2
has yielded a more satisfactory agreement between theory and experiment. The
detailed fitting procedure will be discussed elsewhere. The essential new features of this
analysis are: 1) the addition of the AT
1/2
from iis; 2) the use of k
f
2
E
f
-E
c
(n/n
c
-1) in the
expression for x in Eq. (9b) which guaranties that as thus requiring 3) the
use of a screening wavevector consistent with the two-component model and with the
280
Figure 9. a) Typical data for versus T
1/2
for Ge:Sb [38], Si:P [41], and Si:As [43,76] show upward
(downward) deviations for A
t
<0 (m(n)>0); b) the ratio versus T. For large n/n
c
-1 shows a
quadratic variation in (kT/E
F
)
2
.
screening radius r
s
identified with the insulating side; 4) neglect of intervalley effects [78] and
the use of the single valley Hartree-exchange factor taking account of the A-A sign
error for the d=3 case. The parameters for Si: As and Si:P are the same, but will differ for Ge
cases because of the different values of n/n
c
-1 where A
t
= 0.
The results of this new fit are shown in Fig. 10 showing A
t
(n) versus n/n
c
-1. An excellent
fit is obtained with p~ [D(n,0)=D
o
(n/n
c
-1)
2p
] over the range 0.0008<n/n
c
-1<2.5. The location
of the minimum value of A
t
near n~1.2n
c
is consistent with the scatter in the data. The inset
shows D(n,0) versus n/n
c
-1 obtained from the data using
281
Figure 10. A
t
=A+m(n) versus n/n
c
-1 for SiP [41,42] and Si:As [43,76]. The solid () symbols for the Si:P stress
data [42] dont join well with zero stress data where A
t
=0. The inset show D(n,0)/D
o
versus n/n
c
-l for Si:P and
Si:As yielding scaling exponent 1.00.1
(21)
where m
expt
= A
t,expt
- A and the K= (Ce
2
/h){ ]. With the exception
of the region near 1.01n
c
where m(n) is near zero and the experimental uncertainties in m(n)
are larger D(n,0)/D
o
is very close to linear in n/n
c
-1 over more than three orders of magnitude
in n/n
c
-1. There may be a small increase in the slope above 1 for n/n
c
-1<0.06. These results
establish the result that D(n,0) scales with an exponent at least twice that for in good
agreement with the Einstein relations in (16) and in (14). This should leave no doubt
about the significance of the Einstein relations in interpreting the MIT.
282
The T-dependence of for insulating samples just below n
c
The temperature dependence of insulating samples is more complex than that for metallic
samples and consists of at least 3 different contributions in the range 0.75n
c
<n<n
c
. Two of these
are the Mott and ES VRH contributions while the third comes from the thermal activation of
carriers to itinerant states above E
c
[the process in Eq.(5)]. The difference between the NTD
Ge:Ga results and the doped Si systems is most pronounced on the insulating side. The insu-
lating side appears to be where the effects of doping inhomogeneity are most important. A
second reason for the importance of the VRH data relates to the rapid scaling of the Mott T
o
to zero, which provides a very sensitive measure of can be written as
(22)
where the first term is the activated component the second term is Mott VRH and the
third term is ES VRH. The activated term is characterized by a T-dependent prefactor and an
activation energy E
a
(n,T) that is itself a function of T for n close to n
c
. As and this
term becomes the critical conductivity AT
1/2
at n
c
calculated [61] earlier which suggests q~.
Because E
a
(n) increases rapidly with (1-n/n
c
) this activated term will be unimportant for T<1K
for doped Si and Ge. Castner and Shafarman [13] have done a deconvolution of the first two
terms in (22) for the Si:As data for 4 insulating samples for 10<T<77K. This analysis deter-
mined E
a
(n,T), found q~0.5, and demonstrated the Mott VRH term contains an additional fact-
or (1-f
a
(n,T)) in the prefactor where f
a
is the fraction of donor electrons thermally excited above
E
c
. This extra factor measures the fraction of electrons below E
c
available for Mott VRH. In
the critical regime approaching n
c-
particular attention must be paid to the prefactors in the
VRH terms, because the Mott T
o
and the ES both scale to zero as Also of major
importance is the scaling behavior of the prefactors and
One of the controversies in the MIT field is over the nature of Mott VRH in the critical
regime where the Mott T
o
<T, which implies the hopping energy and the mean
hopping length It is claimed [79] that one cant have Mott VRH in this regime, even
though there is a substantial body of data that is well fit by the Mott law with T
o
<T. This
caused many unresolved discussions with Shapir at Rochester and with referees. Finally, this
provoked me to construct a new theory [80 ] for VRH in the critical regime. This is an exten-
sion of MA hopping theory and minimizes the MA impedance between a pair of states
separated by a distance R and energy An important new feature of crucial import for the
critical regime is the inclusion of spatial dispersion (SD) of the dielectric response
given by
(23)
where is the host dielectric constant, is the static dielectic constant which can
include a large hopping contribution if T isnt low enough the susceptibil-
ity as is a screening length given by where k~4 which yields
r
s
~d~ n
-1/3
. In the critical regime where one obtains for while for rr
s
one has
This feature is crucial in explaining the Ge:Ga ES VRH and scaling of SD of
changes the MA matrix elements leading to a prefactor for the MA resonance energy
W that is independent of R. This leads to a minimum impedance after employing
the Mott ansatz The parameter is the crossover parameter
between conventional LT Mott VRH for z
m
>3 and the new HT Mott VRH for z
m
<1. The Mott
exponent changes to 2/7 for z
m
<1/3. The principal results of this theory are in Table 2. The
283
characteristic length L is a macroscopic length independent of T, such as the voltage lead
spacing. The prefactor is T-dependent for the z
m
>3 usual case, but significantly is independent
of T for the new HT Mott regime where T
o
<T. The other change is the numerical coefficient
in the Mott T
o
is reduced from 18.1 to 1.51. This helps explain not only the very rapid drop of
T
o
observed in the Si:P and Si:As results shown in Fig. 11, but also the minimum observed in
R
h
for Si:As for n~0.95n
c
, which is not predicted by the conventional LT Mott theory. The
ES case yields the expected but yields different results for depending on whether SD is
important or is not important. This permits an alternative explanation for the Ge:Ga results of
Figure 11. Characteristic Mott T
o
(Si:P [80] and Si: As [43]) and ES values versus 1-n/n
c
. The rapid increase
in T
o
for Si:P and Si:As for 1-n/n
c
> 0.04 represents a crossover regime where the coefficient D in T
o
increases
rapidly. The Ge:Ga is linear in 1-n/n
c
. (Copyright by the American Physical Society)
284
the Watanabe et al. [9] for the case with SD. For 0.9n
c
<n<0.99n
c
the Ge:Ga results yielded ES
VRH with as shown in Fig. 11. These authors used the expression with no SD
and inferred v~0.33 and The difficulty with this interpretation is that the Harris [81]
and Chayes et al. [82] criterion requires If one employs the new result with SD this
difficulty is removed (since and one infers v~1. This result is of some importance,
because it agrees with the v inferred for Si:P and Si:As from the scaling of the Mott T
0
with
1-n/n
c
for 1-n/n
c
<0.04. The localization and correlation lengths should have the same exponent
but the result is in agreement with the one-loop result v=1/(d-2). This is surprising
because higher order corrections from the nonlinear sigma model lead to corrections that
reduce v substantially BK [83] have obtained v~3/4 with a two-loop calculation.
The prefactor density dependence has received insufficient attention and is of some
importance. For the Mott case both Si:P and Si:As show a strong dependence on 1-n/n
c
as
shown in Fig. 12. The dashed line shows the theoretical dependence (based on and v=1).
It appears the experimental result is steeper for 1-n/n
c
>0.14, however the dramatic deviation
from the theoretical prediction occurs for 1-n/nc<0.07 where
o, expt
is nearly flat and appears to
approach a constant value. Although there is more scatter the Ge:Ga of is also nearly flat
in this range. One possible explanation for this flattening is doping inhomogeneity.
There is a second instance where SD of is of substantial importance, namely in the
width of the ES Coulomb gap which has measured by tunneling measurements by Massey and
Figure 12. The conductivity prefactor versus 1-n/n
c
for Si:P [80], Si:As [43], and Ge:Ga [8,9]. The dashed
line shows theory prediction for z
m
< 1. In all cases expt. values of as (Copyright by the
American Physical Society)
285
Lee [84] for Si:B. The Coulomb gap width is given by Massey and
Lee reported data at T~2K that showed a decrease in from 0.85 to 0.93n
c
, but the trend
reversed and increased from 0.93 to 0.96n
c
. When is increasing [the second term
in (24) is dominant] the is decreasing with increasing n because of the rapid increase in
with n. However, at some point becomes so large that the first term becomes
more important with now closer to This manifests itself in a broadening of the Coulomb
gap as the first term in (23) takes over in importance.
V. INHOMOGENEITY NEAR THE CRITICAL POINT
Those involved in magnetic phase transition studies have always understood the impor-
tance of uniform temperatures near the critical temperature T
c
and uniform magnetic fields in
magnetic resonance experiments. In MIT studies with doped Si and Ge one is always con-
fronted with doping inhomogeneities that need to be understood. It is of particular importance
to understand how doping inhomogeneities (IN) might affect the scaling exponents. It was
recognized long ago that there were several methods to tune thru the critical point and both
magnetic fields and uniaxial stress have been utilized in tuning the MIT in doped Si and Ge.
The former removes time reversal symmetry and changes the universality class and the scaling
exponent s.Uniaxial stress does not change the universality class, but nevertheless it may
change the scaling exponent because of the introduction of relatively large stress inhomogene-
ity from sample bending in compression experiments with bars with large slenderness ratios.
This has now led two groups to suggest a larger exponent s for both Si:B [85] and Si:P [86] -
this despite the fact that the data exhibited features similar to that of Paalanen et al. [42].
For a completely homogeneous system Doping IN will introduce
a normalized dopant distribution of the form where is
the mean dopant density and is a measure of the width of the variation in n. Strictly speaking
the distribution should be spatially dependent with an n(r) and the integral should be over the
volume of the sample, but for simplicity we consider a case neglecting spatial correlation of
n(r). The average conductivity will be given by For the Mott
VRH case for T
o
<T an expansion yields When t is
an integer the integral yields Hermite polynomials of order t. For t=4 one obtains
(24)
For the t=3 case, relevant to the Mott VRH prefactor [first term in the expansion], the
powers of are reduced to 3 and 1. Higher terms in the expansion increase the value
of t [t = 3 3/4 for the second term for m= and v=1]. The net effect of IN is to reduce the var-
iation with as which is in qualitative agreement with the prefactor for
Mott VRH. The analysis suggests one can obtain a constant as For n>n
c
the important
exponent is t=. For this case the integration from n
c
to [note that for n<n
c
] yields
(25)
where is a Whittaker function. For (25) yields a finite while for small
values of z<1 a term linear in n/n
c
-1 appears of order To the order of terms
linear in z the exponent has increased to 1 and there is an apparent critical density n
c
* given
by Nevertheless for larger values of for n far enough above
n
c
P(n-n) appears like a and the integral yields
286
Thus it appears possible, neglecting correlations in n(r), to have a crossover from s = to s
= 1 if the inhomogeneity is large enough. If one can independently identify n
c
experimentally
then provides a direct measure of the inhomogeneity. The Paalanen et al.
results [42] yield scaling with s~ for n/n
c
-1>3xl0
-4
. This either suggests or that the
result neglecting correlation in n(r) in (25) is based on an assumption that isnt valid.
Stress inhomogenity (SI) from sample bending is a simpler case to treat because the stress
distribution is accurately known and is perfectly correlated spatially. When the mean compres-
sive stress is S at the median plane S varies linearly across the beam cross section [axb,a<b]
between S(1-6d(z)/a) and S(1+6d(z)/a), where d(z) is the deflection at a point z along the beam
given by where d
max
is the maximum deflection and L is the length of the
beam. The stress distribution is and zero
outside this interval. The average conductivity is given by
where the form is that relevant for the Si:B case is Again, in the
limit of large the correction from SI is small (second order in But for small [1-
one finds
(26)
where the z integration is between the two voltage leads at z
2
and z
1
, while.
The effect of SI for a cross section at a given z is to shift S
c
to but the scaling
exponent t+1 in this regime is independent of z. A numerical integration of (25) for several
positions of the d
max
relative to z
1
and z
2
confirms the result that scaling exponent
changes from t to t+1 [t~0.5 to 0.6} in excellent agreement with the uniaxial stress results of
Bogdanovich et al. [85] for Si:B. For larger values of their Si:B results are in good
agreement with the other weakly compensated cases showing s~0.5. SI from bending with
samples with L/a>20 is large and unavoidable. It would be useful to repeat these uniaxial stress
experiments with tension rather than compression.
Although this review has emphasized the weakly compensated cases where s~ it is
worth noting that heavily compensated samples seem to show substantial differences in the
compensation dependence of n
c
(K). Recent results by Rentzsch et al. [87] on NTD
74
Ge-
70
Ge
crystals (Ge:As,Ga) show a strong exponential dependence of n
c
(K) [n
c
(K)=n
c
(0)exp(K/K
o
)
s
,
K
o
=0.40, s=1.67] that is much stronger than chemically doped Ge:Sb [38] and Si:(P,B) [88].
Rentzsch et al. suggest it is far more likely in the chemically doped cases to have spatial
correlation between donors and acceptors than in NTD systems.
VI. DISCUSSION, SUMMARY AND ACKNOWLEDGEMENTS
Many theorists in the MIT field have commented on the difficulty of the problem of
treating both disorder and interactions in a self-consistent manner and there has been criticism
of perturbation theories like that of A-A because of the claim it should break down when
In addition the prevailing viewpoint had been that any Boltzmann result could
not explain the scaling of Furthermore, it was not realized, that unlike magnetic
phase transitions characterized by a correlation length the MIT in doped Si and Ge features
a second divergent length, namely the deBroglie wavelength The
Boltzmann result from iis yields It must be emphasized that the theory of
iis (based on Rutherford scattering) yields the same result in both classical mechanics and QM
for an individual ion. The treatment in section III using the alloy-like scattering theory and the
phase shift scattering approach (valid for kd<1 where d is the range of the potential) with the
Friedel sum rule is essential in obtaining the exponent s~ in the degenerate limit. The notion
287
of the MIT in doped Si and Ge being a special, novel case of alloy theory is one that needs
more attention. The screening implicit in the Mott approach and the Mott criterion is buried
in the phase shifts which have no affect on the scaling exponent s, but do affect the magni-
tude of the prefactor This is consistent with the Anderson statement [89] that a McMillan
type of theory is inescapable at T=0. At T=0 interactions can do nothing but screen the
existing random potential. However, the McMillan approach ignored the Einstein relation
between D and , and the charge neutrality condition that cancels n
a
and N
i
that leads to
B,iis
k
F
.
The irony in the explanations for is that the A-A theory, only first order in the
disorder, appears able to explain all the features of the T-dependence, once a second contribu-
tion AT
1/2
from iis is added to the mix. The metallic data can be explained without any contri-
bution from weak localization theory. Specifically, the features of the data rule out any weak
localization contribution. This not surprising since coherent back scattering and quantum
interference seem to be ruled out when the extent of itinerant electron wavepacket is compara-
ble to and is orders of magnitude larger than the impurity spacing in the critical regime.
Very close to n
c
(very small k
F
) Heisenberg uncertaintly principle supports the notion the
carrier wave packets are huge relative to the mean donor spacing and one shouldnt talk about
scattering from individual impurities. This apparently explains the lack of importance of
scattering by neutral donors (localized electrons) and at the same time explains why the Ioffe-
Regel criterion is not relevant for is independent of the mean-free-
path The scaling of to zero gives strong evidence that as a result
at odds with the older conventional wisdom. In the degenerate regime the experimental data
comparison with the A-A result yields D(n,0) scaling with exponent closely twice that for
consistent with the two Einstein relations [Eq. (16)]. Deviations from T
1/2
behavior
demonstrate for the first time the temperature dependence of D(n,T), which differs from that
for in the same T-regime. This behavior is consistent with the classical relation
and clearly demonstrates the importance of the Einstein
relations in understanding the MIT. This same relation and the data demonstrate that the A-A
derivation of a T
1/3
dependence for is incorrect because of an erroneous as-
sumption. The data demonstrate that as one is leaving the degenerate regime
and approaching the regime where is comparable to kT.
The insulating T-dependence of is more complex because of both Mott and ES
VRH, each of which can exhibit T-dependent prefactors as well as exponentials. In addition,
there is an activated component from itinerant electrons thermally excited above E
c
. Whereas
Mott VRH dominates for n>0.9n
c
for Si MIT systems (for T>0.1K) in the Ge:Ga and Ge:As
cases ES VRH is dominant. The new theory, featuring spatial dispersion of extending
VRH into the critical regime where T
o
<T and predicts no T-dependence of the prefactor
This new theory establishes Mott VRH (but with m~2/7 when z
m
<1) very close
to n
c
, but inhomogeneity may become dominant for 0.95n
c
<n<0.99n
c
.This theory provides a
straight forward explanation of the Mott T
o
(n) scaling for both Si:P and Si:As, in addition to
providing an explanation of the ES VRH for the Ge:Ga results of Watanabe et al. These
three results provide evidence that the localization length exponent v=1.00.1. This theory also
explains the n-dependence of the Coulomb gap width observed in tunneling by Massey and
Lee.
The author is grateful to various graduate students over the years, but particularly to D.
New, W.N. Shafarman, D.W. Koon, R.J. Deri, M. Migliuolo, and V.Zarifis for their careful
studies of MIT systems. NSF support made this work possible, but collaborations with G.S.
Salinger, O. Symko, and J.S. Brooks for measurements below 0.5K were essential for a suc-
cessful outcome. Discussions with D.F. Holcomb and his graduate students of the MIT and our
Si: As results were an essential component of the authors education in the MIT field. In the last
288
five years the author has enjoyed stimulating and invigorating discussions with J.C. Phillips
and D.F. Holcomb. The opportunity to take a semiconductor physics course from J. Bardeen
at the University of Illinois played a substantial role in affecting the new developments in
Section III. Finally, the author remembers with fondness various communications with Sir
Nevill Mott and Sir Nevills trip to Rochester in 1982 to present the first David L. Dexter
Lecture.
REFERENCES
1 .
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
Mott, Sir Nevill (1974), Metal-Insulator Transitions, Taylor and Francis, London.
Anderson, P.W. (1958), Phys.Rev.109, 1492-1505.
Belitz, D. and Kirkpatrick, T.R. (1994), Rev.Mod.Phys.66, 261-380.
Shockley, William (1950), Electrons and Holes in Semiconductors, Van Nostrand, New York.
Mott, N.F. (1949), Proc.Phys.Soc.London, Ser.A62, 416-22.
Mott, N.F. (1969), J.Non-Cryst. Solids 1, 1-17.
Efros, A.L. and Shklovskii, B.I. (1975), J.Phys.C8, L49-L51.
Itoh, K.M., Haller, E.E., Beeman, J.W., Hansen, W.L., Emes, J., Reichertz, L.A., Kreysa, E., Shutt, T.,
Cummings, A., Stockwell, W., Sadoulet, B., Muto, J., Farmer, J.W., Ozhogin, V.I. (1996), Phys.Rev.Lett.
77, 4058-4061.
Watanabe, M., Ootuka, Y., Itoh, K.M. and Haller, E.E. (1998), Phys.Rev.B58, 9851-57.
Herzfeld, K.F. (1927), Phys.Rev.29, 701-05.
Ross, M. and Barbee, J.W. (1995), Metal-Insulator Transitions Revisited, edited by P.P. Edwards and
C.N.R. Rao, Taylor and Francis, London, 43-64.
Edwards, P.P. and Sienko, M.J. (1978), Phys.Rev.B17, 2575-81.
Castner, T.G. and Shafarman, W.N. (1999), Phys.Rev.B60, 14182-196.
Thouless, DJ. (1974), Phys.Rep.13, 95-142.
Phillips, J.C. (1983), Solid State Commun.47, 191-193.
Mott, N.F. (1972), Philos.Mag.26, 1015-26.
Miller, A. and Abrahams, E. (1960), Phys.Rev.120, 745-55.
Fritzsche, H. (1955), Phys.Rev.99, 406-419; (1958) J.Phys.Chem.Solids6, 69-80.
Davis, E.A. and Compton, W.D. (1965), Pbys.Rev.140, A2183-94.
Allen, F.R. and Atkins, C.J. (1972), Philos.Mag.26, 1027-42.
Yamanouchi, C., Mizuguchi, K. and Sasaki, W. (1967), J.Phys.Soc.Japan22, 859-64.
Fritzsche, H. (1978), The Metal Non-metal Transition in Disordered Systems, edited by L.R.Friedman and
D.P. Tunstall, SUSSP Publications, Edinburgh, 193-238
Cohen, M.H. and Jortner, J. (1973), Phys.Rev.Lett.30, 699-702.
Mott, N.F. (1973), Phys.Rev.Lett.71, 466-67.
Mahaffey, D.W. and Jerde, D.A. (1968), Rev.Mod.Phys.40, 710-713.
DAltroy, F.A. and Fan, H.Y. (1956), Phys.Rev.103, 1671-74.
Castner, T.G., Lee, N.K., Cieloszyk, G.S. and Salinger, G.L. (1975), Phys.Rev.Lett.34, 1627-30.
Wegner, F. ((1976), Z.Phys.B25, 327-37.
Abrahams, E., Anderson, P.W., Licciardello, D.C. and Ramakrishnan, T.R. (1979), Phys.Rev.Lett.42, 673-
76.
Kravchenko, S.V., Kravenchenko, G.V., Furneaux, I.E., Pudalov, V.M. and DIorio, M.D. (1984), Phys.
Rev.B50, 8039-42; Popovic, D., Fowler, A.B. and Washburn, S. (1997), Phys.Rev.Lett.79, 1543-46.
McMillan, W.L. (1981), Phys.Rev.B24, 2739-43.
Finkelstein, A.M. (1983), Zh. Eksp, Teor. Fiz.84, 168-89 [JETP57, 97-108 ].
Gorkov, L.P., Larkin, A.I. and Khmelnitskii, D.E. (1979), Zh.Eksp.Teor.Fiz.PismaRed.30, 248-52 [JETP
Lett.30, 228-32].
Bergmann, G. (1984), Phys.Rep.107, 1-58.
Lee, P.A. and Ramakrishnan, T.R. (1985), Rev.Mod.Phys.57, 287-337.
Altshuler, B.L. and Aronov, A.G. (1979), Zh.Eksp.Teor.Fiz.77,2028-44 [JETP50,968-76] ; (1983) Solid
State Commun.46, 429-35; (1985) Electron-Electron Interactions in Disordered Systems, edited by A.L.
Efros and M. Pollak, North-Holland, Amsterdam, 1-153.
Rosenbaum, T.F., Andres, K., Thomas, G.A. and Lee, P.A. (1981), Phys.Rev.Lett.46, 568-71.
Thomas, G.A., Kawabata, A., Ootuka, Y., Katsumoto, S. and Sasaki, S. (1982), Phys.Rev.B26, 2113-19.
Einstein, A. (1905), Ann. Phys. (Leipzig)17, 549-60.
289
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
80.
81.
82.
83.
84.
85.
86.
87.
88.
89.
290
Castner, T.G. (1997), Phys.Rev.B55,4003-06.
Rosenbaum, T.F., Andres, K., Thomas, G.A. and Bhatt, R.N. (1980) Phys.Rev.Lett.45, 1723-26; Rosen-
baum, T.F., Milligan, R.F., Thomas, G.A., Bhatt, R.N. and tin, W.N. (1983), Phys.Rev.B27, 7509-23.
Paalanen, M.A., Rosenbaum, T.F., Thomas, G.A. and Bhatt, R.N. (1982), Phys.Rev.Lett.48, 1284 -87;
Thomas, G.A., Paalanen, M.A. and Rosenbaum, T.F. (1983), Phys.Rev.B27, 3897-900.
Shafarman, W.N., Koon, D.W. and Castner, T.G. (1989) Phys.Rev.B40, 1216-31.
Phillips, J.C., (1983), Phil. Mag.47,407-18; (1988) Phil. Mag.58, 361-367.
Kaveh, M. (1985), Philos.Mag.B52, L1-L8.
Kaveh, M. and Mott, N.F. (1987), Philos.Mag.B55, 1-8.
Field, S.B. and Rosenbaum, T.F. (1985), Phys.Rev.Lett.55, 522-25.
Koon, D.W. and Castner, T.G. (1988) Phys.Rev.Lett.60, 1755-58.
Dai, P., Zhang, Y. and Sarachik, M.P. (1994), Phys.Rev.B49, 14039-42; (1993) Phys.Rev.Lett.70, 1968-71.
Zarifis, V. and Castner, T.G. (1987), Phys.Rev.B36, 6198-201.
Elliott, R.J. (1954), Phys.Rev.96, 266-79.
Yafet, Y. (1963), Solid State Physics, Vol.14, edited by F. Seitz and D. Turnbull, Academic, New York,
pp. 1-98.
Stupp, H., Hornung, M., Lkner, M., Madel, O. and v. Lhneysen, H. (1993), Phys.Rev.Lett.71,2634-37.
Rosenbaum, T.F., Thomas, G.A. and Paalanen, M.A. (1994), Phys.Rev.Lett.72, 121(C).
Castner, T.G. (1994), Phys.Rev.Lett.73, 3600 (C).
Lakner, M. and v. Lhneysen, H. (1993), Phys.Rev.Lett.70, 3475-78.
Shlimak, I., Kaveh, M. Ussyshkin, R., Ginodman, V. and Resnick L. (1997), Phys.Rev.Lett.78, 3978-81.
Sarachik, M.P. and Bogdanovich, S. (1998), Phys.Rev.Lett.78, 3977 (C).
Bhatt, R.N. and Ramakrishnan, T.V. (1983), Phys.Rev.B28, 6091-94.
Chattopadhyay, D. and Queisser, H.J. (1981), Rev.Mod.Phys.53, 745-68.
Mansfield, R. (1956), Proc.R.Soc.London Sect.B69, 76-82.
Friedel, J. (1958), Nuovo Cimento Suppl.7, 287-311.
Castner, T.G. (2000), Phys.Rev.Lett.84, 1539-42.
Kittel, C. (1963), Quantum Theory of Solids, John Wiley and Sons, New York, Ch. 28.
Castner, T.G. (2000), Phys.Rev.Lett.84,2905-08.
Phillips, J.C. (1992), Phys.Rev.B45, 5863-5867; (1998) Proc.Nat.Acad.Sci.95, 7264-69.
Kubo, R. (1957), J.Phys.Soc.Jpn.12, 570-86.
Lee, P.A. (1982), Phys.Rev.B26, 5882-85.
Imry, Y., Gefen, Y. and Bergmann, D. (1982), Anderson Localization, edited by Y. Nagaoka and H. Fuku-
yama, Springer, Berlin, 138-149; Kawabata, A. (1984), J.Phys.Soc.Jpn.53, 1429-1433.
a-Castner, T.G. (1995), Phys.Rev.B52, 12434-38; b-Ref. [40]; c-Ref. [47]; d-Dai, P, Zhang, Y. and
Sarachik, M.P. (1991), Phys.Rev.Lett.66,1914-17 ; e-Ref. [8,9]; f-Hess, H.F.; DeConde, K.; Rosenbaum,
T.F. and Thomas, G.A. (1982) Phys.Rev.B25, 5578-80; g-Brooks, J.S.; Symko, O.G. and Castner,
T.G.(1987), JJour.Appl.Phys.26, Suppl.26-3,721-22; h-Ref. [50].
Katsumoto, S. (1987), J.Phys.Soc.Jpn.56,2259-62.
Kirkpatrick, S. (1973), Rev.Mod.Phys.45, 574-88.
Cohen, M.H., Economou, E.N. and Soukoulis, C.N. (1984), Phys.Rev.B30, 4493-500.
Altshuler, B.L. and Aronov, A.G. (1983), Pisma Zh. Eksp.Teor.Fiz.37, 349-51 [JETP Lett.37,410-413)].
Smith, R.A. (1968), Semiconductors, Cambridge Univ. Press, Cambridge, p.237.
Newman, P.P. and Holcomb, D.F. (1983), Phys.Rev.B28,626-28.
Castner, T.G. (2000), private communication
Bhatt, R.N. and Lee, P.A. (1983), Solid State Commun.48 755-59.
Dai, P., Bogdanovich, S, Zhang, Y. and Sarachik, M.P. (1995), Phys.Rev.B52, 12439-40.
Castner, T.G. (2000), Phys.Rev.B61, 16596-609.
Harris, A.B. (1974), J.Phys.C7, 1671-92.
Chayes, J., Chayes, L., Fisher, D.S. and Spencer, T. (1986), Phys.Rev.Lett.57, 2999-3002.
Belitz, D. and Kirkpatrick, T.R. (1992), J.Phys.Condens.Matter4, L37-L42.
Massey, J.G. and Lee, M. (1996), Phys.Rev.Lett.77, 3399-402.
Bogdanovich, S., Sarachik, M.P. and Bhatt, R.N. (1999), Phys.Rev.Lett.82, 137-40.
Waffenschmidt, S., Pfleiderer, C. and v. Lohneysen, H. (1999), Phys.Rev.Lett.83, 1305-08.
Rentzsch, R., Mller, M., Reich, Ch., Sandow, B., lonov, A.N., Fozooni, P., Lea, M.J., Ginodman, V. and
Shlimak, I. (2000), phys.stat.sol.(b)218, 233-36.
Thomaschefsky, U. and Holcomb, D.F. (1992), Phys.Rev.B45, 13356-62.
Anderson, P.W. (1985), Localization, Interaction and Transport Phenomena, edited by B. Kramer, G. Berg-
mann, and Y. Bruynseraede, Springer, Berlin, p. 12.
METAL-INSULATOR TRANSITION
IN HOMOGENEOUSLY DOPED GERMANIUM
MICHIO WATANABE
Department of Applied Physics and Physico-Informatics,
Keio University, 3-14-1 Hiyoshi, Kohoku-ku,
Yokohama 223-8522, Japan
INTRODUCTION
The metal-insulator transition (MIT) in doped semiconductors with a random distribu-
tion of impurities is a unique quantum phase transition in the sense that both disorder and
electron-electron interaction play a key role (see for example Refs. 1 and 2). The metallic
phase of the transition is characterized by a finite electrical conductivity at T = 0, while the
conductivity in the insulating phase vanishes in the limit of zero temperature. From a theo-
retical point of view, the correlation length in the metallic phase and the localization length
in the insulating phase diverge at the critical point with the same exponent v, i.e., they are
proportional to in the critical regime of the MIT, and the value of v provides
important information about the MIT. Here, N is the impurity concentration and N
c
is the
critical concentration for the MIT. Since direct experimental determination of v is extremely
difficult, researchers have usually determined, instead of v, the value of the conductivity crit-
ical exponent defined by
immediately above N
c
Here, is the conductivity extrapolated to
T = 0 and is a prefactor. Values of v are then obtained assuming the relation
which is valid for three-dimensional systems without electron-electron interaction. With a
number of nominally uncompensated semiconductors has been obtained [2]. One of
the best studied nominally uncompensated semiconductors is Si:P. Rosenbaum et al. studied
both the doping-induced MIT and the uniaxial-stress-induced MIT, and showed that Eq. (1)
describes of Si:P with a single exponent over a wide range
Here, uniaxial stress was used for tuning because fine control of N/N
c
1 is difficult.
In the same material, however, was claimed about ten years later for a narrow regime
N
c
< N < 1.1N
c
by a different group [5]. Moreoever, was reported recently for the MIT
driven by uniaxial stress [6]. As the origin of the discrepancy, inhomogeneous distribution of
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 291
the impurities has been pointed out [7].
For the case of melt- (or metallurgically) doped samples, which have been employed in
most of the previous studies including Refs. 4 6, the spatial fluctuation of N due to dopant
striations and segregation can easily be on the order of 1% across a typical sample for the
four-point resistance measurement that has a length of ~5 mm or larger (see for example
Ref. 8). For this reason, it will not be meaningful to discuss physical properties in the criti-
cal regime (e.g., ), unless one evaluates the macroscopic inhomogeneity in
the samples and its influence on the results. In order to rule out the ambiguity arising from
the inhomogeneity, we prepared
70
Ge:Ga samples by neutron-transmutation doping (NTD)
of isotopically enriched
70
Ge single crystals. The NTD method inherently guarantees the
random distribution of the dopants down to the atomic level [9,10]. We show from the con-
ductivity measurements at T = 0.02 1 K that of the NTD
70
Ge:Ga samples is described
by Eq. ( l ) with over a wide range
In order to determine v without assuming we have analyzed the temperature de-
pendence of the conductivity on the insulating side of the MIT in the context of variable-
range-hopping (VRH) conduction wi t hi n the Coulomb gap [12]. Low-temperature ( T <
0.5 K) conductivity of the insulating
70
Ge:Ga samples obeys
which is predicted by the VRH theory [13], with an appropriate temperature dependence in
the prefactor Magnetic field and temperature dependence of the conductivity of the
samples are subsequently measured in order to determine directly the localization length
and the impurity dielectric susceptibility as a function of N in the context of the theory.
This kind of determination of and was performed for compensated Ge:As by Ionov et
al. [14] They found and with and
respectively, for samples having N up to 0.96N
c
. The significance of their
result is the experimental verification of the relation that had been predicted by scaling
theories [15]. However, the critical exponents of compensated samples are known to be dif-
ferent from those of nominally uncompensated samples [ 11] . Therefore, our determination
of and in nominally uncompensated samples is important.
The previous effort to measure the impurity dielectric susceptibility as a function of N
has also contributed greatly. Hess et al. found in nomi nal l y uncompen-
sated Si:P [16]. Since was determined for the same series of Si:P samples [4], the
relation was again valid. Katsumoto has found and for compensated
Al
0.3
Ga
0.7
As:Si, i.e., again, applies [17]. Thus, in these cases the conclusion
was reached indirectly, by assuming We, on the other hand, determine v directly,
i.e., we do not have to rely on the assumption in order to study the behavior of the
localization length near the MIT.
According to theories [2] on the MIT which take into account both disorder and electron-
electron interaction, the critical exponents do not depend on the details of the system, but
depend only on the universality class to which the system belongs. Moreover, there is an
inequality which is expected to apply generally to disordered systems irrespective
of the presence of electron-electron interaction [18]. Hence, if one assumes which
is derived for systems without electron-electron interaction, violates the inequality.
This discrepancy has been known as the conductivity critical exponent puzzle. Kirkpatrick
and Belitz have claimed that there are logarithmic corrections to scaling in universality classes
with time-reversal symmetry, i.e., when the external magnetic field is zero, and that
found at B = 0, should be interpreted as an effective exponent which is different from a
real exponent satisfying Therefore, comparison of with and without the
time-reversal symmetry, i.e., with and without external magnetic fields becomes important.
We study the MIT of
70
Ge:Ga in magnetic fields up to B = 8 T and show that changes
from 0.5 at B = 0 to 1. 1 at The same exponent is also found in the
magnetic-field-tuned MIT for three different samples, i.e.,
292
where B
c
(N) is the critical magnetic field for concentration N. Moreover, an excellent finite-
temperature scaling [2]
where x/y is equivalent to is obtained with the same value of The phase diagram
on the (N, B) plane is successfully constructed, and we find a simple scaling rule which
would obey and derive from a simple mathematical argument that as
has been observed in our experiment.
EXPERIMENTS
Sample preparation
All of the
7()
Ge:Ga samples were prepared by neutron-transmutation doping (NTD) of
isotopically enriched
7()
Ge single crystals. The basic idea of NTD is as follows. Suppose
that a nucleus in a crystal of a semiconductor captures a thermal neutron. After the capture,
the nucleus is not necessarily stable. If it is stable, the element remains unchanged, but if
it is not, it decays and transmutes into a new element which may act as a dopant. This is
NTD. Practically, a crystal is placed in a nuclear reactor which produces thermal neutrons.
Since the neutron field produced by a reactor is large enough to guarantee a homogeneous
fl ux over the crystal dimensions and the small capture cross section (typically 10
24
cm
2
)
of semiconductors for neutrons minimizes self-shadowing, NTD is known to produce the
most homogeneous, perfectly random distribution of dopant down to the atomic level [9],
As for Ge, there are five stable isotopes:
70
Ge (20.5%),
72
Ge (27.4%),
73
Ge (7.8%),
74
Ge (36.5%), and
76
Ge (7.8%). The numbers in the parentheses represent the natural abun-
dance. These five stable isotopes of Ge undergo the following nuclear reactions after captur-
ing a thermal neutron.
(4)
(5)
(6)
(7)
(8)
Here, is a neutron, EC and denote electron capture and decay, respectively, and the
time in the parentheses represents the half life. Note that the natural Ge form both acceptors
(Ga) and donors (As and Se) after NTD. Empirically, the impurity compensation is known
to affect the value of the critical exponent [11]. To avoid the impurity compensation, we use
isotopically enriched
70
Ge.
The Czochralski grown, chemically very pure
70
Ge crystal has isotopic composition
[
70
Ge]=96.2 at. % and [
72
Ge]=3.8 at. % [10]. The as-grown crystal is free of dislocations, p
type with a net electrically-active-impurity concentration less than 5 10
11
cm
3
. The ther-
mal neutron irradiation was performed with the thermal to fast neutron ratio of The
small fraction of
72
Ge becomes
73
Ge which is stable, i.e., no further acceptors or donors are
introduced. The post-NTD rapid thermal annealing at 650 C for 10 sec removed most of the
irradiation-induced defects from the samples. The short annealing time is important in order
to avoid the redistribution and/or clustering of the uniformly dispersed
71
Ga acceptors. The
concentration of the electrically active radiation defects measured with deep level transient
spectrometry (DLTS) after the annealing is less than 0.1% of the Ga concetration, i.e., the
compensation ratio of the samples is less than 0.001. The dimension of most samples used
for conductivity measurements was 60.90.7 mm
3
. Four strips of boron-ion-implanted re-
gions on a 60.9 mm
2
face of each sample were coated with 200 nm Pd and 400 nm Au pads
293
Figure 1. Electrical conductivity as a function of T
1
/
2
for NTD
70
Ge:Ga. From bottom to top in units of
10
17
cm
3
, the Ga concentrations are 1.853, 1.856, 1.858, 1.861, 1.863, 1.912, 1.933, 2.004, 2.076, 2.210,
2.219, 2.232, 2.290, 2.362, and 2.434, respectively.
using a sputtering technique. Annealing at 300 C for one hour activated the implanted boron
and removed the stress in the metal films.
The concentration of Ga acceptors after NTD is determined from the time of thermal-
neutron irradiation. The concentration is proportional to the irradiation time as long as the
same irradiation site and the same power of a nuclear reactor are employed.
Low-temperature measurements
The electrical conductivity measurements were carried out down to temperatures of
20 mK using a
3
He-
4
He dilution refrigerator. All the electrical leads were low-pass filtered
at the top of the cryostat. The sample was fixed in the mixing chamber and a ruthenium ox-
ide thermometer [Scientific Instrument (SI), RO600A, 1.41.30.5 mm
3
] was placed close
to the sample. To measure the resistance of the thermometer, we used an ac resistance
bridge (RV-Elekroniikka, AVS-47). The thermometer was calibrated against 2Ce(NO
3
)
3

3Mg(NO
3
)
2
24H
2
O (CMN) susceptibility and against the resistance of a canned ruthenium
oxide thermometer (SI, RO600A2) which was calibrated commercially over a temperature
range from 50 mK to 20 K. We employed an ac method at 21.0 Hz to measure the resistance
of the sample. The power dissipation was kept below 10
14
W, which is small enough to
avoid overheating of the samples. The output voltage of the sample was detected by a lock-
294
Figure 2. Conductivity as a function of (a) T
1
/
2
and (b) T
1
/
3
, respectively, near the metal-insulator transition.
From bottom to top in units of 10
17
cm
3
, the concentrations are 1.853, 1.856, 1.858, 1.861, 1.863, and 1.912,
respectively. The upper and lower dotted lines in each figure represent the best fit using the data between 0.05
K and 0.5 K for the first and the third curves from the top, respectively. Each fit is shifted downward slightly
for easier comparison.
in amplifier (EG&G Princeton Applied Research, 124A). All the analog instruments as well
as the cryostat were placed inside a shielded room. The output of the instruments was de-
tected by digital voltmeters placed outside the shielded room. All the electrical leads into the
shielded room were low-pass filtered. The output of the voltmeters was read by a personal
computer via GP-IB interface connected through an optical fiber. Magnetic fields up to 8 T
were applied in the direction perpendicular to the current flow by means of a superconducting
solenoid.
RESULTS AND DISCUSSION
Temperature dependence of conductivity in metallic samples and the critical exponent
for the zero-temperature conductivity
The temperature dependence of the electrical conductivity mostly for the metallic sam-
ples is shown in Fig. 1. The temperature variation of the conductivity of disordered metal is
governed mainly by electron-electron interaction at low temperatures [1], and can be written
as
(9)
where
(10)
Here, is a dimensionless and temperature-independent parameter characterizing the Hartree
interaction and D is the diffusion constant [1], which is related to the conductivity via the
Einstein relation
(11)
295
Figure 3. (a) Conductivity as a function of From bottom to top in units of 10
17
cm
3
, the
concentrations are 1.858, 1.861, 1.863, and 1.912, respectively. The solid lines denote the extrapolation for
finding (b) Zero-temperature conductivity vs the dimensionless distance N/N
c
1 from the critical
poi nt on a double logarithmic scale. The dotted line represents the best power-law fit by
where
where is the density of states at the Fermi level. In various reports such as Refs. 5 and
10, was obtained by extrapolating to T = 0 assuming dependence based on
Eq. (9). One should note, however, that it is sound only in the limit of
where D can be considered as a constant, i.e., m is constant, and that the inequality is no
longer valid as N approaches N
c
from the metallic side since also approaches zero. In
such cases m in Eq. (9) is not temperature independent and may exhibit a temperature
dependence different from To examine this point in our experimental results, we go
back to Fig. 1. We see there that of the bottom five curves are not proportional to
while of the other higher N samples are described by The close-ups
of for the six samples with positive in the scale of and T
1/3
are shown
in Figs. 2(a) and 2(b), respectively. The upper and lower dotted lines represent the best fit
using the data between 0.05 K and 0.5 K for the samples with N = 1.912 10
17
cm
3
and
N = 1.861 10
17
cm
3
, respectively. Each fit is shifted downward slightly for the sake of
clarity. From this comparison, it is clear that a T
1/3
dependence rather than a dependence
holds for samples in the very vicinity of the MIT. The opposite is true for the curve at the top.
This means that the dependence in Eq. (9) is replaced by a T
1/3
dependence as the MIT
is approached.
A T
1/3
dependence close to the critical point for the MIT was predicted originally by
Altshuler and Aronov [21]. They considered an interacting electron system with param-
agnetic impurities, for which they obtained a single parameter scaling equation. At finite
temperatures, they assumed a scaling form for conductivity according to the scaling hypoth-
esis:
(12)
where is the correlation length and is the thermal diffusion length. When
which is equivalent to Eq. (9). In the critical region,
where Eq. (12) should be reduced to
(13)
296
Figure 4. The logarithm of the conductivity as a function of T
1
/
2
for insulating samples. the concentrations
from bottom to top in units of 10
17
cm
3
are 1.717, 1.752, 1.779, 1.796, 1.805, 1.823, 1.840, 1.842, 1.843,
1.848, 1.850, 1.853, 1.856, and 1.858, respectively.
Combining this equation and Eq. ( 11) , they obtained The T
1
/
3
dependence was
also predicted from numerical calculations that consider solely the effect of disorder [22].
It is not clear whether the argument in Refs. 21 or 22 is applicable to the present system
or not, but there is an experimental fact that dependence of conductivity changes to
T
1
/
3
as N
c
is approached. It is important that we find a consistent method that allows the
determination of for both the cases. For this purpose, we follow Altshuler and Aronovs
manipulation [21] of eliminating m and D in Eqs. (9)(l 1) and obtain
(14)
where which is temperature independent. In the limit of
this equation gives the same value of as Eq. (9) does. When it
yields a T
1
/
3
dependence for Thus, it is applicable to both and T
1
/
3
dependent
conductivity. From todays theoretical understanding of the problem, Eqs. (9) and (14) are
valid only for and their applicability to the critical region is not clear, because the
higher-order terms of the function [23] which were once erroneously believed to be zero
do not vanish [24]. Nevertheless, we expect Eq. (14) to be a good expression for describing
the temperature dependence of all metallic samples because it expresses both and T
1
/
3
dependences as limiting forms. Then, based on Eq. (14), we plot for
the four close to N
c
samples in Fig. 3(a). The data points align on straight lines, which
supports the adequacy of Eq. (14). The zero-temperature conductivity is obtained by
extrapolating to T = 0. The curve on the top of Fig. 3(a) is for the sample with the lowest N
297
Figure 5. Conductivity multiplied by T
1
/
3
vs (a) T
1
/
2
and (b) T
1
/
4
. From bottom to top in units of
10
17
c m
3
, the concentrations are 1.848, 1.850, 1.853, and 1.856, respectively.
among the ones showing dependence at low temperatures, i.e., this sample has the largest
value of among samples. The value of obtained for this particular
sample using Eq. (14) differs only by 0.6% from the value determined by the conventional
extrapolation assuming Eq. (9). This small difference is comparable to the variation arising
from the choice of the temperature range in which the fitting is performed. Therefore, the
extrapolation method proposed here is compatible with the conventional method based on
the extrapolation.
Based on this analysis the MIT is found to occur between the first and second samples
from the bottom in Fig. 3(a), i.e., 1.858 10
l7
cm
3
< N
c
< 1.861 10
17
cm
3
. Thus, unlike
the case for Si:P [5,7], N
c
is fixed already within an accuracy of 0.16% and the evaluation
of the critical exponent will not be affected by the ambiguity in the determination of N
c
.
Figure 3(b) shows the zero-temperature conductivity as a function of N/N
c
1 on a
double logarithmic scale. A fit of Eq. (1) (dotted line) is excellent all the way down to
( N/ N
c
1) = 4 10
4
. The fitting parameters are
and We note that is obtained even when we
use only the four samples closest to N
c
for the fitting.
Variable-range-hopping conduction in insulating samples and the critical exponents for
localization length and dielectric susceptibility
The temperature dependence of the conductivity of insulating samples is shown in
Fig. 4. The electrical conduction of doped semiconductors on the insulating side of the MIT
is often dominated by variable-range hopping (VRH) at low temperatures. The temperature
dependence of for VRH is written in the form of
(15)
where p = 1/2 for the excitation within a parabolic-shaped energy gap (Coulomb gap), and
p = 1/4 for a constant single-particle density of states around the Fermi level [13]. The
temperature dependence of contributes greatly to the temperature dependence of
298
Figure 6. T
0
determined by as a function of the dimensionless concentration
1 N/N
c
.
near N
c
because the factor T
0
/ T in the exponential term becomes very small, i.e., the tem-
perature dependencies of the prefactor and that of the exponential term become comparable.
Theoretically, is expected to depend on temperature as
(16)
but the value of r including the sign has not been derived yet for VRH conduction with both
p = 1/2 and p = 1/4.
As we have seen in Fig. 2(b), the temperature variation of the low-temperature conduc-
tivity of the
70
Ge:Ga samples within of N
c
is proportional to T
1/3
. Since both the
T
1/3
dependence of the conductivity and the VRH with p = 1/2 are results of the electron-
electron interaction in disordered systems, they can be expressed, in principle, in a unified
form. Moreover, the electronic transport in barely metallic samples and that in barely insu-
lating samples should be essentially the same at high temperatures so long as the inelastic
scattering length and the thermal diffusion length are smaller than, or at most comparable to
the correlation length or the localization length. So, the temperature dependence of conduc-
tivity at high temperatures should be the same on both sides of the transition. Such behavior
is confirmed experimentally in the present system, i.e., as seen in Fig. 2(b) the conductiv-
ity of samples very close to N
c
shows a T
1/3
dependence at irrespective of the
phase (metal or insulator) to which they belong at T = 0. Based on this consideration we fix
r = 1/3. Figure 5 shows with r = 1/3 for four samples ( N/ N
c
= 0.993, 0.994, 0.996,
and 0.998) as a function of (a) T
1/2
and (b) T
1/4
. All the data points lie on straight lines
with p = 1/2 in Fig. 5(a) while they curve downward with p = 1/4 in Fig. 5(b). This depen-
dence is maintained even when we change the values of r between r = 1/2 and 1/4. Thus
we conclude that the conductivity of all samples on the insulating side for N up to 0.998N
c
is described by the theory for the VRH conduction where the excitation occurs within the
Coulomb gap, i.e., Eq. (15) with p = 1/2.
Based on these findings, we evaluate the N dependence of T
0
in Eq. (15) with p = 1/2
and r = 1/3. Figure 6 shows T
0
as a function of 1 N/N
c
. The vertical and horizontal
error bars have been estimated based on the values of T
0
obtained with r = 1/2 and r =
1/4, and the values of l N/(l.858 10
17
cm
3
) and l N/(1.861 10
17
cm
3
), where
1.858 10
17
cm
3
is the highest concentration in the insulating phase and 1.861 10
l7
cm
3
is the lowest in the metallic phase, respectively. According to theory [13], T
0
in Eq. (15) is
299
Figure 7. (a) Logarithm of vs B
2
at constant temperatures for the sample having N = 1.840 10
17
cm
3
.
From top to bottom the temperatures are 0.095 K, 0.135 K, and 0.215 K, respectively. The solid lines represent
the best fits. (b) Slope vs T
3
/
2
for the same sample. The solid line represents the best fit.
given by
(17)
in SI units, where is the dielectric constant, and is the localization length. Here,
we should note that the condition T < T
0
is needed for the VRH theory to be valid, i.e., T
0
has to be evaluated only from the data obtained at temperatures low enough to satisfy the
above condition. This requirement is fulfilled in Fig. 6 for all the samples except for the one
with N = 0.998N
c
. Concerning this latter sample, we will include it for the determination of
(Fig. 8) and and (Fig. 9) but not for the calculation of the critical exponents.
Our next step is to separate T
0
into and in the framework of the theory of VRH con-
duction with and without a weak magnetic field [13]. For the magnetoconductance
is expressed as
(18)
where is the magnetic length in SI units. According to Eq. (18), the magnetic-
field variation of ln at T = const, is proportional to B
2
, i.e.,
(19)
and the slope C
2
(T) in the above equation is proportional to T
3/2
. In order to demonstrate
that these relations hold for our samples, we show for the N = 0.989N
c
sample
vs B
2
in Fig. 7(a) and C
2
(T) determined by least-square fitting of vs T
3 / 2
in
Fig. 7(b). Since Eq. (18) is equivalent to
(20)
is given by
(21)
In this way we have determined as a function of T
0
for nine samples (Fig. 8). The value
of is almost independent of T
0
, and if one assumes the form of one obtains a small
300
Figure 8.. Coefficient defined by Eq. (20) as a function of T
0
.
Figure 9. (a) Dielectric susceptibility arising from the impurities vs N
c
/N 1. (b) Localization length
vs 1 N/N
c
.
value of from least-square fitting. We determine and
from Eqs. (17) and (21), and show them in Fig. 9 as a function of 1 N/N
c
and N
c
/N 1,
respectively. Here, is the dielectric constant of the host Ge, and hence, is the dielectric
susceptibility of the Ga acceptors. We should note that both and are sufficiently larger
than the Bohr radius (8 nm for Ge) and respectively. According to the theories
of the MIT, both and diverge at N
c
as and
respectively. We find, however, both and do not show such simple dependencies
on N in the range shown in Fig. 9, and that there is a sharp change of both dependencies
at On both sides of the change in slope, the concentration dependence of
and are expressed well by the scaling formula as shown in Fig. 9. Theoretically, the
quantities should show the critical behavior when N is very close to N
c
. So and
may be concluded from the data in 0.99 < N/N
c
. However, the other region
(0.9 < N/N
c
< 0.99), where we obtain and is also very close
to N
c
in a conventional experimental sense.
As a possible origin for the change in slope, we refer to the effect of compensation.
301
Although our samples are nominally uncompensated, doping compensation of less than 0.1 %
may be present due to residual isotopes that become n-type impurities after NTD. In addition
to the doping compensation, the effect known as self compensation may play an important
role near N
c
[26]. It is empirically known that the doping compensation affects the value of
the critical exponents [11]. Rentzsch et al. studied VRH conduction of n-type NTD Ge in the
concentration range of 0.2 < N/N
C
< 0.91, and showed that T
0
vanishes as
with 3 for K = 38% and 54%, where K is the compensation ratio [27]. Since
[Eq. (17)], we find for our NTD
70
Ge:Ga samples = 3.5 0.8 for 0.99 < N/N
c
< 1 and
= 0.95 0.08 for 0.9 < N/N
c
< 0.99. Interestingly, = 3.5 0.8 agrees with 3
found for compensated samples. Moreover, we have recently proposed the possibility that
the conductivity critical exponent 1 in the same
70
Ge:Ga only wi t hi n the very vicinity
of N
c
(up to about +0.1 % of N
c
) [28]. An exponent of = 0.50 0.04, on the other hand,
holds for a wider region of N up to 1.4N
c
as we have seen in Fig. 3(b). Again, 1 near
N
c
may be viewed as the effect of compensation. Therefore, it may be possible that the
region of N around N
c
where v 1 and 1 changes its width as a function of the doping
compensation. In the limit of zero compensation, the part which is characterized by v 1
and 1 vanishes, i.e., we propose v = 0.33 0.03, = 0.620.05, and = 0.500.04 for
truly uncompensated systems and that the relation = v [3] is not satisfied. In compensated
systems, on the other hand, = v may hold as it does in the very vicinity of N
c
. However,
the preceding discussion needs to be proven experimentally in the future by using samples
whose compensation ratios are controlled precisely and systematically.
Metal-insulator transition in magnetic fields
Figure 10 shows the temperature dependence of the conductivity of the sample having
N = 2.004 10
17
cm
3
for several values of the magnetic induction B. Application of the
magnetic field decreases the conductivity and eventually drives the sample into the insulating
phase. This property can be understood in terms of the shrinkage of the wave function due to
the magnetic field. In strong magnetic fields and at low temperatures, i.e., when g
B
B k
B
T,
the conductivity shows another dependence [ 1 ]
(22)
where
(23)
One should note that Eqs. (9) and (22) are valid only in the limits of
(N, 0, 0) or It is for this reason that we have observed
a T
1/3
dependence rather than the dependence at B = 0 in Fig. 2 as the critical point
[ (N,0,0) = 0] is approached from the metallic side. However, Fig. 10 shows that the
dependence holds when B 0 even around the critical point. Hence, we use Eq. (22) to eval-
uate the zero-temperature conductivity (N, B, 0) in magnetic fields. Since m
B
is independent
of B, the conductivity for various values of B plotted against should appear as a group
of parallel lines. This is approximately the case as seen in Fig. 10 at low temperatures (e.g.,
T < 0.25 K).
The zero-temperature conductivity (N, B, 0) in various magnetic fields obtained by ex-
trapolation of ( N, B, T) to T = 0 based on Eq. (22) is shown in Fig. 11. Here, ( N, B, 0) is
plotted as a function of the normalized concentration:
(24)
Since the relation between N and (N,0,0) was established in Fig. 3(b) as (N,0,0) =
(0)[N/ N
C
(0) 1]
0. 50
where N
C
(0) = l.860 10
l7
cm
3
and (0) = 40 S/cm, n is equiv-
alent to N/N
c
(0) 1. Henceforth, we will use n instead of N because employing n reduces
302
Figure 10. Conductivity of the sample having N = 2.004 10
17
cm
3
as a function of T
1/2
at several magnetic
fields. The values of the magnetic induction from top to bottom in units of tesla are 0.0, 1.0, 2.0, 3.0, 4.0, 4.7,
5.0, 5.3, 5.6, 6.0, 7.0, and 8.0, respectively.
Figure 11. Zero-temperature conductivity ( N, B, 0) vs normalized concentration
N/N
c
(0) 1, where (N,0,0) is the zero-temperature conductivity and (0) is the prefactor both at B = 0.
From top to bottom the magnetic induction increases from 1 T to 8 T in steps of 1 T. The dashed curve at the
top is for B = 0. The solid curves represent fits of (N, B, 0) [n/n
c
(B) 1]
(B)
. For B 6 T, we assume
= 1.15.
303
Figure 12. Zero-temperature conductivity (N, B, 0) vs magnetic induction B. From bottom to top, the
normalized concentrations defined by Eq. (24) are 0.04, 0.09, and 0.22, respectively.
the scattering of the data caused by several experimental uncertainties, and it further helps us
concentrate on observing how ( N, B, 0) varies as B is increased. Similar evaluations of the
concentration have been used by various groups. In their approach, the ratio of the resistance
at 4.2 K to that at a room temperature is used to determine the concentration [5]. The dashed
curve in Fig. 1 1 is for B = 0, which merely expresses Eq. (24), and the solid curves represent
fits of
(25)
The exponent (B) increases from 0.5 with increasing B and reaches a value close to unity
at B 4 T. For example, = 1.03 0.03 at B = 4 T and = 1.09 0.05 at B = 5 T. When
B 6 T, three-parameter [ (B), n
c
(B), and (B)] fits no longer give reasonable results
because the number of samples available for the fit decreases with increasing B. Hence, we
give the solid curves for B 6 T assuming (B) = 1.15.
We show (N, B, 0) as a function of B in Fig. 12 for three different samples. When the
magnetic field is weak, i.e., the correction (N, B, 0) ( N, B, 0) ( N, 0, 0) due to B is
small compared with (N,0,0), the field dependence of (N, B, 0) looks consistent with
the prediction by the interaction theory [1],
(26)
In larger magnetic fields, (N,B,0) deviates from Eq. (26) and eventually vanishes at some
magnetic induction B
c
. For the samples in Fig. 12, we tuned the magnetic induction to the
MIT in a resolution of 0.1 T. We fit an equation similar to Eq. (25),
(27)
to the data close to the critical point. As a result we obtain ' = 1 . 1 0.1 for all of the three
samples. The value of ' depends on the choice of the magnetic-field range to be used for
the fitting, and this fact leads to the error of 0.1 in the determination of '. In Fig. 13 we
show that ' = 1 . 1 yields an excellent finite-temperature scaling [Eq. (3)]. Note that the data
both on the metallic side and on the insulating side are included in this scaling plot. Here
304
Figure 13. Finite-temperature scaling plot for the field-induced metal-insulator transition in the
70
Ge:Ga
sample having n = 0.09.
we employ B
c
obtained by fitting Eq. (27), x = 1/2 from the fact that dependence holds
in magnetic fields even around the critical point, and y = x/' = 0.45, i.e., none of them are
treated as a fitting parameter. Hence, Fig. 13 strongly supports ' = 1.1.
From the critical points n
c
(B) and B
c
(n), the phase diagram at T = 0 is constructed on
the (N, B) plane as shown in Fig. 14. Here, n
c
(B) for B 6 T shown by triangles are obtained
by assuming = 1.15. The vertical solid lines associated with the triangles represent the
range of values over which n
c
(B) have to exist, i.e., between the highest n in the insulating
phase and the lowest n in the metallic phase. Solid diamonds represent B
c
for the three
samples in which we have studied the magnetic-field-induced MIT. Estimations of B
c
for the
other samples are also shown by open boxes with error bars. The boundary between metallic
phase and insulating phase is expressed by a power-law relation:
(28)
From the eight data points denoted by the solid symbols, we obtain C
3
= (1.33 0.17)
10
3
and = 2.45 0.09 as shown by the dotted curve. The shift of N
c
in magnetic fields
was studied theoretically by Khmelnitskii and Larkin [29]. They considered a noninteracting
electron system starting from
(29)
where is the correlation length. They claimed that the argument of the function f
2
should
305
Figure 14. Phase diagram of
70
Ge:Ga at T = 0. The solid circles and the open triangles represent the critical
concentrations n
c
, and the solid diamonds and the open boxes represent the critical magnetic induction B
c
.
be a power of the magnetic flux through a region with dimension This means
(30)
where is the magnetic length, and hence, = 1 /2. In order to discuss the shift
of the MIT due to the magnetic field, they rewrote Eq. (30) as
(31)
based on the relation in zero magnetic field
(32)
Here, t is a measure of distance from the critical point in zero field, e.g.,
(33)
The zero point of the function gives the MIT, and the shift of the critical point for the MIT
equals
(34)
Thus, = l/(2v) results. In the present system, however, this relation does not hold, as long
as we assume = v [3]. Experimentally, we find = 2.5, while l/(2v) = l/(2) = 1 for
70
Ge:Ga at B = 0.
Based on the phase diagram we shall consider the relationship between the two critical
exponents: for the doping-induced MIT and ' for the magnetic-field-inducedMIT. Suppose
that a sample with normalized concentration n has a zero-temperature conductivity at B 0
and that [n/n
c
(B) 1] 1 or [1 B/B
c
(n)] 1. From Eqs. (25) and (27), we have two
expressions for
(35)
306
and
(36)
On the other hand, we have from Eq. (28)
(37)
in the limit of (1 B/B
C
) 1. This equation can be rewritten as
(38)
Using Eqs. (35), (36), and (38), we obtain
(39)
Since Eq. (39) has to hold for arbitrary B, the following relations
(40)
and
(41)
are derived.
In Fig. 15 we see how well Eq. (41) holds for the present system. We have already
shown in Fig. 12 that ' = 1. 1 0.1 is practically independent of n. Concerning the exponent
, however, its dependence on B has not been ruled out completely even for the highest B we
used in the experiments. This is mainly because the number of available data points at large B
is not sufficient for a precise determination of . In Fig. 15 the results of the doping-induced
MIT for B 4 T (solid symbols) and the magnetic-field-induced MIT for three different
samples (open symbols) are plotted. Here, we plot vs [n/n
c
(B) 1]/
with = 2.5 and = 1.1 for the doping-induced MIT, and vs [1 B/B
c
(n)]
for the magnetic-field-induced MIT. Figure 15 shows that the data points align exceptionally
well along a single line describing a single exponent = ' = 1.1.
We saw in Fig. 11 that apparently takes smaller values in B 3 T, which seemingly
contradicts the above consideration. We can understand this as follows. We find that the
critical exponent in zero magnetic field is 0.5 which is different from the values of in
magnetic fields. Hence, one should note whether the system under consideration belongs
to the magnetic-field regime or not. In systems where the MIT occurs, there are several
characteristic length scales: the correlation length, the thermal diffusion length, the inelastic
scattering length, the spin scattering length, the spin-orbit scattering length, etc. As for the
magnetic field, it is characterized by the magnetic length When is smaller
than the other length scales, the system is in the magnetic-field regime. As the correlation
length diverges at the holds near the critical point, no matter how weak the
magnetic field is. When the field is not sufficiently large, the magnetic-field regime where
we assume = 1.1 to hold, is restricted to a narrow region of concentration. Outside the
region, the system crosses over to the zero-field regime where = 0.5 is expected. This is
what is seen in Fig. 11.
CONCLUSION
We have measured the electrical conductivity of NTD
70
Ge:Ga to study the metal-
insulator transition, ruling out an ambiguity due to inhomogeneous distribution of impurities.
The critical exponent 0.5 in zero magnetic field for doped semiconductors without impu-
rity compensation has been confirmed. On the insulating side of the MIT, while the relation
307
Figure 15. Normalized zero-temperature conductivity (N, B, 0)/ (n) and (N, B, 0)/ [ (B)] as functions
of [1 B/B
c
(n)] and [n/n
c
(B) 1]
/
respectively, where = 2.5 and = 1.1. The solid line denotes a power-
law behavior with the exponent of 1.1. The open and solid symbols represent the results of the magnetic-field-
induced metal-insulator transition (MIT) in the range (1 B/B
C
) < 0.5 for three different samples (n = 0.04,
0.09, and 0.22) and the doping-induced MIT in constant magnetic fields (4, 5, 6, 7, and 8 T), respectively.
308
2v predicted by scaling theories [15] holds for 0.9 < N/N
C
< 1, the critical exponents for
localization length and impurity dielectric susceptibility change at N/N
C
0.99. The small
amount of doping compensation that is unavoidably present in our samples may be responsi-
ble for such a change in the exponents. We have also measured the conductivity in magnetic
fields up to B = 8 T in order to study the doping-induced MIT (in magnetic fields) and the
magnetic-field-induced MIT. For both of the MIT, the critical exponent of the conductivity
is 1. 1, which is different from the value 0.5 at B = 0. The change of the critical exponent
caused by the applied magnetic fields supports a picture in which varies depending on the
universality class to which the system belongs. The phase diagram has been determined in
magnetic fields for the
70
Ge:Ga system.
ACKNOWLEDGMENTS
This work has been performed in collaboration with K. M. Itoh, Y. Ootuka, and E. E.
Haller. We are thankful to T. Ohtsuki for fruitful discussions, and to S. Katsumoto, B. I.
Shklovskii, M. P. Sarachik, and J. C. Phillips for valuable comments. The author is supported
by Research Fellowship of Japan Society for the Promotion of Science for Young Scientists.
REFERENCES
1. Lee, P.A. and Ramakrishnan, T.V. (1985) Disordered electronic systems, Rev. Mod. Phys. 57, 287-337.
2. Belitz, D. and Kirkpatrick, T.R. (1994) The Anderson-Mott transition, Rev. Mod. Phys. 66, 261-380.
3. Wegner, F.J. (1976) Electrons in disordered systems. Scaling near the mobility edge, Z. Phys. B 25,
327-337; (1979) The mobility edge problem: continuous symmetry and a conjecture, ibid. 35, 207-210.
4. Rosenbaum, T.F. Milligan, R.F. Paalanen, M.A. Thomas, G.A. Bhatt, R.N. and Li n, W. (1983) Metal-
insulator transition in a doped semiconductor, Phys. Rev. B 27, 7509-7523.
5. Stupp, H. Hornung, M. Lakner, M. Madel, O. and Lhneysen, H.v. (1993) Possible solution of the
conduct i vi t y critical exponent puzzle for the metal-insulator transition in heavily doped uncompensated
semiconductors, Phys. Rev. Lett. 71, 2634-2637.
6. Waffenschmidt, S. Pfleiderer, C. and Lhneysen, H.v. (1999) Critical behavior of the conductivity of Si:P
at the metal-insulator transition under uniaxial stress, Phys. Rev. Lett. 83, 3005-3008.
7. Rosenbaum, T.F. Thomas, G.A. and Paalanen, M.A. (1994) Critical behavior of Si:P at the metal-insulator
transition, Phys. Rev. Lett. 72, 2121.
8. Zulehner, W. (1989) Czochralski growth of silicon, in Harbeke, G. and Schulz, M.J. (eds.) Semiconductor
Silicon: Material Science and Technology, Springer-Verlag, Berlin, pp. 2-23.
9. Haller, E.E. Palaio, N.P. Rodder, M. Hansen, W.L. and Kreysa, E. (1984) NTD germanium: a novel
material for low temperature bolometers, in Larrabee, R.D. (ed.) Neutron Transmutation Doping of Semi-
conductor Materials, Plenum, New York, pp. 2136.
10. Itoh, K.M. Haller, E.E. Beeman, J.W. Hansen, W.L. Emes, J. Reichertz, L.A. Kreysa, E. Shutt, T. Cum-
mings, A. Stockwell, W. Sadoulet, B. Muto, J. Farmer, J.W. and Ozhogin, V.I. (1996) Hopping conduction
and metal-insulator transition in isotopically enriched neutron-transmutation-doped
70
Ge:Ga, Phys. Rev.
Lett. 77, 4058-4061.
11. Watanabe, M. Ootuka, Y. Itoh, K.M. and Haller, E.E. (1998) Electrical properties of isotopically enriched
neutron-transmutation-doped
70
Ge:Ga near the metal-insulator transition, Phys. Rev. B 58, 9851-9857.
12. Watanabe, M. Itoh, K.M. Ootuka, Y. and Haller, E.E. (2000) Localization length and impurity dielectric
susceptibility in the critical regime of the metal-insulator transition in homogeneously doped p-type Ge,
Phys. Rev. B 62, R2255-R2258.
13. Shklovskii, B.I. and Efros, A.L. (1984) Electronic Properties of Doped Semiconductors, Springer-Verlag,
Berl i n.
14. Ionov, A.N. Shlimak I.S. and Matveev, M.N. (1983) An experimental determination of the critical expo-
nents at the metal-insulator transition, Solid State Commun. 47, 763-766.
15. Kawabata, A. (1984) Renormalization group theory of metal-insulator transition in doped silicon, J. Phys.
Soc. Jpn. 53, 318-323.
16. Hess, H.F. DeConde, K. Rosenbaum, T.F. and Thomas, G.A. (1982) Giant dielectric constants at the
approach to the insulator-metal transition, Phys. Rev. B 25, 5578-5580.
309
17. Katsumoto, S. (1990) Photo-induced metal-insulator transition in a semiconductor, in Kuchar, F, Hein-
rich, H. and Bauer, G. (eds.) Localization and Confinement of Electrons in Semiconductors, Springer-
Verlag, Berlin, pp. 117-126.
18. Chayes, J.T. Chayes, L. Fisher, D.S. and Spencer, T. (1986) Finite-size scaling and correlation lengths for
disordered systems, Phys. Rev. Lett. 57, 2999-3002.
19. Kirkpatrick, T.R. and Belitz, D. (1993) Logarithmic corrections to scaling near the metal-insulator tran-
sition, Phys. Rev. Lett. 70, 974-977.
20. Watanabe, M. Itoh, K.M. Ootuka, Y. and Haller, E.E. (1999) Metal-insulator transition of isotopically
enriched neutron-transmutation-doped
70
Ge:Ga in magnetic fields, Phys. Rev. B 62, 15817-15823.
21. Altshuler, B.L. and Aronov, A.G. (1983) Scaling theory of Andersons transition for interacting electrons,
JETP Lett. 37, 410-413.
22. Ohtsuki, T. and Kawarabayashi, T. (1997) Anomalous diffusion at the Anderson transitions, J. Phys. Soc.
Jpn. 66, 314-317.
23. Abrahams, E. Anderson, P.W. Licciardello, D.C. and Ramakrishnan, T.V. (1979) Scaling theory of local-
ization: absence of quantum diffusion in two dimensions, Phys. Rev. Lett. 42, 673-676.
24. Bernreuther, W. and Wegner, F.J. (1986) Four-loop-order function for two-dimensional nonlinear sigma
models, Phys. Rev. Lett. 57, 1383-1385.
25. Castner, T.G. Lee, N.K. Tan, H.S. Moberly, L. and Symko O. (1980) The low-frequency, low-temperature
dielectric behavior of n-type germanium below the insulator-metal transition, J. Low Temp. Phys. 38,
447-473.
26. Bhatt, R.N. and Rice, T.M. (1980) Clustering in the approach to the metal-insulator transition, Philos.
Mag. B 42, 859-872.
27. Rentzsch, R. Ionov, A.N. Reich, Ch. Mller, M. Sandow, B. Fozooni, P. Lea, M.J. Ginodman, V. and Shli-
mak, I. (1998) The scaling behaviour of the metal-insulator transition of isotopically engineered neutron-
transmutation doped germanium, Phys. Status Solidi B 205, 269-273.
28. Itoh, K.M. Watanabe, M. Ootuka, Y. and Haller, E.E. (1999) Scaling analysis of the low temperature
conductivity in neutron-transmutation-doped
70
Ge:Ga, Ann. Phys. (Leipzig) 8, 631-637.
29. Khmelnitskii, D.E. and Larkin, A.I. (1981) Mobility edge shift in external magnetic field, Solid State
Commun. 39, 1069-1070.
310
EXPERIMENTAL EVIDENCE FOR FERROELASTIC NANODOMAINS IN
HTSC CUPRATES AND RELATED OXIDES
J. JUNG
Department of Physics, University of Alberta, Edmonton, AB T6G 2J1,
Canada
INTRODUCTION
This paper reviews an experimental evidence for the presence of nanodomains in
HTSC perovskite cuprates and in CMR perovskite manganites. Perovskite materials
belong to a class of ferroelastic crystals [1]. A crystal is ferroelastic if it has two or more
stable orientational states in the absence of a mechanical stress (and an electric field), and
if it can be reproducibly transformed from one state to another of these states by the
application of mechanical stress[2, 3]. Ferroelasticity is a structure-dependent property
and is directly inferable from the crystal structure. In the ferroelastic state the crystal
symmetry is reduced to a subgroup of a higher symmetry class by a small distortion
which is a measure of the spontaneous strain. By the analogy to a ferroelectric case, as
the spontaneous polarization is reoriented by application of an electric field, so in the
ferroelastic case the spontaneous strain is reoriented by application of a mechanical
stress. Reported maximum values of an atomic displacement in a variety of ferroelastic
crystals range from 0.04 to 0.24 nm. Similarly to a ferroelectric crystal which exhibits a
hysteresis of a polarization versus an electric field, a stressstrain hysteresis is manifested
by a ferroelastic crystal. A transformation from one orientational state to the other in
ferroelastic crystals is in general accompanied by the formation of ferroelastic domains
which reflect concentration of strain due to static correlated displacements of atoms from
their periodic lattice sites. For example, transmission electron microscopy (TEM) studies
of a well know ferroelastic crystal of barium titanate (BaTiO
3
) revealed the presence of
domain structures of the size down to 4-6 nanometers [4]. The onset of ferroelasticity, as
a function of temperature or pressure is often accompanied by additional cooperative
phenomena such as ferroelectric or magnetic ordering. Correlation between the
ferroelastic and superconducting transition has been also observed in conventional
superconductors such as V
3
Si and N b
3
Sn [3].
The occurrence of ferroelasticity in perovskite oxides has raised the question
about the role of ferroelastic transitions in the origin of HTSC in cuprates and CMR in
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 311
manganites. This paper analyzes recent experimental evidence for ferroelastic transitions
and nanodomains in HTSC and partly in CMR. The evidence for the ferroelastic
transition in HTSC cuprates includes the measurements of stress-strain properties, helium
ion channeling, and neutron and Raman scattering. The experimental evidence for
ferroelastic nanodomains in HTSC cuprates has been provided by high resolution TEM,
STM, inelastic neutron scattering, and magnetic studies of critical currents.
FERROELASTICITY IN HTSC CUPRATES
One of the first experiments, that have indicated a ferroelastic nature of HTSC
cuprates, were based on stress-strain measurements of Bi
2
Sr
2
CaCu
2
O
x
whiskers of
T
c
=75K by Tritt et al. [5]. 2212 whiskers are flexible ribbon-shaped single crystals with
a c-axis perpendicular to the plane of the ribbon. The a-axis is
the growth direction of the whiskers, perpendicular to the b axis Both a and
b axes lie in the plane of the ribbon. The force (stress) was applied to the sample along
the a-axis, and the displacement (strain) was measured capacitively. The results show a
hysteresis in the stress-strain curves above 270K (Fig. 1), in addition to a maximum of
Youngs modulus at 270K (Fig. 2). This suggests the existence of a structural phase
transition. The data is characteristic of a displacive phase transformation and in
particular of a ferroelastic transformation with the hysteresis which resembles that for a
ferroelastic transition. The hysteresis and a temperature dependence of Youngs modulus
indicate a stress-related formation and dynamics (the hysteresis relaxation time is about
20 seconds) of ferroelastic domain walls.
Figure 1. Stress versus strain at various temperatures for a Bi
2
Sr
2
Ca Cu
2
O
x
whisker: (a) 50, 100 and
200K; (b) 255K; (c) 280K; (d) 290K; (e) 300K; (f) 315K. Hysteresis appears in the stress-strain
relationship above 270K. The direction of the hysteresis loop is indicated by the arrows. The stress and
strain relax towards the middle of the hysteresis loop with a characteristic time of approximately 20
seconds. The shape of the stress-strain curve at 270K gives a value for Youngs modulus of 20 GPa. The
sample has the fol l owi ng dimensions L=0.65mm, A=12m
2
(From Tritt et al [5])
312
Figure 2. The normalized modulus, Y(T)/Y(270K), as a function of temperature for a sample of
Bi
2
Sr
2
Ca Cu
2
O
x
(open circles) and NbSe
3
sample (filled triangles). The value of Y at 270 K for
Bi
2
Sr
2
Ca Cu
2
O
x
and NbSe
3
samples are Y=20 and 91 GPa, respectively (From Tritt et al [5]).
MeV helium ion channeling have been used by Sharma et al. [6] to probe lattice
distortions (static or dynamic) in as a function of temperature and oxygen
doping. Ion channeling provides a direct real space probe of extremely small
(sub-picometer) displacement of atoms in single crystalline materials. Sharma et al
measured the excess lattice distortion above the thermal background u
ex
as a function of
temperature between 30 and 300K (Fig. 3). The magnitude of u
ex
was extracted from the
measured FWHM (full-width-at-half-maximum) of the channeling angular scan. For an
optimally doped YBCO (T
C
=92K) u
ex
shows a drop from room temperature down to
about 230K (T
1
) by about . u
ex
exhibits a cusp (of the magnitude of about
) between 230K and about 140K (T
2
). Near T
c
, u
ex
drops further by about
with a decreasing temperature. The authors have stated that the changes in u
ex
are
indicative of some kind of phase transition in the system, which enhances the fluctuation
effects (static or dynamic). These changes could be attributed to ferroelastic phase
transitions at temperatures and T
3
=T
c
=90K. In the underdoped
YBCO (T
C
=65K) a drop of u
ex
between 300K and T
1
=240K and a cusp between 240K
and T
2
=150K is less pronounced, except a sudden drop of u
ex
at T
c
by about
Similar behavior has been observed in an underdoped YBCO of T
c
=45K. Taking into
account that oxygen deficiency (oxygen vacancies) in YBCO leads to nanoscopic
disorder, the ferroelastic transitions are diffused since they are sensitive to a disorder.
Kaldis et al. [7] observed a displacive structural transformation in the copper-oxygen
planes of YBa
2
Cu
3
O
x
at the underdoped overdoped phase separation line at x=6.95.
They measured, as a function of x, the dimpling in the CuO
2
planes using x-ray
absorption fine-structure spectroscopy (EXAFS) at 25K and the oxygen O(2,3) in-phase
mode Raman shifts. The data (Fig. 4) show for anomalously large static
displacements of the Cu(2) atoms off the O(2,3) layer and a gap in the distribution of the
O(2,3) in-phase Raman shifts. On doping from the underdoped side at x=6.805 up to
x=6.885, the Cu(2) position was found to move along c-axis by about off the
O(2,3) layer. From x=6.895 up to x=6.945 the dimpling of Cu(2) increases further by
313
Figure 3. Temperature dependence of excess atomic displacement above the thermal background u
ex
in
YBa
2
Cu
3
O
7x
. The top and middle figures: u
ex
for an underdoped YBa
2
Cu
3
O
7x
with T
c
of 45K and 65K,
respectively. The bottom figure: u
ex
of an optimally doped YBa
2
Cu
3
O
7x
with T
C
=92.5K. (From Sharma
et al [6]).
another to its maximum value of , almost entirely due to displacements of
the Cu(2) atoms off the Y (yttrium) layer. At the onset of overdoping between x=6.970
and x=6.984 both the Cu(2) and O(2,3) layers shift off the Y-layer reducing the dimpling
to . Earlier work by Conder et al. [8], based on standard refinements of neutron
diffraction patterns (measured at 5K) has also shown a reduction in the dimpling of the
CuO
2
planes by about at x=6.974. The negative direction of this discontinuity
has been attributed by Kaldis et al. to the structural transformation which develops first is
small domains of the crystal. Anomalous softening of the O(2,3) Raman shifts, which
starts at x larger than 6.90, has been found to correlate with the anomalously large
displacements of the Cu(2) atoms off the O(2,3) layer observed by EXAFS around
x=6.95 (Fig. 5). The authors stated that the increase of the dimpling in the CuO
2
planes
softens the Cu(2) O(2,3) bonds and thus may decrease the wave number of the O(2,3)
in-phase vibrations. The drop of the Raman shift by 5 cm
-1
within a narrow
314
Figure 4. Spacing between the Y-Cu(2) and Y-O(2,3) layers, as a function of oxygen concentration x in
YBa
2
Cu
3
O
x
. The spacing was determined from Y-EXAFS at 25K. Vertical arrows indicate the
magnitude of di mpl i ng in the Cu O
2
planes. x
opt
marks the optimum oxygen concentration (From Kaldis et
al [7]).
Figure 5. Raman shifts of the O(2,3) in-phase mode in YBa
2
Cu
3
O
x
for oxygen concentrations between
x=6.438 and 6.984. Dashed horizontal lines indicate the phase boundaries between coexisting phases, the
drawn out horizontal lines (6.95) indicate the miscibility gap in the overdoped regime. The thick drawn
boxes emphasize the sequence of phases occurring on doping (see Ref. 7 for more details) (From Kaldis et
al [7]).
concentration range of gave an evidence that the deformation of the CuO
2
planes is of the displacive type. On the other hand, an increase of the lattice constant in
the c-axis direction is associated with a decrease of the a-axis lattice constant. The
opposite behavior of these deformations suggest a martensitic (ferroelastic) nature of this
phase transformation.
Pulsed neutron diffraction experiments on YBa
2
Cu
3
O
6+z
(z=0.25, 0.45, 0.65, and
0.94) at 15K by Gutmann et al. [9] have revealed that the average copper Cu(2) - apical
oxygen O(4) bond length changes from down to on going from z=0.1
up to z=0.94. The difference in these two bond lengths is . Displacements of
315
Cu(2) in directions out of the a-b plane are also consistent with the presence of diffuse
scattering in the electron diffraction {Etheridge [10]} which originates from the
ferroelastic distortions.
FERROELASTIC NANODOMAINS IN HTSC CUPRATES
One of the first studies that have addressed the presence of nanoscale structural
perturbations (domains) in the copper-oxygen planes of cuprates, were the high resolution
electron microscopy studies of YBCO by Etheridge [10]. They revealed a weak diffuse
scattering in electron diffraction pattern of which arise from static
displacements of atoms from their periodic lattice sites. The local atomic displacements
partition the copper-oxygen planes into cells with dimensions comparable to a coherence
length in the a-b planes (about 2 nm). These features were reported to be intrinsic to the
orthorhombic form of and independent of the fabrication route.
The intensity distribution of the diffuse streaks in the electron diffraction pattern is
characteristic of scattering from atomic displacements and not of scattering from vacant
chain-oxygen sites. It has been determined that there must be a static component to the
atomic displacements, however the presence of a dynamic component to these
displacements has not been found. In order to give the observed dark line contrast along
Figure 6. A schematic illustrating the orientation of the irregular two-dimensional grid of walls of
nanoscopic cells (approx. 2nm in size) in the Cu O
2
planes of the YBa
2
Cu
3
O
6.95
(From Etheridge [10]).
316
axes perpendicular to c, the two-dimensional grids (cells) must be correlated along the
c-axis (Fig. 6). From selected-area electron diffraction patterns, static atomic
displacements have been identified with the geometry consistent with displacements of
the planar copper and apical oxygen atoms that are nearest neighbors along the
displacement directions, namely <2803> and <091>. High resolution images have
revealed that there are connected networks of oxygen-pyramidal planes (which link the a-
b planar oxygen atoms with the apical oxygen atom in the CuO
5
pyramid) at which the
charge density distribution is locally perturbed. These effectively partition each of the
copper-oxygen planes into cells with dimensions of about 2 nm in the a-b plane. These
studies have suggested that the a-b planes of buckle into the
network of slightly misaligned cells in a struggle to relieve internal stresses. One source
of internal stress is the mismatch between natural dimensions of the copper-oxygen
planes and the chain-oxygen and barium-oxygen planes (by natural dimensions it is
meant the dimensions that would minimize the energy of the plane if it were in isolation).
The planar copper atom in the CuO
5
pyramid is considered to be tightly bonded in the
center of the oxygen pyramid so that the pyramid might be treated as a rigid unit (Fig. 7).
As the structure distorts in a response to internal strain it might be expected to yield most
easily at the soft apical oxygen site, possibly causing the CuO
5
pyramid to tilt. These
features suggest the effect of ferroelastic transition.
Modulations along the copper-oxygen chains with a wavelength of approx. 1.5
nm have been inferred from scanning tunneling microscopy (STM) images of cleaved
YBCO crystals by Edwards et al. [11]. This length scale is comparable with the cell
dimensions which suggests that these two phenomena are coupled. This is not surprising,
given that it is the apical oxygen that links the CuO chains to the CuO
2
planes, so the
displacements of the planar copper atoms are coupled to the displacements of the apical
oxygen atoms two uni t cell apart. In turn, the displacements of the apical oxygen atom
distort locally the CuO chain, generating distortions in the chain with a spacing
comparable to the cell structure.
Figure 7. The oxygen-pyramidal planes (delineated by the bold lines) in the average YBa
2
Cu
3
O
6.95
structure (From Etheridge [10]).
317
The electrical transport and magnetic measurements of an optimally doped YBCO
have revealed that the copper-oxygen planes behave like nanogranular superconducting
systems (e.g. a conventional nanogranular niobium nitride (NbN) thin film). In a
nanogranular superconductor with a coherence length comparable to the grain size, the
temperature dependence of the critical current density J
c
is governed by the Josephson
tunnel junctions at low temperatures where the Ginzburg-Landau (GL) coherence length
is smaller that the grain size. At higher temperatures (close to T
c
) where the GL-
coherence length is larger than the grain size, the temperature dependence of J
c
is
determined by the suppression of the order parameter in the grain. At the crossover
Figure 8. (a), (b). (c):Schematic representation of nanostructures in the a-b planes of The
a- and b-axes are approx. 45 degrees relative to the domain walls, which are roughly along (110) directions
(for details see Ref.10). An optimally doped superconductor could have nanostructures as in (a), where the
critical current I
c
at low temperatures is governed by the interdomain Josephson junctions. The temperature
dependence of I
c
is therefore that of Ambegaokar-Baratoff (AB) at low temperatures and that of Ginzburg-
Landau (GL) above approx. 0.85T
c
as described by the Clems model (Ref.12) [(see solid triangles in (d)].
The dotted line in (d) shows I
c
(T) of a single Josephson junction (pure AB dependence). An underdoped
superconductor is shown in (b) where I
c
(T) is governed by the suppression of the order parameter in the
nanodomain. I
c
(T) is governed by the GL dependence see solid squares in (d). When a
superconductor is a mixture of optimally doped and underdoped phases, its nanostructure is described
schematically by a disordered chessboard shown in (c). In this case, I
c
(T) is expressed by open circles
shown in (e). The solid line in (e) is a superposition of two components; an optimally doped one with an
AB-like I
c
(T) (solid triangles) and an underdoped one with a GL-like I
c
(T) (dashed straight line). in (d)
and (e) is the Josephson coupling constant. See Reference 13 for more details.
318
temperature T*, the GL coherence length equals the grain size (for NbN nanogranular
film T* 0.85T
c
[12]). Consequently, at temperatures T T* , J
c
(T) is described by the
Ambegaokar-Baratoff theory and at T>T* by the Ginzburg-Landau one [12] with
J
c
(T) (T
c
T)
3/2
. It has been found that J
c
(T) in YBCO exhibits this type of behavior
[13] (Fig. 8). The measurements of J
c
(T) in the a-b planes of YBCO have been done on
c-axis oriented thin film ring-shaped samples. This geometry allows one to determine
J
c
(T) simply from the radial profile of the axial component of the self-field B
z
(r) of a
maximum (critical) persistent current I
c
. The relationship between I
c
and B
z
(r) is
provided by the Biot-Savart law. The magnitude of the Ginzburg-Landau coherence
length at T* has been used to estimate the size of the grain (cell) in the a-b planes to be
approximately 3-4 nm. The cell size of 3-4 nm includes the width of the cell walls. This
is in rough agreement with the TEM result of Etheridge [10].
The existence of large number of identical regions with diameters of about 3 nm
(which have a relatively high density of low energy quasi-particle states) have been
revealed in Bi
2
Sr
2
CaCu
2
(T
c
=87K) by Hudson et al. [14] using low-temperature
(4.2K) scanning tunneling spectroscopy (STS). The studies have been carried out with a
high resolution scanning tunneling microscope (STM) which simultaneously could
measure, with atomic resolution, both the surface topography and the local density of
states (LDOS) of a material. Following atomic resolution imaging of 130x130 nm
2
area,
the mapping of the differential conductance at zero-bias was performed on the same area.
This kind of a map is a measure of the LDOS of low energy quasi-particles, and in a
superconductor well below T
c
, is expected to show a very low differential conductance.
In contrast to this expectation, a typical zero-bias conductance map of BSCCO revealed a
large number of localized features (domains), with a diameter about 3 nm, which have a
relatively high zero-bias conductance (high LDOS near the Fermi energy) (Fig.9). These
features appear to be randomly distributed. The authors suggested that these LDOS
Figure 9. A 130nm square zero-bias conductance map (from Hudson et al [14]). Quasi-particle scattering
resonance (high LDOS) appear as bright regions approx. 3nm in diameter, owing to their higher zero-bias
conductance.
319
features are caused by quasi-particle scattering from atomic-scale defects or impurities,
with oxygen inhomogeneities being most likely, due to the absence of the magnetic field
effects. In view of the previously described evidence for ferroelastic nanodomains, the
local high conductance features could originate at randomly distributed oxygen deficient
nanodomains.
An indirect but an elegant experimental proof for the presence of static charged
nanodomains has been provided by inelastic neutron scattering measurements of the
temperature and composition dependencies of high energy Longitudinal Optical (LO)
phonons in YBa
2
Cu
3
O
6+x
{Petrov et al [15]}. The measurements of the composition
dependence of the inelastic neutron scattering intensity from YBa
2
Cu
3
O
6+x
with x=0.20,
0.35, 0.60, and 0.93 at 10K was performed for the LO mode along the inplane Cu-O bond
direction at energy transfers between 50-80 meV and momentum transfers Q along (100)
direction from (3, 0, 0) to (3.5, 0, 0) in the unit of the reciprocal lattice vector
(a*=1.63 ). This measurement detects only Longitudinal Optical (LO) phonons.
According to Pintschovius et al. [16] the underdoped (x=0) sample has a dispersionless
LO branch at 75 meV, while doping softens the zone-edge mode down to 55 meV, so one
might expect a continuous softening at the zone-edge as doping level is increased.
Surprisingly, the current experiments {Petrov et al. [15]} show that the LO branch is
always split in two, the high energy branch around 75 meV and the low energy branch
around 55 meV (Fig. 10). Instead of continuous softening the spectral weight is
transferred from the high-energy branch to the low-energy one as doping is increased.
The authors of this result, argued that this can be understood in terms of the two-phase
picture if one associates the high-energy and the low-energy branches with the nano-
phases which have low and high charge densities, respectively. When the charges are
segregated into nanoscopic domains, an increase of the doping level (x) does not change
the local charge density in the domains, but instead increases their total volume fraction,
which in turn causes the spectral weight transfer from the high-energy branch to the low-
energy one. The characteristic Q-dependence of these two LO branches indicates that the
size of the charged domains is nanoscopic. The size of the charged domain roughly
Figure 10. Composition dependence of the inelastic neutron scattering intensity from YBa
2
Cu
3
O
x
single
crystals with x=0.2. 0.35, 0.60, and 0.93 at 10K. (From Petrov et al [15]).
320
estimated from the shape of the dispersion-less portion of the phonon dispersion is about
2x1 nm {McQueeney et al. [17]}, which corresponds to the coherence length of the
localized phonon in the a-b planes. According to Phillips [18] the flatness of the LO
energy versus Q branch at 75 meV occurs because the phonons are localized in two-
dimensional nanodomains in the CuO
2
planes.
FERROELASTIC NANODOMAINS IN MANGANESE PEROVSKITES
Recent neutron scattering experiments {e.g. [19,20]} on manganites pointed out
the presence of microscopic (nanoscopic) phase segregation. The small-angle neutron-
scattering (SANS) studies of La
1x
Ca
x
MnO
3
(x=l/3), at temperatures larger than T
c
,
observed a short (weakly temperature dependent) ferromagnetic (FM) correlation length
which has been attributed to magnetic clusters 1-2 nm in diameter [19]. High-resolution
neutron diffraction and inelastic neutron scattering studies of Pr
0.7
Ca
0.3
MnO
3
[20]
revealed a two-phase segregation. One of the coexisting phases is a ferromagnetic (FM)
metal, while the other is an insulator. This type of electronic segregation cannot become
long-range, due to high Coulomb energy cost. Therefore a microscopically
inhomogeneous state develops, whereby FM clusters, 1-2 nm in diameter, are
interspersed into a charge localized matrix, the latter associated with much larger
lattice distortions than the former. The studies point to intra-granular strain as the main
driving force for phase segregation. These phase separation tendencies influence the
properties of ferromagnetic region by increasing charge fluctuations [21].
REFERENCES
1. Aizu, K. (1969) Possible species of ferroelastic crystals and of simultaneously ferroelectric and
ferroelastic crystals, J. Phys. Soc. Jpn 27, 387-396.
2. Abrahams, S.C., Bernstein, J.L. and Remeika, J.P. (1974) Ferroelastic transformation in samarium
orthoaluminate, Mat. Res. Bull. 9, 1613-1616.
3. Abrahams, S.C. ( 1971) Ferroelasticity, Mat. Res. Bull. 6, 881 -890.
4. Bur si l l , L.A. and Lin, P.J. (1984) Microdomains observed al the ferroelectric / paraelectric phase
transition of barium titanate, Nature 311, 550-552.
5. Tritt, T.M., Marone, M., Ehrlich, A.C., Skove, M.J., Gillespie, D.J., Jacobsen, R.L., Tessema, G.X.,
Franck, J.P., and Jung, J. (1992) Evidence in the elastic properties for a stress-related phase transition
in the high T
c
material: Bi
2
Sr
2
CaCu
2
O
x
, Phys. Rev. Lett. 68, 2531-2534.
6. Sharma, R.P., Ogale, S.B., Zhang, Z.H., Liu, J.R., Chu, W.K., Veal, B., Paulikas, A., Zheng, H. and
Venkatesan, T. (2000) Phase transitions in the incoherent lattice fluctuations in YBa
2
Cu
3
Nature
404, 736-739.
7. Kaldis, E., Rhler, J., Liarokapis, E., Poulakis, N., Conder, K and Loeffen, P.W. (1997) A displacive
structural transformation in the CuO
2
planes of YBa
2
Cu
3
O
x
at the underdoped overdoped phase
separation line, Phys. Rev. Lett. 79, 4894-4897.
8. Conder, K., Zech, D., Krger, Ch., Kaldis, E., Keller, H., Hewatt, A.W. and Jilek, E. (1994)
Indications for a phase separation in YBa
2
Cu
3
in E. Sigmund and K.A. Ml l er (eds), Phase
Separation in Cuprate Superconductors, Springer Verlag, Berlin, pp. 210-224.
9. Gutmann, M., Billinge, S.J.L., Brosha. E.L. and Kwei, G.H. (2000) Possible charge inhomogeneities in
the CuO
2
planes of YBa
2
Cu
3
O
6+x
(x= 0.25, 0.45, 0.65, 0.94) from pulsed neutron diffraction, Phys.
Rev. B 61, 11762-11769.
10. Etheridge, J. (1996) Structural perturbations at intervals of the coherence length in YBa
2
Cu
3
Philos. Mag. A 73, 643-668.
1 1 . Edwards, H.L., Barr, A.L., Markert, J.T. and deLozanne, A.L. (1994) Modulations in the CuO chain
layer of YBa
2
Cu
3
Charge density waves? Phys. Rev. Lett. 73, 1154-1157; Edwards, H.L., Derro,
D.J., Barr, A.L.,Market, J.T. and de Lozanne, A.L. (1995) Spatially varying energy gap in the CuO
chains of YBa
2
Cu
3
O
7x
detected by scanning t unnel i ng spectroscopy, Phys. Rev. Lett. 75, 1387-1390.
12. Clem, J.R., Bumble, B., Raider, S.I., Gallagher, W.J. and Shih, Y.C. (1987) Ambegaokar-Barratoff-
Ginzburg-Landau crossover effects on the critical current density of granular superconductors, Phys.
Rev B 35, 6637-6642.
321
13. Jung, J., Yan, H. , Darhmaoui, H., Abdelhadi, M., Boyce, B., Skinta, J., Lemberger, T. and Kwok, W-
K. (2000) Nanost ruct ures in hi gh temperature superconductors, Proc. of SPIE (i n press); Darhmaoui,
H. and Jung. J. (1996) Crossover effects in the temperature dependence of the critical current in
YBa
2
Cu
3
O
7-
, Phys. Rev. B 53, 14621-14630; (1998) Coherent Josephson nanostructures and the
dissipation of the persistent current in the a-b planes of YBa
2
Cu
3
thin films, Phys. Rev. B 57,
8009-8025; Yan, H., Jung, J., Darhmaoui, H., Ren, Z.F., Wang, J.H. and Kwok, W-K. (2000) Fast
vortex motion and filamentary phase separation in high-T
c
t hi n films, Phys. Rev. B 61, 11711- 11721.
14. Hudson, E.W.. Pan, S.H., Gupta, A.K., Ng, K.W. and Davis, J.C. (1999) Atomic scale quasi-particle
scattering resonances i n Bi
2
Sr
2
Ca Cu
2
Science 285, 88-91.
15. Petrov, Y., Egami, T., McQueeney, R.J., Yethiraj, M., Mook, H.A. and Dogan, F. (2000) Phonon
signature of charge inhomogeneity in high temperature superconductors YBa
2
Cu
3
O
6+x
, Cond.-Mat.
Preprint 0003414 / 25 Mar 2000.
16. Pintschovius, L., Pyka, N., Reichardt, W., Rumiantsev, A.Y., Mitrofanov, N.L, Ivanov, A.S., Collin,
G. and Bourges, P. ( 1991) Lattice dynamical studies of HTSC materials, Physica C 185-189, 156-161.
17. McQueeney, R.J., Petrov, Y., Egami, T., Yethiraj, M., Shirane, G. and Endoh, Y. (1999) Anomalous
dispersion of LO phonons in La
1.85
Sr
0.15
Cu O
4
at low temperatures, Phys. Rev. Lett. 82, 628-631.
18. Phi l l i ps, J.C. (2000) Zigzag filamentary theory of LO phonons in high temperature superconductors,
preprint.
19. De Teresa, J. M. , Ibarra, M.R., Algarabel, P.A., Ritter, C., Marquina, C., Blasco, J., Garcia, J.,
del Moral, Al. and Arnold, Z. (1997) Evidence for magnetic polarons in the magnetoresistive
perovskites. Nature 386, 256-259.
20. Radaelli, P.G., Ibberson, R.M., Argyriou, D.N., Casalta, H., Andersen, K.H., Cheong, S.W. and
Mitchell, J.F. (2000) Mesoscopic and microscopic phase segregation in manganese perovskites,
Cond-Mat. Preprint 0006190 / 12 Jun 2000.
21. Moreo, A., Yunoki , S. and Dagotto, E. (1999) Phase separation scenario for manganese oxides and
related materials, Science 283, 2034-2040.
322
ROLE OF Sr DOPANTS IN THE INHOMOGENEOUS GROUND STATE OF
La
2x
Sr
x
CuO
4
D. HASKEL
1
, E. A. STERN
2
and F. DOGAN
3
1
Experimental Facilities Division, Advanced Photon Source
Argonne National Laboratory, Argonne, IL 60439, USA
2
Department of Physics, University of Washington
Seattle, WA 98195-1560, USA
3
Department of Materials Science and Engineering, University of Washington
Seattle, WA 98195-2120, US A
INTRODUCTION
The role of dopants in high T
c
superconductors is widely seen as being limited to the
introduction of hole carriers into the CuO
2
planes of otherwise insulating parent compounds.
This simplified assumption is partly driven by the lack of information on the local atomic
and electronic structure around dopants. While experimental evidence favoring inhomoge-
neous charge distributions of the doped holes is still mounting, the role that dopants play
in determining this inhomogeneous ground state, if any, is still unclear. Since high T
c
su-
perconductors manifest strong carrier-lattice interactions, as evidenced, e.g., in the presence
of Jahn-Teller distorted CuO
6
octahedra, structural techniques can, in principle, provide in-
formation on the spatial distribution of doped charges through the structural response. The
x-ray absorption fine structure (XAFS) technique is particularly suited for elucidating such
response around dopants, as it is element specific; i.e., by tuning the x-ray energy to a charac-
teristic absorption threshold of a given dopant, it can determine the partial pair correlations
between the dopant and its neighbors. This is of paramount importance since the dopants
substitute at the crystal sites of the majority atoms and therefore techniques that sum over
all pair correlations are dominated by the correlations involving the majority, host, atoms. In
this paper we present evidence that the doped holes are spatially correlated to Sr dopants in
La
2x
Sr
x
CuO
4
. This is manifested as a unique structural response of the Sr-O(2) distance
across the x ~ 0.06 insulator-metal transition, while no such response is observed for the
La-O(2) distance. This result by itself proves that the doped charge density is not uniformly
distributed, and that the dopants play a role in determining the inhomogeneous charge state.
In addition to local effects around dopants, we show evidence for the appearance, with Sr
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 323
doping, of an inhomogeneous structural ground state at This ground state is char-
acterized by the presence of local domains with a larger orthorhombic distortion than the
long-range averaged distortion determined by crystallography.
EXPERIMENTAL
Samples of La
2x
Sr
x
CuO
4
were prepared by (i) conventional solid-state reaction tech-
niques starting from related oxides (Radaelli et al. [1], ) and (ii) by pre-
cipitation from ionic solution starting from related nitrates (Haskel et al. [2], ).
T
c
s were determined by magnetic susceptibility measurements and average crystal structures
were determined by neutron and x-ray powder diffraction (Refs. 1 and 2). The local environ-
ment around La/Sr atoms in La
2x
Sr
x
CuO
4
is quite complex and includes nine oxygen neigh-
bors at six different distances. In a powder XAFS experiment, an angular-average over all the
relative orientations of the electric field vector and the La/Sr-O bond directions is performed.
There is simply not enough information in a powder XAFS experiment to resolve all the dif-
ferent distances. By performing our experiments on c-axis magnetically aligned powders, we
can exploit the angular dependence of the XAFS signal in the anisotropic (layered) cuprates
to measure subsets of the local structure around the absorbing atom allowing a complete de-
termination of the local structure. (For excitation of a 1s core electron, i.e., K-edges, this
dependence is a one, with the angle between the electric field vector and the bond
orientation). Since we are particularly interested in structural ground state properties, we
present here results of measurements performed at T=10K using a closed cycle helium dis-
plex refrigerator. Orientation-dependent spectra were taken by rotating the oriented samples
relative to the electric field vector of the synchrotron radiation. Experiments were performed
at the National Synchrotron Light Source (NSLS, Brookhaven) and at the Advanced Photon
Source (APS, Argonne). La K-edge measurements were performed in transmission geometry,
which determines the energy-dependent absorption coefficient by measuring the attenuation
of the x-ray intensity after passing through the sample thickness. Sr K-edge measurements
were done in both transmission and fluorescence geometries, where the
latter determines the absorption coefficient by measuring the secondary radiation that accom-
panies the de-excitation of the absorbing atom. The reader is referred to Refs. 1-3 for further
experimental details.
RESULTS
Figure 1 shows local Sr-O(2) and La-O(2) apical distances as determined from c-axis po-
larized XAFS experiments at Sr and La K-edges as a function of Sr content in La
2x
Sr
x
CuO
4
at T=10K. The O(2) oxygens form the apices of CuO
6
octahedra and the La/Sr-O(2) api-
cal bonds nearly coincide with the crystallographic c-axis (slightly off c-axis due to
tilts of CuO
6
octahedra and correlated off-center displacements of La/Sr atoms in the La
2
O
2
planes. See Radaelli et al.[1] for a detailed description of the crystal structure). We have
previously reported evidence for the existence of a double site distribution for the apical O(2)
oxygens only near Sr atoms, a distribution that can be described as two Sr-O(2) distances
at 2.55 , respectively, with relative weights that strongly depend on x (Haskel
et al. [4]). These previous results were obtained on samples prepared by solid-state reaction.
The samples prepared by precipitation from solution did not show the broad (split) distri-
bution in the Sr-O(2) distance, indicating a more homogeneous Sr environment in the latter.
Since samples from both families have similar values of T
c
, it is implied that the double site
distribution observed previously must not be related to the mechanism of superconductivity.
The single Sr-O(2) distance found in the samples precipitated from solution, however,
324
Figure 1. Top: Local Sr-O(2) (centroids) and La-O(2) apical distances obtained from c-axis polarized Sr
and La XAFS at their K-edges at T=10K. Sr measurements were done on samples prepared by solid state
reaction (circles) and by precipitation from solution (squares, crosses) and at different synchrotron facilities
(APS, NSLS). Results from both transmission and fluorescence measurements are shown for comparison for
x = 0.04. Crosses are displaced in x for clarity. Bottom: weighted averages of Sr-O(2) and La-O(2) distances as
a function of x and their comparison with the results of crystallography.
agrees well with the weighted average of the two Sr-O(2) distances found previously. This is
shown in Fig. 1 where the centroid of the Sr-O(2) apical distribution is plotted for samples
made by the different methods (x = 0.07, 0.15). Good agreement is also obtained for Sr-O(2)
distances obtained by measuring XAFS at the Sr K-edge in transmission and fluorescence
geometries (x = 0.04) as well as between measurements performed at different synchrotron
facilities
Figure 1 shows that the local Sr-O(2) apical distance is significantly longer than the local
La-O(2) distance. Their weighted average, ([La O(2)]
*
(2 x) + [Sr O(2)]
*
x) / 2, agrees
with the values obtained by crystallography, as expected (lower panel of Fig. 1). The longer
Sr-O(2) local distance is readily explained by the stronger attraction of a negatively charged
O(2) ion to a trivalent La
+3
ion than to a divalent Sr
+2
. This has significant implications
for the local electronic structure of La
2x
Sr
x
CuO
4
, as discussed below. The local La-O(2)
distance is nearly independent of x, and therefore the average expansion of the La/Sr-O(2)
distance determined by crystallography as the Sr content is increased is due to the increase
in weight, with x, of the long Sr-O(2) distance. A more striking observation is the change in
slope that is observed only in the Sr-O(2) distance at x ~ 0.06 but not in the La-O(2) distance.
We recall that in La
2x
Sr
x
CuO
4
an insulator-metal (I-M) transition takes place at
The response of the Sr-O(2) distance to the delocalization of doped holes is indicative of a
spatial correlation between these doped holes and the Sr dopants. This observed change in
slope explains a similar (but opposite) change in slope measured by crystallography in the x
dependence of the Cu-O(2) apical distance (Figure 2 and Ref. [1]). O(2) apicals bridge Cu and
La/Sr atoms in a nearly collinear configuration. Since we know from Sr and La XAFS that
the Sr-Cu and La-Cu distances along the c-axis are nearly identical (within 0.01 , Haskel
et al. [5]) the Cu-O(2) apical distance measured by diffraction is a weighted average of a
majority long Cu-O(2) distance (near La) and a minority short Cu-O(2) distance (near Sr),
325
Figure 2. Cu-O(2) apical distance measured by crystallography (Radaelli et al.[1]). The change in slope at
the I-M transition reflects the similar (but opposite in sign) change in slope observed in the Sr-O(2) apical
distance.
with the latter showing a reversal in slope at the I-M transition.
We now turn to the experimental evidence showing the appearance of nanodomain struc-
ture in La
2x
Sr
x
CuO
4
for The structural phase diagram of La
2x
Sr
x
CuO
4
indicates
a low temperature orthorhombic (LTO) ground state with correlated CuO
6
octahedral tilts
of magnitude ~ 3 about crystallographic axis. Crystallography finds that as the Sr
content is increased the tilt angle of CuO
6
octahedra gradually decreases, becoming zero at
x ~ 0.21 (at T=10K), at which point the second-order phase transition to a macroscopically
tetragonal phase (HTT) is completed. Figure 3 shows the different tilt patterns of CuO
6
oc-
tahedra encountered in La-cuprates. The Sr-doped system only exhibits the LTO and HTT
phases, as determined by neutron diffraction. It is immediately obvious from Fig. 3 that by
measuring Cu-O(1) and Cu-O(2) distances (i.e. the Cu atoms nearest neighbors distances) it
is not possible to determine the direction nor the magnitude of the CuO
6
octahedra tilts. This
is because the tilts are nearly rigid and their magnitude too small to produce any measurable
changes in the near-neighbors distances within the sensitivity of XAFS, ~ 0.01. The O(2)
apical atoms lie in the La
2
O
2
planes and different tilt directions and/or magnitudes result in
significantly different La-O(2) planar radial distribution functions (Figure 3). Therefore by
measuring in-plane polarized XAFS at the La site, we are extremely sensitive to tilt direction
and magnitude. By measuring the local La-O(2) planar distribution of distances as a function
of x we can follow the structural phase transition from the LTO phase (3 distances) to the
HTT phase (a single distance).
Figure 4 shows the results of such measurements. It is immediately obvious that al-
though the local splitting (and therefore the local magnitude of tilt angle) initially decreases
up to x ~ 0.15, the local splitting deviates from the macroscopic value determined by crystal-
lography above this concentration. In particular, whereas the averaged tilt angle goes to zero
at the LTO HTT phase transition (x = 0.21), the local tilts do not vanish.
DISCUSSION
That the Sr-O(2) distance shows a large response to the delocalization of holes at the
I-M transition but the La-O(2) distance does not (Figure 1) is direct evidence that a spatial
correlation exists between the doped holes and the dopants that introduced them. This might
not be surprising, as at low Sr concentrations the dopants potential is poorly screened and it
is energetically favorable for a doped hole to remain in the vicinity of the Sr. At larger dopant
concentrations screening becomes more efficient but remains poor at very short distances, so
the doped holes, even if itinerant, are expected to have significant weight in the vicinity of
326
Figure 3. Top: The different CuO
6
octahedral tilt patterns encountered in La-cuprates together with the resul-
tant La-O(2) planar distances caused by such tilts. Crystallography finds that only the LTO and HTT phases
materialize in La
2x
Sr
x
CuO
4
While XAFS at the Cu sites is nearly insensitive to tilt direction and magnitude,
XAFS at the La sites is very sensitive to both as seen in the very different radial distribution functions of La-O(2)
distances. The splitting in La-O(2) distances is directly proportional to the magnitude of octahedral tilts.
327
Figure 4. La-O(2) planar and La-La planar distances as a function of x determined by in-plane (electric field
parallel to La
2
O
2
planes) polarized La K-edge XAFS in La
2x
Sr
x
CuO
4
at T=10K. Solid lines show the results
of crystallography (Radaelli et al.[1]), with the macroscopically averaged splitting of La-O(2) and La-La planar
distances going to zero at x = 0.21 (the splitting is proportional to the tilt angle of CuO
6
octahedra). The XAFS
results show that, although decreasing up to x = 0.15, the splitting remains present in the local structure (as do
the tilts) even in the nominal HTT phase.
the dopants. The structural response to the change in localization of the holes also shows that
there is a strong local interaction between the doped hole and the lattice. We believe this is the
first direct experimental evidence for a spatial correlation between dopants and doped holes,
although a similar conclusion was derived from the interpretation of NQR data by Hammel
et al.[6].
The local distortion in the Sr-O(2) apical distance has an important role in determining
the local electronic structure of La-cuprates. The electronic orbital character of doped holes
was determined by polarization-dependent x-ray absorption near edge structure (XANES)
measurements, which, at the absorption threshold, determine the density of unoccupied states
(holes) at the Fermi level, projected into the angular momentum states allowed by dipole se-
lection rules. O K-edge (1s) measurements showed that doped holes acquire more out-of-
plane, O 2p
z
orbital character, compared to their in-plane, O 2p
x,y
character, as doping is
increased (Chen et al. [7]). This phenomenon is hard to explain by band structure calculations
using the periodic average structure of crystallography and a rigid band model to account for
the changes in chemical potential with doping. However, the introduction of Sr
+2
dopants
results in a long, local, Sr-O(2) distance which raises the local energy of O 2p
z
orbitals to-
wards the Fermi level, allowing them to become more populated by the doped holes. The
raise in orbital energy is due to both a smaller Madelung energy contribution of the Sr
+2
ions
compared to La
+3
together with an increased overlap of O 2p
z
orbitals with ones
that results from the closer O(2)-Cu distance. The XANES measurement cannot determine
which oxygens in the structure are being populated by the doped holes, as it averages over all
oxygens. Our measurements indicate that the out-of-plane holes are being introduced in O(2)
apicals neighboring the Sr dopant atoms.
328
This result is crucial in understanding the c-axis transport properties of La
2x
Sr
x
CuO
4
,
as O 2p
z
orbitals provide connectivity along the c-axis. For example, the c-axis normal
state conductivity is insulating/semiconducting throughout the superconducting region of the
phase diagram, due to the small overlap of O 2p
z
orbitals (the normal state ab-plane con-
ductivity is metallic). As the Sr content increases, this overlap increases with a concomitant
increase in c-axis conductivity, to result in c-axis metallic conduction at x ~ 0.25. (At this
concentration of dopants, ca. 12% of La sites, a random distribution of Sr atoms results in at
least two Sr atoms per unit cell).
That the structure of La
2x
Sr
x
CuO
4
is composed of nanodomains for is readily
seen from Figure 4. For these concentrations the local structure is different from the aver-
age structure; specifically the local tilt angle of CuO
6
octahedra (and the related splitting of
La-O(2) and La-La planar distances) is larger than its macroscopically averaged value. In
particular, while the long range averaged tilt becomes zero at x ~ 0.21, the local tilt remains,
with a magnitude of comparable size to the one at x = 0.15. This can be explained by the
presence of structural disorder in the form of nanodomains. XAFS obtains local structural
information within a length scale that is determined by the photoelectron mean free path,
~ 10. Diffraction techniques average over much longer length scales. In order for the
macroscopically averaged tilt angle to be smaller than the local tilt the latter has to become
disordered over the length scale measured by diffraction. The tilts, therefore, are locally or-
dered within domains whose size is determined by the correlation length of this ordering.
These domains are at least as big as the XAFS length scale (~ 10 ) as no evidence for local
disorder is found in the XAFS measurements. We do not see the domain boundaries in our
measurements but we can put an upper limit, on the order of 50 , to the domain size. Larger
LTO domains would be visible in the diffraction measurements, but those were not observed.
The presence of nanodomains with size l, 50 , for is a direct consequence
of our measurements.
Most theories aiming at describing the transport properties of La
2x
Sr
x
CuO
4
limit the
Sr dopants involvement to their introduction of hole carriers into the CuO
2
planes. A much
more active role for the dopant sites is postulated by J. C. Phillips in the context of his zigzag
filamentary theory of high T
c
superconductors (Phillips [8]). In this theory dopants are res-
onant tunneling centers that serve the role of providing interlayer connectivity by creating a
percolative current path. A large carrier-lattice coupling at the dopant sites aids in establish-
ing this connectivity. The observation, in our measurements, of a structural response in the Sr
environment (and not in the La one) to hole-delocalization at the I-M transition is a signature
of the inhomogeneity of the wave function of the doped holes, which is peaked in the vicinity
of the dopant sites. This provides some support for the ideas in Ref. [8] which also depend
on the inhomogeneity of the carriers wave function and on a large hole-lattice coupling at
the dopant sites.
The appearance of structural disorder at could be related to the decrease of
T
c
in the overdoped regime; however, it has also been argued that the presence of local,
short-ranged orthorhombic domains with orthorhombic distortion larger than the macroscopic
orthorhombic distortion measured by crystallography favors superconductivity (Phillips [8]).
The role played by disorder introduced with dopants is currently being debated, particularly
how it influences the topology of inhomogeneous charge distributions in high T
c
cuprates
(Hasselmann et al.[9]).
Summary
Our experiments in La
2x
Sr
x
CuO
4
provide evidence for a spatial correlation between the
doped holes and the Sr dopants, in addition to the presence of a large hole-lattice coupling
only in the vicinity of the Sr dopants. The local structure deviates from the macroscopic, aver-
age, structure at indicating the appearance of structural disorder. This disorder is in
329
the form of nanodomains, within which the local tilts are ordered, but tilts become disordered
relative to each other in going from one domain to the other. These results indicate that Sr
dopants might have a much more significant role in determining normal and superconducting
properties of La
2x
Sr
x
CuO
4
than is typically assumed.
Acknowledgments
It is a pleasure to thank V. Polinger, J. C. Phillips, Y. Yacoby, A. R. Moodenbaugh,
D. G. Hinks, A. W. Mitchell, J. D. Jorgensen, F. Perez and M. Suenaga for valuable discus-
sions. This work supported by DOE grant no. DE-FG03-98ER45681 and DOE contract no.
W-31-109-Eng-38.
REFERENCES
1. Radaelli, P.G., Hinks, D. G., Mitchell, A.W., Hunter, B.A., Wagner, J.L., Dabrowski, B., Vander-
voort, K.G., Viswanathan, H.K. and Jorgensen, J.D. (1994) Structural and superconducting properties
of La
2x
CuO4 as a function of Sr content, Phys. Rev. B 49, 4163-4175.
2. Haskel, D., Stern, E.A., Dogan, F. and Moodenbaugh, A.R. (2000) XAFS study of the low-temperature
tetragonal phase of La
2x
Ba
x
CuO
4
: Disorder, stripes, and T
c
suppression at x = 0.125, Phys. Rev. B 61,
7055-7076.
3. Haskel, D., Stern, E.A., Hinks, D.G., Mitchell, A.W., Jorgensen, J.D. and Budnick, J.I. (1996) Dopant
and Temperature induced structural phase transitions in La
2x
Sr
x
CuO
4
, Phys. Rev. Lett. 76, 439-442.
4. Haskel, D., Polinger, V and Stern, E.A. (1999) Where do the doped holes go in La
2x
Sr
x
CuO
4
? A close
look by XAFS High Temperature Superconductivity, AIP Conference Proceedings 483, 241-246.
5. Haskel, D., Stern, E.A., Hinks, D.G., Mitchell, A.W. and Jorgensen, J.D. (1997) Altered Sr environment
in La
2x
Sr
x
CuO
4
, Phys. Rev. B. 56, R521-524.
6. Hammel, P.C., Statt, B.W., Martin, R.L., Chou, F.C., Johnston, D.C. and Cheong, S.W. (1998) Localized
holes in superconducting lanthanum cuprate, Phys. Rev. B. 57, R712-715.
7. Chen, C.T., Tjeng, L.H., Kwo, J., Kao, H.L., Rudolf, P., Sette, F. and Fleming, R.M. (1992) Out of plane
orbital character of intrinsic and doped holes in La
2x
Sr
x
CuO
4
, Phys. Rev. Lett. 68, 2543-2547.
8. Phillips, J.C. (1999) Dopant sites and structure in high T
c
layered cuprates, Philos. Mag. B. 79, 1477-
1498; Phillips, J.C. (1997) Filamentary microstructure and linear temperature dependence of normal state
transport in optimized high temperature superconductors, Proc. Natl. Acad. Sci. USA 94, 12771-12775;
Phillips, J.C. (1999) Is there an ideal phase diagram for high T
c
superconductors?, Philos. Mag. B. 79,
527-536
9. Hasselmann, N., Castro Neto, A.H., Morais Smith, C. and Dimashko, Y (1999) Striped Phase in the
Presence of Disorder and Lattice Potentials, Phys. Rev. Lett. 82, 2135-2138.
330
Universal Phase Diagrams and Ideal High Temperature Superconductors:
J. L. Wagner
1
, T. M. Clemens
1
, D. C. Mathew
1
, O. Chmaissem
2
,
B. Dabrowski
2
, J.D. Jorgensen
3
and D.G. Hinks
3
1
Physics Department, University of North Dakota, Grand Forks, ND 58201
2
Physics Department, Northern Illinois University, Dekalb, IL 60115
3
Science and Technology Center for Superconductivity, and Material
Science Division, Argonne National Laboratory, Argonne, IL 60439
INTRODUCTION
The series of compounds have the highest superconducting
temperatures of any known class of materials with T
c
s of 98, 128, and 135 K for the n= 1,
2, and 3 compounds, respectively. The first member of this series, (Hg-
1201)
1
, has been the most
extensively studied of the Hg-
cuprates and possesses the
simplest structure with T
c
determined by the variable
oxygen defect concentration
Many studies have
investigated the structure and
superconductivity over limited
doping regions,
2 - 6
It has been
reported to span the entire range
of superconducting behavior from
a T
c
~ 0 K in the underdoped
regime to T
c
~0 K in the
Figure 1. Crystal structure of The O3 overdoped regime. Therefore,
site is a partially occupied defect site (see text). Hg-1201 exhibits a remarkable
range of superconducting
properties spanning nearly 200 K,
making it an ideal compound to
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 331
investigate the interplay between doping, atomic structure and physical properties.
The structure of Hg-1201 is show in Fig.l and consists of [Cu-O
2
] planes separated by
a blocking layer composed of planes. Defects within the
layer determine the structural and superconducting properties of this compound. The exact
nature of all defects in the is still controversial. All studies have reported the O3
oxygen atom, shown in Fig. 1, as the predominant defect in the Hg-1201 structure, having
the greatest variation and effect on physical properties. Different groups have also reported
evidence of small amounts of Cu/CO
3
substitution on the Hg-site
1, 7-9
as well as a second
type of oxygen defect, O4, in the plane.
2, 5
However, the concentrations of these
latter defects are small with little variation, making it difficult to establish their role on
physical properties. In this paper, only the O3 defect is discussed.
Even when considering only the O3 defect, significant differences in the functional
and quantitative dependence of T
c
on exist in the literature. Simple parabolic doping
dependence consistent with the universal doping behavior proposed by Presland et al.
10
have been reported.
3, 4, 6
Other studies have shown more complex doping dependence
typified by trapezoidal-shaped curve exhibiting-a plateau across the region of optimal T
c
.
5
In this work, the doping curve consists of three linear regions corresponding to the
underdoped, optimally doped and overdoped regions. A wide variation in the optimal
doping level necessary for maximum T
c
, also exists. Reported values of range
from 0.06 to 0.18 (refs. 1 and 3, respectively), spanning nearly the entire range of doping
concentration for Hg-1201. These quantitative disagreements may be due to uncertainties
in values of obtained through structural refinements, however, the functional differences
in reported doping trends are not easily reconciled.
In the work by Chmaissem et al.,
5
a trapezoidal doping curve was found in which the
maximum T
c
persisted over a range of oxygen contents from to 0.16. Anomalies
in lattice parameters, cell volume and charge transfer were found upon crossing this
plateau. Possible electronic origins for these anomalies were proposed. In particular,
observed anomalies were considered to be a structural response to the proximity of the
Fermi level to features in the electronic structure as the doping is varied near maximum T
c
.
These structural distortions may involve charge ordering within the CuO
2
planes (i.e.,
stripe formation), a change in the defect chemistry, or possible oxygen ordering.
Regardless of origin, the ordering/distortions would then lead to a suppression in T
c
, giving
rise to a plateau rather than the expected parabolic behavior. If this picture is correct, then
the maximum T
c
for Hg-1201 could be considerably higher if these distortions could be
suppressed.
An alternative theoretical view has been proposed by Phillips
11,12
that explains this
novel behavior based upon the formation of filamentary, nanodomain structures in cuprate
materials. In this model, under appropriate conditions the oxygen defect atoms self-
organize to form percolative, metallic nanofilaments embedded in an insulating matrix.
Within these nanofilaments, the dopant-electron/phonon interaction is ~25 times larger
than that of a normal Fermi liquid, thus giving rise to maximum T
c
. This filamentary
structure persists over a range of doping concentrations, resulting in the observed plateau.
This model also predicts a continuos phase transition associated with the formation of
nano-filamentary structures at low defect concentrations, and a discontinuous phase
transition associated with an abrupt disappearance of this filamentary structure and
superconductivity at high concentrations.
Both models predict deviations from simple parabolic doping behavior though arising
from different origins. If present, electronically driven structural distortions will lead to a
suppression of T
c
from that possible in the undistorted structure. The filamentary theory
of Phillips is an enhancement in T
c
arising from the formation the filamentary structure. To
test these theories, it is essential to probe the resulting change in T
c
arising from samples of
differing degrees of disorder and/or structural distortion.
332
In this paper, results of neutron and x-ray diffraction structural experiments are
presented on two series of samples comprised of 34 individual samples, spanning the
complete range of superconducting properties of Hg-1201. Different annealing conditions
were used for these series to achieve underdoped and optimally doped samples. High
pressure, high temperature processing was used to fully overdoped samples to T
c
= 0 K.
Distinct differences in both the structure and superconductivity were found for samples in
the underdoped and optimally doped region. Analysis of these differences show that total
oxygen content alone is not sufficient to determine the superconducting and structural
properties of Hg-1201. Between samples of identical oxygen content, variations in T
c
as
large as 30 K were found.
EXPERIMENTAL
Samples for this work were from two, large master synthesis (~12 g each) that were
then processed after initial synthesis at low temperatures to adjust the doping level Two
series of samples, denoted A and B, were obtained and are comprised of 34 individual
samples spanning the entire underdoped and overdoped regime. Samples within each
series were from a single master synthesis. Results of powder neutron diffraction studies
on series A have been previously published.
5
Both master samples were synthesized in
sealed quartz tubes as described previously.
2
Two different methods of crossing the intermediate underdoped to optimally doped
region (50 K < T
c
< 98 K) were used. For series A, doping was varied by anneals at a
fixed temperature of 400 C in varying oxygen partial pressures, 10
-7
< P(O
2
) < 1 atm. For
series B, samples were annealed in flowing Ar gas (P(O
2
) < 10
-8
atm) at annealing
temperatures from 150 to 400 C. Both series of samples were cooled to room
temperature over time intervals of ~30 min. For strongly underdoped samples (T
c
< 50 K),
samples of series B were annealed in Ar gas at temperatures from 400 to 650 C.
Moderately overdoped samples (T
c
> 50 K) for both series were obtained by annealing
samples in 10 to 150 atm of oxygen at temperatures under 180 C for 100 hrs. Strongly
overdoped samples for series B were obtained from high-pressure anneals up to 6 Gpa at
800 C for 1 hr. No structural differences were found for overdoped samples prepared
using the moderate-pressure/long-time anneals with those made at high-pressure. Details
of the high-pressure work will be described elsewhere.
14
Fully overdoped samples (T
c
= 0
K) could only be obtained by high pressure techniques.
Neutron powder-diffraction data were collected on samples of series A on the Special
Environment Powder Diffractometer (SEPD) at Argonnes Intense Pulsed Neutron Source
(IPNS).
14
X-ray diffraction data were taken for samples of series B on a Philips PW1400
diffractometer with neutron-diffraction data also collected for selected samples in series B.
Structural parameters were refined using the General Structural Analysis Software (GSAS)
suite of programs.
15
T
c
s were determined from ac susceptibility measurements.
RESULTS AND DISCUSSION
Variation in lattice parameters and cell volume with T
c
Annealing experiments were conducted on two series of samples, A and B, utilizing
different conditions to produce moderately underdoped to optimally doped samples. Series
A samples were annealed at a fixed temperature with varying P(O
2
). Series B samples
were annealed at a fixed P(O
2
) with varying temperatures. Similar annealing conditions
333
were used to produce strongly underdoped and overdoped samples in both series. All
samples were found to be single-phase by neutron and x-ray diffraction after anneals.
Figure 2 (a) shows the a lattice parameter vs. T
c
for series A and B. Both series show
a general decrease in a with increasing doping. Similar values of a are found for strongly
underdoped samples (T
c
< 50 K) and overdoped samples, consistent with similar annealing
conditions used to obtain these samples in both series. However, differences in a are found
in the underdoped region as T
c
increases and are most pronounced in the region of optimal
T
c
. Across the region of maximum T
c
, values of a are seen to be discontinuous in series A,
indicating a range of values corresponding to the same maximum T
c
. That is, for the slow-
cooled samples in series A the maximum value of T
c
is nominally constant over a range of
doping levels, producing the plateau behavior in the doping curve as discussed earlier.
5
For samples in series B, this discontinuity is not found.
In both series, doping trends in the underdoped and overdoped regions are distinctly
different. For underdoped samples, a kink in the doping curve appears near 70 K, with
this feature being more pronounced in series A. For overdoped samples, a is found to
smoothly decrease with increased doping.
Figure 2 (b) shows the c lattice parameter vs. T
c
. Values of c are found to be different
for both series throughout the entire doping range, though general trends observed in these
series are similar. For series A, a discontinuity in c is observed at maximum T
c
, like that
observed in the a lattice parameter. No discontinuity is seen for series B. The differences
in doping trends between underdoped and overdoped samples are similar to that found in a-
lattice parameters. The origin for the different c lattice parameters throughout the entire
doping range is not understood, but likely arise from small variations in other structural
defects. This illustrates the importance of using a single, large master batch in which to
conduct these doping experiments.
The unit cell volume vs. T
c
for series A and B are plotted in Figure 2 (c). Differences
in cell volume are seen in the underdoped region with those of series A being on the order
of 0.2% less than those for series B. To within our experimental uncertainty, no difference
in cell volume is detected for overdoped samples.
The dependence of lattice parameters and unit cell volume vs T
c
show distinct
differences between the two series of samples, with the largest variations occurring at
optimal T
c
.
T
c
and volume dependence on
Determination of the excess oxygen is crucial to understand the physical properties
of Hg-1201. Neutron diffraction data were collected for samples in series A and
previously reported.
5
Rietveld analysis of the neutron data allowed determination of the
O3 occupancy with an estimated uncertainty of less than 0.01 oxygen atoms/unit cell
for all samples in series A. For samples in series B, neutron diffraction data were collected
on four samples and x-ray diffraction data were collected for all samples. Using values of
obtained from neutron data, values of were calculated for the remaining samples of
seried B using a linear interpolation of vs. a lattice parameters across the entire doping
range.
The T
c
vs. dependence for the 34 samples in series A and B are shown in Fig. 3. For
series A, the curve exhibits a plateau where T
c
is nominally constant and also corresponds
to maximum T
c
's for this series of samples.
334
Figure 2. (a) a-lattice parameter, (b) c-lattice parameter and (c)
unit cell volume V vs. T
c
for series and Estimated
standard deviations for all quantities are less than symbol size. Lines
are drawn as guide to the eye. An oval is drawn indicating the
samples near maximum T
c
.
For series B, the maximum T
c
occurs at a value of Series B also exhibits a
shoulder in the doping curve for to 0.10 over which T
c
gradually increases from
70 K to 80 K. It is of interest to note that the onset of the shoulder feature in series B
coincides with the beginning of the plateau feature for series A. Both series exhibit regions
where T
c
is nominally constant over a range of doping starting at The onset of
these features corresponds to the kinks seen in curves of lattice parameters and cell volume
vs. T
c
, in which these structural parameters change rapidly over a narrow range of T
c
.
In terms of possible structural distortions occurring as doping increases, it is useful to
plot the unit cell volume vs. as shown in Fig. 4. For samples in series B, the kink
present in the volume vs. T
c
curve is absent and the volume is found to monotonically
decrease with increasing This suggests that the nature of the oxygen defect remains
unchanged and structural distortions are absent as this series is doped throughout its entire
range. For series A, however, this is not the case.
335
Figure 3. T
c
vs. oxygen content for series and
Lines drawn are as guides to the eye. Error bars are shown for
series A only.
For samples in series A, the volume vs. curve has three distinct regions,
corresponding to the underdoped, optimally doped and overdoped regions. In the
underdoped region, volume for samples in series A are less than those found in series B for
a given The decrease in volume with increasing is more rapid and continues until
these samples achieve optimal T
c
at For overdoped samples, the volume
dependence on is the same in both series. For optimally doped samples (i.e. those that lie
on the plateau of the doping curve in Fig. 3) the volume is found to expand as is
increased across the region of maximum T
c
.
From the doping curves for series A and B, it is clear that alone is not sufficient to
uniquely determine T
c
nor the structure of Hg-1201. For example, samples with are
found to have differences in T
c
of 30 K between series A and B. Unlike series A, T
c
and
volume are found to be well defined for a given value of series B. It is therefore of
interest to look at the amount of charge transfer present in the samples of series A.
16
The
structural variables most sensitive to charge transfer are the z coordinates of the Ba and
apical oxygen O2 atoms. These atoms move in opposite directions along the c-axis as
charge is transferred from the [Hg-O
d
] layer to the CuO
2
plane. Therefore the parameter
most sensitive to the charge transfer on the CuO
2
plane is the structural quantity [z(Ba)
z(O2)].
Fig. 5 (a) shows T
c
plotted as a function of [z(Ba) z(O2)] for slow-cooled samples of
series A. In this plot the four samples of series A spanning the plateau in fig. 3 with
0.06 to 0.16 cluster about a single point. This indicates that all samples having the
maximum T
c
in series A have the same charge transfer, regardless of different oxygen
content.
336
Figure 4. Unit cell volume vs. oxygen content for series and
Error bars are shown for series A only.
It has been proposed that as oxygen is added in crossing the plateau for series A, the
charge remains constant as a result of a structural distortion occurring over this range of
oxygen compositions. This was inferred from the discontinuity in unit cell volume shown
in fig. 5 (b). This indicates that while T
c
and charge transfer remains constant over the
plateau, the volume increases as increases. This volume increase can be considered a
distortion, and is thought to be a structural response to the proximity of the Fermi level to
features in the electronic density of states. This distortion then produces the observed
plateau in the doping curve rather than the expected parabolic behavior. Whether this
distortion involves charge inhomogeneities within the CuO
2
planes (i.e., charge stripe
formation), different defect oxygen sites, oxygen ordering, or is purely of electronic origin
is not addressed. Whatever the cause, it was argued that higher T
c
s would be possible if
this distortion were suppressed.
The new data from series B provides a test of this hypothesis, since no such structural
anomalies are observed for these samples as is varied between 0.00 and 0.25. For
overdoped samples in both series, the volume dependence is the same. However,
differences exist for all underdoped and optimally doped samples between these series.
This suggests that all underdoped and optimally doped samples of series A are structurally
different from those of series B despite identical oxygen contents.
In terms of possible distortions present, samples of series B appear to possess less
distortion than those of series A based upon the volume vs. dependence shown in figure
4. However, T
c
values for series B samples are less than those found for series A. That is,
whatever distortions are present across the maximum-T
c
, plateau region in series A,
these distortions actually enhance rather than suppress T
c
.
337
Figure 5. (a) Tc and (b) unit cell volume vs. the structural parameter [z(Ba) z(O2)], a measure of the charge
transfer, for samples in series A.
Therefore, it appears that the trapezoidal doping dependence found for series A is not
readily explained by a structural distortion that competes with superconductivity in the
region of maximum T
c
. Whatever ordering is present, it must account for the constant
charge transfer and optimal T
c
observed for those samples across the plateau region. The
filamentary model of Philips
11,12
readily accounts for these observations, as well as
providing a mechanism capable of producing the high Tcs found in cuprate materials.
CONCLUSIONS
The structural and superconducting properties of have been investigated
over an extended range of doping Two series of samples were measured in
which oxygen content was varied using different annealing conditions. Both series were
338
found to have non-parabolic doping dependence as well as unusual structural phenomena
in the vicinity of maximum T
c
. Samples of identical oxygen content were found to have T
c
variations as large as 30 K. These differences in superconducting and structural properties
are evidence of oxygen ordering present in
This oxygen ordering was observed samples processed at different temperatures, and
was found to enhance superconductivity, consistent with the nano-filamentary models for
superconductivity recently proposed by Phillips. Work is in progress to investigate the
nature of the oxygen ordering present in and means by which to tune this
ordering.
ACKNOWLEDGEMENTS
This work was funded by the Office of Naval Research through grant #N000 14-99-1-0574 (JLW, TMC,
DCM), the US Department of Energy, Division of Basic Energy Science Materials Science, contract No.
W-31-109-ENG-38 (JDJ, DGH), and the National Science Foundations, Office of Science and Technology
Centers grant No. DMR 91-20000 (OC, BD).
REFERENCES
1. S.N. Putlin, E.V. Antipov, O. Chmaissem and M. Marezio, Nature (London) 346, 226 (1993)
2. J.L. Wagner, P.O. Radealli, D.G. Hinks, J.D. Jorgensen, J.F. Mitchell, B. Dabrowski, G.S. Knapp and
M.A. Beno, Physica C 210,447 (1993).
3. Q. Huang, J.W. Lynn, Q Xiong and C.W. Chu, Phys Rev. B52, 462 (1995).
4. A. Fukuoka, A Tokowa-Yamamoto, M. Itoh, R. Usami, S. Adachi, H. Yamauchi and K. Tanabe, Physica
C265, 13 (1996).
5. O. Chmaissem, J.D. Jorgensen, D.G. Hinks, J.L. Wagner, B. Dabrowski and J.F. Mitchell, Physics B
241-243, 805 (1998).
6. A.M. Balagurov, D.V. Sheptyakov, V.L. Aksenov, E.V. Antipov, S.N. Putlin, P.G. Radaelli and M.
Marezio, Phys. Rev. B 59, 7209 (1999).
7. D. Pelloquin, V. Hardy, A. Maignan and B. Raveau, Physica C 273, 205 (1997).
8. P. Bordet, F. Duc, S. LeFloch, JJ. Capponi, E. Alexandre, M. Rosa-Nunes, S. Putlin and E.V. Antipov,
Physica C 271, 189(1996).
9. S.M. Loureiro, E.T. Alexandre, E.V. Antipov, J.J. Capponi, S. de Brion, B. Souletie, J.L. Tholence, M.
Marezio, Q Huang and A. Santoro, Physica C 243, 1 (1995),
10. M.R. Presland, J.L. Tallon, R.G. Buckley, R.S. Liu and N.E. Flower, Physica C 176, 95 (1991).
11. J.C. Phillips, Phil. Mag. B 79, 527 (1999).
12. J.C. Phillips, Filamentary theory of cuprate superconductivity phase diagram and giant electron-phonon
interactions, Superconducting and Related Oxides: Physics and Nanoengineering III, SPIE Proc.,
3481, 87, 1998.
13. J.L. Wagner, J. D. Jorgensen, D.G. Hinks, T.M. Clemens and D.C. Mathew, in preparation.
14. J.D. Jorgensen, J. Faber, Jr., J.M. Carpenter, R.K. Crawford, J.R. Haumann, R.L. Hitterman, R. Kleb,
G.E. Ostrowski, F.J. Rotella and T.G. Worlton, J. Appl. Crystallogr. 22, 321 (1989).
15. A.C. Larson and R.B. VonDreele, General Structure Analysis System, LAUR 86-748 (1986-1990).
16. J.D. Jorgensen, O. Chmaissem, D.G. Hinks, A. Knizhnik, Y. Eckstein, H. Shaked, J.L. Wagner, B.
Dabrowski, S. Short J.F. Mitchell and J.P. Hodges, Novel Structural Phenomena at the Maximum Tc in
123 and Superconductors: Evidence for a Structural Response That Competes With
Superconductivity, Chemistry and Technology of High-Temperature Superconductors and Related
Advanced Materials, Kluwer Acad. Publ. B.V., Moscow, 1998.
339
This page intentionally left blank
COEXISTENCE OF SUPERCONDUCTIVITY AND WEAK FERROMAGNETISM
IN Eu
1.5
Ce
0.5
RuSr
2
Cu
2
O
10
I. FELNER
The Racah Institute of Physics, The Hebrew University,
Jerusalem, Israel 91904
INTRODUCTION
The general antagonism nature between superconductivity (SC) and long range
magnetic order is one of the fundamental problems of condensed matter physics and has
been studied experimentally and theoretically for almost four decades. In conventional s-
wave superconductors, local magnetic moments break up the spin singlet Cooper pairs and
hence strongly suppress SC, an effect known as pair-breaking. Therefore, a level of
magnetic impurity of only 1 %, can result in a complete loss of SC. In a limited class of
intermetallic systems, SC occurs even though magnetic ions with a local moment occupy
all of one specific crystallographic site, which is well isolated and de-coupled from the
conduction path. The study of this class of magnetic-superconductors was initiated by the
discovery of RRh
4
B
4
and RMo
6
S
8
compounds (R=rare-earth), and has been recently
revitalized by the discovery of the RNi
2
B
2
C system. In all three systems, both SC and
antiferromagnetic (AFM) order states coexist. The onset of SC takes place at Tc ~ 2-15 K,
while AFM order appears at lower temperatures (except for DyNi
2
B
2
C), thus, the ratio
T
N
/T
C
is ~ 0.1-0.5. Many of the high Tc superconducting systems (HTSC) contain
magnetic R ions as structural constituents, and are AFM ordered at low temperatures, e.g.
in GdBa
2
Cu
3
O
7
(T
c
= 92 K), and T
N
(Gd)=2.2 K. The R sublattice is electronically isolated
from the Cu-O planes, and has no adverse effect upon the superconducting state.
Much attention has been focused on a phase resembling the RBa
2
Cu
3
O
7
materials,
having the composition R
1.5
Ce
0.5
MSr
2
Cu
2
O
10
(M-2122, M= Nb, Ru or Ta) [1]. The
tetragonal M-2122 structure (space group I4/mmm) evolves from the RBa
2
Cu
3
O
7
structure
by inserting a fluorite type R
1.5
Ce
0.5
O
2
layer instead of the R layer in RBa
2
Cu
3
O
7
, thus
shifting alternate perovskite blocks by (a+b)/2. The M ions reside in the Cu(l) site and
only one distinct Cu site (corresponding to Cu(2) in RBa
2
Cu
3
O
7
) with fivefold pyramidal
coordination, exists. The hole doping of the Cu-O planes, which results in metallic
behavior and SC, can be optimized with appropriate variation of the R/Ce ratio[2]. SC
occurs for Ce contents of 0.5-0.7, where the highest Tc was obtained for Ce=0.5, the
concentration which has been studied here. The Nb-2122 and Ta-2122 materials are SC
with T
C
~28-30 K.
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 341
Coexisting of weak-ferromagnetism (W-FM) and superconductivity (SC) was
discovered about three years ago in R
1.5
Ce
0.5
RuSr
2
Cu
2
O
10
(R=Eu and Gd, Ru-2122)
layered cuprate systems [3-6], and more recently [7] in GdSr
2
RuCu
2
O
8
. In both systems,
the magnetic order does not vanish when SC sets in at Tc, and remains unchanged and
coexist with the SC state. The Ru-2122 materials (for R=Eu) display a magnetic transition
at T
M
= 125-180 K and bulk SC below T
C
= 32-50 K (T
M
>T
C
) depending on oxygen
concentration and sample preparation. Here Tw/Tc ~4, a trend which is contrary to that
observed in the intermetallic systems. The SC charge carriers originate from the CuO
2
planes and the W-FM state is confined to the Ru layers. SC survives because the Ru
moments probably align in the basal planes, which are practically de-coupled from the
CuO2 planes, so that there is no pair breaking. Scanning tunneling spectroscopy and muon
spin rotation experiments have demonstrated that all materials are microscopically uniform
with no evidence for spatial phase separation of superconducting and magnetic regions.
That is, both states coexist intrinsically on the microscopic scale [3].
In the Ru-2122 system, the W- FM state, as well as irreversibility phenomena, arise as
a result of an antisymmetric exchange coupling of the Dzyaloshinsky-Moriya (DM) type
[3] between neighboring Ru moments, induced by a local distortion that breaks the
tetragonal symmetry of the RuO
6
octahedra. Due to this DM interaction, the field causes
the adjacent spins to cant slightly out of their original direction and to align a component of
the moments with the direction of applied field. Below the irreversible temperature (T
irr,
),
which is defined as the merging temperature of the zero-field (ZFC) and field-cooled (PC)
curves, the Ru-Ru interactions begin to dominate, leading to reorientation of the Ru
moments, which leads to a peak in the magnetization curves.
The most remarkable magnetic properties of the Ru-2122-samples are: (a) The
negative magnetic moments in the ZFC branches measured at low applied fields (H) (b)
The ferromagnetic-like hysteresis loops and strong enhancement of coercive field which
appear only in the SC state at T< TC. (c) The so-called spontaneous vortex phase (SVP)
model, which permits magnetic vortices to be present in equilibrium without an external
field. The vortices in the SC planes, are caused by the internal field (higher than
H
c1
) of a few hundreds of G of the FM Ru sublattice. (d) No diamagnetic signal, in the FC
branch (the Meissner state (MS) the conventional signature of a bulk SC), has been
observed. The absence of the MS, may be a result of the SVP, and/or the high Ru magnetic
moment induced by the external field at T
M
, which masks this SC signature. On the other
hand, when Ru is partially replaced by Nb, the small positive contribution of the W-FM Ru
sublattice decreases the internal field and the MS is readily observed.
Hole (or carrier) density in the CuO
2
planes, or deviation of the formal Cu valence
from Cu
2+
, is a primary parameter which affects T
C
in most of the HTSC compounds. The
concentration of charge carriers (p), which may be measured as the effective [CuO
2
]p
charge, can be varied by removal or addition of oxygen. It is well accepted that addition of
hydrogen reduces p in a way very similar to that caused by removal of oxygen, and at high
hydrogen concentrations SC is suppressed and the materials a become semi-conducting and
magnetic. In Ru-2122 the hole doping in the CuO
2
planes can be achieved with appropriate
variation of the oxygen concentration which is obtained by annealing the as prepared
samples (ASP) under oxygen pressure up to 150 atm, or by loading the materials by
various amount of hydrogen. The effect of oxygen treatment is to shift both T
C
and T
M
up
to 49 and 225 K respectively (when annealed under 150 atm.). On the other hand, when
hydrogen atoms are loaded, they occupy interstitial sites and suppress SC and enhance the
W-FM properties of the Ru sublattice. This effect is reversible: namely, by depletion of
hydrogen, SC is restored and T
M
drops back to its original value.
This paper is organized as follows: (a) We first show that in the Ru-2122 system,
both TC and T
M
depend strongly on the oxygen concentration, (b) We present a systematic
study of the effect of hydrogen on both states. (c) We show experimental evidence for the
existence of the SVP by means of magneto-optical imaging. It is shown that below T
C
, at
342
zero applied field, magnetic flux is present in equilibrium in the sample and disappears
above T
C
. (d) The magneto-SC mixed (Ru,Nb)-2122 system is introduced, in which the MS
is readily observed.
EXPERIMENTAL DETAILS
Ceramic samples with nominal composition (Ru-2122) and
(Ru,Nb)-2122) were prepared by a solid state reaction
technique as described elsewhere [3-5] Parts of the ASP sample were re-heated for 24 h at
800 C under various pure oxygen pressures up to 150 atm. and will be identified
according to applied pressure. Determination of the absolute oxygen content in the ASP
material and in the samples annealed under oxygen pressures, is difficult because CeO
2
is
not completely reducible to a stoichiometric oxide when heated to high temperatures.
Figure 1. XAS spectra at the K edge of Ru of Ru-2122 and reference compounds.
Thermo-gravimetric measurements show that the materials are stable up to 600C and
no oxygen weight loss is detected. Above this temperature a small weight decrease begins
and our analysis indicates that the sample annealed at 150 oxygen atm. (150 atm.) contains
~4 at % more oxygen than the ASP sample. Hydrogen charging with several concentrations
(up to 0.28(1) at.% per formula unit) was accomplished by direct contact with high
pressure hydrogen gas at 300 C in a calibrated volume chamber. The hydrogen loaded
samples (Ru-2122H
X
) will be referred according to their hydrogen content. Removal of
the hydrogen was made by re-heating the hydrogenated sample for 10 hours at 250 C at
ambient pressure. Powder X-ray diffraction measurements confirmed the purity of the
compounds (~97%) and indicate within the instrumental accuracy, that all samples
studied, have the same lattice parameters as the ASP material, a=3.846(1) and
c=28.72(1). ZFC and FC dc magnetic measurements in the range of 5-300 K were
performed in a commercial (Quantum Design) super-conducting quantum interference
device (SQUID) magnetometer. Magneto-optical (MO) flux imaging studies have been
carried on polished ceramic sample, using an indicator (iron garnet film) with in-plane
anisotropy and high Faraday rotation angle, which was attached to the sample. The
magnetic flux density in the material (the bright regions) was deduced from the light
intensity depending on the Faraday rotation angle of the indicator, by using a polarizing
microscope.
343
Mossbauer spectroscopy performed at 90 and 300 K on
151
Eu show a single narrow
line with an isomer shift =0.69(2) and a quadrupole splitting of 1.84 mm/s, indicating that
the Eu ions are trivalent with a nonmagnetic J=0 ground state. This is in agreement with X-
ray-absorption spectroscopy (XAS) taken at L
III
edges of Eu and Ce that shows that Eu is
trivalent and Ce is tetravalent. The local electronic structure in several Ru based
compounds was studied by XAS at the K edge of Ru, and the results obtained at room
temperature are shown in Fig. 1. Since the valence of Gd
3+
, Sr
2+
and O
2-
are conclusive, a
straightforward valence counting for GdSr
2
RuO
6
and SrRuO
3
yields Ru, as Ru
+5
and Ru
+4
ions respectively. The similarity between the XAS spectra of Ru-2122 and GdSr2RuO
6
indicates clearly, that in Ru-2212 the Ru ions are in a pentavalent state. It is apparent that
SC in the M-2122 system exists only for pentavalent M ions such as Nb, Ta and Ru.
EXPERIMENTAL RESULTS
(I) The Effect of Oxygen on the SC and magnetic behavior of Ru-2122
The temperature dependence of the normalized resistance R(T) for the ASP and 22
atm. samples (measured at H=0 ) is shown in Fig. 2. The onset of the SC transition for the
ASP (T
C
= 32 (0.5) K) is shifted to 38 K. At high temperatures, a metallic behavior is
observed, and for the ASP sample, an applied field of 5 T smears the onset of SC and shifts
it to 28 K. The SC transition for the ASP sample is more easily seen in the derivative
dR/dT plotted in the inset. At T
C
= 32 K the derivative rises rapidly and does not fall to
zero until the percolation temperature around 19 K is obtained. This behavior is typical for
under-doped HTSC materials, where inhomogeneity in oxygen concentration causes a
Figure 2. Normalized resistivity measured at H=0 of the as prepared ASP Ru-2122 and the sample annealed
under 22 oxygen atmosphere. The inset shows the derivative of the resistivity for the ASP sample.
distribution in the T
C
values. This distribution is also reflected in the broad range of gap
values observed in our STS data, as shown below. The dependence of T
C
on the applied
oxygen pressure obtained from resistivity measurements, is presented in Fig. 3, exhibiting
a monotonic increase from 32 to 49 K.
R(T) curves of the 75 atm. sample measured at various applied fields are shown in
Fig. 4. In contrast to the ASP sample, an applied field of 5 T only smears the onset of SC at
46 K, but does not shift it to lower temperatures. The temperature dependence of dR/dT at
344
Figure 3. The effect of the annealing oxygen pressure on T
C
.
H=0 T, shows two peaks (Fig. 4 inset). This provides clear evidence for the two major SC
phases having T
C
at 32 and 46 K, where the latter is below the percolation threshold. This
is consistent with the STM data shown in Fig. 5.
The spatial distribution of the SC gap on the surface of the ASP and 75 atm.
samples are exhibited by the histograms in Fig. 5(a) and (b). The gaps were extracted by
fitting the Dynes function [8] to tunneling I-V curves acquired at various lateral tip
positions. In the ASP sample, the I-V curves show an ohmic gap-less structure, and the
values of range mainly between 3 and 5.5 meV. This broad distribution probably results
from spatial variations in hole-doping, and is consistent with the broad SC transition
exhibited in Fig. 2. In Fig. 5 (b), two peaks are clearly observed in the distribution,
showing the existence of two SC phases, (a trace of the higher T
C
phase is already present
Figure 4. Normalized resistivity measured at various magnetic applied field of Ru-2122 sample annealed
under 75 oxygen atmosphere. The inset shows the derivative of the resistivity curve at zero applied field.
345
in the ASP sample). The ratio between the large and small gap values is 1.45, in agreement
with the ratio between the two peaks extracted from Fig. 4 (inset). Note that the small-gap
phase (lower T
C
) is dominant, consistent with the fact that the higher T
C
phase should be
below the percolation threshold.
Figure 5. Histograms showing the spatial distribution of the SC energy gaps for the ASP sample (a) and the
sample annealed at 75 atm. oxygen (b). Inset: Two tunneling dI/dV vs. V curves obtained on the annealed
sample, one taken on a region of small gaps (dotted), the other on a large-gap region (solid).
Generally speaking, in Ru-2122, the temperature dependence of the magnetization
=M/H) at low applied fields is composed of three contributions: (a) a negative moment
below T
C
due to SC state, (b) a positive moment due to the paramagnetic effective
moment of Eu (or Gd) and (c) a contribution from the ferromagnetic-like behavior of the
Ru planes. ZFC and FC magnetic measurements for all samples were performed over a
broad range of applied magnetic fields, and typical M/H curves measured at 50 Oe., of the
Figure 6. ZFC and FC susceptibility curves measured at 50 Oe for The inset shows
the typical ferromagnetic hysteresis loop at 5 K.
346
ASP and 75 atm. are shown in Figs.6-7. At 50 Oe. the diamagnetic signal due to high
shielding fraction (SF) of the SC state dominates, and the net moment at low temperatures
is negative. The weak ferromagnetic component of Ru and the high paramagnetic effective
moment of Eu
3+
do not permit a quantitative determination of the SF from these curves. T
C
can also easily be determined from the deflection points in ZFC curves. The two curves
merge at T
irr
=92 and 137 K, respectively, indicating the effect of oxygen on the magnetic
behavior of the Ru sublattice. Note, that T
M
(Ru) is not at T
irr
. The curves do not lend
themselves to an easy determination of T
M
(Ru), and T
M
(Ru)=122 and 168 K, were
obtained directly from the temperature dependence of the saturation moment (Ms),
discussed below. Isothermal magnetization measurements at various temperatures indicate
that the Ru moment saturates around 5 kOe, therefore at this applied field, both, the
anomalies and the irreversibility are washed out [3].
Figure 7. ZFC and FC susceptibility curves for Ru-2122 sample annealed under 75 oxygen atmosphere
measured at 50 Oe.
Since, SC is confined to the CuO
2
planes; therefore, all the magnetic anomalies in
Figs. 6-7 are related to the Ru-O planes. The irreversibility at T
irr
arises as a result of an
antisymmetric exchange-coupling of the DM type between neighboring Ru moments,
induced by a local distortion that breaks the tetragonal symmetry of the RuO6 octahedra.
Due to this DM interaction, the external field causes the spins to cant slightly out of their
original direction and to align a component of the moments with the direction of H. At low
temperatures, the Ru-Ru and/or Eu(Gd)-Ru interactions begin to dominate, leading to
reorientation of the Ru moments, and the peak in the ZFC branch is observed. The exact
nature of the local structural distortions causing this reorientation is not presently known.
and we assume that the magnetic DM exchange coupling (as well as the SC behavior) in
the Ru-2122 system are extremely sensitive to oxygen concentration.
The isothermal magnetization (M(H)) curves measured at various temperatures can
also be divided into three parts. Fig. 8 shows the low field part for both the ASP and 75
atm materials measured at 5 K. The negative moments of the virgin curves increase up to
50 and 170 Oe, and the estimated SF (taking into account contributions from Ru and Eu
3+
)
are ~30% and 65%, for the ASP and 75 atm samples respectively. All M(H) curves below
T
M,
, are strongly dependent on the field up to 4-5 kOe, until a common slope is reached
(Fig. 9 inset). M(H) can be described as: where M
s
(the saturation
moment) corresponds to the W-FM contribution of the Ru sublattice, and is the linear
paramagnetic contribution of Eu and Cu. M
s
decreases with increasing T, and becomes
zero at T
M
(Ru)=122(2) and 168(2) K for the ASP and the 75 atm. samples respectively
347
Figure 8. The isothermal magnetization as a function of the applied field in the low fields limit for the ASP
compound and the sample annealed at 75 atm.
(see Fig. 11). For the ASP sample the M
s
values for the APS sample at 5 K) are
larger than for the 75 atm. material. These values, are smaller than the fully saturated
moment expected for low-spin state of Ru
5+
, i.e., for g=2 and S=0.5. This means
that in the ordered state, some canting on adjacent Ru spins occurs, and the saturation
moments at low temperatures are not the full moments of the Ru
5+
ions.
In the intermediate applied field region, a ferromagnetic-like hysteresis loop is opened
(Fig. 6 inset) from which the two characteristic parameters: the coercive field (H
C
) and the
remanent moment (M
rem
) can be deduced. Fig. 9 shows, that for both materials, H
C
disappears around T
C
, (and not at T
irr
and/or at T
M
) which strengthen our experimental
evidence for the spontaneous vortex phase, described below. M
rem
(T) for both samples
disappears at T
irr
(not shown). Below 20 K, the M
rem
values for the ASP material are a bit
higher than those of the 75 atm sample, (0.37 and 0.26 at 5 K), but for 20 <T the M
rem
values for both samples are similar.
For the 150 atm. sample we obtain T
C
= 49 K (Fig. 3). The ZFC and FC branches
measured at 50 Oe merge at T
irr
=178 K, and we obtain T
M
(Ru)=225 K. The STM results
Figure 9. The temperature dependence of the coercive field (H
C
) for the ASP and the sample annealed at 75
atm. The inset shows the M(H) curve up to 50 kOe for the ASP material.
348
for this material are similar to the 75 atm. sample (Fig. 5b), but exhibit an increase of the
relative abundance of the large gaps.
Mossbauer effect studies (ME) on
57
Fe doped samples has been proved to be a
powerful tool in the determination of the magnetic nature of the Fe site location. When the
ions of this site become magnetically ordered, they produce an exchange field on the Fe
ions residing in this site. The Fe nuclei experience a magnetic hyperfine field leading to a
sextet in the observed ME spectra. As the temperature is raised, the magnetic splitting
decreases and disappears at T
M
.
Figure 10. Mossbauer spectra of 0.5 %57Fe doped in Ru-2122 (R=Gd) below and above T
M
(Ru).
The main effect to be seen in Fig. 10 is that the ME spectra of Ru-2122 (R=Gd),
consist of one site only, below and above T
M
(Ru)=175 K. A least square fit to the
spectrum at 180 K yields an isomer shift (IS) of 0.30(1) mm/s (relative to Fe metal) with a
line width=0.35 mm/s, and a quadrupole splitting of 1.00(1) mm/s. We
attribute this doublet to paramagnetic Fe ions in the Ru site. This interpretation is
consistent with: (i) the similarity of the chemical properties of Fe and Ru (Ru resides
below Fe in the periodic table), (ii) and with the fact that in most HTSC materials the Fe
atoms are found to occupy predominantly the Cu(l) sites which is equivalent to the Ru site
in the Ru-2122. At low temperatures, all spectra display magnetic hyperfine splitting,
which is a clear evidence for long-range magnetic ordering. The fitting parameters of the
single sextet obtained at 4.1 K are: IS= 0.40(l)mm/s, H
eff
(0)=467(3) kOe and an effective
quadrupole splitting value of Using the relation:
we obtained for the Ru site a hyperfine field orientation, relative
to the tetragonal symmetry c axis. As the temperature is raised, H
eff
decreases and
disappears completely at T
M
(Ru)= 1875(5) K. H
eff
values obtained at 110, 130, 150 K
and 160 K are: 399(3), 358(5), 312(2) and 279(5) kOe. The variation of the normalized
H
eff
(T)/H
eff
(0) values, as a function of the reduced temperature is exhibited in Fig. 11.
It was shown theoretically [3], that for all Fe-doped YBa
2
Cu
3
O
7
(as well as M-2122)
materials, when Fe reflects the magnetic behavior of Cu(2) sites, the normalized H
eff
values
349
Figure 11. The temperature dependence of (I) the normalized hyperfine field acting on Fe in the Ru sites for
R=Gd (left scale) and (II) the reduced magnetic moment of Ru deduced from magnetic measurements for
R=Eu (right scale). The dashed line is the universal theoretical curve for Fe-Cu exchange strength. Note the
deviation of the experimental data from the universal curve.
fall on one universal curve, regardless of Fe or oxygen concentration and whether Y is
replaced by Pr. The model assumes that the temperature dependence of magnetization of
Cu(2) and Fe
3+
as a probe, behaves like a spin 1/2 and spin 5/2 systems and that the Fe-Cu
exchange is only 26% of the Cu-Cu exchange strength. The dashed line in Fig. 11 is the
universal theoretical curve calculated in this way. The deviation of the experimental data
from this universal curve is our supporting evidence that the magnetic sextet in Fig. 10 is
due to Fe in the Ru site. Moreover, the fact that both the reduced magnetic moment
obtained directly from the M(H) curves for R=Eu , and the data obtained from ME for
R=Gd, lie on the same curve, indicates clearly that the Ru-Fe and Ru-Ru exchange strength
are quite similar.
(II) The effect of hydrogen on the SC and magnetic behavior of Ru-2122
We have demonstrated [5] that Ru-2122H
0.35
, the effect of hydrogen is to suppress SC
and to enhance the W-FM properties of the Ru sublattice (T
M
increased to 225 K). The
hydrogen atoms reside in interstitial sites, and their effect is reversible. Namely, by
depletion of hydrogen, SC is restored and T
M
drops back to 122 K. This is in contrast to the
behavior observed in all other HTSC materials, in which charging and/or depletion of
hydrogen is irreversible and destructive. Two scenarios that could lead to this phenomenon
are: (a) in addition to the change of p in the Cu-O
2
planes, there is a transfer of electrons
from hydrogen to the Ru 4d sub-bands, resulting in an increase in the Ru moments, and
hence to enhance the magnetic parameters; (b) the enhancement arises from a change of the
anti-symmetric exchange coupling of the DM type between the adjacent Ru moments,
discussed above, which causes the spins to cant out of their original direction to a larger
angle. The question arises whether T
C
decreases monotonically with hydrogen
concentration until SC is suppressed completely, or whether small hydrogen quantities
induce phase separation. In this case, T
C
may remain unchanged (32 K), but the fraction of
the SC phase reduces with increasing the hydrogen concentration, thus smearing the
transition, until SC is globally suppressed.
To address this question, the magnetic measurements of the hydrogen charged Ru-
2122H
X
materials are shown in Figs 12-14. Fig. 12 shows ZFC and FC branches of Ru-
2122H
0.07
(H=0.07at %), measured at 50 Oe. and for the sake of clarity, we display again
350
Figure 12. ZFC and FC susceptibility curves for ASP and Ru-2122H0.07.
the data of the ASP sample. Here again, the magnetic properties due to the Ru are all
enhanced, as compared to the ASP material. The ZFC and FC values are much higher, and
both T
irr
and T
M
(Ru) are shifted to 167(2) and 225(2) K respectively. Similar T
irr
and T
M
values have been also obtained for samples with H>0.07, whereas for the Ru-2122H
0.03
these values are close to those of the APS sample. In contrast to the ASP sample, the ZFC
signal has a large contribution from the positive moments due to the W-FM state, which
mask the negative contribution due to the SC state. Hydrogen induces high porosity in
these materials, affecting the macroscopic transport measurements, therefore, we could not
extract the SC state properties, neither directly from the magnetic ZFC curves, nor from
four point resistivity measurements. These features were studied by the STM technique.
The picture emerges from the STM measurements [6] is that hydrogen doping indeed leads
to phase separation. Even at very low doping (Ru-2122H
0.03
), insulating regions start to
form. As doping is increased, the density and size of the insulating regions increase, until
they coalesce and the sample becomes globally insulating.
Typical hysteresis loops opened below 5 kOe and are shown in Fig. 13. For Ru-
2122H
0.07
both M
s
, and M
rem
values are higher than for the ASP
Figure 13. The low range of the hystersis loops at 5 K for the ASP material and for Ru-2122H0.07
351
Figure 14. ZFC susceptibility curves for various hydrogen loaded samples.
sample. In the limit of uncertainty the H
C
values do not change much. Fig. 14 presents the
ZFC curves obtained for all hydrogen loaded samples studied. For the ASP and the Ru-
2122H
0.03
samples, the peak is around 80 K, and for the samples with H>0.14 at. the peaks
are shifted to about 160 K. For the intermediate hydrogen concentration the (Ru-
2122H
0.07
), a superposition of both peaks is observed which leads to a somewhat flat curve.
Regeneration was made on the sample with the Ru-2122H
0.14
and the magnetization
measurements (carried out on powdered sample) prove that: (i) SC is restored, (ii) the
peak in ZFC curve is shifted back to 80 K and (iii) Tirr ~92 K. Thus, all the enhanced
parameters of the Ru-2122H
X
presented in Figs. (9-11), are reduced to the original values
of the ASP compound [6].
Figure 15. Magneto-optic images of Eu-2122 measured at 10 and 90 K and H=0 and 50 Oe. Note the bright
area below T
C
at H=0.
352
(III) Spontaneous Vortex Phase in Ru-2122
Fig. 15 shows MO images for Ru-2122 measured at 10 K (below T
C
) and 90 K at
external fields of zero and 50 Oe. The triangular dark regions in the left and right hand
sides of the pictures are due to magnetic domains of the indicator with reversed
magnetization direction, and are not related to the flux density of the Ru-2122 sample. The
bright area at 10 K and H=0 observed in the central part of the picture, indicates clearly
that magnetic flux (vortices ?) is present in the sample in equilibrium. The internal field
(of a few hundreds G) induces these flux lines in the SC planes (the mixed state), without
an external field. At 90 K (T
C
<T< T
M
) no flux density is observed (dark area). On the
other hand, under H= 50 G, the flux density persists up to T
M
, and the images obtained at
10 and 90 K are quite similar. Since the internal field associated with the is greater
than H
C1
but smaller that H
C2
a SVP is obtained. This observation demonstrate
unambiguously its appearance as predicted in Ref. [4]. A detailed study of MO images on
samples with various oxygen concentrations, as well as on the mixed (Ru,Nb)-2122
system, is now under investigation.
(IV) The Meissner-State (MS) in (Ru,Nb)-2122
ZFC and FC magnetic measurements for the (Ru,Nb)-2122 sample annealed under 50
atm. of oxygen, were performed over a broad range of applied magnetic fields, and typical
M/H vs T curves measured at 10 Oe., and the isotherm magnetization measured at various
temperatures, are shown in Figs.16-18. The onset of the SC transition is shifted to T
C
=41
K (confirmed also by four point resistivity measurements), and the ZFC and FC branches
merge at T
irr
=142 K. All M (H) curves below T
M
, are strongly dependent on the field until
a common slope is reached (Fig. 17). At 5 K, a value which is much smaller
than obtained for the ASP sample, M
s
decreases with T, and becomes zero at
T
M
(Ru)=160(2). Thus, reduction of the Ru content does not change the typical coexistence
of SC and W-FM found in the Ru-2122 system. (The enhancement of T
C
and T
M
is
consistent with Fig. 3 and related to the extra oxygen content in this material). The clear
peak at T
C
in both the ZFC and FC branches in (Ru,Nb)-2122 (Fig. 16), and the negative
signal in the ZFC curve below T
C
are quite evident. On the other hand, in contrast to Figs.
6-7, the typical expulsion of magnetic flux lines at T
C
(the MS) in the FC curve, as well as
the flatness at low temperatures, are readily observed. The small positive contribution of
the W-FM Ru sublattice to the total magnetic moment decreases the internal field, and as a
Figure 16. ZFC and FC susceptibility curves measured at 50 Oe. for Ru,Nb -2122. Note the flux expulsion at
353
T
C
.
Figure 17. Isothermal magnetization curves of (Ru,Nb) -2122.
result the SVP is weaker. Therefore, the typical MS is clearly observed, although the M/H
values in the FC branch are positive. Therefore, we may conclude that the absence of the
MS in Ru-2122 is caused either (or both) by the significantly high Ru moment
contribution to the overall moment, and by the enhanced internal fields which lead to the
SVP.
The hysteresis loop at 5 K, at low applied fields is shown in Fig. 18. Note (a) the
increase of negative moments up to 200 Oe. in the virgin curve, and (b) the particular
hysteresis loop obtained at low applied fields. In contrast to the typical FM hysteresis loop
obtained for Eu-2122 shown in Figs. 6 (inset) and 13) this loop is a superposition of SC
and W-FM properties of the sample. Here again, the SC properties are not masked by the
reduced W-FM features.
Figure 18. A superposition of SC and W-FM properties in the hysteresis loop at 5 K for (Ru,Nb) -2122.
DISCUSSION
We suggest two scenarios that could lead to the observed phenomena. A central
assumption is that Ru in Ru-2122 orders magnetically at elevated temperatures, and bulk
SC is confined to the CuO
2
planes. Both sublattices are practically decoupled, and thus the
present system is the first magnetic-superconducting system in the HTSC based materials.
Supporting evidence for this interpretation is (1) the high SF obtained for R=Eu in the SC
354
state, and (2) the overlapping of the two normalized curves exhibited in Fig. 11. The
second interpretation invokes analogy to inhomogeneous materials, e.g., the reason for the
two physical phenomena are grains with different oxygen concentrations, part of them are
SC and the rest magnetic. Moreover, one may argue that the magnetic anomalies exhibited
in Figs. 6-7 are due to SrRuO
3
impurity phase which is ferromagnetically ordered at 165
K(l 1). In order to reconcile these arguments we have prepared pure and Fe doped SrRuO
3
samples, and measured their magnetic and ME properties. The measured curve (at 10
Oe) is a typical one obtained for a ferromagnetic-like sample and in the ME spectra at
T>90 K, there is no sign whatsoever of magnetic order, indicating a weak coupling Fe -
Ru. Those measurements are completely different from the data presented in Figs 6-7. In
addition, our STM topography and spectroscopy measurements, described above, are not
consist, to say the least, with the mixed granular magnetic/SC picture.
The physical behavior of the oxygen and hydrogen charged Ru-2122 YBa
2
Cu
3
O
7
and
YBa
2
Cu
4
O
8
which have been studied extensively. Hole (or carrier) density in the CuO
2
planes, or deviation of the nomimal Cu valence from Cu
2+
, is a primary parameter which
governs T
C
in most of the HTSC compounds. Changes in the SC properties of YBa
2
Cu
3
O
7
can be induced by either (I) removing oxygen or by (ii) hydrogen loading which is a
destructive and irreversible. By depletion of hydrogen the crystal structure is destroyed
and SC is not restored. It appears that in Ru-2122, the ASP compound, is under-doped, due
to the fact that (a) annealing under high oxygen pressure shifts T
C
to higher temperatures
(Figs. 2-3), and (b) the effect of an applied field is to reduce T
C
. It is not clear yet whether
optimum doping is obtained with the 150 atm. sample. On the other hand, the influence of
hydrogen on Ru-2122 is reversible and not destructive, which means that hydrogen
changes the hole density of the Cu-O
2
planes, either by increasing or decreasing the ideal
effective charge of the planes. Depletion of hydrogen leads to the original charge density
and SC is restored. This is reflected in both the macroscopic magnetization studies and in
the SC gap distribution extracted from STM result. Data for the regenerated sample are not
presented here.
Oxygen pressure, as well as hydrogenation, enhance T
M
and changes other W-FM
characteristic features of the APS material. This effect, which was also observed in several
rare-earth based intermetallic hydrides, is probably an electronic effect. As described
above, in addition to the change in the hole density of the Cu-O planes, there is a transfer
of electrons from hydrogen (or oxygen) to the Ru 4d sub-bands, resulting in an increase of
the exchange interactions between the Ru sublattice and hence to an increase in T
M
of the
materials. An alternative way is to assume that the change (enhancement in the case of
hydrogen) in M
sat
, and M
rem
arises from an alternation of the anti-symmetric exchange
coupling of the DM type between the adjacent Ru moments, which causes the spins to cant
out of their original direction with a smaller (or larger) angle and as a result, a different
component of the Ru moments forms the W-FM state. However, this scenario cannot
reconcile the higher T
M
observed in all oxygen and hydrogen loaded materials. The exact
nature of the local structure distortions causing the W-FM behavior in this system, as well
as the oxygen and/or hydrogen location in the matrix, are not presently known and neutron
diffraction studies are now being carried out to address these points. Since hydrogen
loading affects both (SC and W-FM) phenomena, and the original behavior is restored
when hydrogen is depleted, we tend to believe that H atoms occupy interstitial sites close
to these planes, presumably inside the Sr-O planes.
In conclusion, we have shown that both SC and weak-ferromagnetism coexist in RU-
2122 and are an intrinsic property of this system. In contrast to other intermetallic
magnetic-SC systems, the present materials exhibit magnetic order well above the SC
transition (T
M
/T
C
~4). We attribute the magnetic order to the Ru sublattice, whereas SC is
confined to the CuO2 planes. Both sites are practically decoupled from each other.
355
ACKNOWLEDGMENTS
This research was supported by the BSF(1998) and the Klachky Foundation.
REFERENCES
1. Cava, R.J., Krajewski, J.J., Takagi, H., Zandbergen, H.W., Van Dover, R.B., Peck Jr, W.F. and Hessen,
B. (1992) Superconductivity at 28 K in a cuprate with niobium oxide intermediary layer, Physica C
191, 237-242.
2. Bauernfeind, L., Widder, W. and Braun, H.F. (1995) Ruthenium-based layered cuprates,
RuSr
2
LnCu
2
O
8
and RuSr
2
Ln
1+x
Ce
1x
Cu
2
O
10
(Ln=Sm, Eu and Gd), Physica C 254, 151-158.
3. Felner, I., Asaf, U., Levi, Y. and Millo, O. (1997) Coexistence of magnetism and superconductivity in
R
1.4
Ce
0.6
RuSr
2
Cu
2
O
10
(R=Eu and Gd), Phys. Rev. B 55, R3374-R3377.
4. Sonin, E.B. and Felner, I. (1998) Spontaneous vortex phase in a superconducting weak ferromagnet,
Phys. Rev. B 57, R14000-R14003.
5. Felner, I., Asaf, U., Goren, S. and Korn., C. (1998) Reversible effect of hydrogen on
superconductivity and weak ferromagnetism in R
1.4
Ce
0.6
MSr
2
Cu
2
O
10
(M=Nb and Ru), Phys. Rev. B 57,
550-556.
6. Felner, I., Asaf, U., Levi, Y. and Millo, O. (2000) Tuning of ferromagnetic behavior by oxygen and
hydrogen in R
1.5
Ce
0.5
RuSr
2
Cu
2
O
10
, Physica C (in press).
7. Pringle, D.J., Tallon, J.L., Walker, B.G. and Trodahl, H.J. (1999) Oxygen isotope effect on the critical
and Curie temperature and Raman modes in the ferromagnetic superconductor RuSr
2
GdCu
2
O
8
, Phys.
Rev. B 59, R11679-R11682.
8. Dynes, R.C., Narayanamurti, V. and Garno, J.P. (1978) Direct measurement of quasiparticle-lifetime
broadening in a strong-coupled superconductor, Phys. Rev. Lett. 41, 1509-1512.
356
QUANTUM PERCOLATION IN HIGH Tc SUPERCONDUCTORS
V.DALLACASA
Laboratory for materials analysis, Department of Science and
Technology, University of Verona, Strada Le Grazie, Verona, Italy
INTRODUCTION
Order and periodicity of perfect crystals have been the leading aspects of solid state
physics since 1970. The band theory, elaborated as an outcome of quantum mechanics,
explains in a nice way the metallic character of copper as well as the one of semi-metal as
graphite and of semiconductors as germanium and silicium. This theory asssumes that there
is a regular arrangment of atoms in space and the potential experienced by any electron is
periodic in space. However, disordered solids are the rule and not the exception and the
science of materials and researchers have to deal massively with methods of treatment both
theoretical and experimental appropriate. The corresponding electronic properties of such
materials have for a long time discouraged researchers who have preferred to reduce, when
possible, their study to methods employing some remnant form of order. This is even more
true if one thinks to the immense impact played by crystalline materials of high purity in
the electronic industry, but this does not mean that the disordered systems can be neglected.
Despite their complexity arising from the spatial inhomogeneity of the potential seen
by each electron, one can argue that the strongest the disorder the more uniform the crystal
will appear in the average, since a global property like, say the conductivity, may be
thought to result from repeated motion of electrons through inequivalent sites. This has as a
consequence that the conductivity will have eventually comparable order of magnitude in
all systems, a distinct feature with respect to a crystalline and pure system where it may
easily vary by many orders of magnitude from metals, to semiconductors to insulators. The
reason may be traced back to the fact that electronic states tend to be localized by disorder
in stochastic positions in space and hence the movement will take place predominantly to
nearest sites where localization occurs. It is a sort of quantum space on which electrons are
permitted to move. On the contrary, the infinite extension of the wavefunction in a regular
system will give the electrons chance to hop to more distant sites, compatibly with the
scattering mechanisms involved. There is a consequence on the transport properties: in a
regular medium the presence of phonons will be an adverse mechanism of movement,
while in conditions of localization the opposite will take place, with the phonons aiding the
proceess by supplying the energy necessary to overcome the energy difference between any
two localized states. In an otherwise perfect medium there will be a diffusive motion
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 357
through space, while in a disordered space only difficult hops will occur only with the
assistance of phonons.
For an ensemble of sites of localization arranged stochastically in space, one basic
problem is then be to understand the elementary hop between two sites, this may be
modeled as a tunnelling process assisted by phonons. But actually, the more difficult
problem is to follow the evolution of iterative and succesive hops. Percolation theory has
proved of great value, since it has been shown that a problem like this is dominated by
critical paths corresponding to easy transit between sites and since necessarily this problem
if of statistical nature, we can understand why there will be at a macroscopic scale, greater
independence on hopping details like molecular structure, crystal structure etc. On the
contrary, great dependence on the dimensionality is expected since the hopping diffusion is
expected to be quite different in one, two or three dimensions.
LOCALIZATION AND PERCOLATION
When the degree of disorder is sufficiently strong, the electron wavefunctions will be
localized, i.e they will decay exponentially from some point in space where
is the localization length and r the distance from the point. The existence of localized
states in presence of disorder is a consequence of the Ioffe-Regel criterium. In order that a
state described by a wave packet be extended troughout the whole system a necessary
condition is that the distance over which it loses coherence, i.e the mean free-path, be
longer than the interatomic spacing , i.e otherwise the state should be localized. As a
function of energy then, states must change their character and the critical energy E
c
at
which this occurs is the mobility edge. Thus the mobility edge marks the transition from a
metal to an insulator because for extended states the conductivity will have metallic
character, while for localized states it will vanish at zero temperature.
The scaling theory of localization of non interacting electrons and numerical estimates
[1] have established that for dimensions d =3 there is transition between extended and
localized states as the strength of disorder increases and correspondingly there is a
localization threshold On the contrary all states are localized in d=l and d=2 , which may
be restated saying that the carrier density has a threshold at p
c
= The localization length
is predicted to diverge when going from the localized side to the extended one, i.e on
crossing the mobility edge, as where the critical exponent In general,
as the carrier density increases, the Fermi energy approaches the threshold and hence the
localization length tends to diverge in the vicinity of a MIT.
The theory of percolation aims to obtain quantitative estimates for the properties of a
disordered system. In classical percolation [2] one considers a periodic lattice of sites each
of which can be randomly occupied with probability p or empty with probability 1 p.
Clusters, i.e a group of occupied sites containing neighboring occupied sites, will then be
present. As the concentration increases from zero, larger and larger clusters appear. The
mean size of these clusters grow with p and diverges at a well defined critical
concentration p
c
. For there exists an infinite cluster, which connects the two sides of
an arbitrary large sample. In the limit of an infinite lattice the value of the percolation
threshold p
c
is sharply defined. The infinite cluster percolates through the lattice with finite
probability. For p geater than p
c
the infinite cluster coexists with smaller finite clusters
which join it as p is further increased. If the above probabilities are referred to occupancy
of a nearest bond, rather than a site, we have bond percolation.
Although the percolation problem is easily defined it cannot be solved exactly.
However, it has quite interesting properties, i.e universality (independance of details) a
358
result of self-similarity, that is invariance under the change of length scale. These features
are closely related to the special geometric structure of the infinite cluster at p
c
, which
exhibits self-similarity, that is invariance under the change of length scale. The main appeal
of percolation theory is based on the possibility of correlating the geometrical and
topological properties with transport properties such as the electric conductivity, noise and
optical properties.
Quantum percolation can be formulated in terms of a tight binding one electron
hamiltonian [1] on a regular lattice. As in the classical case, we can define site and bond
percolation .The main concern is the location of the percolation threshold p
c
below which
all eigenstates of the hamiltonian are localized. The quantum threshold is greater than its
classical counterpart since the existence of an infinite cluster is a necessary but not a
sufficient condition for states to be extended. The most fundamental question of percolation
theory is to predict the critical value of the percolation threshold, namely the value of the
concentration at which an infinite network is first formed in the infinite lattice. Results for
some type of lattices are reported [2] in Table 1. The numbers in parenthesis refer to
quantum site and bond percolation thresholds [1] which in the case of two-dimensional
lattices diverge according to the scaling hypotesis.
The mean cluster size S , i.e the average of the size of clusters around randomly selected
occupied states in the lattice and the correlation length defined as the root mean square
average distance between two randomly selected occupied sites in the same cluster diverge
on approaching the threshold from below as and for p very
close to p
c
, p
c
> p. Values of the critical exponents are reported [2] in Table 2.
The exponents reported are the same for all two-dimensional and three-dimensional lattices.
This universality is what makes percolation theory appealing; there will be universality
behaviour of all systems irrespective of their details.
One notes that both the localization length and the correlation length have exponents of
order 1. If interactions between particles are taken into account these exponents turn out to
1/ 2. When discussing percolation due to localization and to electron grains in granular
metals, this difference will be evidenced.
TRANSPORT PROPERTIES
It is now widely accepted that the parent state of the high T
c
materials is an insulator
showing long-range antiferromagnetism and that the doping process on such a primitive
structure introduces holes/electrons either by cation substitution or by oxygen intercalation
359
or by a combination of them [3]. When holes are sufficient in number, usually already at
very low doping level of order 0.05 holes per Cu in the CuO
2
planes, they destroy the
antiferromagnetic state and superconductivity appears. Despite the disappearance of the
long-range order, strong spin correlations persist up to quite large doping levels, even when
the maximum Tc is obtained, the so-called optimum-doped region, and further. The passage
to superconductivity occurs at a critical doping where an insulator-to-metal transition first
occurs followed by the superconducting state. The critical temperature increases with
doping from the value zero up to a maximum value and then decreases with the further
doping until its total disappearance again at un upper critical value of the doping. In the
electron-doped system like Nd
2x
Ce
x
CuO
4y
the phase diagram is quite similar, the
disappearance of the critical temperature at the higher doping still occurs, whereas there can
be some coexistence of antiferromagnetism and superconductivity [3]. It has become
costumary to call underdoped those specimens with doping lower that the optimal doping at
which the maximum T
c
is observed and overdoped those with higher number of carriers. It
is generally agreed that the overdoped region is most similar to a Fermi liquid while the
underdoped phase is dominated by insulation of the Mott-Hubbard type at the lowest
doping levels.
A notable feature of the phase diagram is that the superconducting phase occurs close to the
insulating phase and that on increasing the number of carriers one has a transition from an
insulator to a metal-superconductor. Also noticeable is the fact that the appearance of
superconductivity takes place at the lowest T
c
values already in the insulating state. Similar
data have been obtained in the La
2x
Sr
x
CuO
4
, in YBa
2
Cu
3
O
7y
, Bi
2
Sr
2
Ca
1x
Y
x
Cu
2
O
8+y
,
Nd
1+x
Ba
2x
Cu
3
O
7y
and many others. The system appears to bifurcate between an insulating
ground state and a superconducting state with no normal metallic state in between.
The report by Bednorz and Muller [4] of superconductivity in Ba
x
La
5-x
Cu
5
O
5(3y)
(x=1
and x=0.75,y=0) showed that superconductivity could be obtained in an insulating state.
They attributed the onset of superconductivty in the 30K range to granularity and
percolation and concluded that grains of dimension 100A should exist in their sample. The
resistivity in these samples has metallic character at higher temepartures, at temperatures
slightly higher than T
c
onset shows an upturn and on lowering further the temperature it
undergoes a substantial drop into the superconducting state. It would become clear later on,
through extensive transport studies in a variety of cuprates, that in fact superconductivity
evolves from an insulating state in which localization induced by disorder and electron-
phonon interactions seem to play a major role.
In the low-doping region [5,6,7,8]the temperature dependence of the resistivity in the
normal state can be fitted by the law: where the exponent usually
assumes values ranging from to (Table 3.). By increasing the number of holes there is
a progressive tendency towards a metallic state with the resistivity losing the
semiconducting behaviour and acquiring a monotonic tendency to rise with temperature.
Close to optimum doping the resistivity shows a linear dependence in the normal state
which may persist even at temperatures as high as T=1000K At still increased doping the
transition superconducting temperature disappears while the resistivity tends to acquire a
superlinear dependence of the form T
152
[9] which is interpreted as the appearance of a
conventional Fermi liquid.
A number of studies of transport properties as a function of temperature and doping [10,11]
have indicated the progressive decrease of the characteristic temperatureT
0
as a function of
doping on approaching the metal-insulator transition (Table 4)
In the underdoped and optimum doped region (with the exclusion of the hard insulating
phase, where it can be shown that the exponential law, with and with
360
suitable changes of T
0
, is a fair representation of data even close to the superconducting
state, and including also the linear behaviour.
These results are usually interpreted in terms of phonon-assisted hopping models, either of
the Mott type or of the Efros-Shklovskii type (see next paragraph), enphasizing thus the
role of phonons. In such a parametrization the observed decrease of the characteristic
temperature T
0
can be traced back to the increase of the localization length As an order
of magnitude one finds that in the insulator with a density of states at the Fermi
level in the metallic state and close to the MIT values can be
found. Another possible parametrization of data can be achieved through the theory of
granular metals (see nest paragraph); in this case the role of the localization length is
assumed by the grain radius and similar orders of magnitudes for the latter are found.
In underdoped materials there are striking deviation from the linear behaviour. The
resistivity assumes values less than linear and eventually can turn upwards at the smaller
temperatures if superconductivity does not set in. We refer to ref. [9] for a summary of
behaviour in La
2x
Sr
x
CuO
4
. At the lowest temperatures the resistivity shows
semiconducting behaviour, a signature of localization. At intermediate temperatures the
resistivity increases superlinearly up to a certain temperature , where a break occurs in
the slope and the temperature dependence becomes almost linear. A reduced resistivity in
YBa
2
Cu
4
O
8
, in YBa
2
Cu
3
O
6+x
and underdoped Hg1223 showing similar deviations from the T-
linear law have also been observed [3]. There is now some consensus on the fact that these
361
deviations can be attributed to the opening of a pseudogap in the normal state, to which
corresponds a reduced scattering rate. It is interesting to refer ref. [12] which reports the
temperature obtained through various probes including Hall coefficient, static
susceptibility and Knight shift and of course the resistivity and infrared measurements.
PERCOLATION IN HIGH T
c
SUPERCONDUCTORS
The idea of a quantum percolation model can be traced back to the most peculiar
property of high Tc cuprates, namely that small changes in composition or structure can
change the material from semiconducting to metallic in the normal state and that there is a
metal-insulator transition in the transport properties evolving from a hard insulating phase
as the doping level is increased towards the critical point. Although being a probe of the
average charge structure, the resistivity as a function of temperature and eventually of a
magnetic field, is certainly a decisive parameter and in fact it is usually the first
characterization made of samples.
Phillips suggested a filamentary model [13] in which the metallic cuprate planes are
broken up into metallic domains and interlayer defects provide electrical bridges which
give the CuO
2
layers metallic character. In their absence the layers would be striclty two
dimensional and hence insulating, as predicted by the scaling theory, as a result of
localization from disorder in 2D. In this quantum percolation theory the density of
electronic states near the Fermi level is the contribution of localized and extended parts and
only the extended states can become superconducting. In this modified electronic structure,
as compared to a normal metal, called the X phase superconductivity is the result of the
electron phonon interaction. As a result of a sufficient density of defects, the coupling in the
planes can be sufficiently high to produce a high T
c
, yet the lattice instabilities that would
accompany such coupling if only extended states were present is in reality restrained by the
cage of the localized states in the planes [14]. Within the filamentary model, normal state
resistivities, optimized superconductivity with Tc and the lowest temperature for which
linearity holds are explained [15].
Direct evidence of charge domains can be obtained from direct imaging in diffraction
or similar studies. Early investigation of local structure through electron microscopy
indicated i.e, inhomogeneous distribution of oxygen in the form of blocks, of typical linear
size 100A, with different oxygen concentration in YBa
2
Cu
3
O
7x
(x>0.5) single crystals
[16].
Vacancy-ordering effects and the existence of their domains have been predicted
theoretically and revealed experimentally. Computer model simulations have suggested the
formation of oxygen-ordered domains in various cuprates including YBa
2
Cu
3
O
6+x
[17].
Electron diffraction studies in YBa
2
Cu
3
O
7x
[18] and electron microscopy in YBa
2
Cu
3
O
6.7
[19] have confirmed the presence of short-range oxygen-vacancy ordering within the
CuO
2
planes with dimensions of the order 100A .
Mesot et al. [20], on employing inelastic neutron scattering in ErBa
2
Cu
3
O
x
(6<x<7)
have shown that the energy spectra can be interpreted as a superposition of three stable
states, corresponding to three different types of clusters in the CuO
2
planes, two metallic
and one semiconducting with weight depending on oxygen concentration, with
superconductivity resulting from the formation of a two-dimensional network. The
localization range is estimated to be a few unit cells and the concentration of the two-
plateau structure of Tc predicted by means of a bond percolation model..
From neutron powder diffraction data and pair distribution analysis, local tilted
directions of the CuO
6
octahedra have been inferred, with local displacements ordered over
362
lengths of the order of 10A, this local structure being the same as the one of
antiferromagnetic spin correlations and of the superconducting coherence [21].
Coherence lengths for magnetic fluctuations of the order 10-20A and the existence of
locally ordered mesoscopic domains are obtained in the La
2x
Sr
x
CuO
4
system with
0.02 < x < 0.08 through the measurements of the
139
La spin-lattice relaxation rates vs.
temperature [22].
1/f noise measurements on copper oxide superconductors [23,] have evidenced that in
the normal state the noise is higher, typically 7-10 orders of magnitude, than that of a
conventional material. However, there are cases, like Tl
2
Ba
2
CaCu
2
O
8
thin films, were the
magnitude of the 1/f noise spectral density is much lower than in other cuprates [24]. The
important role played by morphology has been discussed by the authors just mentioned.
They make the assumption that the noise is due to tunneling transport across domain
boundaries of percolation grains. From the fitting of experimental data with tunneling
transport models these domains are found to have dimension of 100A and a wall thickness
10A, i.e much smaller than the one of domains and grains existing in the oxides even in the
single crystal form, which have spacing of order 1 or of defects, like twins, which have
spacing of the order 1000A. These studies, together with the resistivity studies, indicate that
the inhomogeneites are intrinsic to the microscopic structure and that superconductivity can
occur in microscopically small regions of space. In the following Table 5 a resume of
parameters for the model is given.
Support to the idea that superconductivity can occur in reduced space comes from the
huge experimental evidence in various systems which exemplify a number of physical
situations of reduced space. Metal clusters and metallic particles are examples. The
superconducting properties of metal clusters have been of interest for many years. Zeller
and Giaver [25] studied isolated, superconducting Sn clusters embedded in an oxide layer.
They found an increase of the superconting transition temperature with decreasing cluster
size, exceeding the one of bulk Sn, for clusters smaller than 40A.. Granular films composed
of Bi clusters have also been found superconducting by [26] Wei, while the bulk metal Bi is
not.. They find that superconductivity disappears for cluster size above 200A. Although in
principle cluster superconductivity in such systems could be explained as surface
superconductivity, normal coupling with phonons within the clusters can also lead to results
in agreeement with the observations. Thus, this is an interesting evidence for the
appearance of superconductivity in small systems.
In general the superconductivity of small metallic particles is a well established
phenomenon and the interest in the corresponding materials has always been high.
Although there is an intrinsic lower l i mi t of size for the superconductivity to occur, due to
the strong fluctuations expected for the order parameter, such l i mi t can be estimated around
363
20-25A, so for larger sizes of order 100A or so there is a concrete possibility for
superconductivity to develop.
Further experimental and theoretical studies support the viability of a percolative
nature of the charge carriers in cuprates. Experimental evidence of granular
superconductivity has been reported in YBaCuO system by Cai et al. [27] , through the
measurement of anomalous transient voltage excursions; the conclusions of these studies
are that granular superconductivity in this material can exist with Tc as high as 160K for
grains of the order 1000A..
Percolation models based on the existence of an intragrain superconducting transition
and a coherence transition induced by the Josephson coupling between neighboring grains
have been advanced, for example, to explain of the whole resitivity curves obtained in
YBaCuO ceramics [28].
A percolation model also has been introduced by Hizhnyakov et al.[29] to calculate
the critical magnetic cluster size prior to the disappearance of the antiferromagnetic long
range order in La
2x
(Ba,Sr)
x
CuO
4
and YBa
2
Cu
3
O
6+x
and to describe the transition from the
antiferromagnetic to the metallic or superconducting state. This work interestingly shows
that percolative models can be extended to the region of the phase diagram where
antiferromagnetism and its fluctuations exists.
The limiting form of reduced space is the unit cell and several experimental groups
have found superconducting behaviour in one-unit cell thick YBa
2
Cu
3
O
7
films with T
c
around 20K-30K [30]. Models based on tunneling coupling of the Josephson type of two
adjacent CuO2 planes have been used to explain the data [30]
Phase separation of charge can also be considered as a percolation effect . In this
respect there are a number of experimental evidences of phase electronic separation in
cuprates, i.e in La
2x
Sr
x
CuO
4
and By measuring the infrared
activity of photoinduced carriers, Kim et al. [31] found evidence of phase separation of the
photocarriers into metallic domains in the CuO
2
planes and of their condensation in a
superconducting state with T
c
= 40K. It is experimentally now well established from
Mossbauer spectroscopy, microwave absorption, neutron scattering and EPR [32] that
phase separation occurs through domains of conducting and superconducting type. Using
experimental data obtained from electrical resistivity, susceptibility and magnetization
measurements, the phase separation in and La
2x
Sr
x
CuO
4
has been demonstrated
to be of electronic and percolative type [32,33].
The glassy behaviour exhibited by the high Tc oxides can be taken as evidence for
intragrain Josephson junctions arising from well identified metallic-superconductor grains
separated by thin insulating barriers [34]. The interesting aspect is that glassy behaviour is
exhibited in single phase samples and even in single crystals. It has been suggested that
granular behaviour can result from the short coherence length of cuprates and the suggested
order of magnitude of these domains is 100-1000A with barriers of the order of 10A [34].
HOPPING CONDUCTIVITY WITHOUT INTERACTION
In the case of phonon-assisted hopping [35] first gave and expression for the
conductivity. Their expression derived by the golden rule, essentially contains a coupling
parameter which is assumed a deformation potential term, an overlap integral of the
localized states and the phonon population factor. For the purpouse of the present paper,
their expression can be reelaborated to arrive at a simpler formula. When the overlap
between wave functions is suffciently small, i.e the localized states are sufficiently distant
apart with respect to the radius of the localized states, taking account that the number of
364
carriers participating in the hopping process is one arrives at an expression of the
form:
(1)
for the conductivity of a two-site process, where E is the coupling parameter, N
F
the density
of states at the Fermi energy, R the separation of the centres of localized orbitals, the
energy difference between states that has to be supplied by phonons, the localization
radius and n
q
the phonon population at wavevector q .
There are two useful limiting forms of this equation for low and high temperatures. We
obtain
(2)
(3)
Eq. (1) relating to a two-site process, has to be integrated over the whole process of
subsequent hops.
Miller and Abrahams showed that the current can be reduced to a problem of equivalent
resistors connecting sites, and the conductivity is related to an integral over all conducting
paths, a problem connected with a kind of percolation problem in which the paths carrying
the current can be evaluated as a critical percolative problem.
A procedure, alternative to the integration over paths has been suggested by Mott. It is a
saddle-point method which amounts to resolve the integration on selecting those paths
where the transition probability per unit time has a maximum value. From this one can
deduce simple rules for calculating the conductivity under more complex situations.
Mott has argued that a simple geometric argument can be established to obtain a
relationship between the parameter and R. It amounts to require that the total number of
states within a hopping sphere This gives in three
dimensions. Similar reasoning can be applied to lower dimensionality.
Then the saddle point maximation procedure leads to the results:
(4)
(5)
where and is a factor of order one.
Motts argument holds for a constant density of states at the Fermi level. There have been
various attempts to remove such an assumption, thereby improving on the Motts law, on
assuming more general dependencies of the density of states at the Fermi level on energy
and also avoiding the geometrical argument. These modifications usually change the pre-
exponential of the law, while leaving the exponential factor unchanged.
From eqs. (2-3) one finds that as a result of the phonon population at high and low
temperatures, the conductivity at low T turns out to be higher than the one at high T by the
factor This result is common to other models (see i.e granular metals) and has
relevance in connection with the pseudogap behaviour of the conductivity.
365
The functional form of variable range hopping may be modified close to a metal-insulator
transition, the fractal nature of the wavefunction leading to a possible modification of the
pre-exponential. We shall discuss this problem in connection with the granular metal
models, which share close resemblance with the localization problem
HOPPING CONDUCTIVITY WITH INTERACTION. COULOMB GAP
Motts law in its various forms is due for neutral centres. The effects of the Coulomb
repulsions have been considered by Efros and Shklovskii [36]. Their results for the
conductivity can be deduced by noting that in such a case the activation energy between
two centres is given by the expression where R is the separation of sites and a
suitable static dielectric constant. On using the maximation procedure similar to the Motts
law one gets the result
(6)
(7)
with and The factor A(T) may include corrections due to fractal
nature of wavefunctions. In fact the same caution for the pre-exponential has to be used in
the case of vicinity of the MIT like in Motts law. The results above are shown to arise from
a modification of the density of states from the coulomb repulsions, leading to a density of
states vanishing at the Fermi level, the so-called coulomb gap. Efros and Shklovskii have
shown that the density of states follows the law The constant A can be
deduced on imposing that i.e the density of states goes over continuosly to its
umperturbed value for eenrgies greater than the gap. One has for instance in
three dimensions. An equation which generalizes Motts argument is for an arbitrary
dependence of the density of states on energy is This is due for the
particular case of 3D. In lower dimensions suitable modifications of the volume element in
t hi s equation has to be carried out. For the case of a constant density of states
one finds back Motts result for and in thecase of the coulomb gap one ends with the
expression of the coulomb activation energy.
In general, a cross-over occurs from the Motts law to the Efros-Shklovskii law on
varying the temperature, since at higher T the activation energy for the neutral case
becomes smaller than the coulomb activation energy and thus the law is confined to
smaller T. Aharony et al. [37]. have discusseed the cross-over on combining the two
activation energies for the charged and neutral cases. It follows from these treatments a
form of universality or scaling in the cross-over region. If the combination of the two
activation energy is written as Aharony et al. [37] find that such a cross-
over occurs at the temperature with the conductivity behaving as:
In i.e a universal function of x = T/ T
x
. It is shown by
Aharony et al. that the Efros-Shklovskii and Motts results are limiting cases of this
expression when T << T
x
and T >> T
x
with In with and
366
and where for T T
x
and f or T T
x
. Recently it was
shown that Efros-Shklovskiis and Motts activation energies can be viewed as two terms of
a multipolar expansion of the polarization charge on the sites, the R
1
term being the charge
term and the R
3
term the quadrupolar term [38]. A R
2
term, which is viewed as the
counterpart of the cubic term in two dimension, can be identified with the dipolar term of
the expansion. A discussion of the cross-over including also the dipolar term has been
given, showing that universality is maintained on rescaling T
x
[38].
GRANULAR METALS
The granular metals are materials composed of metals and insulators. These materials,
known as cermets, were originally used as resistors due to their high resistivity at low
temperatures, but they are now widely studied for their unique properties even at the
submacroscopic scale 10-100A.
There are three regimes of interest. The metallic regime occurs when the metal
fraction x is large, the metallic grains touch and a metallic continuum exists with dielectric
inclusions. The dielectric regime occurs when there is an inversion of the former where
small metallic particles are dispersed in a dielectric continuum. Finally there is a transition
regime corresponding to an intermediate state.
In the dielectric regime with metallic islands dispersed in the dielectric the electrical
conduction results from tunneling processes from one island to the other and thermal
activation. The carriers are generated on removing an electron/hole from a metallic island
and moving it to the other. Thus a pair of charged grains is created with a cost in energy of
the order where d is radius of the grains and s is the wall thickness
among grains and where the factor F takes account of the particular form and distance of
the grains. For suffciently large distances F = 1 leaving a purely coulombic barrier, but at
closest distances the effects of the wall thickness among the grains may radically change
F . Sheng et al. [39], taking the charging energy into account , have proposed that the
resistivity at low applied fields should obey a law of the form In The hypotesis
underlying this result is a homogeneity rule implying that to ensure spatial homogeneity
of the samples it should hold that d/s=C=const. The dielectric regime thus is characterized
by small radius of the grains and corresponding small intergrain wall thickness. This law
for the conductivity is the consequence of a distribution of insular radius and an average of
the conduction paths over the grain dimensions.
In the transition region the metallic particles grow and there appear interconnections
between them. At a critical composition it first appears a metallic continuum and
conductivity is due to an infinite diffusive process.
We review the theory of Sheng et al. [39], Our presentation of the problem makes use
of concepts of the formulation of the tunnelling procesess introduced by Neugebauer and
Webb [40]. The model assumes that there are a large number of metal islands with a
relatively small number of them charged. The equilibrium concentration of these charges is
maintained thermally. The probability that an electron will tunnel from one negatively
charged island i to a neighbouring neutral island j is proportional to the density of states
at each island and to the transmission coefficient, i.e
(8)
367
where D is a diffusion constant and the Fermi functions f and 1 f take account that the
initial state is full and the final one is empty. Both the forward and backward (with respect t
o the direction of the field) probabilities are taken into account as well as the energy shift
induced by the field energy eV and the charging energy E
c
. The net transition probability
wi l l be P = P
+
P

. The conductivity is related to P by the relation On

assuming that both N
F
and D have a weak energy dependence, P can be obtained from a
straighforward integration of (. . . ). One finds at low fields
(9)
We shall first discuss results valid in the case when the barrier thickness is large. The
transmission coefficient can be obtained from the tunneling across the walls as
on assuming thus that where is a barrier parameter. The
charging energy is given by ,using the homogeneity condition, and the
distribution probability of the sizes of the islands can be taken of the form
[39] P(s) = As
n
exp(s/s
0
) with n a suitable power (Sheng et al. assume n=1) and s
0
the
average wal l size.
At low tempeartures E
c
/ kT 1 one then finds from eq. (9)
(10)
where The upper l i mi t can be pushed to at small temperatures and the
integral involved in t hi s expression can be reduced to a form calculable in terms of Bessel
functions whith v = n .
Thus at low T such that T
0
/ T 1 we find the resistivity
( 1 1 )
At high temperatures the integral of the conductivity can be reduced to the the form, for
sufficiently high n
(12)
Thus, for T T
0
the lower l i mi t of the integral can be push down to zero and the result is
again expressable in terms of Bessel functions. We get
(13)
where f is a number depending on the order of the Bessel function.
There is a close mathematical analogy between the hopping model and the model of
granular metals: in the former carriers make percolative transitions between localized states
, in the latter the transitions occur between grains. In both models there is a tunneling factor
as well as a thermal factor arising from an energetic barrier to overcome. Both models are
368
based on random motion with the conductivity being an average through a distribution of
hopping sites or grain dimensions.. The conclusions for the two models are identical with
the obvious significance of terms.In partciular we note the linear relation (13) for the
resistivity with an intercept different from zero at T=0.
Numerical simulation performed by Zhou et et. [41] indicates that the power of the
exponential law is obtained as an intermediate behaviour of the high T simply activated
behaviour and the -like Motts exponential law. This result indicates that while in hopping
processes usually the low temperature region is dominated by the T
1/2
law, for the granular
metals it is dominated by the T
1/4
law.
In the transition region we expect and and one can assume
that to account for power-law fall of the correlations with distance and
emphasizing the increased probability of percolating paths with small s .
At sufficiently small T we shall again find the conductivity behaving as eq. (..) i.e
dominated by the exponential dependence; but at large T the leading term of the integral
will be
(14)
Since, due to normalization requirements for the distribution function, one has we
end with a superlinear behaviour of the resistivity with T. The physical origin of it is the
spatial variation of the percolation probability at small s.
SUPERCONDUCTIVITY
The anomalous phenomenology of high temperature superconductors forced a
reconsideration of Fermi liquid behaviour, casting doubts on the the assumption of pairing
resulting from phonon coupling and suggesting the investigation of a number of purely
electronic or mixed models. In the light of the quantum percolation model, however,
phonon coupling occurring in a finite volume is worth of attention. As a result of the small
coherence length in the cuprate superconductors there can be much more sensitivity to
structural changes, local structures and even imperfections. Such a coupling may arise in
localization regions either within the localization radius or in metallic clusters or grains
within the metallic grains of finite volume.
It turns out that, as experiments suggest, the relevant scales for optimal superconductivity
temperatures are of the order 100A.
In order to understand superconductivty coupling in these conditions, under the
assumption that superconductivity is an effect of the domain interior and not of its surface,
attempts were done to assimilate domains in first approximation to a small metallic particle
and on concentrating mainly on the influence of the domain size on the electron system.
The idea leading to this assumption is two-fold. First, the phonon indirect coupling of
electrons is inversely proportional to the volume, secondly within the BCS theory adapted
to the case of the finite dimension of localized states (or percolating clusters) the cutoff of
momenta for which results in the separation of the levels of
the finite cluster approximating such finite space and if this exceeds the attraction range,
only levels at the Fermi enrgy will contribute, enhancing T
c
The change of the phonon
frequency, as compared to the bulk, can be taken into account by the introduction of an
effective Debye frequency.
369
The validity of the BCS theory rests upon the fact that the fluctuatios of the order parameter
within the domain is negligible. The minimal allowed size and turns out to be of order 25A.
Another minimal length is the coherence length, typically 10 A. Thus for values of
interest 100A of the domain size, these conditions are met.
No domain interaction is considered likewise; predictions then are made only on the Tc
onset corresponding to superconductivity in single uncorrelated domains and not on the
lower Tc at which coherence between domains occurs.
One finds increased critical tempeartures as compared to the bulk, due to the inverse
dependence of the pair interaction on the volume and to quantization of levels.
We present results for the Cooper problem of non interacting carriers. The study of the
Cooper pairing problem in the finite system can be approached via a numerical procedure,
assuming the usual form of the interaction where is the step
function and are the energy levels of the box representing the domain, while M
2
and
are the electron-phonon matrix elements and the phonon cut-off frequency. The form of the
coupling, essential for the results, i s M
2
=M
2
/ N where N is the number of cells within
the domain and M
2
refers to the coupling in the bulk . For a system confined to a cube of
length L there will be a set of degenerate states whose separation is aroung
the Fermi surface.
The Cooper problem leads to the eigenvalue equation
(15)
This equation can be solved using the form of the levels of a box.. The outcome of this is a
single bound state and positive eigenvalues. For sufficiently dense levels within the
attraction range (high L),the numerical result go over to the analytical Cooper result.
The l i mi t depends on the strength of the coupling. An increase of Tc on decreasing the
length L is obtained, the effects becoming stronger at smaller lengths. In general, one finds
a region at large L in which T
c
is almost constant and approaches the Cooper limit, and a
region at small L in which a rapid increase of Tc occurs as L decreases. At the minimum
length of the numerical procedure one can find Tc of the order of some 100K under
conditions of weak coupling The cross-over can be estimated to occur when the separation
of levels becomes countably small within the attraction range.
The asymptotic behaviour at small and large lengths can be obtained analytically. The
procedure can be implemented directly on the BCS equations rather than on the Cooper
problem. The limiting results for Tc are
(16)
(17)
where the limiting length is given by
Similar calculation can be carried out for the gap The second of these equations
indicate a scaling of the critical temperature with the inverse dimension of the percolating
clusters and a direct relation with the Fermi level. These are two competing factors: when
370
the carrier density increases, the Fermi energy increases while N decreases. The result will
be a compromise with the Tc exhibiting a maximum at some carrier density.
Results for interacting electrons within the coulomb gap can be obtained by solving
the BCS equations with the density of states of the gap: for and
for with where n is the static dielectric constant.
For the results indicate a proportionality of the critical temperature to the
coulomb gap, i.e which displays a similar dependence on the carrier density as in
the non-interacting case at small L.
It appears quite natural to assume that L correspond to the coherence length of the
quantum percolation states, i.e the mean square radius of the percolating-localized clusters.
We then can take below the percolation threshold in which is a critical
exponent. Thus assuming that the Fermi energy scales as for a two-dimensional
electron gas we get the formula as a convenient parametrization
formula of the critical temperature.
This equation predicts in particular that in the underdoped region
where m* is the carrier effective mass, a result in agreement with the findings of [42]. At
larger n the critical temperature will be a dome-shoped curve with a maximum at
and
Analisis [42]of avalilable data in and
systems as well as and Cheverel phases have indicated that the critical exponent
can be accurately determined. The data strongly indicate an extrapolation of to zero for a
critical number of holes in the planes, i.e the existence of a threshold.
From the fitting procedure values are obtained with the exclusion of the Chevrel
systems for which is obtained. The percolative threshold value established by these
data is per plane.
Thus, there is a threshold which indicates a three-dimensional character of the percolative
network This means that although the planes are expected to provide the
superconducting carriers, the critical temperature is influenced by out-of-plane effects[13].
Low Tc systems as the SrTiO and Chevrel systems appear to display characteristics similar
to the high materials, perhaps indicating that the mechanism analized here is not peculiar
to cuprates. The value of the critical exponent indicates coulomb effects, i.e an interacting
carrier gas. This result is in agreement with quantum percolation with interaction and an
analysis of Kaveh and Mott [43]who have indicated how this critical exponent changes
from the value 1 to the value in the vicinity of a superconducting state as a result of
coulomb interactions. For the Cheverel systems we get indicative, on the contrary of a
non-interacting carrier gas. The value of the threshold agrees with the one expected for
simple cubic strucures although the complicated nature of the unit cell of cuprates is
difficult to reduce to a simple known lattice, for which percolation thresholds have been
calculated.
The typical percolation sizes can be estimated around L=100A depending of the values
typically used for the Fermi energy.
Previous results of the effect of altered wavefunctions as a result of reduced
dimensions have been considered in connection with superconductivity in thin films; the
gap parameter is found to increase with decreasing film thickness [43]. The application of
the BCS theory to finite sysytems has been considered in connection with its mathematical
limit when the size becomes infinite [44].
371
SUMMARY AND CONCLUSIONS
We have reviewed a quantum percolation model in which charge clusters below the
percolation threshold undergo phonon-asssited hopping processes. We have indicated
theoretical and experimental evidence of the existence of such clusters and discussed two
possible mechanisms which appear to be consistent with the phenomenology of high Tc
superconductors. These are: a phonon-assisted hopping model relying on Andersons
localization by disorder and a granular metal model in which hopping occurs on charged
grains. In both models the role of coulomb interactions occurring during hops between
localized wavefunctions or charged grains has been emphasized. The principal parameters
of quantum percolation is the localization length/grain radius, which defines the value of
the gap parameter whose increase on increasing the carrier concentration describes the
evolution of transport properties in the whole region of the phase diagram. At very low
doping, the insulating phase can be described by variable range hopping of the Motts type.
For larger doping, in the overdoped and optimum doped region eqs. (6) and (7) relative to
coulomb hopping or (11) and (12) for the granular metals description apply. One finds a
low temperature regime eq.(7) with an insulating exponential variation at the lowest
temperatures, followed by a superlinear behaviour at moderately higher temperatures with
the curves appear to rise linearly with temperature with zero intercept at the origin;
a typical high limit to this behaviour being 300K. In the high temperature range the
curves are described by eq. (6) and are dominated by the pre-exponential so that a linear
relation with T occurs, but with an intercept at the origin proportional to as evidenced by
eq. (13), so the curves appear as if there were a lower slope at high T. Since decreases as
the relevant length increases, on increasing the carrier concentration the linear behaviour
will be observed in a larger temperature interval down to zero and it will progressively lead
to the disappearance of the superlinear behaviour These features are found to describe
quite accurately the superlinear and linear behaviour of the resistivity in the underdoped an
optimum doped regions of the phase diagram and give a plausible origin to the pseudogap
as observed in transport properties.. It follows that the existence of the coulomb gap in the
underdoped region may also explain why a gap is detected in the normal state by external
probes, like photoemission, giving the impression of persistence of the superconducting gap
inside the normal state.
The high critical temperatures are understood within the quantum percolation model as a
result of electron-phonon coupling in the microscopic clusters of charge. Two competing
factors affect it, namely the Fermi energy leading to its increase and the length of the
percolating cluster, leading to a decrease with carrier density. The result is a maximum
allowed value. The relevant cluster size at which is of the order required in the cuprates
is L=100A. The proportionality of to the coulomb gap establishes a situation in which
there is a coincidence of the gap in the superconductor, arising from the Cooper pairing,
and in the normal state, in which it corresponds to the vanishing of the density of states at
the Fermi level.
REFERENCES
1. Soukoulis C.M. and.Grest G.S. (1991) Localization in two-dimensional quantum percolation,
Phys.Rev.B 44,4685-4688.
2. Aharony A.and Stauffer D. (1987) Percolation, Encyclopedia of Physical Science and Technology 10,
226-245.
372
3. Y.Iye (1992) Transport properties of high T
c
cuprates, in D.M.Ginsberg (ed.), Physical Properties of
high temperature Superconductors III, World Scientific, pp.285-361.
4. Bednorz J.G. and Muller K.A. (1986) Possible high T
c
superconductivity in the Ba-La-Cu-O systems,
Z.Phys. B-Condensed Matter 64,189.
5. Dabrowski B.,. G. Hinks D., Jorgensen J.D., Kalia R.K., Vashishta P., Richards D.R., Marx D.T. and
Mitchell A.W. (1988) Variable-range hopping conduction in Ba
1-x
K
x
BiO
3-y
system, Physica C 156,24-
26.
6. Leew D.M., Mutsaers C.A.H.A., Steeman R.A., Frikkee E. and Zandbergen H.W. (1989) Crystal
Structure and electrical conductivity of Physica C 158, 391-396.
7. Mandal P., Poddar A.and Ghosh B. (1991) Variation of T
c
and transport properties with carrier
concentration in Y- and Pb-doped Bi-based superconductors, Phys. Rev.B 43,13102-13111.
8. Dallacasa V. and.Feduzi R. (1993) The localized character of the metallic state ih high T
c
superconductors, Journal of Alloys and Compounds 195,531-534.
9. Battlog B., Hwang H.Y., Takagi H., Cava R.L. and Kwo J. (1994) Normal state phase diagram of
(La,Sr)
2
CuO
4
from charge and spin dynamics, Physica C 235-240, 130-133.
10. Ellman B., Haeger H.M., Katz D.P., Rosenbaum T.F., Cooper A.S. and Espinosa G.P. (1989) Transport
studies of La
2-x
Sr
x
CuO
4
near the insulator-metal-superconductor transition, Phys. Rev. B 39, 9012-9016.
11. Mandrus D., Forro L., Kendziora C.and Mihaly L. (1992) Resistivity study of Bi
2
Sr
2
Ca
1-x
Y
x
Cu
2
O
8
single crystals, Phys.Rev.B 45,12640-12642.
12. Timusk T. and Statt B. (1999) The pseudogao in high-temperature superconductors: an experimental
survey, Rep.Prog.Phys. 62, 61-122.
13. Phillips J.C. (1999) Is there an ideal phase diagram for high-temperature superconductors?, Phil.Mag.B
79, 527-536.
14. Phillips J.C. (1989) Quantum percolation and lattice instabilities in high-Tc cuprate superconductors,
Phys.Rev.B 40, 8774-8779.
15. Phillips J.C. (1997) Filamentary microstructure and linear temperature dependence of normal state
transport in optimized high temperature superconductors, Proc.Natl.Acad.Sci.USA 94, 12771-12775.
16. Barabanenkov Yu.A., Zakharov N.D., Kotyuzhnanskly B.Ya., Meleshina V.A., Svistov L.E. and
Shapiro A.Ya (1989) Oxygen distribution in YBa
2
Cu
3
O
7-x
, Zh. Eksp.Teor. 96, 2133-2139.
17. De Fontaine D., Wille L.T. and Moss S.C. (1987) Stability analysis of special-point ordering in the
basal plane in YBa
2
Cu
3
O
7-6
, Phys.Rev.B 36, 5709-5712.
18. Werder D.J., Chen C.H., Cava R.J. and Battlog B. (1988) Diffraction evidence for oxygen-vacancy
ordering in annealed superconductors, Phys.Rev.B 37, 2317-2319.
19. Chen C.H., Werder D.J., Schneemeyer L.F., Gallagher P.K. and Waszcazak J.V. (1988) Observation of
oxygen-vacancy-ordered domains in oxygen-deficient single crystal YBa
2
Cu
3
O
6.7
, Phys.Rev.B 38,
2888-28991.
20. Mesot J.., Allenspach P., Staub U., Furrer A. and Mutka H. (1993) Neutron spectroscopic evidence for
cluster formation and percolative superconductivity in ErBa
2
Cu3O
x
, Phys.Rev.Lett.70, 865-868.
21. Billinge S.J.L., Kwei G.H. and Takagi H. (1994) Local octahedral tilts in La
2-x
Ba
x
CuO
4
: Evidence for a
new structural length scale, Phys.Rev.Lett.72, 2282-2285.
22. Cho J.H., Borsa F., Johnston D.C. and Torgeson D.R. (1992) Spin dynamics in La
2-x
Sr
x
CuO
4
from
139
La NQR relaxation: Fluctuations in a finite-length-scale system, Phys.Rev.B 46, 3179-
3182.
23. Ong Yi, Misra A.,Crooker P.P. and Gaines J. (1991) 1/f noise and morphology of single
crystals, Phys.Rev.Lett.66, 825-829.
24. Misra A., Song Yi, Crooker P.P. and Gaines J. (1991) 1/f noise in Tl
2
Ba
2
Ca
1
Cu
2
O
8
thin films: Influence
of crystal structure, Appl.Phys.Lett.59, 863-865.
25. Zeller H.R. and Giaever I. (1969) Tunneling, zero-bias anomalies and small superconductors,
Phys.Rev.181, 789-799.
26. Weitzel B. and Micklitz H. (1991) Superconductivity in granular systems built from well-defined
rhombohedral Bi clusters:evidence for Bi-surface superconductivity, Phys.Rev.Lett.66, 385-388.
27. Cai X., Joynt R. and Larbalestier D.C. (1987) Experimental evidence for granular superconductivity in
Y-Ba-Cu-O at 100K to 160K, Phys.Rev.Lett. 58, 2798-2801.
28. Rosenblatt J., Raboutou A., Peyral P. and Lebeau C. (1990) Intragranular and intergranular transitions
in Y-Ba-Cu-O ceramics, Revue Phys. Appl. 25, 73-78.
29. Hizhnyakov V. and Sigmund E. (1988) High-T
c
superconductivity induced by ferromagnetic clustering,
Physica C 156, 655-666.
30. Bandte C. and Appel J. (1994) Superconductivity in the CuO
2
bilayer, Physica B 194-196, 1359-1360.
31. Kim Y.H., Cheong S.W. and Fisk Z. (1992) Phase separation of charge carriers in La
2
CuO
4
, Physica C
200, 201-206.
32. Kremer R.K., Hizhnyakov V., Sigmund E., Simon A. and Muller K.A. (1993) Electronic phase
separation in La-cuprates, Z. Phys:B 91, 169,174.
373
33. Sigmund E., Hizhnyakov V., Kremer R.K. and Simon A. (1994) On the existence of percolative phase
separation in high-T
c
cuprates, Z. Phys.B 94 17-20.
34. Deutscher G. (1988) Superconducting glass and related properties, Physica C 153-155, 15-20.
35. Miller A.and Abrahams E. (1960) Impurity conduction at low concentrations, Phys.Rev. 120, 745-755.
36. Efros A.L. and Shklovskii B.I. (1975) Coulomb gap and low temperature conductivity of disordered
systems, J.Phys.C: Solid State Phys. 8, L49-51.
37. Aharony A., Zhang Y. and Sarachik M.P. (1992) Universal crossover in variable range hopping with
coulomb interactions, Phys.Rev.Lett. 68, 3900-3903.
38. Dallacasa V. (1998) Internal field-assisted thermally activated hopping and tunnelling in insulators and
composite materials, J.Phys.:Condens.Matter 10, L409-416.
39. Sheng P., Abeles B. and Arie Y. (1973) Hopping conductivity in granular metals, Phys.Rev.Lett. 31, 44-
47.
40. Neugebauer C.A. and Webb M.B. (1962) Electrical conduction mechanism in ultrathin, evaporated
metal films, J.Appl.Phys. 33, 74-82.
41. Zhou M., Sheng P., Chen L. and Abeles B. (1992) Numerical simulation of hopping conductivity in
granular metal films, Phil.Mag.B 65, 867-871.
42. Kaveh M. and Mott N.F. (1992) Metal-insulator transition near a superconducting state, Phys.Rev.Lett.
68, 1904-1907.
43. Paskin A. and Singh A.D. (1965) Boundary conditions and quantum effects in thin superconducting
films, Phys.Rev.140, A1965-1967.
44. Henley E.M., Kennedy R.C. and Wilets L. (1964) Finite superconductors and their infinite volume
limit, Phys. Rev.135, A1172-1174.
374
SUPERSTRIPES
Self organization of quantum wires in high T
c
superconductors
A. BIANCONI
1
, D. DI CASTRO
1
, N. L. SAINI
1
and G. BIANCONI
2
1
Unit INFM, Dipartimento di Fisica, Universit di Roma La Sapienza,
00185 Roma, Italy
2
Department of Physics, Notre Dame University,
46556 Notre Dame, Indiana
INTRODUCTION
High T
c
cuprate perovskites provide an exotic superconducting phase at half way
between absolute zero temperature and room temperature. Conventional superconductivity
appears in metals with a very high charge density and near absolute zero temperature. In
these materials the electrons in the normal phase, above the critical temperature T
c
can be
considered as free particles following the Fermi statistics (fermions) being in the high-
density limit and at low temperature. The electrons are described by a single particle
wavefunction that gives the probability to find an electron in the point r. Below
T
c
electron pairs condense into a single quantum state. The condensate is described by the
order parameter where gives the density of condensed pairs and is the phase.
This macroscopic quantum state is characterized by exceptional manifestation of the
quantum order: zero resistivity [1], perfect diamagnetism [2], quantization of magnetic flux
[3,4] and quantum interference effects [5]. The wavefunction of the condesate decays
exponentially as we go from the surface of the material to the vacuum with the Pippard
coherence length [6] and the magnetic field decays exponentially [7] as we go from the
surface into the material with the London penetration length
The formation of the condensate made of electron pairs has been described by the
BCS theory [8]. The key point of the BCS theory is that the formation of the condensate is
due to the fact that electrons actually are not free particles but they are interacting; however
the interaction is much smaller than the Fermi energy. In this weak coupling limit the
interacting electrons are replaced by Landau quasiparticles. The very small electron-
electron attraction triggers the formation of pairs of quasiparticles, with zero momentum
and zero spin. The standard BCS theory assumes that the electron-phonon interaction
provides the mechanism for the pairing however the pairing can also be mediated by
electronic excitations (excitonic or plasmon mechanisms) in the low density limit.
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 375
In the weak coupling limit the critical temperature T
c
is related with the energy
needed to break the pair
(1)
where is the superconducting energy gap. The critical temperature (and the gap) is
given by:
(2)
where T
F
is Fermi temperature, is the wavevector of electrons at the Fermi
level, is the coherence length of the condesate that is related with the size of the pair and
f is a measure of the deviation from the weak coupling limit
The BCS approximations are valid for a 3D metal with critical temperature close to
zero Kelvin. In the strong coupling limit the critical temperature for the many body
superconducting phase remains low since the pairs form Bose particles at high temperature
but the phase coherence of the Bose condensate occurs only at low temperature.
The discovery of high T
c
superconductivity [9] in copper oxide perovskites, with a
record of T
C
~150K in Hg Ba
2
Ca
2
Cu
3
O
8+y
[10] has clearly shown that the
superconducting condensate can be formed beyond the standard BCS approximations.
The mechanism driving the superconducting state from the range 0<T
c
<23K of metals
and alloys to the high temperature range 20<T
C
<150K of doped cuprate oxide perovskites,
i. e., enhancing critical temperature by a factor ~10, is the object of this work.
The cuprate perovskites are heterogeneous materials formed by
1) metallic bcc CuO
2
layers intercalated between
2) insulating fcc AO
1-x
layers (A=Ba,Sr,La,Nd,Ca,Y) [11-13] and
3) charge reservoir layers where the chemical dopants are stored.
Therefore the 3D superconducting order is realized in a superlattice of 2D metallic
layers where the layers are separated by a spacing of the order of the Pippard coherence
length. The pairs can jump between the layers while the single electrons are confined in the
layers. In fact the pairs have a hopping of the order of between the 2D layers, where
D
o
is the density of states at the Fermi level, t
p
is the single particle transverse hopping
between the layers.
The critical temperature of the 3D condesate is controlled by the parameters of the 2D
electron gas: the density where r
s
is the electron density parameter and a
B
is
the Bohr radius; the Fermi wavevector and the Fermi energy E
F
(Ry)
=2/(m
*
r
s
2
), and it is given by:
(3)
Uemura et al. and Keller et al. [14-16] have measured from the London
penetration depth in different cuprate perovskites at a fixed doping, showing the linear
relation for T
c
versus
376
Within the BCS approximations the critical temperature increases by increasing the
electron-electron attraction, and by decreasing the size of the pairs, i. e., the coherence
length of the superconducting phase.
However the BCS approximations breaks down in the strong coupling regime where
the electron-electron attraction is larger than the Fermi energy. In this extreme limit all
electrons form localized pairs (LP). These local pairs are formed below the high
temperature T
p
however the superconducting critical temperature T
c
occurs at low
temperature where the local pairs Bose condense and in the strong coupling the factor
f>>1. Therefore the critical temperature T
c
reaches a maximum in a optimum intermediate
coupling (OIC) regime where [17].
THE METALLIC HETEROGENEOUS PHASE
The heterogeneous structure of a Cu
2+
cuprate perovskite is shown in Fig. 1. The CuO
2
layers form a fcc layer of a tetragonal structure with crystallographic axis a
t
=b
t
=3.94 .
The Cu ion form a square pyramid or bipyramid with planar Cu-O distance R=1.97 and
axial Cu-O(A) distance 2.4-2.6 due to cooperative Jahn Teller effect for the Cu
2+
3d
9
ion
that removes the degeneracy of the and orbital.
The bcc CuO
2
layers are intercalated between insulating rocksalt fcc AO layers, This
second material fit in the heterostructure by rotating its orthorhombic axis a
0
=b
0
by 45
0
and
for the distance
The electronic structure of the CuO
2
plane is a charge transfer insulator with a half
filled valence band. The covalency of the Cu-O bond is very high and the single hole per
Figure 1. The heterostructure of a Cu
2+
cuprate perovskite and the molecular orbitals forming the electronic
structure of the CuO
2
plane.
377
Cu ion is both in the orbital and/or in the molecular orbital combination of the
oxygen 2p orbitals of local bi symmetry There is a strong
local Coulomb repulsion for two holes in the same Cu 3d orbital, U
dd
~6 eV, that gives a
Mott-Hubbard gap for the charge transfer where indicates a
hole in the orbital.
The gap for the electron transfer of a hole from the Cu(3d
9
), or to the oxygen orbital
is smaller than the Hubbard gap. This charge transfer gap is given
by where in the final state configuration there is a Coulomb repulsion U
dL
between a hole on Cu and a hole on the nearest oxygen, and is the energy separation
between O(2p) and orbital. The relevant local inter-atomic Coulomb repulsion
U
dL
~2 eV has been determined by joint x-ray photoemission and x-ray absorption and the
optical gap for the insulating compound since [18].
The metallic phase in the CuO
2
plane is obtained by two separate steps in the design of
the material: first, chemical dopants that play the role of acceptors and pump electrons from
the CuO
2
plane, are introduced in the charge reservoir blocks; second, multiple
substitutions of metallic ions A (A=Ba,Sr,La,Nd,Ca,Y ) in the rocksalt layers are made in
such a way to change the average ionic radius of the rocksalt layers. In cuprates with
multiple CuO
2
layers the rocksalt layers between the copper planes loose completely their
oxygen ions. Doping introduces holes in the O 2p orbital and a single hole remains in
the Cu site [19]. However the symmetry of the molecular orbital for the added hole have a
mixed symmetry with a component with local a
1
symmetry
mixed with These results have
shown that the symmetry of the doped holes is not the pure m"=2 symmetry of the
antiferromagnetic insulator at half filling. Therefore the doped holes in the metallic phase
are associated with a local lattice distortions (LLD) mixing states with different orbital
momentum. These LLD distortions are expected for a pseudo Jahn Teller electron lattice
interaction of the doped holes [23]. Here the key point has been to show that the electronic
correlation lower the Jahn-Teller energy separation between the states with and
symmetry from about 1.5 eV to about 0.5 eV since the Coulomb repulsion U
dL
for
the configuration is much smaller than for
The 2D electron gas in the CuO
2
plane of cuprate perovskites is therefore a strongly
correlated electron gas described by the Hubbard Hamiltonian. Moreover there is a relevant
electron lattice interaction of the type of cooperative pseudo Jahn-Teller coupling of
charges with Q
2
-type local modes. This can be described by the Holstein Hamiltonian with
a next-near neighbour hopping integral t. Therefore its metallic phase is described by the
Hamiltonian:
(4)
The first two terms describe the itinerant charges in a 2D square lattice simulating the
CuO
2
plane where t is the electron transfer integral between nearest-neighbor sites <i,j> and
t` is the electron transfer integral between next-nearest-neighbor sites <<i,j>>, is
the local electron density, denotes the electron creator operator at site i.
378
The third term is the Hubbard Hamiltonian describing the electronic correlation in the
CuO
2
plane. The Hubbard term induces a mass renormalization of a factor of the order of 5
giving an effective mass in the direction, m
*
/m
0
~2.
The coupling of the charges with local lattice distortions (LLD) of the CuO
4
unit can
be described by the Holstein Hamiltonian (H
ph
+H
I
). The position of the lattice site is
indicated by R
i
and represents the creation operator for phonon with wavevector q, is
the frequency of the optical local phonon mode and g indicates the coupling of the charge
with this local lattice mode. This term describes the weak electron-phonon interaction
while for g sufficiently large the charge is coupled with local lattice distortions.
The local lattice distortion Q follows the equation
therefore the electron-lattice interaction provides a force
that induces a displacement of the equilibrium position. The
local lattice distortion (LLD) appears when becomes larger than zero energy vibration
amplitude. In the present square lattice it is possible to identify the electron lattice coupling
constants where d=2 for a 2D electron gas. We have
found that in the cuprate perovskites, while the first coupling constant is in the weak
coupling limit, the second one is in the intermediate-strong coupling limit
and it is expected to give local lattice distortions. In this situation we are in an intermediate
regime where charges trapped into LLD coexist with itinerant charges. This situation is
expected to occur in special cases in the intermediate electron-lattice coupling regime.
The experimental evidence for LLD due to pseudo JT electron-lattice interaction (JT-
LLD) was provided by the presence of two different types of doped holes in the oxygen
orbital [20-23]:
of partial a
1
symmetry, mixed with (orbital angular momentum
and
of b
1
symmetry mixed with (orbital angular momentum
The pseudo JT-LLD should be associated with the doped holes since the Q
2
type
lattice distortion forms molecular orbital of mixed and angular momentum.
The search for these local lattice distortions motivated the Rome group to solve the
incommensurate structural modulation of the CuO
2
plane in Bi
2
Sr
2
CaCu
2
O
8.2
(Bi2212) by
joint single crystal x-ray diffraction and EXAFS. We have found in 1992 that the pseudo
JT-LLD get self organized in linear arrays, i.e., stripes [24-26]. The co-existing itinerant
particles form rivers of charges and at the Erice workshop in 1992 [25] the scenario of
superconducting stripes, where the free charges move mainly in one direction, like the
water running in the grooves of a corrugated iron foil , was introduced for the first time in
the field of high T
c
superconductors.
A heterogeneous phase of the matter is a generic phenomenon following doping of a
high correlated antiferromagnetic insulating electronic system. The formation of a
microscopic electronic phase separation with the formation of metallic droplets in a
antiferromagnetic background was first shown in doped magnetic semiconductors [27].
Experimental evidence that at very low doping in the cuprates the doped holes segregate
into strings of charges that play the role of domain walls between anti ferromagnetic
domains, forming a glassy phase of strings, has been reported [28]. At higher doping if the
counterions are mobile the system is unstable toward a macroscopic phase separation
between macroscopic metallic domains and insulating antiferromagnetic domains [29,30].
379
The high T
c
superconductors are a special case of heterogeneous doped magnetic
superconductors since in the metallic droplets we have the coexistence of doped charges in
the weak coupling limit, phase A, with doped charges associated with local lattice
distortions associated with pseudo Jahn Teller electron lattice interaction phase B. There is
now clear experimental evidence that there are two types of doped charges in the cuprates
[31]. The ordering of charges trapped by the pseudo Jahn-Teller LLD with an associated
modulation of the orbital angular momentum gives stripes and orbital density waves.
The phase diagram of the metallic phase of high T
c
superconductors is usually given
as a function of hole doping, measuring the distance from the antiferromagnetic (AF)
insulating Mott Hubbard phase. In the two components 2D electron fluid it is necessary to
measure the actual density of the itinerant component by using the electron density
parameter r
s
measured by the Hall effect at low temperature.
The phase diagram of La
2-x
Sr
x
CuO
4
as a function of electron density parameter r
s
is
shown in Fig. 2 (lower panel).
Figure 2. Phase diagram of the La
2-x
Sr
x
CuO
4
(lower) and (upper) as a function of electron density
parameter r
s
of the itinerant 2D electron gas measured by Hall effect. Usual notations are used to denote
different regimes of the phase diagram. The structural phase transition boundary between the orthorhombic
and tetragonal phase is shown.
The system La
2-x
Sr
x
CuO
4
shows a complex phase diagram typical of a glassy system
due to the random distribution of countercharges (Sr ions) in the block layers. The AF
phase appears in the range r
s
>37, and a spin glass phase appears for 37>r
s
>15. The metallic
phase, 15>r
s
>5, shows a typical glassy phase with several crossover temperatures that
depend on the measuring time of each experimental probe.
380
To understand the basic physics of the metallic phase of cuprate superconductors we
need to study a simple crystalline system. This is provided by (and
where the itinerant holes in the CuO
2
plane are compensated by the
negative charges carried by the mobile interstitial oxygen in the LaO layers (and in the
BiO charge reservoir layer) that can get ordered. There is no frustrated phase separation
regime in due to mobile counterions, therefore it does not show the spin glass
phase of the doped Mott insulator in the range 37>r
s
>15, where it shows the expected phase
separation below about 300K between an insulating doped AF lattice (r
s
~37) and a metallic
phase (r
s1
~12).
This first superconducting phase with T
c
=32 K shows the universal 1D dynamical spin
fluctuations below T
sdw
~60K and the 1D stripes [32], CDW and/or ODW, indicated by the
ordering of local lattice distortions of the CuO
6
octahedra (tilting) below a critical
temperature T
c0
=190 K [33], as shown in Fig. 2 (upper panel).
A second stable phase with the highest critical temperature appears at 5<r
s2
<6. In this
phase, oxygen ordering and charge and orbital stripes in the CuO
2
plane have been found
by several experimental techniques probing different physical parameters: NMR revealing
two different Cu sites [34], EXAFS solving the local CuO
4
rhombic distortions,
characteristic of the pseudo Jahn Teller polarons below 150 K [35], x-ray [36] and electron
diffraction [30]. The universal 1D dynamical spin fluctuations at and
have been also observed in this regime below 60 K by inelastic magnetic neutron
scattering (37).
In a recent work we have identified the short range incommensurate charge ordering
in the CuO
2
plane reflected by a pattern of diffuse x-ray scattering peaks with wavevector
q
CDW
=(0.208b*,0.29c*) and a coherence length of about 350 . The anharmonicity of this
modulation is evident from large intensity of higher harmonics. By cooling the sample the
diffuse charge ordering peaks show a temperature dependence reported in Fig. 3, where the
square root of the intensity of the second harmonic peak is plotted.
Figure 3. Temperature dependence of the order parameter for the stripe formation with wavevector
in La
2
CuO
4.1
. The fit shows a critical temperature T
co
=190 K for charge ordering.
In the second harmonic peaks we can well separate the charge ordering from the 3D
oxygen ordering peaks. The square root of the intensity plotted in Fig. 3 gives the direct
measure of the density of charge ordered in the CuO
2
plane with wavevector q, that
is the order parameter for the charge ordered phase. The solid line is a fit to the
381
experimental intensities with an expression which represents a typical second
order phase transition with T
co
~188 K. This effect is clearly due to formation of charge
stripes in the CuO
2
plane since the oxygen mobility is frozen below 200K. In fact the 3D
oxygen ordering has already been established at higher temperature (in the range 270-230
K) in the system as evidenced by temperature evolution of the resolution-limited diffraction
peaks. The second harmonic of the charge modulation at 0.416b* has the same wavevector
of the nesting vector at 2k
F
~0.4b* or observed by Saini et al. [18] in the
Fermi surface of Bi2212 that induces the suppression of the spectral weight at selected
spots in the k space and gives a broken Fermi surface. The 3D ordering of dopants
stabilizes the orthorhombic phase, as shown in Fig. 1, and the symmetry of the CuO
2
plane
is broken. The ordering of stripes in the b direction gives a superconducting phase in a
broken symmetry, with a broken Fermi surface. The Fermi surface is therefore formed not
by closed circles but by segments and the "mini-gaps" in the density of states due to the 1D
superlattice of stripes [19-20] give origin to the "pseudo-gap" scenario. In this stripe
scenario the amplification of the superconducting critical temperature is realized by tuning
the Fermi level at a "shape resonance" of the superlattice [19-20]. The pairing mechanism
is mediated by charge fluctuations in a metal with the anomalous dielectric constant typical
of an anomalous Fermi fluid at r
s
>4 [17].
Figure 4. The critical temperature of different cuprate perovskites at fixed doping at (open circles) and
(open squares) as function of the ratio /m*, where is the condensate density and m* the effective
mass. The dashed line shows the calculated T
c
using formula (5) for the coherence length of the order of
the distance between particles <d>=2r
s
a
B
.
In this regime a generalized BCS scheme remains valid and the dynamical pairing
described by the BCS is still possible up to the point where the size of the pairs of quasi-
particles is of the order of the wavelength of the electrons at the Fermi level In this
regime the coexistence of local pairs (bosons) and fermions in the normal phase is
expected. The proximity to Wigner localization gives a metal with a negative dielectric
permitivity predicted by Dolgov and Ginsburg, needed for high T
c
with The highest critical temperature is reached for close to the average
382
distance between two electrons <d>=2r
s
a
B
in a 2D electron gas. This is the highest critical
temperature possible in an extended BCS scheme in the intermediate coupling:
(5)
where is measured in
-2
, and m* is the effective mass. In this limit the critical
temperature depends only on the ratio and using the phenomenological value, f=2,
from tunneling data at optimum doping, the critical temperature has been calculated as a
function of in ref. 17. The predicted temperature is plotted in Fig. 4 and it is in very
good agreement with the experimental data.
The expected density of charge carriers is constant but the condensate density is
different in different cuprate perovskites. Therefore there is a term in the Hamiltonian of
the metallic phase of the CuO
2
layers, that changes the electronic structure and drives the
critical temperature and the condensate density. The object of this work is the identification
of the term in the Hamiltonian that drives the CuO
2
plane to the optimum intermediate
coupling regime giving high T
c
superconductivity.
THE PHASE DIAGRAM OF BI2212
After the discovery of HTcS the physics of cuprates was described by a generic phase
diagram where the critical temperature is plotted as a function of doping i. e., a
measure of the charge density and the distance from the Mott Hubbard insulator at By
increasing the system goes through quite different states. At low doping the doped holes
form a disordered electron glass. At very high doping a normal metallic phase appears. The
high T
c
superconducting phase appears between these two phases. In 1993 we presented the
phase diagram for Bi2212, shown in Fig. 5. The doping was measured in units of
Figure 5. The phase diagram for Bi2212. The charge density is measured in units of the critical doping
for the expected commensurate polaron crystal (CPC) also if in this family by changing the doping we
do not cross the CPC phase. We observe 1) an insulating phase A of an electron gas in the localization limit
(k
F
1<1) for 2) a homogeneous metallic phase B at and 3) a region of co-existence of the two
different phases, a 1D-polaronic incommensurate charge density waves and the superconducting
phase.
383
the critical density for the insulating commensurate JT charge ordered crystal at in
La
1-x
Ba
x
CuO
4
This electronic crystal is in competition with superconductivity and it is at
the origin of the huge suppression of the superconducting critical temperature in La
1-
x
Ba
x
CuO
4
at x=l/8. The long range Coulomb interaction between charge trapped in pseudo
JT-LLD is expected to play a key role in the formation of the ordered phase of localized
charges. Therefore we call this phase a generalized Wigner commensurate JT polaron
crystal (CPC).
The CPC was not observed at in the phase diagram of Bi2212 shown in
Fig. 5 where only a weak minimum of T
c
appears at By decreasing the temperature,
below about T*=T
CO
=130 K, the system form an inhomogeneous phase where a one-
dimensional (1D) incommensurate polaronic charge density wave (ICDW) or polaronic
stripes coexist with free carriers. In the underdoped phase, the charge density
of free carriers is smaller than that of JT polarons in the ICDW. In the high doping phase,
the charge density of free carriers is larger than that of JT polarons in the ICDW.
This particular ICDW does not suppress but promotes the pairing of the free carriers below
T
c
. In fact in this unexpected metallic phase of condensed matter, a superlattice of quantum
mesoscopic stripes of width L, the chemical potential is tuned to a "shape resonance". The
"shape resonance" occurs when the de Broglie wavelength of electrons at the Fermi level
The measure of the stripe width L in 1993 has established the presence of the
"shaperesonance" at optimum doping. At the shape resonance the chemical potential
is tuned to the bottom of a superlattice subband and therefore to a narrow peak in the
density of states (DOS). At optimum doping a BCS-like superconductivity is observed
and the highest T
c
is reached where the chemical potential is tuned to this narrow
DOS peak via the calculated "shape resonance" effect on the superconducting gap. A patent
has been granted with priority date 7 Dec 1993 for a method of T
c
amplification via the
"shape resonance" effect in new materials formed by a superlattice of quantum wires. A
very good agreement with experimental data has been found for the calculated critical
temperature plotted as a solid line in Fig. 5 assuming the pairing mechanism mediated by
charge fluctuations in a superlattice of quantum wires at the "shape resonance" [17].
Figure 6. The average <Cu-O> bond length measured by Cu K-edge EXAFS as a function of the average
radius <r> of metal ions in the rocksalt layers.
384
THE MICRO STRAIN QUANTUMCRITICAL POINT
The electron-lattice interaction g of the pseudo JT type in the cuprates is controlled by
the static distortions of the CuO4 square plane and the Cu-O(apical) distance. In fact
where Q is the conformational parameter for the distortions of the CuO
4
square, like the rhombic distortion of CuO
4
square; is the dimpling angle given by the
displacement of the Cu ion from the plane of oxygen ions; and is the JT energy splitting
that is controlled by the Cu-O(apical) bond length.
Figure 7. The critical temperature for charge ordering and the superconducting critical temperature T
c
as
function of the micro strain at optimum doping
There is an external field acting on the CuO
2
plane of the cuprates that controls
via the micro-strain of the CuO
2
lattice due to the compressive stress generated by the
lattice mismatch between the metallic bcc CuO
2
layers and the insulating rocksalt fcc AO
layers [40,41]. The bond-length mismatch across a block-layer interface is given by the
Goldschmidt tolerance factor where [r(A-O)] and [r(Cu-O)]=d
0
are
the respective equilibrium bond lengths in homogeneous isolated parent materials A-O and
CuO
2
[42]. The hole doped cuprate perovskite heterostructure is stable in the range 0<t<0.9
that corresponds to a mismatch of
We have focused our attention to 3 main cuprate perovskite systems 1) Hgl212, 2)
Bi2212 and 3) La
2
CuO
4+y
. In these materials the hole concentration in the CuO
2
plane is
controlled by the oxygen doping in the charge reservoir blocks. The stress due to the
mismatch or the chemical pressure acting on the CuO
2
plane is controlled by the average
ionic radius in the rocksalt layers. The stress increases going from Ba to Sr to La. We have
measured the average Cu-O bond lengths by Cu K-edge EXAFS, a local structural probe,
and shown in Fig. 6 as a function of average ionic radius of metallic ions in the rocksalt
layers <r>. Decreasing <r> is equivalent to a anisotropic chemical pressure acting on the
CuO
2
plane. We define the local or micro strain of the CuO in plane bond length
<Cu-O>/d
0
), where d
0
=1.97 is the equilibrium Cu-O distance at doping in many
different systems.
385
In the cuprate perovskites the micro-strain e drives the system to a quantum critical
point for the formation of a superlattice of quantum stripes. The stripes of local
lattice distortions are detected by x-ray diffraction above a critical micro-strain
We have plotted in Fig. 7 the variation of the critical temperature for charge ordering
T
co
and the superconducting critical temperature T
c
as function of the micro-strain at
optimum doping
Figure 8. The superconducting critical temperature T
c
plotted in a half-tone scale (from T
C
=0K, black, to T
c
~ 135K, white) as a function of the micro-strain and doping The maximum
Tc
occurs at the critical point
and
In Fig. 8 we report the critical temperature T
c
in a color plot (the critical temperature
increases from black, T
c
=0K, to white, the maximum T
c
~ 135K) as a function of the micro-
strain and doping for all superconducting cuprate families. The figure shows that the
maximum T
c
occurs at the critical point
From these data we can derive a qualitative phase diagram for the normal metallic
phase of all cuprate perovskites that give high T
c
superconductivity that is shown in Fig. 9.
This phase diagram solves the long standing puzzle of the phase diagram of the
normal phase of the cuprates. There was a hidden physical parameter, the micro-strain, that
triggers the electron-lattice interaction at a critical value for the onset of charges trapped
into pseudo JT-LLD. The doping of the strained antiferromagnetic lattice forms both free
carriers and charges trapped into the JT-LLD above the critical micro-strain For as
it was discussed for the case of oxygen doped Bi2212 and La 124, the systems show a quasi
first order phase transition as a function of doping.
The quantum critical point QCP is well defined at constant finite doping as a function
of the micro-strain as it is shown in Fig. 8. Direct experimental evidence for quantum
critical local lattice fluctuations has been obtained by measuring the dynamical fluctuations
of the Cu-O bond at a high temperature T
H
>T
CO
in all families of cuprates (T
H
~200K).
In conclusion we have deduced a phase diagram for the superconducting phases where
T
c
depends from both doping and micro-strain. The anomalous normal phase of cuprate
superconductors is determined by an inhomogeneous phases with co-existing polaronic
stripes and itinerant carriers that appears for an electron lattice interaction larger than a
critical value. Fluctuations of lattice-charge stripes appear in this critical regime. The
micro-strain drives the electron lattice interaction to a QCP of a quantum phase transition
[43]. Near this QCP the stripes get self organized in a superlattice of quantum wires of
charges trapped into JT-LLD that co-exist with free carriers. This superlattice forms an
386
array of superstripes where the chemical potential is tuned to a shape resonance. The plot
reaches the highest temperature at the critical point
Figure 9. The phase diagram of the normal phase of doped cuprate perovskites as a function of micro-strain
on the Cu-O(planar) bond and doping. The high T
c
superconductivity occurs in the region of quantum
fluctuations around the micro-strain quantum critical point QCP.
This research has been supported by INFM, by MURST - Programmi di Ricerca
Scientifica di Rilevante Interesse Nazionale, and by "Progetto 5% Superconduttivit del
CNR.
REFERENCES
1. H. Kamerlingh Onnes Comm. Phys. Lab. Univ. Leiden Nos. 119, 120, 122 (1911).
2. W. Meissner & R. Ochsenfeld Naturwiss 21,787 (1933).
3. F. London Superfluids J. Wiley and Sons Inc. New York, 1950 Vol. 1 pag. 152
4. R. Doll & M. Nabauer Phys. Rev. Lett. 7, 51 (1961); B. S. Deaver Jr. and W. M. Fairbank Phys. Rev.
Lett. 7, 43 (1961).
5. B. D. Josephson Physics Lett. 1, 251 (1962).
6. P. B. Allen & B. Mitrovic Solid State Physics 37, 1 (1982).
7. H. London, & F. London Proc. Roy. Soc. (London) A149, 71 (1935); Physica 2, 341 (1935).
8. J. Bardeen, L. N. Cooper, and J. R. Schrieffer Phys. Rev. 108, 1175 (1957).
9. J. G. Bednorz, & K. A. Mller Z. Phys. B 64, 189(1986); J. G. Bednorz and K. A. Mller Rev. Mod.
Phys. 60 565(1988).
10. C. W. Chu et al. Nature 365, 323 (1993).
11. J. -M. Triscone, et al Phys. Rev. Lett. 63, 1016 (1989).
12. Qi Li, T. Venkatesan & X. X. Xi Physica C, 190, 22 (1991).
13. M. Holcomb, J. P. Collman, & W. A. Little Physica C, 185-189, 1747(1991)
14. Y. J. Uemura et al. Phys. Rev. Letters 62, 2317 (1989); Phys. Rev. B 38, 909 (1988); Phys. Rev. Lett.
66, 2665 (1991);Nature 364, 605 (1993).
15. H. Keller, et al. Physica (Amsterdam) 185-189C, 1089 (1991); T. Schneider & H. Keller Phys. Rev.
Lett. 69, 3374 (1992);
16. Ch. Niedermayer et al. Phys. Rev. Lett. 71, 1764 (1993).
387
17. A. Bianconi Sol. State Commun. 91, 1 (1994).
18. A. Bianconi, A. Clozza, A. Congiu Castellano, S. Della Longa, M. De Santis, A. Di Cicco, K. Garg, P.
Delogu, A. Gargano, R. Giorgi, P. Lagarde, A. M. Flank, and A. Marcelli in International Journal of
Modern Physics B 1, 853 (1987).
19. A. Bianconi, A. Congiu Castellano, M. De Santis, P. Rudolf, P. Lagarde, A. M. Flank, and A. Marcelli,
Solid State Communications 63, 1009 (1987).
20. M. Pompa, S. Turt, A. Bianconi, F. Campanella, A. M. Flank, P. Lagarde, C. Li, I. Pettiti, and D.
Udron, Physica C185-189, 1061-1062 (1991).
21. M. Pompa, P. Castrucci, C. Li, D. Udron, A. M. Flank, P. Lagarde, H. Katayama-Yoshida, S. Delia
Longa, and A. Bianconi, Physica C184, 102-112 (1991).
22. A. Bianconi M. Pompa, S. Turtu, S. Pagliuca, P. Lagarde, A. M. Flank, and C. Li Jpn. J. Appl. Phys. 32
(Suppl. 32-2) 581(1993).
23. Y. Seino, A. Kotani, and A. Bianconi J. Phys. Soc. Japan 59, 815 (1990).
24. A. Bianconi, S. Delia Longa, M. Missori, I. Pettiti, and M. Pompa in Lattice Effects in High-T
c
Superconductors, edited by Y. Bar-Yam, T. Egami, J. Mustre de Leon and A. R. Bishop; World
Scientific Publ., Singapore, (1992) pag. 6.
25. A. Bianconi, in "Phase Separation in Cuprate Superconductors" ed. by K. A. Muller & G. Benedek,
World Scientific Singapore (1993) p. 352; ibidem pag. 125.
26. A. Bianconi, S. Delia Longa, M. Missori, I. Pettiti, and M. Pompa and A. Soldatov Jpn. J. Appl. Phys.
32 suppl. 32-2, 578(1993).
27. E. L. Nagaev Sov. Jour. JEPT Lett. 16, 558 (1972); V. A. Kaschin and E. L. Nagaev Zh. Exp. Teor.
Phys. 66, 2105 (1974).
28. J. H. Cho, F. C. Chu and D. C. Johnston, Phys. Rev. Lett. 70, 222 (1993).
29. P. C. Hammel, A. P. Reyes, Z. Fisk, M. Takigawa, J. D. Thompson, R. H. Heffner, S. Cheong, and J. E.
Schirber Phys. Rev. B 42, 6781 (1990).
30. J. C. Grenier, N. Lagueyte, A. Wattiaux, J. -P. Doumerc, P. Dordor, J. Etourneau and M. Pouchard,
Physica C 202, 209. (1992).
31. D. Mihailovich, and K. A. Mller, High T
c
Superconductivity: Ten years after the Discovery (Nato ASI
Series-Applied Sciences, Vol. 343) ed E. Kaldis, E. Liarokapis, K. A. Mller, (Dordrecht, Kluwer) pag.
243-256.
32. B. O. Wells, Y. S. Lee, M. A. Kastner, R. J. Christianson, R. J. Birgeneau, K. Yamada, Y. Endoh, and
G. Shirane Science 277, 1067-1071 (1997).
33. X. Xiong, P. Wochner, S. C. Moss, Y. Cao, K. Koga and N. Fujita, Phys. Rev. Letters 76, 2997 (1996).
34. B. W. Statt, P. C. Hammel, Z. Fisk, S. W. Cheong, F. C. Chou, D. C. Johnston and J. E. Schirber, Phys.
Rev. B 52, 15575(1995).
35. P. G. Radaelli, J. D. Jorgensen, R. Kleb, B. A. Hunter, F. C. Chou and D. C. Johnston, Phys. Rev. B 49,
6239(1994).
36. A. Lanzara, N. L. Saini, A. Bianconi, J. L. Haemann, Y. Soldo, F. C. Chou, and D.C. Johnston Phys.
Rev.B 55, 9120 (1997).
37. Y. S. Lee, R. J. Birgeneau, M. A. Kastner, Y. Endoh, S. Wakimoto, K. Yamada, R. W. Erwin, S. -H.
Lee and G. Shirane Phys. Rev. B 60, 3643 (1999).
38. A. Bianconi, D. Di Castro, G. Bianconi, N. L. Saini, A. Pifferi, M. Colapietro F. C. Chou & D. C.
Johnston Physica C Proc. of the M2S Houston Conf. 2000.
39. A. Bianconi, M. Missori, N. L. Saini, H. Oyanagi, H. Yamaguchi, D. H. Ha, Y. Nishihara J.
Superconductivity 8, 545 (1995); A. Bianconi, M. Missori, H. Oyanagi, H. Yamaguchi, D. H. Ha, Y.
Nishihara, and S. Delia Longa Europhysics Lett. 31, 411 (1995);A. Bianconi, N. L. Saini, T. Rossetti,
A. Lanzara, A. Perali, M. Missori, H. Oyanagi, H. Yamaguchi, and Y. Nishihara, D. H. Ha, Phys. Rev.
B 54, 12018 (1996); A. Bianconi, M. Lusignoli, N. L. Saini, P. Bordet, . Kvick, and P. Radaelli Phys.
Rev. B 54, 4310 (1996).
40. C. N. R. Rao and A. K. Ganguli Chem. Soc. Rev. 24, 1 (1995).
41. P. P. Edwards, G. B. Peakok, J. P. Hodges, A. Asab, and I. Gameson in, High T
c
Superconductivity:
Ten years after the Discovery (Nato ASI, Vol. 343) ed E. Kaldis, E. Liarokapis, and K. A. Mller,
(Dordrecht, Kluwer) (1996) 135.
42. J. B. Goodenough Supercond. Science and Technology 3, 26 (1990); J. B. Goodenough and A.
Marthiram J. Solid State Chemistry 88, 115 (1990).
43. S. Sachdev Quantum Phase Transitions (Cambridge Univ. Press, New York, 1999).
388
ELECTRON STRINGS IN OXIDES
F.V.Kusmartsev
School of MAP, Loughborough University, LE11 3TU, and
Isaac Newton Institute for Mathematical Sciences
20 Clarkson Road, Cambridge, CB3 0EH, UK,
Landau Institute, Moscow, Russia
INTRODUCTION
The electronic phase separation scenario originally introduced by Nagaev into the physics
of magnetic semiconductors[1] is very popular now as a possible scenario for a creation of
a colossal magnetoresistance phenomenon[2-7]. The idea of string formation also belongs
to this stream flow[8,9]. The phenomenon of the colossal magneto-resistance observed in
hole doped manganites [10-14]. (see, also references inside) may have a natural explanation
in such a novel framework. The strings are arising in solids with narrow bands[8,9] due to
an effect of a self-traping(ST). For a single particle the self-trapping phenomenon has been
intensively studied in the past (see, for example, in Ref.[15-22].
Since the mass of electrons is a lot smaller than the atomic masses the electrons move
much faster than the atoms. Therefore, conventionally, an adiabatic approximation for the
motion of the atomic lattice is assumed. If we are interested in stationary points of the adi-
abatic potential the kinetic energy of the atoms may be neglected. The state of a charged
particle together with the surrounding static deformation associated with a stationary point
of the adiabatic potential was originally called a deformon and was later conventionally
termed a self- trapped (ST) state which is analogous to a conventional polarons[18-22]. The
effective radius of this deformation cloud is the radius of the self-trapped state.
Such an effect is strongly enhanced for low dimensional systems or for systems with
strong electron-phonon interactions [17]. In 3D crystals with wide bands the self-trapping
process is associated with a penetration of the electron-phonon system through a self-trapped
barrier[23-25]. while in 1D system the self- trapped barrier is absent. In 2D systems the self-
trapping arises only if the electron-phonon coupling is greater than some threshold value[17].
The origin of ST barrier in 3D systems and the ST threshold are due to a competition (or a
balance) between the kinetic and potential energies of the electrons(holes)[23].
The most interesting self-trapped phenomena to date include self-trapped excitons ob-
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phi l l i ps and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 389
served in rare gas solids (see, Ref.[17-26] and references therein). The creation of such states
gives rise to an electronic mechanism of the defect creation in solids[24,27]. That is the cre-
ation of self-trapped excitons by light causes structural defects in solids. The self-trapping
terminates the tails in the electronic density of states in disordered semiconductors[25,28].
Finally the self-trapping gives rise to a deformation arising in vortex cores[29] observed re-
cently in superconducting Nb by polarized neutron scattering[30].
In the case of the many-particle self-trapping (a strings formation) the situation changes
drastically since the potential deformation or polarization energy as well as a Coulomb energy
are dominant. The self-trapping process is barrierless in such a state the deformation energy
takes maximal possible value that causes a localization of many particles in a potential of a
linear shape. Here, our main result is that the phenomenon of electronic string-defect forma-
tion - a multi-particle self-trapping[8,9] may also arise in manginites due to an interaction of
electrons(holes) with acoustical and Jahn-Teller phonons.
Note that this phenomenon is arising only in solids with narrow bands where the kinetic
energy of electrons is limited by the small width of the band. Therefore, due to the small
kinetic energy of electrons (holes) in a single deformation potential well many charged par-
ticles may be self-trapped. In this case the ST deformation well is self-consistently created
by all the particles trapped by this well. Although the shape of this well may be arbitrary, it
is primarily dictated by a Coulomb repulsion between the particles. The contribution of the
electron kinetic energy may be important for the formation of conducting string associated
with long-range electron-phonon interaction with polar or longitudinal optical phonons[9].
Qualitatively, if the deformation potential well traps M electrons(holes), the lattice deforma-
tion Q is proportional to M and consequently the elastic energy of the lattice is proportional
to Q
2
~ M
2
. Hence, the electron kinetic energy of the self-trapped particles and, therefore,
the lattice adiabatic potential of the ST state decreases as ~ E
P
M
2
, where E
p
is a polaron
shift. However, this decrease is opposed by the Coulomb repulsion between the particles
trapped by the string well, which energy has an additional log M multiplier[8,9] so that the
Coulomb energy is approximately equal to ~ VM
2
log M, where V is a constant of the inter-
site inter-electron Coulomb repulsion. A balance between these energies which is determined
by a minimisation of the total energy gives a stationary many particles ST string state where
the string length N ~ M ~ exp (E
p
/V 1/2), (see, for example, and compare with Ref.[8].
The precise dependence of the Coulomb energy of the trapped particles depends strictly
on the shape of the self- trapped spot. When all particles well separated, of course, the
Coulomb energy is minimal. But in this case the deformation energy and the adiabatic po-
tential are not of the lowest one. For the multi-particle self-trapping spot the lowest Coulomb
energy corresponds to a linear chain when all the self-trapped particles are located along a
single line. We call such a many-body self-trapping object as a self-trapped string.
We describe here an electron(hole) string consisting of M trapped particles and with the
length Na created due to electron-phonon interaction with Jahn-Teller and acoustical phonons
in a hypercubic d dimensional lattice where a is a characteristic distance between nearest-
neighboring atoms. In a doped antiferromagnet such a string may have a lower energy than
M spatially separated self-trapped particles. Such ST strings in a hypercubic lattice may
take arbitrary nonlinear configurations, which correspond to different highest values of the
adiabatic potential. Indeed, recent neutron scattering and electron microscopy experiments
[12-14] indicate on electronic phase separation and stripe or string ordering in semiconduct-
ing La
1x
Ca
x
MnO
3
and analogous compounds which may be related to the effect of string
formation as described in the present paper.
With the increase of the doping the forming insulating electron strings may be ordered
into the parallel stripes[12-14] If the material consists of layers with ordered magnetic mo-
ments like LaMnO
3
then the threshold of a bond or a site percolation in the 2D plane (a single
layer) is equal to x
c
= 0.5. Since the strings are charged objects then it is very probable that
at such a percolation threshold the percolating charged strings will be ordered. As a result
390
the system of charged parallel stripes as it was observed in the recent experiments[12] will be
formed. However for a proper comparison with existing experimental data[10-14] an inves-
tigation of the possible microscopic forms of electronic phase separation (like, the described
string formation) in the framework of the double exchange model of spinful fermions having
orbital degrees of freedom and interacting with Jahn-Teller phonons is urgently needed. The
present work may serve as an attempt in this direction.
HAMILTONIAN
We consider a Hamiltonian of charged spinful fermions having orbital degrees of free-
dom and interacting with acoustical and Jahn-Teller phonons on a d dimensional hypercubic
lattice:
(1)
where is a matrix of the electron hopping-integrals, the operator creates (destroys)
a fermion at a lattice site i, n
i
, is a density operator which is expressed via the occupation
number operator as and the operator is an operator of the
creation (destruction) of an branch of phonons. We consider Jahn-Teller and acoustical
phonons. The summations in eq.(l) extend over the lattice sites i and -as indicated by <
i, j > -over the associated next nearest sites j. The matrix elements of the electron-phonon
interaction is equal to
(2)
For the acoustical phonons (the dispersion relation where s is a velocity of
sound) and for dispersionless optical (Holstein) phonons the function
const while for the Jahn-Teller phonons The weak
dispersion is associated with the elastic interaction between neighboring octahedra.
To describe these multi-electron strings we employ a variational Hartree-Fock(HF) many-
body wave function and consider adiabatic and antiadiabatic approximations, separately.
ADIABATIC APPROXIMATION
First for an adiabatic limit it is convenient to approach this problem working in the first
quantization form (see, for example, in Ref.[8,9] To build up such HF approximation for
the ST phenomenon we have to find relevant single particle wave functions. The discrete
Schrdinger equations describing a single electron(hole) interacting with a lattice potential
matrix associated with the strain deformation, and Jahn-Teller distortions in a
tight-binding model on a hypercubic lattice has the form:
(3)
where is a hopping integral matrix, for similicity we will consider here only a diagonal
matrix The operator is a lattice version of the Laplacian operator which for the
hypercubic lattice is defined as:
(4)
391
where the summation is carried out over all the nearest-neighbor sites around the n-th site;
is the wave function of the pth self-trapped electron(hole) on the nth site associated
with the orbital and we use the units where a = 1.
JAHN-TELLER PHONONS AND STRING SOLUTIONS
When an electron or a hole is interacting with Jahn-Teller distortions of the lattice there
may arise the similar electron strings. The Jahn-Teller distortions are typical for oxides and
manganites such as La
2
CuO
4
and La
2
MnO
3
or other their relatives. In both cases there are
oktahedron configurations of CuO
6
. At the presence of the cubic symmetry for the ions
Cu
2+
they are associated with the degenerate Jahn-Teller term E
g
. The presence of atoms
La descreases the cubic symmetry to tetragonal, due to Jahn-Teller effect and, in principle,
removes the degeneracy. In this case the Jahn-Teller Hamiltonian of the electron-phonon
interaction has the form[31,32]:
(5)
where this matrix acts on the coefficients and of the two-component electron wave
function expanded with respect to the symmetry basis The
value D is the constant of deformation potential created by the defortmation
where are unit basis vectors of the hypercubic lattice. This model is very
different from Holstein-Hubbard model used, for example, in Ref.[33].
With the use of this Hamiltonian, the Shrdinger equations have the form:
(6)
(7)
With the aid of these Shrdinger equations and taking into account the Hamiltonian of the
elastic and Jahn-Teller lattice distortions which has the form:
(8)
where in the elastic interaction between neighboring octahedra the values are defined
as and < n, m> are the
nearest neighbor pairs of octahedra along the direction and (see, for example in
Ref.[32]).
We may built up the total adiabatic potential of the lattice, which also includes the
Coulomb and exchange repulsion between the self-trapped electrons(holes) as an extra term
V
HF
which will be explicitely given lately:
(9)
where the Hartree-Fock term must be modified properly to take into account the two compo-
nent many body wave function.
392
In adiabatic approximation neglecting the atomic kinetic energy after a minimization of
H with resepect to Q
3,n
and Q
2,n
we have
(10)
(11)
and by a minimization of H with respect to the deformation we obtain the system of
discrete equations describing elastic strain deformations created by M electrons(holes) in an
approximation of isotropic elastic medium:
(12)
where the index p indicates a summation carried over all M trapped electrons(holes). After
the solution of these equations with respect to uknown deformations and next substitution
these found expressions into the Shrdinger equations obtain a new system of complicated
nonlinear equations. The system is immediately simplified if we put K' = 0. Then we have
the following system of nonlinear equations:
(13)
(14)
where and another
couple of equations for the complex conjugation of these wave function. This system of four
equations may be simplified to the case to a single component nonlinear Shrdinger equa-
tion. So, for example, if we put while we recover the conventional nonlinear
Shrdinger equations[8,9]. Another self-consistent substitution is which sympli-
fies this system to the same conventional discrete nonlinear Shrdinger equations discussed
in Refs.[8,9]. In general these two solutions are following from a self-consistent substitution
which remains the invariant the strain tensor: and The value of
the parameter may be found by a minimization of the total energy associated with Jahn-
Teller distortions which gives that the value
After the complete exclusion of phonon variables the extremal points (minima and max-
ima) of total adiabatic potential, the Hamiltonian H
total
are determined then with the aid of
the following nonlinear Schrdinger equations(NSE):
(15)
where the operator is defined as and the coupling constant, c', is defined
as c' =
The described wave functions associated with fermions self-trapped into the string cor-
respond to the following eigenvalues:
(16)
where d is the dimension of the hypercubic lattice containing the string and the constant J
N
is
a dimensionless integral. When the string is embedded into a 3D atomic lattice the integral,
J
N
, takes the form:
(17)
393
where is a 3D lattice Greens function for a Jahn-
Teller deformations and the value
In the limit and b 1 the NSE allows the following assymptotically exact
string solutions, in which the M electron(holes) are trapped by N neighboring sites with equal
probability, 1/N:
(18)
where k
x
is the momentum of the pth electron. We assume that the string is oriented in
the x direction and is located on the sites starting from n
x
= 1 to n
x
= N. Inside the string
each trapped electron(hole) has a free motion along the string with the momenta k oriented
in the direction of the string. Such plane waves localized inside the string correspond to the
following eigenvalues, E, of NSE:
(19)
where d is the dimension of the hypercubic lattice containing the string and in this solution
we have to define the new coupling constant
The electron (hole) momentum k is simply quantized if we assume that the trapped cur-
rent carriers are spinless fermions and that their wave function satisfies periodic boundary
conditions (PBC) along the string. If we employ other boundary conditions for the trapped
electrons (for example, open boundary conditions) the main result will not be changed drasti-
cally, although in this case the electron density along the string may become inhomogeneous.
With the use of PBC the electron momenta along the string are quantized: k
nx
= /(aN).
With the use of these eigenvalues and the Pauli exclusion principle we calculate the as-
symptotically exact expression for the adiabatic potential A
N, M
, describing M trapped elec-
trons(holes):
(20)
From this expression for A
N,M
(d) one sees that for a single electron (hole), i.e. M = 1,
the lowest energy corresponds to the string with one site, i.e. N = 1. The existence of such
a single site state was noticed by Rashba and Holstein[15-22] but from different arguments.
With the increase of the number of trapped particles, M, while the deformational energy
increases ~M
2
, the value of the adiabatic potential A
N,M
(d) decreases ~ M
2
. This indi-
cates on the electronic phase separation. However, such an electron(hole) phase separation
is strongly prevented by Coulomb forces between the trapped current carriers. The energy of
the Coulomb repulsion is minimal for the maximal separation of individual electrons (holes).
The adiabatic potential of M separated particles is equal to
(21)
From the comparison of these eqs. for J and A
sep
we may conclude that the energy of
the single string A
N,M
(d) coincides with A
sep
if the number of trapped particle, M, is equal to
the length of the string, N.
In the case if we would have spinful particles and the double occupancy will be allowed
the energy of a string with a greater density of particles, with M > N, will be lower than A
sep
since A
N,M
(d) decreases faster with M than the energy of the individual self-trapped parti-
cles A
sep
. This indicates that the separate individual self-trapped particles may be unstable
and so an electron string may be created. The less dense strings (with M < N) correspond
to metastable minima. All these results are based on the assymptotically exact solutions
obtained in the limit
394
LONG-RANGE COULOMB FORCES
A Coulomb repulsion between the electrons(holes) increases when the interparticle dis-
tance decreases and, therefore, is acting against the density increase, i.e. against the phase
separation. However, it may not completely overcome the phase separation but only stabilizes
the strings of a finite length. With the Coulomb interaction taking into account the model of
non-interacting electrons, whose effective interaction was initially introduced only by lattice
vibrations, is modified by an addition of a new term in the Hamiltonian associated with the
two particle Coulomb interaction. The new modified model (with the Coulomb interaction)
may be treated within the Hartree-Fock approximation. Then the extremal points minima and
maxima) of adiabatic potential A is determined with the aid of appropriate Hartree-Fock equa-
tions, which consist of the eqs (6) with the addition of appropriate Hartree-Fock terms. At
large number of the particles trapped (Mc/t >> 1) these modified equations have exact string
solutions, in which the M electrons are trapped by N neighboring sites with equal probability,
1/N. The solution (the many-body wave function) has the form of a Slater determinant of
free single particle wave functions.
The eigenvalues of these new equations modified by Hartree-Fock terms are different, of
course, from those presented in eq.(8), have more complicated and tedious form. However,
the total energy (adiabatic potential) is modified only by an addition of the Hartree-Fock term
V
HF
obtained self-consistently with the use of the obtained, assymptotically exact, solutions
of Hartree-Fock equations. This term depends solely on N and M and is given by:
(22)
where and the paramethter is the effective dielectric constant which may be equal
to a static or a high-frequency dielectric constant depending on the ionicity of the solid. So
for solids like metallic oxides we may take as due to a strong participation of polar
(longitudinal optical) phonons into the screening of the electron-electron interaction. Then,
the function V
HF
may be approximated as
(23)
and behaves similar to that obtained in the electrostatic approximation[9]:
(24)
The total energy consisting of the adiabatic potential A
NM
(d) and the energy of the Coulomb
repulsion V
HF
equals:
(25)
where A
NM
(d) is defined in eq. for adiabatic potential.
From the presented expression for E
S
, one sees that for a string of fixed length N the
total energy always has a minimum given by
(26)
The optimal number of particles trapped in the string of fixed length N is determined by a
minimization of E
S
with respect to M and is given by the eq.:
(27)
395
ANTIADIABATIC APPROACH
The similar expression for the length of the string valid even beyond adiabatic limit may
be obtained[9] with the use of Lang-Firsov unitary transformation[34] which transforms the
Hamiltonian, H, into the form:
(28)
where S= n
i i
[w
i
(q)b
q
h.c.], the hopping integral = t exp( [w
i
(q)w
j
(q)]b
q
h.c.),
the polaron shift and the effective inter-particles interaction is
(29)
Then, with the use of this expression the total energy of the string when M = N, and when
t 0, is equal to
(30)
The minimization of this expression with respect to the value M gives the equation for the
number M
The same result may be obtained for the dispersionless optical (Holstein) phonons. How-
ever for Jahn-Teller phonons (in a contrast with acoustical phonons) there arises a weak dis-
persion like (with q independent matrix element
there a weak attraction between particles on next-neighboring sites will be gener-
ated. This will give an extra contribution into the total energy, as:
(31)
The minimization of this expression with respect to the value M gives the length of the string
N = M as
(32)
This equation gives much longer value for the length of the string than
it was recently estimated in the same model in Ref.[35]. The different value for the estima-
tion of N in Ref.[35] originates in incorrect approximation for the total energy, where the
contribution from the polaron shift equal to cM/2 has been missed.
From these two approaches (adiabatic and antiadiabatic) we obtain that the minimum of
the total energy corresponds to a string of arbitrary length N and with M trapped particles,
the logarithm of which is proportional to the electron-phonon interaction, c, or to the polaron
shift E
p
and is inversely proportional to the intersite Coulomb repulsion between holes,
c
.
On the other hand the total energy E
S
(M) at large values of M decreases strongly with M
and increases with N. In the case when the double occupation of the sites is prohibited, the
number M can not be larger than N. Then the minimum energy E
S
corresponds to the relation
N = M. With the use of this relation and eq. for M the expression for E
Smin,
is simplified to
the form:
(33)
The comparison of these equations indicates that a string with M trapped charged par-
ticles may have a lower energy than the total energy of M separated self-trapped particles if
< c. However, the optimal length becomes very small N < 1, that indicates that in this
396
case there is realized a marginal extremum with N = 1. In the opposite case when the
smooth minimum of the total energy associated with string does exist but corresponds to a
metastable state.
THE STRING ENERGY FOR JAHN-TELLER PHONONS
With the use of the many body wave function obtained in the course of exact solution in
the limit of strong coupling with phonons here we have estimated an expectation value of the
Hamiltonian H. These calculations have been done in two steps. Using this many-body wave
function, first, we calculated the one body and the pair correlation functions. Then with the
use of the adiabatic approximation we have excluded slow (classical) phonon variables to get
an expression for adiabatic potential E
S
including the Coulomb and exchange energies (see,
for details, Refs[8,9]. The calculated expression of the total energy E
S
per particle has the
form:
(34)
where d is a dimension of the hypercubic lattice, the value n is the electron(hole) doping
inside the string: n = M/N and the value with a as an interatomic distance;
in the Hamiltonian the coupling constant of interaction with acoustical phonons
c = D
2
/K where D is a deformational potential and K is an elastic modulus or for Jahn-Teller
phonons we define c = / ( c
1 1
c
12
). The first three terms in the r.h.s. of this eq. are asso-
ciated with electron kinetic energy while the last two terms in the r.h.s. of the same eq. are
associated with the energies of electron-phonon and electron-electron interactions, respec-
tively (see, for comparison, in Ref.[8]). This expression represents a variational estimation of
the total energy of M particles self-trapped into a string of length N valid for a wide range of
values of c/t since it was obtained on the basis of an exact solution found in the limit of very
strong coupling c/t 1. Therefore, in the framework of this variational approach we may
get a reliable estimation of the number of particles, the length and the energy of an electron
string valid for a wide range of the parameters of the Hamiltonian such as a coupling constant
c, the bandwidth t and the characteristic Coulomb energy Here the values M and n are
variational parameters. The optimal number of particles trapped into the string of fixed length
N is determined by a minimization of E
S
/M with respect to M and is given by:
(35)
After a substitution of this expression into eq. for E
S
we get the dependence E
S
= E
S
(n) on
the doping of the string n = M/N. Depending on the relation between the values of t, c and
e
c
there may exist one or two types of solutions which correspond to two different types of
strings: when n = 1 we define an insulating string and when n < 1 we define a conducting
string. When c ~ t > the conducting string may be in a ground state. Then the number
of particles trapped into the string is described by last eq. and the value of the string doping
must be determined numerically by next minimization of the total energy with respect to the
value n. When the coupling constant c is very large (c t and c> ) the obtained eq. is not
applicable, since the associated solution describing a conducting string disappears while the
other solution associated with the marginal extremum n = 1 and describing insulating strings
still exists. The insulating strings have been already discussed in previous sections.
ANTI-FERROMAGNETIC CORRELATIONS
Additional factors which may enhance the string formation are the polaron effect[9]
397
which arises in polar semiconductors and the exchange next-neighbor spin-spin interaction
which arises in doped antiferromagnet.
The metastable minimum associated with the deformational string may become an ab-
solute minimum in the doped antiferromagnet. For a single hole in the antiferromagnet there
is an increase in the exchange energy equal to 2dJ
ex
, where J
ex
is an exchange constant. For
M separated holes this energy increase is equal to 2dMJ
ex
. On the other hand for M holes
trapped in a string such increase in exchange energy is equal to J
ex
(2dM M+1). Therefore,
the total energy of the deformational string in a doped antiferromagnet is described by the eq.:
(36)
where the value M is defined by eq. for M. The comparison of this expression with the
total energy of M separated self-trapped particles indicates that the strings may have a lower
energy if the following inequality holds:
(37)
Thus, the exchange interaction significantly improves the physical conditions required for
string formation in doped antiferromagnets. Therefore, if this conditions holds at low tem-
peratures the M separated particles will condense into a string configuration. For an arbitrary
number of particle in the system there may be created many strings or an array of these
electron strings. This array may be in ordered or in disordered state. Probably, at some
critical concentration there arise a percolation between these strings and the strings may be
ordered into charged stripes. In general this percolation picture reminds the filamantary mi-
crostructures suggested by Phillips[36] with the difference, however, that there the filaments
are conducting, like conducting strings decribed in Ref.[9].
STRINGS IN IONIC SOLIDS
In previous sections we have discussed the string solutions found for deformational type
of strings and also in the case when the electron(hole) is interacting with Jahn- Teller phonons
[31]. The obtained results are applicable, in general, for any type of short-range electron-
phonon interaction as, for example, with Holstein optical phonons. The number of particles
in the string is defined above and nearly equal to the number of sites in the string. The string
length depends on the type of the string and for conducting strings must be estimated by a
minimization of E
S
(n) with respect to n, numerically. For each type of phonons which have
a short-range interaction with electrons(holes) the coupling constant in eqs. above must be
defined, respectively, while the main eqs. for a number of particles and the length of the
string remain the same (for more details, see Ref[8,9])
The case when a single electron or hole is interacting with polar phonons, i.e. with lon-
gitudinal optical phonons with frequency coo (and with the constant of the electron- phonon
interaction ( see, for example, in Refs.[15-22]) is relevant and im-
portant to most oxides having a considerable amount of ionic bonding. Here the value of to-
tal energy including the Coulomb and exchange contributions from the long-range Coulomb
forces between fermions may be calculated analogously to the case of short-range electron-
phonon interaction presented above (see, also for example, in the Refs[8,9] That is, first,
with the aid of the Hartree-Fock many-body wave function of the M self-trapped particles
(1, 2, . . . , M) (see, eq. for the many body function) we have calculated the pair and off-
diagonal correlation functions, and then with the use of these functions the dependence of the
total energy on n and M having the form:
(38)
398
where we have introduced the notations The first two terms
in the r.h.s. of this eq. are associated with the electron kinetic energy while the other terms in
the r.h.s. of this eq. are associated with the energies of electron-phonon (~ E
p
) and electron-
electron (~ E
c
)interactions, respectively.
A minimization of this expression with respect to M and n gives an estimation for the
length of the string N and the number of particles M trapped into the string. For the value of
M we get the analytic expression:
(39)
The value of the doping n may be calculated numerically. In the limit of a low density n 1
the values of M and N or n may be presented by the analytic formulae:
(40)
where The total energy of the string per electron equals j
string
= 2d 2 + 2/N
nE
c
. To be in a ground state this string energy must be smaller than the energy of an individual
polaron j
p
equal to 2dE
p
. The comparison of these two energies gives the precise criterion
for the string formation. The conducting string corresponds to the ground state iff
(41)
which roughly means that the polaron shift must be smaller than the string bandwidth 2t.
Thus, we arrive at the conclusion that in oxide compounds with ionic bonding the forma-
tion of highly conducting electron strings created by a polarization potential is possible. The
string length is typically much larger than the number of self-trapped holes, which is deter-
mined by the dielectric constants of the solid. Below in this table we present the parameters
for the electron strings calculated only taken the longitudinal optical phonons into account.
Table 1. Approximate number of electrons M trapped by a string con-
sisting of N lattice sites in Oxides due to a polarization
Compound La
2
MnO
3
La
2
CuO
4
TiO
2
SrTiO
3
WO
3
M 3 7 15 34 63
N 5 40 40 40 160
APPLICATION TO HTSC
Thus, we arrive at the conclusions that in polar oxide materials, like HTSC there may
arise electronic strings which are linear multi-particle electronic molecules. At low tem-
peratures the electron strings may be ordered in CuO planes creating a nematic liquid crystal.
The striped phase in HTSC and in manganites observed in numerous experiments[41-47]
may correspond to such a liquid crystal of conducting strings. With the doping of an antifer-
romagnet La
2
CuO
4
there arises only the change in the distance between the strings while the
structure of the strings (like, the string doping n or the length N) is not changed. The metal-
lic stripe phase arises due to a correlated percolation over these strings when a density of
such strings will be larger than the percolation threshold. For square lattices the percolation
threshold is well known and is equal to x
c
~ .5 . Then, using this value and our estimation for
the string doping in La
2
CuO
4
as n = 7/40 we may readily get the hole doping = nx
c
.09
of the antiferromagnet La
2
CuO
4
at which the metallic stripe phase may arise. The spin-spin
399
and hole-spin correlations will of course slightly modify this result. It seems that our con-
clusion about the important contribution of the phonons into the origin of the stripe phase
is confirmed in recent experiments which discover a huge influence of isotope effect on the
critical temperature of the stripe ordering and strong lattice fluctuations in YBCO which may
be associated with the dynamics of the strings. With the isotope changes the structure of indi-
vidual strings is changed (for example, the strings become shorter) and, therefore, the critical
temperature of the stripe ordering must change.
In summary, we find that in oxides HTSC there may arise highly-conducting electron
strings which are linear electronic molecules. Note that to find such molecules we have to
treat the kinetic and potential energies of electrons on equal footing. A single electronic
molecule has a cigar shape with the length of the order of 10-20 nanometers and consisting
of 7-10 holes. For other oxides the string parameters will not be changed as much. It is also
very natural that such polymeric electron molecules may form a liquid crystal which may
be associated with the stripe phase of HTSC.
The described linear strings with M = N have a much lower energy than 2D sheet con-
figurations like a single circular or square spot consisting of M sites with M self-trapped
electrons (holes). It is clear that while in the single sheet spot of arbitrary shape with the
same number of particles, the phonon contribution remains the same the contribution of the
Coulomb interaction increases compared to the string shape. The string may not only have a
linear form but may also be bent, curved or even create a closed loop. Such curved configu-
rations will probably correspond to low energy excitations of the string.
Thus, we arrive at the conclusion that in narrow band doped antiferromagnet there may
arise an electronic phase separation in a form of linear electronic defects-strings. That is the
motion of free particles becomes unstable with the formation of strings which are linear multi-
particle objects. These strings are created by a deformation potential and have a length equal
to the number of self-trapped particles, which is determined by the elastic and deformational
constants of the doped antiferromagnet. Our findings are probably relevant to stripe formation
observed in HTSC[37-47] While here we have discussed insulating strings the stripes may
have an origin in other type of the hole conducting strings arising for a small hole doping as
described in Ref[9]. With an increase of the hole doping the concentration of these strings
increases and they may form either a nematic liquid crystal or a superlattice of stripes.
STRINGS IN MANGANITES
Although the precise qualitative and quantitative picture of physics in manganites must
be worked out in the framework of a double exchange model a qualitative wave-hand scenario
of the Colossal Magnetoresistance effect may be readily given. Indeed, these materials con-
sist of magnetic ions embedded into a 3D atomic perovskite lattice which may be described
by a double exchange model with a strong Hunds coupling. From a general point of view
there the parameter and the onsite Coulomb repulsion and/or Hunds interaction are
very large so that as an plausible approximation we may consider only spinless fermions asso-
ciated with the charge degrees of freedom induced by an infinite strong Hunds coupling. Let
us estimate the length of these strings in the manganite materials. If we take in a conventional
way the elastic module is such a material of the order c
11
= 1110
10
erg/cm
3
, the interatomic
distance a 4 then we get K ~ 4.1eV. The deformation potential may be conventionally
approximated as D e
2
/a = 3.4eV then for the electron-phonon coupling we obtain that
c = D
2
/K = 2.5eV. If we take the dielectric constant for LaMnO
3
found by an analysis of
experimental data in (see, Refs[10,l 1]) as =5, then we get that V e
2
/ = 0.68eV.
This estimation gives that the length of the string will be of the order 10 interatromic dis-
tances. Then, if we suggest that there is a coexistence of a droplets with Fermi liquid of the
polarised strongly-correlated individual particles with the regions of the phase consisting of
400
insulating strings similar to that which was observed in Ref.[12], then CMR phenomenon
could be described as follows.
At lower temperatures the ground state energy may correspond to a droplets with Fermi
liquid of the polarised strongly-correlated individual particles. Each droplet is in highly con-
ducting state coexisting with insulating domains consisting of the described strings.
When the temperature rises the area associated with the domains consisting of insulating
electron strings with M = N increases since they are associated with the metastable minima.
The conductivity drops since such strings are not conducting. At higher temperatures the
strings begin to evaporate and the conductivity associated with mobile polarons[15-22] in-
creases. A magnetic field will remove the barriers between conducting domains so improve
the conductivity.
It is also interesting to note that a qualitatively similar instability of small polarons and
the formation of a string-type trap has been found in another system of soft polymers[48-50]
but in the framework of a very different model of electrons interacting with rotating dipoles.
They found that the highly conducting strings termed superpolarons are more stable than
single polarons or bipolarons. The criteria for the string-superpolaron formation are that (a)
the polymer has very different static and high-frequency dielectric constants (we also need
such a condition) and (b) it is very soft (shear elastic constants are very small) so that an
excess charge orientates all dipoles around the string.
In a summary the multi-electron self-trapping phenomenon of the electronic string-
defect creation has been described in the framework of a simple self-consistent electron-
phonon models of interacting spinless fermions. This is a many-body (Hartree-Fock) gen-
eralization of the Pekar-Rashba model for the single particle self-trapping. The electron
phase separation is associated with the appearance of the multi-electron cigar-shaped local-
ized string- droplets. Note that for the first time we have shown that in systems with a narrow
band the electron-phonon interaction may play a very significant role in the electronic phase
separation associated with the strings.
Acknowledgments
I am very grateful to A. Bianconi, C.H. Chen,C. Di Castro, M. Grilli, S. Kivelson, P.
Littlewood, D. Edwards, G. Gehring, V. Emery, E.I. Rashba, Danya Khomskii, A.S. Alexan-
drov, V.V. Kabanov, H.S. Dhillon and other participants of the workshop ECRYS-99 and on
strongly correlated electrons in Isaac Newton Institute (Cambridge) for illuminating discus-
sions. The work has been supported by Isaac Newton Institute, University of Cambridge.
REFERENCES
1 . E.L. Nagaev, Sov. Jour.- JETP Lett., 16, 558 (1972); V.A. Kaschin and E.L. Nagaev, Zh. Eks. Teor. Fiz.,
66, 2105(1974)
2 . A. Moreo, S. Yunoki, and E. Dagotto, Science 283,2034 (1999)
3. D. Arovas and F. Guinea, Phys. Rev. B58, 9150(1998)
4 . D.I. Khomskii, Physica B 280-288, 325 (2000)
5 . G.S.Uhrig and R. Vlaming, Phys. Rev. Lett. 71, 271 (1993)
6 . M. Y. Kagan, D.I.Khomskii and M.V. Mostovoy, Eur. Phys. J. B12, 217 (1999)
7 . L.P. Gorkov and V.Z. Kresin, JETP Letters 67, 985 (1998)
8 . F.V. Kusmartsev, J. de Physique IV,9, Pr10-321, (1999)
9 . F.V. Kusmartsev, Phys.Rev.Lett. 84, 530, 5026 (2000); and submitted (2000)
10 . 3. K. Ahn et al . Phys. Rev. B 58, 3697 (1998)
11. 4. Y. Okimoto et al Phys. Rev. B 55,4206 (1997)
12 . M. Uehara, S. Mori, C.H. Chen and S. W. Cheong, Nature, 399, 560 (1999)
13 . S. Mori, C.H. Chen and S. W. Cheong, Nature, 392, 473 (1998)
14 . Y. Moritomo, A. Asamitsu and Y. Tokura, Phys. Rev. B56, 12190 (1997)
401
15 . E.I. Rashba, Opt. Spectr. 2, 78, 2, 88 (1957)
16. Y. Toyoazawa, Prog. Theor. Phys. 26, 29 (1961)
17 . E.I. Rashba, in: Excitons, ed. by E.I.Rashba and M.D. Sturge, North-Holland (Amsterdam) 1982, p.543.
18 . S.I. Pekar, Untersuchungen ber die Elektronentheorie Kristalle, Akademie Verlag, Berlin, 1954; see,
also in J. Appel, Solid State Phys. 21, 193 (1968); Y.B. Levinson and E.I. Rashba, Rep. Progr. Phys. 36,
1499 (1973); G D Mahan, Many Particle Physics, Plenum, NY-London, pp. 487 (1981); D. Emin, Phys.
Today, 35, 34 (1982); A.S.Alexandrov and N.F.Mott, Polarons and Bipolarons (WS, Singapore) 1995
19 . J. Yamashita and T. Kurosawa, J. Phys. Chem. Solids 4, 34 (1958).
20 . T. Holstein, Ann. Phys. 8,325, 343 (1959)
21. D.M. Eagles, Phys. Rev. 130, 1381 (1963); [Erratum ibid, 132, 2800 (1963)]. The recent development of
the small polaron theory is presented in the paper by A.S. Alexandrov and P.E. Kornilovich, Phys. Rev.
Lett. 82, 807(1999).
22 . V.V. Kabanov and O.Yu. Mashtakov, Phys. Rev. B47, 6060 (1993); A.J. Millis, R. Mueller and B.I.
Shraiman, Phys. Rev. B54, 5389 (1996); K. Yonemitsu, A.R. Bishop, J. Lorenzana, Phys. Rev. B47,
8065(1993)
23 . F. V.Kusmartsev and E.I.Rashba, Self-Trapping from Degenerate Bands (Spin S= 1) and related phenom-
ena, Zh.Eksp.Teor.Fiz., 86, (1984), 1142, or Sov.Phys. Journal JETP, 59, (1984), 668.
24. F. V.Kusmartsev and E.I.Rashba, Self-Trapping in crystals and nonlinear wave processes: Self- Trapping
Barrier for Plasma Caviton. Zh. Eksp. Teor. Fiz., 84, (1983), 2064, or Sov.Phys. Journal-JETP 57 1202
(1983) ,(1983), 300.
25 . F.V.Kusmartsev, Multi-phonon theory of the absorption of light in nonpolar crystals, Phys.Rev. B43
(1991) 1345.
26 . A.S.Alexandrov and N.F.Mott, Polarons and Bipolarons (WS, Singapore) 1995
27 . F. V.Kusmartsev and E.I.Rashba, Self-Trapping of Excitons and Lattice Defect Production in Solid Rare
Gases, Chech.Journal of Phys., B32 54 (1982)
28 . F. V.Kusmartsev and E.I.Rashba, Disappearing of Fluctuation Levels and Abrupt End of Density State
Tail, Fiz. andTekn. Poluprovod, 18,(1984),691,or Sov.Phys.Semicond., 18,(1984), 420.
29 . F. V.Kusmartsev, K.U. Neumann, O. Schrpf and K.R. A. Ziebeck, Self-Trapping of Electrons by Vortices,
Europhys. Lett. 42 547 (1998)
30. K.U. Neumann, F. V.Kusmartsev, H.J. Lauter, O. Schrpf and K.R.A. Ziebeck, Experimental observation
of lattice distortion due to a flux line lattice in niobium. Eur. Phys. J. B1 5 (1998)
31. L.P.Gorkov and A.B.Sokol, Pisma Zh. Eksp. Teor. Fiz. 46, 333 (1987)
32. Z. Popovic and S. Satpathy, Phys. Rev. Lett. 84, 1603 (2000)
33. G. Seibold, C. Castellani, C. Di Castro and M. Grilli, Phys. Rev. B58, 13 506 (1998)
34. I.G.Lang, Y.A.Firsov and JETP, 16, 1301 (1963)
35. Recently we become aware that the similar result for the string length has been obtained also with the use
of Lang-Firsov transformation by Alexandrov and Kabanov in the case of dispersive phonons, Phys. Rev.
Lett., submitted (2000)
36. J.C. Phillips, Proc. Natl. Acad. Sci. USA 94, 12 771 (1997)
37. J.R.Zaanen and O. Gunnarson, Phys. Rev. B40, 7391 (1989)
38. U. Lw, V. J. Emery, K. Fabricius and S.A. Kivelson, Phys. Rev. Lett. 72, 1918 (1994)
39. V.J. Emery, and S.A. Kivelson, Nature (London) 374, 434, (1995); V.J. Emery, S.A. Kivelson, and O.
Zachar, Phys. Rev, B56, 6120, (1997).
40. V.J. Emery, S. Kivelson and H.Q. Lin, Physica B163, 306, (1990); Phys. Rev. Lett. 64, 475, (1990)
41. A. Bianconi, Phys. Rev. B54, 12018 (1996); M.v. Zimmermann et al, Eur.Phys. Lett. 41, 629 (1998).
42. T.R.Thursten et al, Phys. Rev. B40, 4585 (1989).
43. J.M. Tranquada, Nature (London) 375, 561, (1995).
44. A. Bianconi et al, Phys. Rev. Lett. 76, 3412 (1996); and see references therein.
45. H.A. Mook, P. C . Dai, S.M. Hayden, G. Aeppli, T. G. Perring and F. Dogan, Nature, 395 580 (1998)
46. N.L.Saini, J.Avila, A.Bianconi, A.Lanzara, M.C.Asensio, S.Tajima, G.D.Gu and N.Koshizuka, Phys.
Rev. Lett. 79, 3467(1997)
47. M.Hennion, F. Moussa, G. Biotteau, J. Rodriguez-Carvajal, L. Pinsard and A. Revcolevschi, Phys. Rev.
Lett, 81 1957 (1998)
48. L.N. Grigorov, Makromol. Chem., Macromol. Symp. 37, 159 (1990).
49. L.N. Grigorov, V.M. Andreev, and S.G. Smirnova, Makromol. Chem., Macromol. Symp. 37, 177 (1990).
50. L.N. Grigorov, Pis ma Zh. Tekh. Fiz. 17(5) (1991) 45. [Sov. Tech. Phys. Lett. 17,368(1991)].
402
HIGH-TEMPERATURE SUPERCONDUCTIVITY
IS CHARGE-RESERVOIR SUPERCONDUCTIVITY
JOHN D. DOW
1
, HOWARD A. BLACKSTEAD
2
,
and DALE R. HARSHMAN
1,3
1
Department of Physics, Arizona State University,
Tempe, Arizona 85287-1504 U.S.A.
2
Department of Physics, University of Notre Dame,
Notre Dame, Indiana 46556 U.S.A.
3
Permanent address: Physikon Research Corporation,
P.O. Box 1014, Lynden, Washington 98264 U.S.A.
INTRODUCTION
Surprisingly, although high-temperature superconductivity is now virtually fifteen years
old, no agreement has been reached concerning the nature of either the phenomenon or the
theory that describes it. In this paper, we present evidence that such superconductivity orig-
inates from holes in the charge-reservoirs of the crystal structures and does not necessarily
require cuprate planes, as has been widely assumed.
An overview of high-temperature superconductivity and the nature of its origin is pre-
sented. The following critical facts will be discussed here: (i) PrBa
2
Cu
3
O
7
superconducts
at 90 K when its crystal structure has no Pr
Ba
defects; (ii) the chemical trends for super-
conductivity in PrBa
2
Cu
3
O
7
, Pb
2
Sr
2
Pr
0.5
Ca
0.5
Cu
3
O
8
, and Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
indicate
that the holes responsible for the primary superconductivity reside in the charge-reservoir
layers and not in the cuprate-planes (and so materials without cuprate-planes can be high-
temperature superconductors); (iii) there are no n-type high-temperature superconductors;
(iv) in the Nd
2z
Ce
z
CuO
4
homologues, the superconductivity proceeds through interstitial
oxygen ions paired with Ce ions; (v) the magnetic rare-earth ions in the superconductors are
potential Cooper pair-breakers, but for many materials only the L=0 magnetic ions in contact
with the superconducting condensate are pair-breakers, because the L 0 rare-earth ions often
experience crystal-field splitting of their electron energy levels; and (vi) Cu-doped Sr
2
YRuO
6
is a high-temperature superconductor having no cuprate planes.
PhaseTransitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 403
Figure 1. The c-axis parameter of (rare-earth)Ba
2
Cu
3
O
7
compounds against rare-earth radius. Note that the
traveling solvent floating zone (TSFZ) grown PrBa
2
Cu
3
O
7
superconducts, but flux grown PrBa
2
Cu
3
O
7
and
CmBa
2
Cu
3
O
7
(which have short c-axes) do not.
PrBa
2
Cu
3
O
7
PrBa
2
Cu
3
O
7
had long been regarded as a material that did not superconduct, until (i)
it was predicted to superconduct [1,2]; (ii) it was shown to exhibit granular superconduc-
tivity [35]; and (iii) bulk superconductivity was demonstrated in it as well [613]. De-
spite this evidence, some scientists still believe that PrBa
2
Cu
3
O
7
does not superconduct,
although the PrBa
2
Cu
3
O
7
materials that do not superconduct all have anomalously short
c-axes, and hence do not follow the trends for c-axis length versus ionic radius obeyed by
superconducting (rare-earth)Ba
2
Cu
3
O
7
compounds. (See Fig. 1.) The short c-axes are symp-
tomatic of PrBa
2
Cu
3
O
7
material containing Pr
Ba
(antisite Pr-on-a-Ba-site) defects. (See Figs.
2 and 3.) The PrBa
2
Cu
3
O
7
materials that do superconduct have longer c-axes than non-
superconducting PrBa
2
Cu
3
O
7
[14,15].
The important facts for superconducting PrBa
2
Cu
3
O
7
(and NdBa
2
Cu
3
O
7
) compounds
are that perfect PrBa
2
Cu
3
O
7
(and NdBa
2
Cu
3
O
7
) are 90 K superconductors, while the same
materials with numerous Pr
Ba
or Nd
Ba
defects have relatively short c-axes and do not super-
conduct (or have depressed transition temperatures).
Defects such as Pr
Ba
(and Nd
Ba
) destroy superconductivity because they are magnetic
pair-breakers. The failure of Pr
Pr
and Nd
Nd
to destroy superconductivity, and the location
of the Pr
Ba
and Nd
Ba
defects, namely in or near the charge-reservoir (CuO and BaO) layers,
indicates that the holes responsible for the primary superconductivity do not reside in the
cuprate-planes, as widely assumed.
TRENDS
Further evidence against the usual picture which places the superconducting holes in the
cuprate-planes is provided by examining the trends in critical temperatures for PrBa
2
Cu
3
O
7
(T
c
90 K [13]), for Pb
2
Sr
2
Pr
0.5
Ca
0.5
Cu
3
O
8
(T
c
60 K [16]), and for Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
404
Figure 2. Crystal structure of imperfect (non-superconducting) PrBa
2
Cu
3
O
7
with a defect Pr
Ba
and an O(5)
oxygen. Another Pr
Ba
is needed to balance charge.
Figure 3. The crystal structures of (a) perfect PrBa
2
Cu
3
O
7
with Tc 90 K; (b) Pb
2
Sr
2
Pr
0.5
Ca
0.5
Cu
3
O
8
(Pr
0.5
Ca
0.5
-PSYCO) with Tc 60 K, and (c) Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
with T
c
30 K. Note that the layers sur-
rounding the cuprate-planes in all three crystal structures are almost the same suggesting that the cuprate-
planes are not the primary superconductors.
405
Figure 4. Crystal structure of Nd
2z
Ce
z
CuO
4
with an interstitial oxygen. The interstitial oxygen is needed to
provide a potential sufficient to ionize Ce to Ce
+4
.
(T
c
30 K [17]) all of which have similar crystal structures adjacent to their cuprate-
planes, with a rare-earth on one side of a cuprate-plane, and either SrO or BaO on the other.
(See Fig. 3.) Consequently we must conclude that either (i) the local chemistry of the cuprate-
planes is unimportant in determining T
c
or (ii) the primary superconducting layers in high-
temperature superconductivity are not the cuprate-planes. Based on arguments given below,
we conclude that the cuprate planes do not determine T
c
because they are not the primary
superconducting layers.
THERE ARE NO n-TYPE HIGH-TEMPERATURE SUPERCONDUCTORS
Pb
2
Sr
2
(rare-earth)Cu
3
O
8
, namely (rare-earth)-PSYCO, has the property that it should be
a p-type high-temperature superconductor for most rare-earth ions co-doped with Ca: (rare-
earth)
1x
Ca
x
, for x 0.5. In fact such p-type materials do indeed superconduct [16,18,19].
It is widely believed (based on questionable analyses of the superconductivity of
Nd
2z
Ce
z
CuO
4
) that Pb
2
Sr
2
(rare-earth)Cu
3
O
8
should be an n-type high-temperature super-
conductor once the crystal-field potential at the rare-earth site is strong enough to ionize the
rare-earth to the (rare-earth)
+4
charge-state, as does occur for Am and Ce, whose ionization
potentials to Am
+4
and Ce
+4
are small enough in magnitude. In fact, perfect Am-PSYCO
and Ce-PSYCO must be n-type but do not superconduct at all, although they apparently do
conduct [16,20].
Indeed, the claims that Nd
2z
Ce
z
CuO
4
(Fig. 4) is an n-type superconductor, although
widely believed, also have serious problems associated with them: (i) The computed po-
tential at the rare-earth site is too weak (by 7 V) to ionize Ce to Ce
+4
[21]; (ii) p-type
high-temperature superconductivity has been observed in Pr
2z
Ce
z
CuO
4
by Brinkmann et al.
[22]; and (iii) attempts to make rectifying p-n junctions of Nd
2
_
z
Ce
z
CuO
4
/YBa
2
Cu
3
O
7
failed
[23]. These problems, which compromise claims of n-type superconductivity, caused us to
propose that interstitial oxygen ions paired with Ce ions actually dope the Nd
2z
Ce
z
CuO
4
406
Figure 5. Crystal structure of (a) Gd
2z
Ce
z
CuO
4
(T
c
=0 for this Gd compound; but T
c
for Pr
2z
Ce
z
CuO
4
is 24 K) and (b) Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
which is a superlattice of Gd
2z
Ce
z
CuO
4
and
/SrO/NbO
2
/SrO/CuO
2
/ layers.
homologues p-type, causing (rare-earth)
2z
Ce
z
CuO
4
compounds to superconduct: the super-
conducting Nd
2z
Ce
z
CuO
4
homologues are actually p-type [24,25].
Gd
2z
Ce
z
CuO
4
COMPARED WITH Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
is a superlattice of Gd
2z
Ce
z
CuO
4
with the additional layers
/SrO/NbO
2
/SrO/CuO
2
/. (See Fig. 5.) Consequently from a cuprate-plane theory viewpoint,
one would expect that Gd
2
_
z
Ce
z
Sr
2
Cu
2
NbO
10
will superconduct only if Gd
2
_
z
Ce
z
CuO
4
also
superconducts (and vice versa). However, Gd
2z
Ce
z
CuO
4
does not superconduct, while
Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
does superconduct [26].
Indeed, for all the rare-earth ions that form both (rare-earth)
2z
Ce
z
CuO
4
compounds
and the corresponding (rare-earth)
2z
Ce
z
Sr
2
Cu
2
NbO
10
superlattice materials, except for the
Gd-based (and Cm-based) materials, both superconduct. The sole exception occurs for the
L=0 ions Gd (and Cm): Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
superconducts but Gd
2z
Ce
z
CuO4 does not.
(Cm
2z
Th
z
CuO
4
, which involves L=0 Cm, also does not superconduct.)
If we assign the different behaviors of Gd
2z
Ce
z
CuO
4
and Cm
2z
Th
z
CuO
4
(which do
not superconduct) and Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
(which does superconduct) to the L=0 char-
acter of J 0 Gd and Cm, then we must propose that pair-breaking Gd in superconducting
Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
is not a nearest-neighbor to the superconducting layer, although Gd
in non-superconducting Gd
2z
Ce
z
CuO
4
is which is why Gd breaks pairs in the latter ma-
terial and not in the former. This means that SrO is the nearest potentially superconducting
layer to Gd in Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
, while Gd
2
O
2
is the layer that would superconduct
in Gd
2z
Ce
z
CuO
4
if Gd were replaced by an L 0 ion such as Nd [27]. Putting a layer of
magnetic L=0 ions in or adjacent to the SrO layer can kill the superconductivity, provided
the superconducting condensate is in the charge-reservoir layers (as we propose), not in the
cuprate-planes (as widely believed currently). Magnetic L=0 ions, such as Gd and Cm are not
crystal-field split and are Cooper pair-breakers which destroy superconductivity. In contrast
L 0 ions are often crystal-field split, which can cause them to lose their pair-breaking ability.
407
Figure 6. Idealized crystal structure of one-quarter of the unit cell of Sr
2
YRuO
6
. Copper doping of the Ru-site
produces superconductivity. Not shown are the rotations of the oxygen octahedra.
Ba
2
GdRu
1u
Cu
u
O
6
COMPARED WITH Sr
2
YRu
1u
Cu
u
O
6
Ba
2
GdRuO
6
and Sr
2
YRuO
6
(both doped with Cu on Ru sites) are interesting compounds
[2833] because (i) they have the same crystal structure, but (ii) Cu-doped Ba
2
GdRuO
6
does
not superconduct, while Sr
2
YRuO
6
does. (For the crystal structure, see Fig. 6.) We pro-
pose that this is because the two isostructural compounds are two-layer compounds (like
Nd
2z
Ce
z
CuO
4
) and so a magnetic ion that is not crystal-field split, such as L=0 Gd or Cm,
will destroy the superconductivity in both layers: Cu-doped Ba
2
GdRuO
6
consequently does
not superconduct, but Sr
2
YRuO
6
does (at T
c
45 K [31]).
Sr
2
YRu
1u
Cu
u
O
6
: CUPRATE-PLANE-LESS SUPERCONDUCTIVITY
Sr
2
YRu
1u
Cu
u
O
6
superconducts at 45 K, but also has two antiferromagnetic ordering
temperatures: one at 86 K which has been identified as due to Cu, another at 23 K to
30 K due to Ru [31]. This means that when the material superconducts, (i) its Cu is already
antiferromagnetic, and (ii) below 23 K the Ru is also antiferromagnetically ordered. (The
material is too pure to permit the assumption that the Cu dopant forms cuprate planes.)
The conventional concept is that magnetic layers do not superconduct. Therefore
Sr
2
YRu
1u
Cu
u
O
6
, with its magnetic Ru and Cu ions on its Ru sites, must superconduct
in the non-magnetic layers, and not in the magnetic YRu
1u
Cu
u
O
4
layers. Namely, the holes
carrying the superfluid density must reside in the SrO layers.
Recent muon spin relaxation (SR) studies [31-34] lend further support to the assign-
ment of the superconducting hole condensate to the SrO layers. The Sr
2
YRu
1u
Cu
u
O
6
mate-
rial contains oxygen ions in both its (SrO)
2
and YRu
1
_
u
Cu
u
O
4
layers. Since positive muons
tend to stop near the negatively charged oxygen ions, one would expect to observe one mag-
netically distinguishable muon site for each of the various different local environments capa-
ble of trapping a
+
particle: we observed two such sites, each with a radius of order 0.5 .
Above 30 K, the SR spectra (acquired in a field of 500 Oe transverse to the muon beam)
408
Figure 7. Muon spin rotation relaxation rate (in (s)
1
) versus temperature (in K) for the two muon sites,
measured in a 500 Oe field (perpendicular to the initial muon spin polarization direction), after reference [31].
The data were analyzed assuming exponential forms for the relaxation functions. (The open triangles correspond
to data taken at 500 Oe after cooling in zero field.) We attribute the fast relaxation to the YRuO
4
layer, and the
slower one to the SrO layer.
exhibit a single slowly relaxing component with a Larmor precession frequency correspond-
ing to the applied field. However, below 30 K the Ru ions begin to order, revealing two
distinct components, one fast-relaxing and the other slow-relaxing. (See Fig. 7.) The fast-
relaxing signal, which accounts for about 90% of the muons, also shows a dramatic increase
in its muon precession frequency, corresponding to a local field of about 3 kOe. The remain-
ing component, with the markedly slower relaxation rate, also exhibits a slight diamagnetic
shift of the average muon precession frequency. These data indicate that the muon site asso-
ciated with the slowly relaxing component experiences a near-zero net local magnetic field in
the ordered state. (See Fig. 8.)
Neutron powder diffractometry measurements, acquired on the same samples as the SR
data, indicate that the Ru moments order ferromagnetically in the YRu
1u
Cu
u
O
4
planes, with
the net polarization reversing direction from one YRu
1u
Cu
u
O
4
layer to the next along the
c-axis. This polarization reversal produces a net zero field in the SrO layers. In contrast,
the local field in the YRu
1u
Cu
u
O
4
layers is necessarily non-zero due to the Ru and the Cu
moments.
From this, we can unambiguously attribute (i) the fast-relaxing component observed in
the SR spectra to muons stopped in the YRu
1u
Cu
u
O
4
layers, and (ii) the slowly-relaxing
signal to muons stopped in the SrO layers. These identifications are consistent with the bond-
valence sum calculations [35], which show that (i) the oxygen ions in the SrO layers have
stretched bonds (which implies that those oxygen ions are positively charged with respect to
O
2
), and (ii) the oxygen ions in the YRu
1u
Cu
u
O
4
layers are virtually fully charged to O
2
.
The positive muons tend to stop near the more negatively charged oxygen, which is why
90% stop in the YRu
1u
Cu
u
O
4
layers. The relaxation rate increase and the diamagnetic
shift for the SrO component are both consistent with the presence of vortices, as determined
by first cooling in zero magnetic field, and then by applying a 500 Oe magnetic field. (See
409
Figure 8. Muon spin rotation frequency versus temperature for the SrO layers of Sr
2
YRuO
6
(doped with Cu
on Ru sites) after reference [31]. The large error bars below 30 K indicate detection of a flux lattice (because
the detected spot sometimes is somewhere in a vortex). The open triangle represents a zero-field datum.
Fig. 7.) The observed increase in relaxation rate at low temperatures indicates the presence
of vortices. Moreover, the flux lattice was observed to be only weakly pinned, which is
consistent with a very short c-axis flux-line correlation length, such as what one would expect
for a set of isolated sheets of pancake vortices. Interestingly, bulk superconductivity is not
evident in the muon data until the Ru ions order, suggesting that the fluctuating paramagnetic
Ru spins may act to suppress the superconductivity above 30 K.
SUMMARY
By placing the charge-carrying holes in the charge-reservoirs, rather than in the cuprate-
planes, we predicted that PrBa
2
Cu
3
O
7
would superconduct and it does. We also pre-
dicted the superconductivity of three more compounds that have since been found to have at
least granular superconductivity: Gd
1.6
Ce
0.4
Sr
2
Cu
2
TiO
10
[36,37], Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
[21,38], and Eu
1.5
Ce
0.5
Sr
2
Cu
2
TiO
10
[36,39]. The chemical trends for PrBa
2
Cu
3
O
7
,
Pb
2
Sr
2
Pr
0.5
Ca
0.5
Cu
3
O
8
, and Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
) suggest that the cuprate-planes are not
the main generators of superconductivity while the fact that Sr
2
YRuO
6
doped with Cu su-
perconducts (and essentially at the same temperature as GdSr
2
Cu
2
RuO
8
and
Gd
1.5
Ce
0.5
Sr
2
Cu
2
RuO
10
) lends credence to the idea that the superconducting holes reside
in the SrO layers of all three of these compounds, which is certainly true for Sr
2
YRuO
6
.
The evidence is that there are no superconductors that can be made both n-type and p-
type, because n-type high-temperature superconductors (at least of this class of materials)
do not exist. Although Gd
2z
Ce
z
Sr
2
Cu
2
NbO
10
is a natural superlattice of Gd
2z
Ce
z
CuO
4
and layers of /SrO/NbO
2
/SrO/CuO2/, the fact that Gd
2z
Ce
z
CuO
4
does not superconduct,
while its superlattice does, indicates that the superconductivity originates in the charge reser-
voirs, not in the cuprate-planes. Ba
2
GdRuO
6
doped with Cu does not superconduct because
410
of the Gd, which is an L=0 magnetic pair-breaker. Finally, Cu-doped Sr
2
YRuO
6
is a su-
perconductor with an onset temperature of T
c
45 K and with the main superconductivity
being in its SrO layers. Logical extension of this idea to (rare-earth)Sr
2
Cu
2
RuO
8
and (rare-
earth)
1.5
Ce
0.5
Sr
2
Cu
2
RuO
10
(which both superconduct at 45 K), strongly suggests that the
superconducting holes are also carried by the SrO layers of these materials as well.
ACKNOWLEDGMENTS
We are grateful to the U. S. Office of Naval Research (Contract N00014-98-10137), for
their support.
REFERENCES
1. Blackstead, H.A. and Dow, J.D. (1993) Tb doping of YBa
2
Cu
3
Superlatt. Microstruct. 14,231-236.
2. Blackstead, H.A. and Dow, J.D. (1995) Role of Ba-site Pr in quenching superconductivity of
Y
1

y
Pr
y
Ba
2
Cu
3
O
x
and related materials, Phys. Rev. B 51, 11830-11837.
3. Blackstead, H.A., Chrisey, D.B., Dow, J.D., Horwitz, J.S., Klunzinger, A.E., and Pulling, D.B. (1994)
Evidence of superconductivity in PrBa
2
Cu
3
O
7
, Physica C 235-240, 1539-1540.
4. Blackstead, H.A., Chrisey, D.B., Dow, J.D., Horwitz, J.S., Klunzinger, A.E., and Pulling, D.B. (1995)
Superconductivity in PrBa
2
Cu
3
O
7
, Phys. Lett. A 207, 109-112.
5. Blackstead, H.A., Dow, J.D., Chrisey, D.B., Horwitz, J.S., McGinn, P.J., M. A. Black, Klunzinger, A.E.,
and Pulling, D.B. (1996) Observation of superconductivity in PrBa
2
Cu
3
O7, Phys. Rev. B 54, 6122-6125.
6. Zou, Z., Oka, K., Ito, T., and Nishihara, Y. (1997) Bulk superconductivity in single crystals of
PrBa
2
Cu
3
O
x
, Jpn. J. Appl. Phys., Part 2. 36, L18-L20.
7. Luszczek, M., Sadowski, W., and Olchowik, J. (1997) Abstract in 5th International Conference on Ma-
terials and Mechanisms of Superconductivity: High Temperature Superconductors, Abstract Book, p.
143.
8. Sadowski, W., Luszczek, M., Olchowik, J., Susla, B., and Czajka, R. (1997) Traces of superconductivity
in Pr-Ba-Cu-O system, Molec. Phys. Rpts. 20, 213-215.
9. Usagawa, T., Ishimaru, Y., Wen, J., Utagawa, T., Koyama, S., and Enomoto, Y. (1997) Superconductivity
in (110) PrBa
2
Ca
3
thin films pseudomorphically grown on (110) YBa
2
Cu
3
single crystal sub-
strates, Jpn. J. Appl. Phys. 36, L1583-L1586, find superconductivity in (110) but not (001) films; see also
Usagawa, T., Ishimaru, Y., Wen, J., Utagawa, T., Koyama, S., and Enomoto, Y, (1998) Non-linear I-V
characteristics and proximity effects for (001) PrBa
2
Cu
3
YBa
2
Cu
3
bi-layered structures grown
on YBa
2
Cu
3
single crystal substrates, Appl. Phys. Lett. 72,1772-1774.
10. Cooley, J.C., Hults, W.L., Peterson, E.J., Dow, J.D., Blackstead, H.A., and Smith, J.L. (1998) Supercon-
ducting PrBa
2
Cu
3
O
x
powders, Bull. Amer. Phys. Soc. Q35.07 (March, 1998).
11. Hults, W.L., Cooley, J.C., Peterson, E.J., Smith, J.L., Blackstead, H.A., and Dow, J.D. (1998)
PrBa
2
Cu
3
O
7
polycrystalline superconductor preparalion, Intl. J. Mod. Phys. B 12, 3278-3283.
12. Shukla, A., Barbiellini, B., Erb, A., Manuel, A., Buslaps, T., Honkimki, V., and Suortti, P. (1999) Hole
depletion and localization due to disorder in insulaling PrBa
2
Cu
3
A Compton scattering study,
Phys. Rev. B 59, 12127-12131.
13. Araujo-Moreira, P.M., Lisboa Filho, P.N., Zanetti, S.M., Leite, E.R., and Ortiz, W.A. (2000) Supercon-
ductivity in siniered-polycryslalline PrBa
2
Cu
3
Physica B 284-288, 1033-1034.
14. Harris, V.J., Fatemi, D.J., Browning, V.M., Osofsky, M.S., and Vanderah, T.A. (1998) Extended X-ray
absorption fine structure measurements of non-superconducting PrBa
2
Cu
3
O
6.9
: Evidence against Ba site
Pr substitution, J, Appl. Phys. 83, 6783-6785, reported a c-axis length of 11.71 , in a material which
did not superconduct, because it had c < 11.72 .
15. Skanthakumar, S., Lynn, J.W., Rosov, N., Cao, G., and Crow, J. E. (1997) Observation of Pr magnetic
order in PrBa
2
Cu
3
O
7
, Phys. Rev. B 55, R3406-R3409 reported c=l 1.703 for PrBa
2
Cu
3
O
7
which did
not superconduct (because c < 11.72 ) and also had 5% BaCuO
2
impurity phase, resulting from Pr
substitution on the Ba-site.
16. Skanthakumar, S. and Soderholm, L. (1996) Oxidation state of Ce in Pb
2
Sr
2
Ce
1x
Ca
x
Cu
3
O
8
, Phys. Rev.
B 53, 920-926.
17. Blackstead, H.A., Dow, J.D., Felner, I., Luo, H., Pulling, D.B., and Yelon, W.B. (1998) Evidence of
granular superconductivity in Pr
2z
Ce
z
Sr
2
Cu
2
NbO
10
, Intl. J. Mod. Phys. B 12, 3074-3079 find T
c
28 K
in Pr
2z
Ce
z
Sr
2
Cu
2
NbO
10
.
411
18. Blackstead, H.A., Dow, J.D., and Pulling, D.B. (2000) Particle-hole doping asymmetry in
Pb
2
Sr
2
YCu
3
O
8
-class superconductors, Phys. Rev. B , in press.
19. Blackstead, H.A. and Dow, J.D. (1999) PSYCO homologues: p-type superconduct, but n-type do not, J.
Low Temp. Phys. 117, 557-561.
20. Soderholm, L., Williams, C., Skanthakumar, S., Antonio, M.R., and Conradson, S. (1996) The synthesis
and characterization of the superconductor-related compound Pb
2
Sr
2
AmCu
3
O
8
, Z. Physik B 101, 539-
545.
21. Blackstead, H.A. and Dow, J.D. (1998) Predicted properties of Nd
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
, Phys. Rev. B
57, 10798-10813.
22. Brinkmann, M., Rex, T., Steif, M., Bach, H., and Westerhalt, K. (1996) Residual resistivity and oxygen
stoichiometry in Pr
2-x
Ce
x
Cu single crystals, Physica C 269, 76-82.
23. Mao, S.N, Xi, X.X., Li, Q., Takeuchi, I., Bhattacharya, S., Kwon, C., Doughty, C., Walken-
horst, A., Venkatesan, T., Whan, C.B., Peng, J.L. and Greene, R. L. (1993) Superconducting
Y
1
Ba
2
Cu
3
O
7x
/Nd
1.85
Ce
0.15
CuO
4y
bilayer thin films, Appl. Phys. Lett. 62, 2425-2427.
24. Blackstead, H.A. and Dow, J.D. (1995) Location of the root of superconductivity in La
2
_ Sr CuO
x
and
Nd
2z
Ce
z
CuO
x
,Philos. Mag. B 72, 529-534.
25. Blackstead, H.A. and Dow, J.D. (1999) Comparison of bulk R
2
_
z
Ce
z
CuO
4
with superlattice
R
2z
Ce
z
CuO
4
/SrO/NbO
2
/SrO/CuO
2
/, Phys. Rev. B 60, 13051-13055.
26. Blackstead, H.A., Dow, J.D., Felner, I., Jackson, D.D., and Pulling, D.B. (1999) Superconductivity of
Pr
2z
Ce
z
Sr
2
Cu
2
NbO
10
, High Temperature Superconductivity, ed. by S. E. Barnes, J. Ashkenazi, J. L.
Cohn, and F. Zuo, (Amer. Inst. Phys. Conf. Proc. 483, American Institute of Physics, Woodbury, New
York, 1999), pp. 191-194.
27. Gd
+3
has angular momentum L=0, and so does not experience crystal-field splitting.
28. Wu, M.K., Chen, D.Y., Chien, F.Z., Sheen, S.R., Ling, D.C., Tai, C.Y., Tseng, G.Y., Chen, D.H., and
Zhang, F.C. (1996) Anomalous magnetic and superconducting properties in a Ru-based double per-
ovskite. Z. Physik B102, 37-41.
29. Chen, D.Y., Chien, F.Z., Ling, D.C., Tseng, J.L., Sheen, S.R., Wang, M.J., and Wu, M.K. (1997) Super-
conductivity in Ru-based double perovskite the possible existence of a new superconducting pairing
state, Physica C 282-287, 73-76.
30. Fainstein, A., Winkler, E., Butera, A., and Tallon, J. (1999) Magnetic interaction and magnon gap in the
ferromagnetic superconductor RuSr
2
GdCu
2
O
8
, Phys. Rev. B 60, R12597-R12600.
31. Blackstead, H.A., Dow, J.D., Harshman, D.R., DeMarco, M.J., Wu, M.K., Chen, D.Y., Chien, F.Z.,
Pulling, D.B., Kossler, W.J., Greer. A.J., Stronach, C.E., Koster, E., Hitti, B., Haka, M. and Toorongian,
S. (2000) Magnetism and superconductivity in Sr
2
YRu
1
_
u
Cu
u
O
6
and magnetism in Ba
2
GdRu
1
_
u
Cu
u
O
6
,
Eur. Phys. J. B 15, 649-656.
32. Harshman, D.R., Kossler, W.J., Greer, A.J., Stronach, C.E., Koster, E., Hitti, B., Wu, M.K., Chen, D.Y.,
Chien, F.Z., Blackstead, H.A., and Dow, J.D. (1999) Muon spin rotation in Sr
2
YRu
1x
CU
x
O
6
, Intl. J.
Mod. Phys. B 13, 3670-3677.
33. Harshman, D.R., Kossler, W.J., Greer, A.J., Stronach, C.E., Koster, E., Hitti, B., Wu, M.K., Chen, D.Y.,
Chien, F.Z., Blackstead, H.A., and Dow, J.D. (2000) Location of the superconducting hole condensate in
Sr
2
YRu
1u
Cu
u
O
6
by u
+
SR, Physica B 289-290, 360-364.
34. Blackstead, H.A., Dow, J.D., Harshman, D.R., Yelon, W.B., Chen, M.X., Wu, M.K., Chen, D.Y.,
Chien, F.Z., and Pulling, D.B. (2001) Magnetically ordered Cu and Ru in Ba
2
GdRu
1
_
u
Cu
u
O
6
and in
Sr
2
YRu
1u
Cu
u
O
6
, to be published.
35. Brown, I.D. (1980) Structure and Bonding in Crystals, Vol. II, edited by M. OKeefe and A. Navrotsky,
pp. 1-20 (Academic Press, New York, 1980).
36. Blackstead, H.A., Dow, J.D., Heilman, A.K., and Pulling, D.B. (1997) Prediction that
Cm
2z
Th
z
Sr
2
Cu
2
NbO
x
and Gd
2z
Ce
z
Sr
2
Cu
2
TiO
10
will superconduct, Solid State Commun. 103, 581-
584.
37. Blackstead, H.A., Dow, J.D., Goldschmidt, D., and Pulling, D.B. (1998) Observation of predicted super-
conductivity in Gd
2z
Ce
z
Sr
2
Cu
2
TiO
x
with x 10, Phys. Lett. A 245, 158-162.
38. Blackstead, H.A., Dow, J.D., Felner, I., Yelon, W.B., Chen, M., Luo, H., and Pulling, D.B. (2000) Detec-
tion of the nearly invisible defect that disrupts bulk superconductivity in Pr
1.5
Ce
0.5
Sr
2
Cu
2
NbO
10
, Phys.
Rev. B 62, 1244-1251.
39. Blackstead, H.A., Dow, J.D., Felner, I., and Pulling, D.B. (2000) Observation of superconductivity in
Eu
1.5
Ce
0.5
Sr
2
Cu
2
TiO
10
, Phys. Rev. B 61, 6303-6306.
412
ELECTRONIC INHOMOGENEITIES IN HIGH-T
C
SUPERCONDUCTORS
OBSERVED BY NMR
J. HAASE
1,2
, C.P. SLICHTER
1
, R. STERN
1,*
, C.T. MILLING
1
,
and D.G. HINKS
3
1
Department of Physics and Materials Research Laboratory, University of
Illinois at Urbana-Champaign, Urbana, Illinois 61801-3080.
2
2. Physikalisches Institut, Universitt Stuttgart, Pfaffenwaldring 57, 70569
Stuttgart, Germany.
3
Materials Science Division, Argonne National Laboratory, Argonne,
Illinois 60439.
INTRODUCTION
The pairing mechanism for the electrons in high-temperature superconductors (HTSC)
is still elusive. The original ideas which led to the discovery focused on strong electron-
lattice interactions like Jahn-Teller distortions [1]. However, certain properties (like s-
wave pairing or the isotope effect) expected for such processes within traditional pictures
are lacking. Although various theories of superconductivity emerged, none of them seems
widely accepted up to now. It is quite clear that dynamic antiferromagnetic spin
fluctuations [2, 3] are present in the materials, since they derive from antiferromagnets by
hole doping. The study of these spin fluctuations and their relation to superconductivity is
an active field of research [4]. Since the early days of HTSC it was argued that the hole
doping in the two-dimensional antiferromagnetic background (the exchange interactions
are much smaller between Cu-O planes) may not lead to a homogeneous electronic fluid
[5-7]. Instead, conducting domains of holes might alternate with antiferromagnetic
domains (e.g., one-dimensional arrays of charge in the two-dimensional antiferromagnetic
background) [8]. Over the years more and more experiments turned up evidence for
inhomogeneous charge and spin structures [8-17], including atomic distances, static and
dynamic spin or charge variations, so-called stripes. When superconductivity is suppressed
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 413
by co-doping with other atoms static stripes appear [8], similar to static stripes that are
incommensurate with the underlying crystal structure in non-conducting mixed valance
compounds. The question arises whether they could be present as dynamic stripes. If so,
they are candidates for a new pairing agent for HTSC since they provide low energy
excitations.
We will review some nuclear magnetic resonance (NMR) data and show new evidence
that electronic inhomogeneities exist, even for samples with the highest critical
temperature, T
c
.
Nuclei that posses magnetic dipole and electric quadrupole moments interact with
local magnetic fields and electric field gradients, respectively, causing a splitting of the
nuclear spin states which is detected by NMR. In non-magnetic materials the electric
quadrupole interaction can dominate in zero magnetic field and nuclear spin transitions can
be observed (nuclear quadrupole resonance, NQR). Application of an external magnetic
field introduces a well defined Zeeman splitting for nuclei without quadrupole moments.
For quadrupolar nuclei the resonance frequencies depend, in addition, on the internal
electric field gradient and its orientation with respect to the magnetic field [18].
Time dependent local fields can induce transitions among the nuclear levels and cause
relaxation, the so-called spin-lattice relaxation (for NMR and NQR). Local field changes
due to nearby nuclear dipole moments (nuclear spin-spin interactions), often used for
distance determination, can also be used to study the electronic properties if the inter
nuclear coupling is amplified by the electron spin excitations.
From the NQR frequency one can determine the static electric field gradient at the
nuclear site and from its spin-lattice relaxation one obtains information about the
fluctuations of magnetic or electric fields. Similar information is available from NMR
measurements in strong magnetic fields, where the electric quadrupole interaction causes
shifts of the Zeeman levels, however, NMR experiments give even greater insight and
allow the application of various techniques which have made NMR so successful.
A most important NMR parameter is the chemical or orbital shift K
L
. It is caused by
the unquenching of orbital angular momentum of the electrons due to the external field
B
ext
. Electron spin excitations of low energy can also cause a change of the local magnetic
field (static or dynamic). These effects can be rather strong if the electronic coupling to the
nucleus (hyperfine coupling) is large. These so-called Knight shifts or spin shifts will be
denoted with K
S
. Both shifts measure the deviation of the local magnetic field B
loc
at the
nucleus from the external field and are quantified by comparison with a reference sample
of preferably small shifts. If the nucleus of the reference sample has the frequency v
ref
in
the field B
ext
, we write v = (1 + K)v
ref
, where K = K
L
+ K
S
is the sum of both shift tensors.
The contribution of each component is not known a priori. One has to invoke models and
measure, e.g., the temperature dependence of the shift, in order to decompose it. The NMR
shifts measure the orbital and spin susceptibilities if the local field changes are
proportional to the polarization created by the external field. Both parameters have proven
to be very useful tools in determining chemical bonding and electronic structure.
The presence of a quadrupole interaction complicates the understanding of the NMR,
but by the same virtue, it is another source of information about the structure of the
material. The strength of the quadrupole interaction does not depend on the external field.
By choosing various laboratory fields (up to 20 T) one can change the size of the Zeeman
term. The two limiting cases of a vanishing external magnetic field (NQR) and a strong
magnetic field (such that the quadrupole interaction is only a small perturbation of the
Zeeman interaction - quadrupole perturbed NMR) can be treated easily. We will mostly be
interested in the latter case since one can obtain all information from the high field spectra.
In such a case, the single Zeeman transition for an I>l/2 nucleus splits into 2I transitions.
414
Figure 1. (a) Zeeman levels of an I = 5/2 nuclear spin in a strong magnetic field with perturbation from
quadrupole interaction. The unperturbed distance between neighboring levels is given by the frequency v
0
.
The quadrupole splitting v
Q,z
depends on the orientation of the electric field gradient with respect to the
magnetic field (z-direction). Corresponding NMR spectra for a distribution of (b) electric field gradients and
(c) local magnetic fields. The latter affects all transitions equally, whereas the former broadens the outer
satellites twice as much as the inner ones and leaves the central transition unaffected to first order.
The splitting depends also on the orientation of the electric field gradient (EFG) tensor
relative to the applied magnetic field. Most nuclear spins are half integer (I = 3/2, 5/2, 7/2)
and in such a case there is the so-called central transition (m=l/2) which is only affected
by the quadrupole interaction in second order. We will simplify the situation by using
single crystals or oriented powder (oriented along the crystal c-axis). With the external
magnetic field in c-direction, the Hamiltonian is axially symmetric so we can neglect the
2
nd
order quadrupole interaction and find a simple formula for the 2I transition frequencies,
(1)
where n denotes the particular transitions and has values for I = 3/2 of n = -1, 0, +1, and for
I = 5/2 of n = -2, -1, 0, +1, +2. The central transition, n = 0, measures the magnetic shifts
only, whereas the satellite transitions are affected by both the magnetic shifts and the
quadrupole splitting v
Q
. This situation is depicted in Fig. 1. Higher order effects from
crystal misalignment or other changes of the EFG orientation can be detected by field
dependent measurements: while the magnetic shifts are proportional to the field, the
second order quadrupole effects are inversely proportional to the field.
In order to have a feeling for the properties that can be measured with NMR, we
mention that an electron in the ground state will produce an electric field gradient in z-
direction of
(2)
415
where is the polar angle and r the distance from the nucleus. Also, external charges from
the crystal electric field will influence the electric field gradient. The calculation of the
EFG can often be obtained with quantum chemical methods.
The orbital or chemical shift measures the availability of excited states
(3)
where is the electron gyromagnetic ratio, L
z
is the z-component of the angular
momentum and the difference E
n
-E
0
measures the energy to the excited state n [19]. The
orbital shift is therefore typically temperature independent. It is large if unoccupied states
are nearby in energy.
Figure 2. Structure of La
2-x
Sr
x
CuO
4
. The Cu-O a-b planes are formed by Cu and planar oxygen. The distance
from Cu to the apical oxygen is much larger than that to the planar oxygen. The La atoms are replaced by Sr
to achieve conductivity/superconductivity. There is only one unique Cu-O plane separated by apical oxygen
and La/Sr atoms.
Lastly, we write down an expression for the spin shift.
(4)
where A
zz
denotes the hyperfine coupling constant and the electron spin polarization in
z-direction of the electron to which the nucleus couples. The latter can be strongly
temperature dependent, as for an isolated electron spin or for a superconductor with spin
singlet pairing, or, almost temperature independent as for a metal. The spin polarization in
a magnetic field is often given by the spin susceptibility.
We see that all three parameters give very important information about the electronic
structure of the material. In a similar fashion one can write down equations for the nuclear
spin relaxation. The time dependence of the processes and the coupling matrix elements
will determine the prevailing relaxation mechanism.
416
EXPERIMENTAL ASPECTS OF NMR OF HIGH-T
C
SUPERCONDUCTORS
A typical high-temperature superconductor, La
2-x
Sr
x
CuO
4
(LSCO), is shown in Fig. 2
(for structural details see [20, 21]). For x = 0 this material is an antiferromagnet with a
Neel temperature of about T
N
= 320 K. At a Sr concentration of x = 0.02 the Neel order
vanishes and around x = 0.05 the material becomes superconducting at very low critical
temperature T
c
. For x = 0.15, T
c
has a maximum of 38 K and starts to decrease further with
increasing x. La
2-x
Sr
x
CuO
4
is a particularly simple since the Cu-O planes are unique and
there are no so-called Cu-O chains in this material.
The relevant nuclear parameters for NMR in La
2-x
Sr
x
CuO
4
are shown in Tab. 1. One
realizes that all nuclei have quadrupole moments. The NMR sensitivity is high for a large
resonance frequency (v = B) and high abundance of the particular isotope. We see from
Tab. 1 that both Cu isotopes provide good sensitivity. NMR on the
I7
O isotope is not
readily performed due to the low abundance and one works preferably with enriched
samples.
139
La NMR is easy to perform in contrast to that of Sr.
From the low crystallographic symmetry of the materials one expects strong
quadrupole interactions. Early experiments showed that the quadrupole splitting is around
30 MHz for Cu, a few hundred kHz for O and a few MHz for La. Therefore, zero field
NQR measurements are readily performed on both Cu isotopes and on La, but not on O.
Table 1. Selection of nuclear parameters for NMR in La
2-x
Sr
x
CuO
4
.
Isotope Spin Gyromagnetic ratio
in MHz/T
Quadrupole moment
Q in 10-
24
cm
2
Natural
abundance
Pin %
63
Cu
65
Cu
17
O
139
La
87
Sr
3 / 2
3 / 2
5 / 2
7/2
9/2
11.28
12.09
5.77
6.01
1.84
-0.210
-0.195
-0.026
0.20
0.15
69.09
30.91
0.037
99.91
7.02
For NMR experiments various complications arise: (1) The rather large quadrupole
coupling for Cu which causes an orientation dependent splitting would result in huge
distributions for powder samples. Therefore, if single crystals are not available, a magnetic
grain alignment is performed. The powder is mixed with epoxy which is then cured in the
magnetic field. Clearly, such a process which is based on the magnetic anisotropy of the
grains will not be very accurate and will produce a distribution of angles between the
crystal c axis and the external magnetic field. (2) Even for aligned material there are a
great many resonance frequencies due to (i) the various transitions of the quadrupole
perturbed NMR, cf. Eq. (1), (ii) the 2 different isotopes for Cu, (iii) the 2 different oxygen
sites, and (iv) the presence of non-equivalent sites (see below) for a given isotope due to
doping. (3) As we will see later on, the distribution of local fields can be quite large in
superconductors. This adds to the complication with the various resonances and the
difficulty of alignment so that severe resolution problems exist in cuprates.
In order to overcome these problems we have introduced new NMR methods [22]
which are indeed very helpful. The basic idea is simple: For quadrupole perturbed NMR
417
the various transitions belong to nuclear spin flips between the 2I eigenstates, cf. Fig. 1.
Before the experiment starts, these will be populated according to the Boltzmann factor.
Given that the splittings are very small, the level population will be a linear function of the
magnetic quantum number. Now, with a pulse that precedes the normal experiment one
can change the level population at will, e.g., a selective inversion pulse on the upper most
transition (Fig. 1) will change the intensities observed in a succeeding (before relaxation
sets in) usual NMR experiment. This is of great importance since the effective interactions
are different for the different nuclear transitions (the central transition is only affected by
the magnetic shift, cf. Fig. 1). This way one can study the correlation between magnetic
shift and electric field gradient, obtain single isotope spectra (Fig. 3), or, measure the
magnetic shift for overlapping apical and planar oxygen signals. The difference of these
methods compared to ordinary NMR is that we irradiate two different frequencies, a
special kind of NMR double resonance [18].
Figure 3. (a) Upper satellite spectrum of La
1.85
Sr
0.15
CuO
4
at 8.3 T and 300 K. (b) Difference between the
regular spectrum and one where the NMR experiment was preceded by an inverting pulse on the
63
Cu central
transition. One notices the absence of
65
Cu signal intensity and the lower frequency tail from not well
aligned grains.
BASICS OF NMR IN THE NORMAL STATE
Application of NMR to the newly found HTSC [23] revealed many important details.
We will focus on the superconducting materials (but should perhaps mention that, e.g., La
NQR is a wonderful probe of the phase transition to Neel order [24]). We will start with
summarizing results that seem to be unique to the various materials, and, one can guess
that the study of the low energy electron spin excitations (from electrons near the Fermi
surface) through spin shift and relaxation measurements are of special interest, as for the
classical superconductors [25]. In order to separate the spin shift from the magnetic shift
one has to do temperature dependent studies of the magnetic shift assuming that the model
of a temperature independent orbital shift is correct. For such experiments, preferably the
central transition is observed, n = 0 in Eq. (1).
We start with the Cu NMR. The observed magnetic shift for c||B
0
(crystal c-axis
parallel to the magnetic field in z-direction) turns out to be rather large but temperature
independent. For c B
0
the magnetic shift decreases with temperature, this effect is stronger
for doping levels below that with highest T
c
(optimal doping). Roughly speaking, the total
magnetic shift is about twice as big with c||B
0
at 300 K. An appealing way to disentangle
418
both shift contributions is by assuming that the spin shift accidentally vanishes for c||B
0
.
For c B
0
one assumes that the temperature independent component (inferred from
comparison with, e.g., Y as an internal standard) represents the orbital shift. This analysis,
especially the vanishing spin shift for c||B
0
, may seem somewhat artificial, however, it
yields 3 results which fit very well other experiments, (i) The ratio of the orbital shift for
both orientations thus obtained is about 4 and therefore agrees with a simple model for a
Cu
2+
state (3d
9
) . From the mixing of the orbitals xy and yz with x
2
-y
2
due to L
2
or L
x
,
respectively, we would expect a factor of 4 from Eq. (3) for vanishing orbital energy
differences between xy and yz [19]. (ii) The temperature dependence of the spin shift
(c B
0
) thus obtained is proportional to that of the static spin susceptibility measured with a
magnetometer [26, 27]. (iii) For the planar oxygen resonance where one expects small
orbital shifts the temperature dependence of the Knight shift agrees with that determined
for Cu.
The question arises whether one can understand the observed spin shift data? Without
going into details about the history of the explanation of the electron nuclear coupling [23,
28, 29], we present the hyperfine Hamiltonian [30, 31] which is believed to be correct for
the cuprates. It assumes a single electronic fluid to which all nuclei couple such that we
have the following expressions for the spin shift, cf. Eq. (4).
The Cu nucleus couples to the onsite electron spin as well as to that of the four Cu
neighbors,
(5)
The planar O nucleus couples to two neighboring Cu electron spins,
(6)
where B, are the hyperfine coupling coefficients for the alignment of the sample
(note that B is isotropic). For La
2-x
Sr
x
CuO
4
we can also write down a similar expression for
the apical O nucleus,
(7)
For the latter material we have the following values for the hyperfine coupling
constants, A
c
= -41.8, B = 11.5, C
c
= 7.44, E
c
= 2.22 in units of 10
7
Hz [32, 33].
For a uniform spin polarization we have
(8)
where is the uniform spin susceptibility. As explained above, one can fit the shift data
with Eq. (8) and determine the hyperfine coupling constants. It is interesting to note that
for the various materials one finds A
c
+ 4B 0, which comes as a surprise. Another
interesting aspect is the decrease of the spin shift as the temperature is lowered at
temperatures which are far above the critical temperature for underdoped materials. For a
metal one would expect a temperature independent spin polarization and thus a
temperature independent spin shift. Classic superconductors show a drastic decrease in
spin shift only below T
c
where the electron spins pair up in singlets with vanishing
magnetic moment (freezing out of spin degrees of freedom). The fact that the spin
excitations decrease already above T
c
has been named the spin gap or pseudo gap
419
effect. Other methods also find a decrease in spectral weight already far above T
c
, a
phenomenon which is not understood.
At this point we could talk about spin lattice relaxation or spin-spin relaxation.
However, these topics are somewhat more involved and not very well understood. So we
will only give a brief summary [23, 33-35].
First, one finds that the dominant relaxation mechanisms for Cu and planar oxygen are
magnetic (but not exclusively [36]). It is interesting to note that due to the different
structure of the Zeeman vs. quadrupole Hamiltonian the transition matrix elements for
nuclear spin flips (allowed transitions) are different. Assume we have created a non-
thermal equilibrium population in the nuclear spin system, similar to the one mentioned
earlier (a selective inversion pulse on one transition). Then, the actual time dependence of
all the population numbers will depend on the allowed transitions. One can perform a
mode analysis for various initial conditions and thus find out whether the relaxation
process is magnetic or quadrupolar in origin.
Second, the relaxation rates for Cu and planar O are very different. The time
dependent local magnetic field fluctuations are much stronger at the Cu nuclei. One can
understand such differences by assuming different form factors for both sites and
antiferromagnetic spin fluctuations. From Eq. (5) and (6) it becomes obvious that an
antiparallel electron spin polarization for neighbors will shield the planar O nucleus from
fluctuating magnetic fields but not the Cu. Indeed, a nearly quantitative understanding
could be reached [3, 37]. However, since it seems to be proven lately that the low energy
spin fluctuations are incommensurate with the underlying lattice, the oxygen relaxation
should be faster than actually observed. This problem is still not quite understood.
Third, the pseudo gap effect for the shifts is also present for the spin-lattice relaxation,
as it should, since the scattering rate of electrons which is involved in magnetic nuclear
relaxation will be diminished as well if electronic spin degrees of freedom freeze out.
INHOMOGENEITIES IN La
2-x
Sr
x
CuO
4
FROM NMR
It was clear from the very beginning that the HTSC must be inhomogeneous to some
extent since they derive from the stoichiometric parent compounds by doping. In systems
where the dopant can move, like oxygen doped La
2
Cu it was soon observed that phase
separation can occur into hole rich and hole pure regions. It is still a matter of debate what
the details of such ordering are (e.g., for oxygen ordering in the chains of YBa
2
Cu
3
The situation is quite different in La
2-x
Sr
x
CuO
4
(LSCO) since the Sr atoms are
distributed in the lattice at high temperatures and are immobile certainly below ambient
temperature. Apart from the doping induced lattice changes there is a phase transition from
a high temperature tetragonal phase to a low temperature orthorhombic phase [21]. This
phase transition temperature decreases with the doping level, but, it does not seem to be
related to the onset of superconductivity.
Let us review some NQR results. Early experiments [38,39] in LSCO already revealed
that there were 2 inequivalent Cu sites in the Cu-O plane, cf. Fig. 3, they were called the A
site and B site. It was noticed that the number of B sites is roughly the same as that of Sr
atoms. It was concluded that this site has to be related to the dopant, perhaps a Cu site
close to it. However, this interpretation was challenged [40] in the following years when it
was discovered that the oxygen doped material also showed two lines but with different
intensities. The idea was put forward that the second line might be a result of a certain
amount of trapped holes the number of which increases with doping. However, with the
new NMR methods we found that the magnetic linewidth of both lines in LSCO is similar,
the same is true for their shifts [41]. Together with theoretical calculations [42] it seems
420
quite convincing that the Cu B site is the Cu position which is bridged by an apical oxygen
to a Sr atom (we also find an apical oxygen B site). The second site in the oxygen doped
material must be related to the oxygen dopant and accidentally similar in frequency to that
in LSCO.
Figure 4. Sketch of the doping dependence of the NQR frequency v
Q
(or NMR satellite splitting) and the full
width at half height of the resonance line at 300 K.
We will now focus on the intense Cu A site only, since we think that the debate over
the origin of the Cu B site is settled (we will give more evidence below). We now look at
the NQR parameters at 300 K as a function of doping in Fig. 4.
The electric field gradient increases smoothly with doping, 21MHz [43].
On the other hand, the distribution of field gradients increases drastically from about 60
kHz for the undoped material to about 2.2 MHz near the onset of superconductivity. As the
doping increases further the linewidth remains nearly constant. Although one might expect
an increase in the distribution of the EFG upon doping due to lattice inhomogeneities, the
actual data are surprising: (i) One would expect a doping dependence of the EFG
distribution for x > 0.05. (ii) From the doping dependence of the mean frequency one
concludes that a 2.2 MHz linewidth would correspond to a variation in doping of about
0.1 which for small dopings is bigger than the average doping level. From NMR satellite
transition measurements for c||B
0
and c B
0
one can estimate the asymmetry of the EFG
since its largest component is along the crystal c-axis. We find that the EFG remains
axially symmetric which seems to contradict random lattice distortions.
Very recently, it was discovered [13] that the Cu NQR intensity for the underdoped
materials is lost below a given temperature which is similar to the charge ordering
temperature observed with inelastic neutron scattering in similar co-doped materials. If the
local electric field gradient undergoes slow (on the time scale of measurement, i.e., ~1 )
variations with large amplitudes, this can wipe out the NQR signal. Another possibility for
the intensity to disappear is by slowly fluctuating electronic moments, e.g., the transition
into a spin glass state, since the hyperfine fields at the nucleus are very large. NMR
experiments at sufficiently low temperatures, such that the low energy excitations are
frozen out, can reveal the order in the system. At the present time it seems likely that both
fluctuations are involved, but it is not yet certain [44].
Above the doping level of x = 0.12 no loss of intensity is observed, instead, we
observe temperature dependent changes in all the local fields. One of the strongholds of
NMR is that we can measure the local fields at various locations in the unit cell and
421
compare them with each other. We will illustrate the foregoing. First let us compare the
local field distributions, in terms of the second moments, of the central transitions of the
planar Cu and oxygen nuclei in optimally doped LSCO. This is shown in Fig. 5. As the
temperature is lowered from 300 K the inverse root of the second moments decreases. This
increase in linewidth is similar for both nuclei down to about 80 K where the oxygen line
starts to narrow. That is indeed to be expected if the broadening is caused by spin effects
(pseudo gap effect). We also realize that the Cu linewidth keeps increasing, which tells us
that it cannot be dominated by spin effects. Could long range doping variations, or grain
effects cause this behavior? In order to investigate that problem one can perform a Spin-
Echo-Double-Resonance (SEDOR) experiment [18] which makes use of the short-range
inter nuclear magnetic coupling (dipolar and indirect nuclear coupling). Here, one looks at
a part of the broad magnetic
63
Cu lineshape with a usual spin echo experiment. For
comparison, in a similar, second experiment one irradiates in addition a selective inversion
pulse to some of the
65
Cu nuclei, i.e., of the other isotope. (Both isotopes lineshapes are
similar and the difference in the resonance frequencies of the 2 isotopes central transition
is much larger than their linewidth.) This additional pulse flips some
65
Cu nuclear moments
and will change the local field at all its neighbors, thus preventing them from contributing
to the
63
Cu spin echo (destruction of the spin echo for the neighbors). If the large Cu
linewidth were caused by long wavelength magnetic field variations, neighboring Cu
nuclei would feel the same local field and thus in the SEDOR experiment described, the
destruction of the
63
Cu spin echo would only occur for the part of the
63
Cu lineshape that
corresponds to the position of the selective inversion pulse of the
65
Cu.
We observed experimentally that no matter where the selective inversion pulse is
irradiated, the destruction of the spin echo occurs at all parts of the
63
Cu line. This tells us
that the strong local field distributions occur over rather short distances and not at long
distances such as between different grains in the powder.
Figure 5. Inverse root second moments of the
63
Cu and
17
O central transition lineshape at 8.3 T for optimally
doped LSCO (full lines are guides to the eye).
Before we draw conclusions about the origin of the field distributions, we compare the
planar oxygen data with those for the apical oxygen. We use Eq. (6) and Eq. (7) and
assume that a variation in the spin polarization can be written as
422
(9)
where has zero mean and varies between the Cu sites and creates the linebroadening.
Introducing Eq. (9) into Eq. (6) and Eq. (7), we see that the root second moments of the
spin shifts contain the variance and correlation function between neighbors, i.e.,
and While the variance can be obtained from the apical oxygen linewidth,
by plotting the second moments of the planar oxygen versus that of the apical oxygen we
can determine the electronic correlation function for neighbors. This is shown in Fig. 6.
From the slope of the line in Fig. 6 we find,
(10)
independent on temperature. This result comes as a surprise since one would expect a ratio
near minus one for a predominantly antiferromagnetic electronic spin response.
We can write down equations for the second moment of the Cu lineshape which will
include correlations between more distant neighbors, cf. Eq. (5). If we only measure one
more parameter, the Cu linewidh for this alignment, we cannot solve for all the correlation
functions. However, we can try to find the maximum possible spin contribution to the local
field at the Cu from that measured at the planar and apical oxygen position. By doing so,
we find that the Cu local field from spin effects is much too small to explain the observed
width at any temperature. This fact, together with the lack of narrowing of the Cu line at
lower temperatures clearly states that the spin effects are not responsible for the (short
wave length) Cu local field variations.
Figure 6. Second moment of the central transitions of the planar
17
O line versus that of the apical
17
O line
(contributions from nuclear dipole interactions subtracted) for temperatures between 100 K and 300 K at 8.3
T (the solid line is linear fit).
The only explanation left for the Cu linebroadening is thus a large scale orbital shift
variation, which comprises almost the total orbital shift range at lower temperatures. Such
effects are very unusual since the orbital structure typically remains unaltered for these
423
small temperature changes. (Contributions from second order quadrupole coupling are
ruled out by studies of the magnetic field dependence of the linewidth.)
Figure 7. Apical oxygen NMR at 8.3 T (c||B
0
). Shown are the two satellite transitions with n = +2 at 300 K
and n = 2 at 80 K. This reveals the temperature independent broadening which is also the same for both
transitions (the satellites with n = lwhich are not shown are half as wide, cf. Fig. 1; the sharp features to the
right of the main line are caused by incomplete subtraction of the planar oxygen resonance).
The large distributions of the EFG for the Cu, which can be measured with NQR and
satellite NMR, are found to be temperature independent at optimal doping. At the apical
oxygen site, cf. Fig. 7, one also finds a temperature independent distribution of the EFGs
but much larger, 12 % of v
Q,z
for oxygen (only 3.2 % for Cu). The origin of these
distributions are still unclear, but suggest the involvement of the 3 z
2
-r
2
orbital for Cu.
For the planar oxygen the situation is quite different as can be seen in Fig. 8. We
notice that the linewidths of the various satellite transitions are not the same but increase as
the temperature is lowered, however, the apparent asymmetry of the whole set of lines
remains unchanged. From comparison with Fig. 1 it is obvious that the found spectrum is
neither caused by a mere quadrupolar nor magnetic shift distribution. The full oxygen
spectrum must result from an interplay between variations of both the local magnetic field
and electric field gradient such that, e.g., for the n = -1 satellite both shifts oppose each
other whereas they add for the n = +1 line. It turns out that we can understand the
experimental findings with a simple theory where we assume that K
S
and the EFG at the
planar oxygen are linear functions of a hidden parameter h, such that
(11)
and h is distributed over a range of values. Then one can reproduce the data and determine
the ratio R/M. It is found that between 300 K and 100 K the ratio R/M is about 2 for 8.3 T.
It is inversely proportional to the applied field, since M is proportional to the field. As one
approaches T
c
, M decreases rapidly (it describes the broadening of the central transition, cf.
Fig. 5) and the full spectrum become symmetric. This shows that at the planar oxygen site
the electric field gradient is a linear function of the spin shift (a correlation of spin and
charge). Also, one notes from Fig. 8 that these effects are quite strong.
For other doping levels such detailed analysis is not yet available. A few more results
on the planar oxygen modulation are shown in Fig. 9 for other Sr dopings.
424
Figure 8. Total planar oxygen lineshapes at 300 K and 100K at 8.3 T.
It is seen that the correlated modulation of the EFG and the spin shift at the planar
oxygen site depend on the doping level x. The clear asymmetry found for the optimally
doped sample, shown in Fig. 8, is not observed for the x = 0.10 sample. One also sees a
slight increase in the local field distribution. For the overdoped sample, x = 0.20, we
observe a tremendous increase in linewidth as the temperature is lowered, the asymmetry
seems to increase as well, indicating an enhanced correlation between EFG and spin shift.
Figure 9. Total planar
17
O spectra at 300 K and 80 K of La
2-x
Sr
x
CuO
4
for x = 0.10 and 0.20 at 8.3 T.
DISCUSSION
A direct consequence of Sr doping of La
2
CuO
4
is the appearance of the planar Cu and
apical oxygen B sites. Together with the facts that the Cu B site shows similar local field
distributions as the Cu A site (as well as modulations of the spin-spin interaction which we
only mention here), and, that the measured shifts (and spin-lattice relaxation rates) can be
explained by quantum-chemical calculations [45] we believe that the B sites are caused by
changes of the location of the apical oxygen which bonds to Sr and Cu. This result is
perhaps not surprising but has to be taken into account for the interpretation of other
structural data which find ambiguous mean Cu-apical O distances [16, 46].
425
The tremendous increase of the EFG distribution (300 K) at Cu which is nearly
independent of doping for x > 0.05 is surprising. The distribution is well approximated by
a Gaussian even for low dopings where one might expect two components from Cu near
the dopant and those very far from it. The found distribution of apical oxygen positions
[20, 21] which does not seem to depend on temperature agrees with these findings.
The loss of Cu NQR intensity and the formation of static magnetic moments below a
certain temperature for underdoped LSCO suggest a transition into a spin and/or charge
ordered state and the freezing of stripes was proposed [13] to be a likely scenario. For
optimally doped LSCO (x = 0.15) there is no transition into a charge or spin ordered state.
However, we showed evidence for correlated modulations at the various points in the unit
cell. Most surprisingly it involves a short range orbital shift modulation which is strongly
temperature dependent and increases as the temperature is lowered. It seems to be
correlated with a modulation of the electron spin susceptibility which was measured at the
oxygen positions. This modulation also increases as the temperature is lowered (apart from
the pseudo gap effect). The electron spin correlation function is almost zero (this does not
mean that there are no correlations, an additional wavelength for the spin fluctuations apart
from the antiferromagnetic one could give a similar result). Lastly, the total NMR
spectrum of the planar oxygen reveals very convincingly the correlated modulation of the
susceptibility with that of the charge (electric field gradient).
Neutron scattering [9, 14] finds incommensurate spin fluctuations for all dopings. On
the underdoped site of the phase diagram the incommensurability is proportional to the
doping x and saturates for the highest T
c
. However, elastic peaks [47] are only found for
the underdoped materials. This change in the behavior from the presence of static spin
structures to an incommensurate susceptibility might be connected with the NMR/NQR
results. A modulation of the static susceptibility will not result in elastic neutron scattering.
Such a modulation could be observed only if a magnetic field were present.
One might think about alternative concepts for the interpretation of the local field
distributions in the optimally doped material. The presence of localized holes (< 0.5 %)
whose presence has been suggested by various authors could have an effect on the local
fields. They should represent a polarizable medium which changes the electronic spin
susceptibility [48, 49]. For small concentrations of such holes one expects a magnetic spin
shift broadening (if the correlation length of the electron spin fluctuations is not too large)
particularly near such holes. While such a scenario might explain the doping independent
EFG distribution at the Cu and the temperature dependence for the oxygen linewidth, there
are various problems: (i) The correlation function between neighbors should still be close
to 1 and not near zero as measured, (ii) It does not explain the correlation between the
oxygen magnetic shift and the temperature dependent EFG at the planar oxygen. (iii) It
does not explain the Cu orbital shift distribution and its temperature dependence. One can
find more arguments against an explanation of the results which has localized holes as the
only agent.
It should be said that the local field distributions differ among the various cuprates.
All the doped La
2
CuO
4
materials seem to have on average larger distributions than the
YBa
2
Cu
3
family of materials, in particular the stoichiometric compound YBa
2
Cu
4
O
8
.
Nevertheless, the temperature dependences of the linewidths we find are similar to those
presented for LSCO. Although a detailed analysis of the local field distributions in other
cuprates is not available, we found also short wavelength spatial modulations which seem
to exhibit similar properties as for LSCO. The planar
17
O spectrum for an (optimally
doped) YBa
2
Cu
3
O
6 . 95
sample, shown in Fig. 10, reveals the similarities most clearly.
An appealing explanation for the fact that the local field distributions are different in
various samples could perhaps start with the idea of the pinning of stripes: The degree of
crystallographic disorder might produce pinning potentials for otherwise rapidly
426
fluctuating dynamic electronic structures from which NMR would only measure the time
average.
Figure 10. Planar
17
O NMR spectrum at 100 K and 8.3 T for YBa
2
Cu
3
O
6.95
. The solid line is a fit using a shift
and EFG correlation according to Eq. (11), R/M = 1.8 (there are 2 non-equivalent planar oxygen sites in this
material).
NMR IN THE SUPERCONDUCTING STATE
The superconducting state bears consequences for NMR [25] due to the changes in
low energy excitations and flux quantization, e.g., the spin shift disappears and additional
local field variations are acquired for NMR. This allows the study of quasiparticle
excitations (which was so important for the proof of the BCS theory [50]) and of fluxoid
properties [51].
At present, only the basic features have been investigated (singlet pairing, d-wave
symmetry). Nevertheless, there are indications for unusual behaviour, e.g., the local field
distribution for Cu is bigger than that expected from the magnetic field distribution due to
the fluxoid lattice. More experiments are needed and a better understanding of the normal
state has to be reached.
CONCLUSIONS
We reviewed recent experimental data and showed new evidence in support of the
idea that structural disorder due to doping is not the mere cause of the large local field
variations observed with nuclear magnetic resonance. The above data showed that there is
also disorder present among the electrons which engage in conductivity
(superconductivity) and that these inhomogeneities are of short range. The data do not
allow us to give a detailed picture of the underlying causes nor can we draw conclusions
from them about the possible connection between inhomogeneous electronic structure and
superconductivity. More experiments and a better understanding of the results obtained
with other methods will help to better identify the cause and role of inhomogeneous
electronic structures.
427
Acknowledgement
This work was supported by The Science and Technology Center for
Superconductivity under NSF Grant No. DMR 91-20000 and the U.S. DOE Division of
Materials Research under Grant No. DEFG 02-91ER45439. We would like to thank N.J.
Curro, T. Imai, P.C. Hammel, S. Kos, A.J. Leggett, P.F. Meier, K.A. Mller, D.K. Morr,
D. Pines, R. Ramazashvili, J. Schmalian for helpful discussions. CPS would like to thank
D. MacLaughlin for a fruitful discussion of tests of the modulation of the orbital shift. J.H.
acknowledges the support by the Deutsche Forschungsgemeinschaft and D.G.H support
from DOE - Basic Energy Sciences under Contract No. W-31-109-ENG-38.
REFERENCES
*Permanent address: National Institute of Chemical Physics & Biophysics, Akadeemia tee 23, 12618 Tallinn,
Estonia
1. Bednorz, J.G. and Mller, K.A. (1986) Possible High T
C
Superconductivity in the Ba-La-Cu-O System
Z. Phys. B 64, 189-193.
2. Rossat-Mignod, J., Regnault, L.P., Jurgens, M.J., Vettier, C., Burlet, P., Henry, J.Y., and Lapertot, G.
(1989) Inelastic Neutron Scattering Study of YBa
2
Cu
3
O
6+x
, Physica C 162, 1269-1270.
3. Pines, D., (1998) The Spin Fluctuation Model for High Temperature Superconductivity. Progress and
Prospects, in NATO ASI on The Gap Symmetry and Fluctuations in High-Tc Superconductors, edited by
J. Bok and G. Deutscher.
4. Bourges, P., Sidis, Y., Fong, H.F., Regnault, L.P., Bossy, J., Ivanov, A., and Keimer, B. (2000) The Spin
Excitation Spectrum in Superconducting YBa
2
Cu
3
O
6.85
, Science 288, 1234-1237.
5. Zaanen, J. and Gunnarsson, O. (1989) Charged Magnetic Domain Lines and the Magnetism of High-T
c
Oxides, Phys. Rev. B 40, 7391-7394.
6. Poilblanc, D. and Rice, T.M. (1989) Charged Solitons in the Hartree-Fock Approximation to the Large-
U Hubbard Model, Phsy. Rev. B 39, 9749-9752.
7. Emery, V.J., Kivelson, S.A., and Lin, H.Q. (1990) Phase Separation in the t-J Model, Phys. Rev. Lett.
64, 475-478.
8. Tranquada, J.M., Sternlieb, B.J., Axe, J.D., Nakamura, Y., and Uchida, S. (1995) Evidence for Stripe
Correlations of Spin and Holes in Copper Oxide Superconductors, Nature 375, 561-563.
9. Yamada, K., Lee, C.H., Kurahashi, K., et al. (1998) Doping Dependence of the Spatially Modulated
Dynamical Spin Correlations and the Superconducting Transition Temperature in La
2-x
Sr
x
CuO
4
, Phys.
Rev. B 57, 6165-6172.
10. McQueeney, R.J., Petrov, Y., Egami, T., Yethiraj, M., Shirane, G., and Endoh, Y. (1999) Anomalous
Dispersion of LO Phonons in La
1.85
Sr
0.15
CuO
4
at Low Temperatures, Phys. Rev. Lett. 82, 628-631.
11. Mook, H.A., Dai, P., Dogan, R., and Hunt, R.D. (2000) One-Dimensional Nature of the Magnetic
Fluctuations in YBa
2
Cu
3
O
6.6
, Nature 404, 729-731.
12. Lee, Y.S., Birgenau, R.J., Kastner, M.A., Endoh, Y., Wakimoto, S., Yamada, K., Erwin, R.W., Lee, S.-
H., and Shirane, G. (1999) Neutron-Scattering Study of Spin-density Wave Order in the
Superconducting State of Excess-Oxygen-Doped La
2
Cu Phys. Rev. B 60, 3643-3654.
13. Hunt, A.W., Singer, P.M., Thurber, K.R., and Imai, T. (1999)
63
Cu NQR Measurement of Stripe Order
Parameter in La
2-x
S
x
CuO
4
, Phys. Rev. Lett. 82, 4300-4303.
14. Kimura, H., Hirota, K., Matsushita, H., et al. (1999) Neutron-Scattering Study of Static
Antiferromagnetic Correlations in La
2-x
Sr
x
Cu
1-y
Zn
y
O
4
, Phys. Rev. B 59, 6517-6523.
15. Bianconi, A., Valletta, A., Perali, A., and Saini, N.L. (1998) Superconductivity of a Striped Phase at the
Atomic Limit, Physica C 296, 269-280.
16. Haskel, D., Stern, E.A., Hinks, D.G., Mitchell, A.W., and Jorgensen, J.D. (1997) Altered Sr
Environment in La
2-x
Sr
x
CuO
4
, Phys. Rev. B 56, R521-R524.
428
17. Bozin, E.S., Kwei, G.H., Takagi, H., and Billinge, S.J.L. (2000) Neutron Diffraction Evidence of
Microscopic Charge Inhomogeneities in the CuO
2
Plane of Superconducting La
2-x
Sr
x
CuO
4
(0<x<0.30),
Phys. Rev. Lett. 84, 5856-5859.
18. Slichter, C.P. (1996) Principles of Magnetic Resonance, Springer Verlag, New York.
19. Pennington, C.H., Durand, D.J., Slichter, C.P., Rice, J.P., Bukowski, E.D., and Ginsberg, D.M. (1989)
Static and Dynamic Cu NMR Tensors of YBa
2
Cu
3
Phys. Rev. B 39, 2902-2905.
20. Cava, R.J., Santoro, A., Jr, D.W.J., and Rhodes, W.W. (1987) Crystal Structure of the High-Temperature
Superconductor La
1.85
Sr
0.l5
CuO
4
above and below T
c
, Phys. Rev. B 35, 6716-6720.
21. Radaelli, P.G., Hinks, D.G., Mitchell, A.W., Hunter, B.A., Wagner, J.L., Dabrowski, B., Vandervoort,
K.G., Viswanathan, H.K., and Jorgensen, J.D. (1994) Structural and Superconducting Properties of La
2-
x
Sr
x
CuO
4
as a Function of Sr Content, Phys. Rev. B 49, 4163-4175.
22. Haase, J., Curro, N.J., Stern, R., and Slichter, C.P. (1998) New Methods for NMR of Cuprate
Superconductors, Phys. Rev. Lett. 81, 1489-1492.
23. Slichter, C.P. (1993) Experimental Evidence for Spin Fluctuations in High Temperature
Superconductors, in Strongly Correlated Electronic Materials, Addison-Wesley,
24. MacLaughlin, D.E., Vithayathil, J.P., Brom, H.B., et al. (1994) Abrupt but Continuous
Antiferromagnetic Transition in Nearly Stoichiometric La
2
Cu Phys. Rev. Lett. 72, 760-763.
25. MacLaughlin, D.E. (1976) Magnetic Resonance in the Superconducting State, Solid State Physics,
Academic Press, Inc., New York.
26. Takigawa, M., Reyes, A.P., Hammel, P.C., Thompson, J.D., Heffner, R.H., Fisk, Z., and Ott, K.C.
(1991) Cu and O NMR Studies of the Magnetic Properties of YBa
2
Cu
3
O
6.63
(T
c
=62K), Phys. Rev. B 43,
247-257.
27. Alloul, H., Ohno, T., and Mendels, P. (1989)
89
Y NMR Evidence for a Fermi-liquid Behavior in
YBa
2
Cu
3
O
6+x
, Physical Review Letters 63, 1700-1703.
28. Mehring, M. (1992) What Does NMR tell us about the Electronic State of High-T
c
Supercondurctors?,
Appl. Magn. Reson. 3, 383-421.
29. Brinkmann, D. and Mali, M. (1994) NMR-NQR Studies of High-Temperature Superconductors, NMR
Basic Principles and Progress 31, 171-211.
30. Shastry, B.S. (1989) t-J Model and Nuclear Magnetic Relaxation in High-T
c
Materials, Phys. Rev. Lett.
63, 1288-1291.
31. Mila, F. and Rice, T.M. (1989) Analysis of Magnetic Resonance Experiments in YBa
2
Cu
3
O
7
, Physica C
157, 561-570.
32. Haase, J., Slichter, C.P., Stern, R., Milling, C.T., and Hinks, D.G., Spatial Modulation of the NMR
Properties of the Cuprates; in 6th International Conference on Materials and Mechanisms of
Superconductivity and High-Temperature Superconductors, Houston, TX, 2000).
33. Zha, Y., Barzykin, V., and Pines, D. (1996) NMR and Neutron-Scattering Experiments on the Cuprate
Superconductors: A Critical Re-examination, Phys. Rev. B 54, 7561-7574.
34. Walstedt, R.E. and Cheong, S.W. (1996)
63,65
Cu Indirect Spin-Spin Coupling in La
1.85
Sr
0.15
CuO
4
, Phys.
Rev. B 53, R6030-R6033.
35. Curro, N.J., Imai, T., Slichter, C.P., and Dabrowski, B. (1997) High-temperature
63
Cu(2) Nuclear
Quadrupole and Magnetic Resonance Measurements of YBa
2
Cu
4
O
8
, Phys. Rev. B 56, 877-885.
36. Suter, A., Mali, M., Roos, J., and Brinkmann, D. (2000) Charge Degree of Freedom and the Single Spin-
Fluid Model in YBa
2
Cu
4
O
8
, Phys. Rev. Lett. 84, 4938-4941.
37. Millis, A.J., Monien, H., and Pines, D. (1990) Phenomenological Model of Nuclear Relaxation in the
Normal State of YBa
2
Cu
3
O
7
, Phys. Rev. B 42, 167-178.
38. Kumagai, K. and Nakamura, Y. (1989) Evidence of Magnetic Order of Cu Moments in the
Superconducting and Normal Region of La
2-x
Sr
x
CuO
4
Observation of Cu-NQR for 0.12<x,0.4, Physica
C 157, 307-314.
39. Yoshimura, K., Imai, T., Shimizu, T., Ueda, Y., Kosuge, K., and Yasuoka, H. (1989) Copper NMR and
NQR in La
2-x
Sr
x
CuO
4
- No Evidence for Coexistence of Magnetic Ordering and Superconductivity down
to 1.3K, J. Phys. Soc. Jap. 58, 3057-3060.
429
40. Hammel, P.C. (1998) Localized Holes in Superconducting Lanthanum Cuprate, Phys. Rev. B 57, R712-
R715.
41. Haase, J., Stern, R., Hinks, D.G., and Slichter, C.P., (2000) On the Structre of the Cu B-Site in La
2-
x
Sr
x
CuO
4
, in Stripes and Related Phenomena, edited by A. Bianconi and N. L. Saini (Plenum Press, New
York, 2000) (in press).
42. Stoll, E.P., Plibersek, S., Renold, S., Claxton, T.A., and Meier, P.F., (2000) Interpretation of NQR
Spectra in Doped La
2
CuO
4
, in Major Trends in Superconductivity in the New Millennium, Klosters, CH.
43. Ohsugi, S., Kitaoka, Y., Ishida, K., Zheng, G., and Asayama, K. (1994) Cu NMR and NQR Studies of
High-T
c
Superconductor La
2-x
Sr
x
CuO
4
, J. Phys. Soc. Jap. 63, 700-715.
44. Curro, N.J., Hammel, P.C., Suh, B.J., Hcker, M., Bchner, B., Ammerahl, U., and Revcolevschi, A.
(2000) Inhomogeneous Low Frequency Spin Dynamics in La
1.65
Eu
0.2
Sr
0.15
CuO
4
, cond-mat/9911268
45. Hsser, P., Suter, H.U., Stoll, E.P., and Meier, P.F. (1999) First-Principles Calculations of Hyperfine
Interactions in La
2
CuO
4
, Phys. Rev. B 61, 1567-1579.
46. Bianconi, A., Saini, N.L., Lanzara, A., Missori, M., Rossetti, T., Oyanagi, H., Yamaguchi, H., Oka, K.,
and Ito, T. (1996) Determination of the Local Lattice Distortions in the CuO
2
Plane of La
1.85
Sr
0.15
CuO
4
,
Phys. Rev. Lett. 76, 3412-3415.
47. Wakimoto, S., Shirane, G., Endoh, Y., et al. (1999) Observation of Incommensurate Magnetic
Correlations at the Lower Critical Concentration for Superconductivity in
La
2-x
Sr
x
CuO
4
(x=0.05), Phys. Rev. B 60, R769-R772.
48. Morr, D., Schmalian, J., Stern, R., and Slichter, C.P. (1998) Probing the Susceptibility in Cuprates
using Ni Impurities, Phys. Rev. B 58, 11193-11196.
49. Haase, J., Morr, D.K., and Slichter, C.P. (1999) Real Space Electron Spin Susceptibility and NMR
T
2G
for Incommensurate Spin Fluctuations in Cuprates, Phys. Rev. B 59, 7191-7195.
50. Hebel, L.C. and Slichter, C.P. (1959) Nuclear Spin Relaxation in Normal and Superconducting
Aluminum Phys. Rev. 113, 1504-1519.
51. Curro, N.J., Milling, C.T., Haase, J., and Slichter, C.P. (2000) Local Field Dependence of the
17
O Spin
Lattice Relaxation and Echo Decay Rates in the Mixed State of YBa
2
Cu
3
O
7
, Phys. Rev. B , (August
2000).
430
TAILORING THE PROPERTIES OF HIGH-T
C
AND RELATED OXIDES:
From Fundamentals To Gap Nanoengineering
DAVOR PAVUNA
Department of Physics - IPA
Ecole Polytechnique Federale de Lausanne
CH - 1015 Lausanne EPFL, Switzerland
Email: pavuna@epfl.ch
In a direct analogy with a successful (Al)GaAs band-gap engineering, I discuss an
equivalent nanotechnology: the advanced heteroepitaxy of layered oxides enables us to
nano-engineer desired superconducting gap and/or insulating barrier. However, in most
oxides, the progress is somewhat hindered by intrinsic materials problems: growth induced
disorder, difficult local doping and non-homogeneous oxygen distribution. As textbook
understanding of the fundamentals is obviously needed, I also discuss electronic properties
across the electronic phase diagram: the pseudogap controversy, metal-insulator transition
and anomalous transport. As we gradually solve remaining obstacles we will be able to
fully exploit the potential of layered cuprates (and related oxides) and tailor their electronic
(and/or magnetic) properties at will. New concepts and applications will emerge from such
an integrated nano-engineering technology in the 21
st
century, yet for real ambient
technology we still require colossal superconductors with
INTRODUCTION: TAILORING THE PROPERTIES OF LAYERED OXIDES
Some 65,000 papers after the striking discovery of high-T
c
superconductivity in
cuprates [1], we are still trying to fully understand these versatile solids and control their
properties. Complete understanding of fundamentals of a given class of electronic materials
often results in a successful new technology. That may require huge interdisciplinary effort
of the whole generation of scientists and engineers. As the main topic of this workshop is
the role of self-organisation in complex electronic materials, it is useful to begin with
applications of high-T
c
oxides and compare them to prominent electronic materials, like Si
or GaAs. Pure Si is an electronically clean solid that, thanks to the skillful use of doping
and SiO
2
insulating oxide, can be functionally nanoengineered (down to about 3nm) and
integrated into large scale multi-transistor chips, GaAs is optically clean material in which
the direct band-gap and desired AlGaAs heterostructure properties can be modified at will
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 431
(and calculated in advance) by a suitable MBE hetero-epitaxy [2]. As a result, the bandgap
engineering photonic technology produces its own original, archetype device - the laser of
a desired wavelength.
Although many of the leaders of the semiconductor industry or analysts from the Wall
Street consider the field of high-T
c
superconductivity an investment risk, the applications
are advancing successfully and may even dominate some technologies of the 21st century.
Note that the successful Si-based technology has so far accumulated more than 10
7
men-
years of know-how, III-V (GaAs) photonic technology 10
6
men-years, while all
superconductivity hasnt even reached the 10
5
men-years. We clearly need at least another
decade of intensive R & D before giving any definite conclusions to the global media.
Especially so, as superconducting oxides have their own archetype device (ultrafast
Josephson switch, and/or RSFQ-logic) while the magnetic oxides (for example,
manganites) provide some of the most versatile magnetic memories [4]. Moreover, there is
no doubt that the in-depth understanding of the fundamentals [3-7] of this field
(superfluidity included) will be relevant to many branches of advanced science and
technology in the 3
rd
millennium.
To illustrate the concept of high-T
c
oxide nano-engineering I now use a direct
analogy with the bandgap engineering of (Al)GaAs lasers [2]: the direct bandgap, and
consequently the emiting wavelength, is altered by varying the Al content in an optically
clean, epitaxially grown AlGaAs heterostructure. An equivalent conceptual approach to
layered high-T
c
(and related) oxides requires, at the very least, the following:
i) Atomically flat heteroepitaxy of electronically clean constituent blocks (sub-layers)
and clean interfaces. Here, considerable progress hase been made [4,8].
ii) Very precise control of the stoichiometry and local carrier doping, across the whole
electronic (magnetic) and crystallographic phase diagram on nano-, meso- and
macroscopic scale. Controlled local doping still poses formidable challenge in some
oxides. Several reports in this workshop clearly illustrate the difficulties.
iii) Reproducible variation of the superconducting gap (or the pseudogap in some cases)
and/or of the thin insulating barrier for Josephson junctions [4,7]. Again, as shown
by other contributors in this workshop, this is often a non-trivial task, given the
growth-induced disorder and the self-organisation in some oxides.
iv) Suitable integration [8] into novel devices and systems that range from array of
Josephson junctions, RSFQ-logic [2] to magnetic storage and nanomagnetism.
In photonic technology an engineer designs the characteristics of the functional
AlGaAs heteroepitaxy, say, a laser with wavelength of 1.5m, the computational physicist
solves the apropriate Schroedinger equation for a predetermined model-structure and the
crystal-grower delivers it by using a computer-controlled molecular beam epitaxy [2]. One
can consider this whole technology as an applied quantum mechanics. There is effectively
a one-to-one correspondence between the calculated and the fabricated structure and the
corresponding, designed and obtained laser properties.
At present, this is already possible to achieve with some layered oxides. the
nanoengineered superconductivity in BSCCO cuprates and nanoengineered magnetism
in manganites [4,8]. However, complete variation of the doping of an YBCO-123 film is
often hindred by oxygen mobility and by incipient, possibly intrinsic growth-induced
disorder [28]. Namely, even as insulators, most of the technologically interesting transition
metal oxides are not even electronically clean (like pure Si), let alone optically clean as
GaAs waveguides. Furthermore, upon doping, these artificially obtained ionic metals, show
the distribution of oxide vacancies which leads to localized states and/or to (nano-)phase
432
separation [9]. That in turn poses a challenge to our understanding of the true
electrodynamic response as well as correct intepretation of the microscopic origin of the so-
called pseudogap [3,10], the exact role of point defects, twins and grain boundaries, the
symmetry of the gap-parameter [3] and ultimately the understanding of the puzzling
difference between hole- and electron doped cuprates [6]. As oxide nano-engineering
progress is discussed by leading resarchers at length elsewhere [4], I will now discuss
some of the anomalies in the electronic properties of high-T
c
cuprates.
ANOMALIES IN ELECTRONIC PROPERTIES OF HIGH-T
C
OXIDES
To fully understand all difficulties associated with the proposed heterostructure
nanoengineering of layered oxides one should realize that they are i) usually truly
thermodynamically stable only as insulators, ii) as illustrated in Table 1, highly anisotropic
ternary and quaternary solids (except for the isotropic Ba
1x
K
x
BiO3), and iii) artificial
ionic metals (and superconductors below T
c
) obtained by doping of the parent insulating
compound. Their properties are complex as compared to usual normal metals and alloys [3-
18]. For example, there is well documented existence of the so-called pseudogap [10,14],
distinct from the true superconducting, condensation gap [11]. Furthermore, in the
underdoped regime and under very high magnetic fields some of these oxide
superconductors exhibit phase transition directly into the insulating state [12], although
this could be due to the granularity of the sample.
As illustrated in Figure 1 the structure of a high-T
c
oxide can be represented as a
layered structure that consists of two CuO
2
planes separated by a spacer. Between these bi-
layers are interlayer regions which, in the case of YBa
2
Cu
3
O
7
, correspond to the CuO
chains [6]. In general, most cuprate superconductors can be discussed in terms of the
block-reservoir and the doped CuO
2
plane and evidently they can be artificially
constructed by block-by-block epitaxial growth. Recent advent of combinatorial
chemistry enables a fast search for new promising phases [8].
charge reservoir
CuO
2
plane
spacer
CuO
2
plane
charge reservoir
Figure 1. A schematic model unit of double CuO
2
layer cuprate.
Before discussing the electronic phase diagram of cuprates I strongly emphasize that
even the very best single crystals of HTSC oxides often contain various defects and
imperfections like oxygen vacancies, twins, impurities ... These imperfections are not only
relevant to their physical properties but possibly even essential to their basic
thermodynamic stability. It may well turn out that various imperfections found in HTSC
433
crystals are intrinsic to these materials. That may be due to the fact that even their parent
insulators can never be made optically clean. Therefore it is more difficult to achieve true
one-to-one correspondence between the computational solid (as used by most mean-field
theories) and the actual sample, made and measured by experimentalists. This gets worse
upon doping. Note that to this day some of the highest T
c
phases, like Hg-cuprates, were
synthesized only as a small single crystals <1mm in diameter and majority of reported
measurements were performed mainly on LSCO, YBCO and BSCCO crystals.
Table 1. Most representative HTSC compounds. The index n refers to the number of CuO
2
superconducting
layers within a given crystallographic structure. m refers to the number of chains in the structure; m=1.5
corresponds to the case of alternating chains.
It is indeed important to understand that the materials science of HTSC oxides is a
non-trivial pursuit and that the understanding of phase diagrams, crystal chemistry, of the
preparation and stability of layered oxides requires an in-depth study and often hands-on
experience in the laboratory. The advancement of our understanding of physics, chemistry
434
and appearance of applications therefore depends very much on the advancements in
materials research [4,8] that is still very much in progress.
Figure 2. Schematic phase diagram of cuprate superconductors. Various authors attach different name or
significance to various regions. Note that the intermediate phase, left of the maximum of T
c
(and below T
*
) is
a highly controversial subject [9-19].
True textbook understanding [6] of measured properties of high-T
c
oxides and their
electronic phase diagram (Figure 2) still presents a major challenge, despite of a remarkable
progress in both, sample preparation and advanced experimental techniques [3-19]. One of
the reasons for this is that as we dope these layered oxides, we do not encounter only rather
complex electronic phases; the underlying crystallographic (structural) and metallurgical
phase diagrams of these quaternary solids are often even more complex and the disorder
and (possibly intrinsic) inhomogeneities clearly play an important role. Consequently, it is
difficult to determine which properties are intrinsic to these solids and which are induced
by disorder as one varies the doping in real samples. Most reports in the literature have
initially ignored inhomogeneity effects in samples yet recent results are clear: these effects
are real. That partly explains why simple mean field theories cannot describe phenomena
observed in these solids and why experimentalists and especially chemists often ignore
most advanced theories.
At present, most experiments seem to indicate that at the left hand side of the phase
diagram we have 2D antiferromagnetic insulator while, at the right hand side, highly
overdoped 3D perovskites tend to exhibit more Fermi-liquid-like properties. There is an
agreement on the existence of the Fermi surface in the optimally doped and overdoped
samples (see Figure 3) but no definite agreement on the fine features [16]. Still, while many
fine details remain to be clarified, especially in the underdoped and overdoped samples,
there is now well established evidence, shown by most experiments, of the existence the
'pseudogap', T
*
. The 'pseudogap' was originally introduced to cuprates already in 1987 by
435
Figure 3. Experimentally determined Fermi surface of Bi-2212: filled squares experimental Fermi surface
locations; open squares Fermi surface obtained by symmetry operations; closed circles superlattice band
crossings along the M-Z and Y high symmetry directions ; open circles locations in the Brillouin zone
where the ARPES spectra were taken [15].
Phillips [10] and subsequently by Friedel [13] in a different context from presently
dominant idea of a pre-formed pairs in the normal state [7,9,14], Some authors consider
that direct observation of the pseudogap can be made by ARPES, yet this can be
misleading [16] and possibly wrong due to the disorder effects [15,16,27]. Deutscher
approached the pseudogap problem somewhat differently [11]. He has compared gap
energies, measured by different experimental techniques, and has shown that these reveal
the existence of two distinct energy scales: p and c. The first, determined either by angle-
resolved photoemission spectroscopy or by tunnelling, is the single-particle excitation
energy - the energy (per particle) required to split the paired charge-carriers that are
required for superconductivity. The second energy scale is determined by Andreev
reflection experiments, and Deutscher associates it with the coherence energy range of the
superconducting state: the macroscopic quantum condensate of the paired charges. In the
overdoped regime, p and c converge to approximately the same value, as would be the case
for a BCS superconductor where pairs form and condense simultaneously. In the
underdoped regime, where the pseudogap is measured, the two values diverge and p (T
*
) is
larger than c (T
c
) [11]. Phenomenologically, this indeed corresponds to the schematic
phase diagram given in Figure 2. However, very different explanations of T
c
and T
*
are
given by theories that consider the phonon mechanism of superconductivity [10,17] as
compared to more unconventional approaches [18-20].
Metal-insulator transition transition is not understood even in simpler solids like
Si:P so it is not surprising that it remains highly disputed topic in high-T
c
oxides [3,7,9,10],
where anomalous behavior was reported by Boebinger et.al [12]: deep in the underdoped
regime LSCO and BSCCO samples directly transit from superconducting into the
insulating state under very high magnetic field. This, however, could be due to the
granularity of the sample so more work is needed until the experimental situation is
clarified. There is also evidence for formation of stripes [5] in some underdoped perovskites
yet the exact role of stripes in the overall HTSC scenario is still unclear. While some
authors consider them crucial for the pairing mechanism, other consider them only as a
stabilizing feature within a low-dimensional system. What seems clear though is that the
appearance of stripes is a consequence of self-organisation.
436
In the superconducting state the agreement exists that the charge q = 2e and that in
most phases holes are the carriers (see Table 2), that the pairing is singlet and that the
symmetry of the gap is not a conventional, isotropic s-wave. In highly doped samples,
measured properties often appear as BCS-like [6], yet this is also highly disputed by some
authors. Namely cuprates are highly anisotropic, the coherence length is very short,
and there are only Cooper pairs per coherence volume; much less than in the
conventional BCS model [6]. In the underdoped regime the disagreement among
researchers is complete and more careful research is needed to clarify the controversies. In
the underdoped to optimally doped samples, majority of experiments indicate a dominant d-
wave symmetry [3]. There seem to be, however, some notable exceptions [3]: there is no
evidence for d-wave component in the electron doped cuprates (Maryland Center), Sharvin
experiments on LSCO give a finite minimum gap (Deutscher et al [3]) and electronic
Raman experiment on Hg-2201 compound (Sacuto et al [3]) is not compatible with d-wave
but rather with extended s-wave (with nodes). Only few experiments were performed
systematically in the overdoped regime [15,16].
While the main characteristics of the (anomalous) transport in cuprates seem to be
well established [18,21], several recent results [22-24] pose a new challenge to our
understanding. Slightly underdoped, disordered Sr
2
RuO
4y
, superconducting (T
C
=0.9K)
perovskite without Cu, exhibits linear resistivity over three decades of temperatures
[22,25], up to 1050K, yet the temperature dependence of the Hall coefficient is similar to
what was measured in cuprates [22]. This suggests that the linear temperature dependence
of resistivity is not an exclusive signature of the anomalous normal state of high-T
c
cuprates but rather of layered oxides in general, especially single layer perovskites, possibly
independently of the magnitude of the superconducting temperature. What is really striking
is that very clean Sr
2
RuO
4
crystals, grown by zone melting, show usual T
2
term at low
temperatures [26] implying that the linear resistivity can be introduced by the growth-
induced disorder. Note that deliberately introduced disorder (by electron bombardment of
samples [15,16,27]) introduces the pseudogap into ARPES spectra of optimally doped
BSCCO-2212 samples which show no ARPES pseudogap before bombardment [27]. While
in-depth analysis requires more experimentation, obviously the disorder plays some role in
the appearance of the pseudogap [16]. The ongoing pioneering photoemission experiments
on in-situ grown films (other than BSCCO-2212 compound) may provide better insight to
such controversies [28].
Let us also note that in single crystals of T1-2212 [T
c
=111K
exhibits usual linear behavior and follows generally metallic-like, positive slope.
However, there is a clear crossover of to semiconductor-like behavior close to T
c
and,
for the first time, above 500K. Under high pressures (<15 kbar) the magnitude of
strongly decreases, yet slope does not change [23]. That suggests pressure
independent out-of-plane mechanism like in resonant tunneling in quasi-one-dimensional
organic conductors, proposed by Weger [24]. Above 500K the hopping is activated hence
the measured crossover in [23]. In general, the out-of-plane transport in layered
oxides is not settled, partly due to very different theoretical approaches [18-26].
CONCLUDING REMARKS
As can be seen from other contributions in this workshop, new superconducting
phases are still being discovered [29] so it is premature to propose very definite conclusions
in this field. Even more so as there is now some new evidence for the co-existence of
superconductivity and magnetism in oxides. This may provide further possibilities for
creative nanoengineering of new functional heterostructures. As there are many puzzling
437
results and open questions related to understanding of high-T
c
and related oxides, we still
need many more systematic experiments on very carefully prepared and characterised
samples of both, under- and over- doped, films and crystals. Moreover, to achieve mature
oxide nano-engineering technology we have to demonstrate all milestones discussed in the
introduction, above all the complete control of the local doping and a full mastery of all
length scales in the material. In short we have to be able to alter at will the electronic
dispersion relation, E(k), within oxide heterostructure. Such work is currently in progress
worldwide [4] and there are oxide electronics companies [8] that are developing promising
oxide heterostructures and new electronic devices.
Future History of Superconductivity: New Millennium Perspective
Figure 4. Future history of high-T
c
superconductivity: Note that superconducting materials above the
boiling temperature of water, colossal superconductors with are required for a genuinely useful
superconductor technology at ambient temperatures.
However, at a dawn of the new millennium, it is probably important to emphasize that,
whatever the progress in understanding of physics, we still need new, colossal
superconductors with T
c
450K, rather than 300K as generally assumed. T
c
450K is
needed as most devices usually operate at temperatures roughly 2/3 of T
c
[6]. For
commercial ambient technology we do require such colossal superconductors. As we
progress in quantum-nanoengineering of layered oxides and the controlled layer-by-layer
epitaxial growth of new phases we will eventually create new artificially layered
superconducting oxides (ALSO), and ultimately, in an analogy with manganites (that are
very similar to cuprates) - first the giant and ultimately the colossal
superconductors (see Figure 4). Although it may seem optimistic, let us note that are no
known fundamental physics limits to the appearance of superconductivity up to, at least,
Debye temperature range i.e. ~500K. Therefore, the opportunity and the means to realize
such remarkable materials and technology are within reach.
438
ACKNOWLEDGEMENTS
I gratefully acknowledge important contributions of my numerous friends and
colleagues, especially of all co-authors listed in references 27 and 28. I also acknowledge
the support by the EPFL and the Swiss National Science Foundation.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
J. G. Bednorz and K.A. Mller, Z Phys. B 64, 189 (1986).
B.E.A. Saleh and M.C. Teich, Fundamentals of Photonics , J. Wiley (1991).
J. Bok, G. Deutscher, D. Pavuna, S.A. Wolf (eds.) The Gap Symmetry and Fluctuations in High-T
c
Superconductors, Proc. NATO-ASI B371 Kluwer (1998).
D. Pavuna and I. Bozovic (eds.), Superconducting and Related Oxides : Physics and Nano-engineering
I, II, III, IV SPIE volumes 2058, 2697, 3481, 4058 SPIE, Bellingham (1994, 1996, 1998, 2000).
A. Bianconi (ed.), Stripes I & II, special issue of J. of Superconductivity (1997, 1999).
M. Cyrot and D. Pavuna, Introduction to Superconductivity and High-T
c
Materials, World Scientific,
London, Singapore, New Jersey (1992).
S. Barnes et. al. (eds.) High Temperature Superconductivity (eds..) CP483 American Institute of
Physics (1999).
Ivan Bozovic, Oxxel GmbH (Bremen), private communication, see www.oxxel.de for further details on
atomic layer-by-layer MBE growth and promissing devices.
See the Proceedings of the Klosters (April 1-10) Millenium Superconductivity Symposium, J. of
Superconductivity incl. Novel Magnetism vol. 13 (5) (2000).
J. C. Phillips, Phys. Rev. Lett. 59 1856 (1987) and the paper in this volume.
G. Deutscher, Nature 397, 410 (1999)
G. S. Boebinger et al , Phys. Rev. Lett. 77 5417 (1996).
J. Friedel, Physica C 153-155, 1610(1988).
B. Batlogg and C. Varma, Physics World 13 (2), 33 (2000) and refs therein.
I. Vobornik, Investigation of the Electronic Properties and Correlation Effects in the Cuprates and in
Related Transition Metal Oxides, D.Sci. Thesis, EPFL (1999).
D. Pavuna, I. Vobornik, G. Margaritondo, J. of Superconductivity 13 (5) 749 (2000).
J. Bok. and J. Bouvier in ref. 2 and ref. 9.
P.W. Anderson, Theory of High-T
c
Superconductivity in Cuprates, Princeton (1997).
D. Ariosa and H. Beck, Int. J. Mod. Phys. B 13 3472 (1999).
B. Normand, D.F. Agterberg, T.M. Rice, Phys. Rev. Lett. 82 (1999).
N.P. Ong, Science 273, 321 (1996) and references therein.
H. Berger, L. Forr and D. Pavuna, Europhysics Letters 41 (5), 531 (1998).
J.P. Salvetat et. al., unpublished (2000); L. Forr, Int. J. Mod. Phys. B 8, 829 (1994).
M. Weger, J. Phys. Colloq. C 6, 1456 (1978).
D. Pavuna, L. Forr, H. Berger, p. 412 in ref. 7.
A.P. Mackenzie et.al. Phys. Rev. Lett. 76 3786 (1996).
I. Vobornik, H. Berger, D. Pavuna, M. Onellion, G. Margaritondo, F. Rullier-Albenque, L. Forr, M.
Grioni, Phys Rev. Lett. 82, 3128 (1999).
M. Abrecht, T. Schmauder, D. Ariosa, O. Touzelet, S. Rast, M. Onellion and D. Pavuna, in print in
August issue of Surface Review and Letters (2000).
See articles by I. Felner et. al., and by J. Dow et. al. in this NATO-Kluwer volume.
439
This page intentionally left blank
DESIGNING PROTEIN STRUCTURES
HAO LI
1
, CHAO TANG
2
, and NED S. WINGREEN
2
1
Department of Biochemistry and Biophysics,
University of California, San Francisco,
513 Parnassus Avenue, Box 0448,
San Francisco, CA 94143-0444
2
NEC Research Institute, 4 Independence Way,
Princeton, NJ 08540
Recent work on lattice models of protein folding has identified a designability principle
for structures [1,2]. It is of course well known that some sequences make better behaved
proteins than others. The designability principle suggests that some structures make for better
behaved proteins as well. A goal of current research is to apply this principle to the design of
novel protein folds. Here we briefly discuss the designability principle and its application to
the design problem.
Essential characteristics of natural, water-soluble proteins are thermodynamic stability
of the ground state, stability against mutations of the amino-acid sequence, and relatively fast,
unassisted folding into the ground state. It is known that only a small fraction of all possible
polypeptide sequences have all these properties [3]. Natural protein sequences have clearly
been carefully selected. There is considerable evidence as well that natural protein structures
have been selected from among all possible folds of a polypeptide chain. There are estimates
that out of the huge number of possible folds, only 2000 are in use by living organisms [4].
Are some structures more apt to give protein-like behavior than others?
In our attempt to answer this question, we have focused on the concept of designabil-
ity. Specifically, the designability of a structure is the number of sequences which have that
structure as their unique ground state. Designability as a concept can therefore be applied
to real polypeptide chains as well as to lattice models of proteins. In lattice-model studies
we always find a small group of structures with designabilities much higher than the aver-
age. Furthermore, these highly designable structures are the ground states of the sequences
with the most protein-like properties, and the structures themselves have striking geometrical
regularities reminiscent of real proteins. It is this connection between the highly designable
structures and the most protein-like sequences which we refer to as the principle of des-
ignability.
Since it is very difficult to determine the designability of real polypeptide structures,
we have concentrated instead on simplified lattice models. In these models, a protein is a
Phase Transitions And Self-Organization in Electronic and Molecular Networks
Edited by J. C. Phillips and M. F. Thorpe, Kluwer Academic/Plenum Publishers, 2001 441
chain of monomers, restricted to fall on the sites of a 2D or 3D square lattice (see Fig. 1). A
structure is the path a chain takes through the sites of the lattice. For many of our studies,
we considered only the compact structures for chains of a given length. For example, the
compact structures of a chain of 27 monomers in three dimensions are all paths filling a
3x3x3 cube (Fig. l(a)). Instead of the twenty amino acids appearing in nature, our model
sequences consist of only two types of monomers: H for hydrophobic, and P for polar. A
sequence is therefore simply a string of Hs and Ps, e. g., HHPHHHPPPHPPPHH..., and
models of this type are called HP models [5].
Figure 1. Structures with the largest designability in 3D 3x3x3 (A) and 2D 6x6 (B) lattice models. The black
balls represent H monomers and the gray balls represent P monomers for particular HP sequences which fold
into the structures.
The thermodynamic and dynamic properties of a lattice-model protein depend on the
choice of an energy function. That is, an energy must be defined for each sequence in each
possible structure. We focus here on a theoretically convenient choice which reproduces the
behavior of more realistic energy functions. Specifically, the sites of each compact struc-
ture are divided into two types, surface and core. For example, in Fig. l(b), the innermost
square of 16 sites is defined to be the core. The energy of a sequence folded into a given
compact structure is taken to be -1 times the number of hydrophobic (H) monomers falling
on core sites. This definition allows a particularly convenient geometrical formulation of the
designability of structures.
442
Since a structure consists of only two types of sites, each structure can be represented as
a string of 1s and 0s, 1 for a core site and 0 for a surface site. Similarly, each sequence can
be represented as a string of 1s and 0s, 1 for a hydrophobic (H) monomer and 0 for a polar
(P) monomer. The energy of a sequence folded into a structure is therefore given by
This energy can be rewritten as
Since the last two terms are effectively constants is constant for a given sequence and
is constant for compact structures, which all have the same number of core sites - only the
first term determines which structure will be the ground state of a given sequence.
The important observation is that the first term in the energy, is a Hamming
distance in the space of strings of 1s and 0s. Each vertex in this string space corresponds to a
possible sequence, and a subset of the vertices correspond to the allowed compact structures.
The energy of a particular sequence folded into a particular structure is just the distance from
the sequence's vertex to the vertex for the structure. The designability of a structure, i. e.
the number of sequences with that structure as a unique ground state, is then just the number
of vertices which lie closer to that structure than to any other. A schematic depiction of this
geometrical construction of designability is shown in Fig. 2.
In the space of strings, some regions have a dense population of structures, while some
regions are very scarcely populated with structures. It is precisely in the scarcely populated
regions of string space that one finds the highly designable structures. It is natural that in
regions with very few structures, the volume lying closest to each structure will be large, and,
by the geometrical construction, the number of vertices in each volume is just the designabil-
ity of the structure.
Within this framework it is easy to see how designability is related to mutational and
thermodynamic stability. By definition a sequence which folds into a given structure is sta-
ble under mutation if the mutated sequence also folds uniquely into the same structure.
Since the designability of a structure is just the total number of associated sequences, it is ob-
vious that high designability corresponds to high mutational stability. The additional insight
provided by the geometrical construction is that since the sequences associated with a given
structure occupy a connected volume in the space of strings, they are generally all connected
by point mutations. The thermodynamic stability of sequences associated with highly des-
ignable structures follows from the small density of competing structures. For the sequences
associated with any one structure in string space, the nearby structures are low lying excited
states. Since the highly designable structures fall in regions with a small density of neighbors,
their associated sequences have a small density of low-lying excited states. This is exactly the
requirement for thermodynamic stability of the sequences. Fast folding, in turn, correlates
well with thermodynamic stability.
Finally, what produces the geometrical regularities observed in highly designable struc-
tures? A recent study [6] has demonstrated that for 6x6 structures, as in Fig. l(b), designabil-
ity is strongly correlated with mirror symmetry. A plausible explanation is that designable
structures are made from repeated designable substructures. Current work aims at pinning
down this connection.
The discussion so far has centered on lattice models of proteins. How could the principle
of designability be tested, and perhaps applied, in the context of real polypeptide chains?
a sequence specifically for this structure, and demonstrate that the sequence folds into the
structure in the laboratory.
Currently, attempts to design novel protein structures by other schemes have not been
successful. The designability principle helps clarify why design has proven so difficult. In
443
Figure 2. Schematic plot of string space. The shaded volume contain the sequences which lie closer to the
central structure than to any other, and hence have that structure as their unique ground state. Degenerate
structures have the same strings and hence cannot be the unique ground state of any sequence.
the lattice models, the vast majority of structures fall in densely occupied regions of structure
space. The sequences associated with these structures are always thermodynamically unsta-
ble because of their high density of low-lying excited states. While real polypeptide chains
cannot simply be mapped onto a space of strings, one can still argue that a space of structures
exists, with some regions densely occupied and some regions sparsely occupied. Previous
attempts to design novel structures have, unfortunately, focused on optimizing the sequence
for an arbitrary chosen target structure. But an arbitrarily chosen structure is generally unsta-
ble, even for the optimal sequence, because it is surrounded by a high density of competing
structures. The principle of designability makes it clear that one must first select a highly
designable structure as a target.
Current work is aimed at theoretically generating novel protein structures of high des-
ignability. There are several foreseeable difficulties. First, real polypeptide chains have a
large number of degrees of freedom. A typical globular protein consists of 100-200 amino
acid residues, and each residue has two free bond angles. Second, the energy function for real
polypeptide chains is not well known. Indeed, even the approximate functions used in sim-
ulations are very complicated, involving van der Waals interactions, Coulomb interactions,
hydrogen bonds, and hydrophobic interactions.
Nevertheless, we believe a balance can be achieved between fidelity to nature and com-
putational practicality. Specifically, the degrees of freedom of the chain can be greatly re-
stricted while still allowing realistic backbone configurations. Furthermore, the designability
444
of structures is in general not very sensitive to the choice of energy function. In the event,
we anticipate being able to generate short, i. e. 20-30 monomer, highly designable protein
segments in the near future. The generation of highly designable large structures remains
challenging, but will likely prove possible as well.
REFERENCES
1. Li, H., Helling, R., Tang, C., and Wingreen, N. (1996) Emergence of preferred structures in a simple
model of protein folding, Science 273, 666-669.
2. Li, H., Tang, C., and Wingreen, N.S. (1998) Are protein folds atypical?, Proc. Natl. Acad. Sci. USA 95,
4987-4990.
3. Davidson, A.R. and Sauer, R.T. (1994) Folded proteins occur frequently in libraries of random amino
acid sequences, Proc. Natl. Acad. Sci. USA 91, 2146-2150.
4. Chothia, C. (1992) Proteins. One thousand families for the molecular biologist, Nature 357, 543-544;
Govindarajan, S., Recabarren, R., and Goldstein, R.A. (1999) Estimating the total number of protein
folds, Proteins 35, 408-414.
5. Dill, K.A. (1985) Theory for the folding and stability of globular proteins, Biochemistry 24, 1501-1509;
Lau, K.F. and Dill, K.A. (1989) A lattice statistical mechanics model of the conformational and sequence
spaces of proteins, Macromolecules 22, 3986-3997.
6. Wang, T., Miller, J., Wingreen, N.S., Tang, C., Dill, K.A., unpublished.
445
This page intentionally left blank
PARTICIPANTS
J. Acrivos
Department of Chemistry
San Jose State University
San Jose, CA 95192-0101
USA
jacrivos@email.sjsu.edu
J. Banavar
Department of Physics
Pennsylvania State University
University Park, PA 16802
USA
jayanth@phys.psu.edu
A. Bianconi
Unit INFM
Dipartimento di Fisica
Universit di Roma La Sapienza
00185 Roma
ITALY
antonio.bianconi@roma1.infn.it
Punit Boolchand
University of Cincinnati
Electrical & Computer Engineering Dept.
PO Box 210030
Cincinnati, OH 45221-0030
USA
pboolcha@ececs.uc.edu
T.G. Castner
Department of Physics
University of Massachusetts, Lowell
1 University Ave.
Lowell, MA 01854
USA
hayden-tg@mercury.mv.net
A.J. Coleman
Dept. of Mathematics and Statistics
Queens University
Kingston, ON K7L 3N6
CANADA
colemana@post.queensu.ca
V. Dallacasa
Dept. of Science and Technology (Mat.
Analysis Lab.)
University of Verona
Strada Le Grazie
I-37134 Verona
ITALY
dallacasa@beta.sci.univr.it
Martin Dove
Department of Earth Sciences
Cambridge University
Downing Street
Cambridge CB2 3EQ
ENGLAND
martin@esc.cam.ac.uk
John Dow
6031 East Cholla Lane
Scottsdale, AZ 85253
USA
cats@dancris.com
J. Dyre
IMFUFA
Roskilde University
POB 260
DK-4000 Roskilde
DENMARK
dyre@ruc.dk
447
M. Dzugutov
Department of Numerical Analysis
Royal Institute of Technology
100 44 Stockholm
SWEDEN
mik@pdc.kth.se
Hellmut Eckert
Institut fr Physikalische Chemie
Westflische Wilhelms Universitt
Mnster
Schlossplatz 7
D-48142 Munster
GERMANY
eckerth@uni-muenster.de
A.L. Efros
Department of Physics
University of Utah
201 JFB
Salt Lake City, UT 84112
USA
efros@top.physics.utah.edu
1. Felner
Racah Institute Physics
The Hebrew University
Jerusalem 91904
ISRAEL
israela@vms.huji.ac.il
D. Georgiev
Dept. of ECECS
University of Cincinnati
Cincinnati, OH 45221-0030
USA
dgeorgie@ececs.uc.edu
G.N. Greaves
Department of Physics
The University of Wales
Aberystwyth
Ceredigion SY23 3BZ
UNITED KINGDOM
gng@aber.ac.uk
J. Haase
2. Physikalisches Institut
Universitt Stuttgart
Pfaffenwaldring 57
70569 Stuttgart
GERMANY
j.haase@physik.uni-stuttgart.de
D. Haskel
Argonne National Laboratory
Exp. Facilities Div., Adv. Photon Source
APS, Bldg. 401
Argonne, IL 60439
USA
haskel@aps.anl.gov
Jan A. Jung
Department of Physics
University of Alberta
Edmonton, AB T6G 2J1
CANADA
jung@phys.ualberta.ca
Richard Kerner
Lab. de Gravitation et Cosmologie
Relativistes
Univ. P. & M. Curie BC 142.
4 P1. Jussieu
75252 Paris 05
FRANCE
rk@ccr.jussieu.fr
F.V. Kusmartsev
School of MAP
Loughborough University
Loughborough, Leics LE11 3TU
UNITED KINGDOM
f.kusmartsev@lboro.ac.uk
G. Lucovsky
Department of Physics
North Carolina State University
Raleigh, NC 27695-8202
USA
gerry_lucovsky@ncsu.edu
M. Micoulaut
Lab. Physique Theor. Liquides
Univ. P. & M. Curie, Boite 121
4 Pl. Jussieu
75252 Paris Cedex 05
FRANCE
mmi@ccr .jussieu.fr
G.G. Naumis
Instituto de Fisica
UNAM
A.P. 20-364
0-1000 Mexico D.F.
MEXICO
naumis@fenix.ifisicacu.unam.mx
448
D. Pavuna
Department of Physics - IPA
Ecole Polytechnique Federale de
Lausanne
CH 1015 Lausanne
SWITZERLAND
pavuna@epfl.ch
J.C. Phillips
Lucent Technologies
Bell Labs Innovations
600 Mountain Ave.
Murray Hill, NJ 07974-0636
USA
jcphillips@bell-labs.com
J.-L. Pichard
Dept. LEtat Condense, CEA
Cent. Etud. Saclay
F-91191 Gif-s.-Yvette
FRANCE
pichard@drecam.saclay.cea.fr
E.A. Stern
Department of Physics
University of Washington,
Box 351560
Seattle, WA 98195
USA
stern@ofisa.phys.washington.edu
M.F. Thorpe
Physics and Astronomy Dept.
Michigan State University
East Lansing, MI 48824
USA
thorpe@pa.msu.edu
John Wagner
Physics Department
University of North Dakota
Grand Forks, ND 58201
USA
john_wagner@und.nodak.edu
Y. Wang
Department of Physics
Osaka University
1-1 Machikaneyama, Toyonaka
Osaka 560-0043
JAPAN
wang@phys.sci.osaka-u.ac.jp
M. Watanabe
Nanostructure Physics, KTH
Lindstedtsvagen 24
SE-100 44 Stockholm
SWEDEN
michio@nanophys.kth.se
N. Wingreen
NEC Research Institute
4 Independence Way
Princeton, NJ 08540
USA
wingreen@research.nj .nec.com
R. Zallen
Department of Physics
Virginia Tech
Blacksburg, VA 24061-0435
USA
rzallen@vt.edu
449
This page intentionally left blank
INDEX
Activation energy, 102, 106
Adiabatic approximation, 391
Agglomeration model, 172
Agglomeration theory, 144, 157, 171
Alternative gate dielectrics, 204
Amino acids, 442
Amorphisation, 226, 233
Amorphous metals, 38
Amorphous solid, 161
Anderson localization, 265
Anderson transition, 263
Anharmonic, 105
Antiferromagnetic correlations, 397
Antiferromagnetic spin fluctuations, 413
Applied stress backbone, 57
Atomic scaffolding, 225
Average cluster size, 178
Average coordination, 123
Avrami relation, 232
Berlinite, 237
Binary glass, 65
Binodal points, 179
Boltzmann factors, 174
Bond bending, 47, 70
Bond bending networks, 50
Bond ionicity, 190, 200
Bond stretching, 47
Bragg diffraction, 209
Bragg peaks, 209
Broken symmetry, 3
Busbar geometry, 56
Calorimetry, 72, 143
Central Force Network, 46, 49
Ceramics, 232
Chalcogenide glasses, 6566, 123, 161, 172
Channel openings, 226
Charge domains, 362
Charge transfer, 332
Chemical ordering, 135
Chemical phase separation, 197
Chemical threshold, 124
Clathrasil, 238
Clausius-Mossotti equation, 265
Collapse, 236
Colossal magnetoresistivity, 209
Compressibility, 235
Connectivity, 60, 65, 181
Conductivity threshold, 40
Connectivity percolation, 1, 44
Constraint theory, 162
Constraints, 4, 46
Continuous random networks, 43, 192, 206
Continuous transition, 2
Correlated electrons, 247
Correlation length, 291
Correlation matrix, 23, 27
Corrugated, 379
Coulomb gap, 247, 250, 264, 366
Coulomb glass, 248
Coulomb interactions, 264, 267, 444
Covalent bonded atoms, 162
Covalent glasses, 43
Covalent network, 163
Critical exponents, 292
Critical point, 286
Critical temperature, 376, 414
Critical volume fraction, 38
Crossover from strong to weak disorder, 248
Crystalline silicon, 190
Crystallization, 143
Cuprate perovskites, 375
Debye behavior, 102
Debye relaxation, 102
Defects, 332
Degenerate structures, 444
Density matrices, 23
Devices, 432
Dielectric constant enhancement, 200
Dielectrics, 190
Diffuse scattering, 213
Diffusion coefficient, 117
451
Disordered systems, 1
Displacive phase transformation, 312
Displacive structural transformation, 313
Distortions, 332
Dopants, 323
Doped antiferromagnet, 390, 398
Doped semiconductors, 291
Dynamical ergodicity, 112
Edwards-Anderson order parameter, 249
Effective medium approximation, 1
Einstein relations, 277, 282
Elastic energy, 106
Elastic properties, 55
Electron gas, 383
Electron glass, 248
Electron strings, 389
Electronegativity, 190
Electron-electron interaction, 247, 291
Electronic properties, 433
Electron-phonon system, 389
Elementary building blocks, 177
Energy barriers, 116
Enthalpic rigidity, 66
Epitaxial growth, 433
Ergodic di ffusi on, 114
Ergodicity breaking, 119
Exotic superconducting phase, 375
Expansion coefficient, 235
Fermats Last Theorem, 12
Fermi energy, 319
Fermi liquid, 2
Fermi surface, 12, 32, 382
Ferroelastic nanodomains, 311, 316, 321
Ferroelasticity, 311312
Ferroelectricity, 209
Filamentary metals, 10
Filamentary, 1
Finite-temperature scaling, 293
First-order transition, 2
Fixed positive charge, 199
Floppy, 85
Floppy modes, 44, 89, 161162, 168, 242
Floppy regions, 88
Floppy units, 88
Gate dielectric materials, 205
Generic networks, 45
Geometrical regularities, 441
Germanium, 291
Glass preparation, 86
Glass transition, 71, 101, 254
Glass transition temperature, 65, 123, 143, 175
Glassy state, 111
Global chaotic connectivity, 113
Granular metals, 367
Hall number, 272
Hartree-exchange, 281
Heteroepitaxy, 431
High-temperature superconductivity, 403, 413
Hopping conductivity, 364, 366
Hydrogen, 350
Hydrogen bonds, 444
Hydrophobic interactions, 444
Hyperfine coupling, 414
Ice I, 237
Impurities, 433
Impurity bands, 1
Impurity dielectric susceptibility, 292
Inelastic neutron scattering, 362
Inelastic scattering cross-section, 212
Inhomogeneity, 292
Insulator-metal transition, 247, 323
Integral operator, 23
Interface properties, 189
Interface, 200
Interfacial fixed charge, 200
Interfacial limitations, 204
Intermediate phase, 1, 5, 43, 50, 52, 54, 65, 74
Ion channeling, 313
Ionic solids, 398
Irreversibility, 238
Isothermal magnetization, 347
Isotopes, 293
Jahn-Teller, 323, 392
Josephson junctions, 432
Kauzmann paradox, 111
Knight shifts, 414
Lagrangian, 44
Lagrangian bonding constraints, 66
Landau-Ginzburg theories, 2
Lattice instabilities, 33
Lattice models, 442
Layered oxides, 431
Liquid-glass transition, 111, 143
Local bonding, 189
Local distortion, 328
Local structure, 209210
Localization, 358, 390
Localization length, 248, 291
Localized states, 11
London penetration length, 375
Loopless networks, 58
Low temperatures, 295
Luminescent ions, 135
Magic angle spinning, 124
Magnetic behavior, 350
Magnetic moments, 390
Magnetic vortices, 18
Manganese perovskites, 321
Manganites, 311
Maximal homogeneity, 173
Mean coordination, 56
452
Mean-field, 66
Mean-field constraint theory, 85
Mechanical rigidity, 242
Medium range order, 127, 157
Meissner state, 353
Melting, 233, 235
Metal-insulator transition, 263, 291
Metallic heterogeneous phase, 377
Minimal local fluctuations, 174
Mode-coupling theory, 112
Modulus, 313
Molecular dynamics, 117
Molecular orbitals, 377
Mssbauer spectra, 138
Mssbauer spectroscopy, 344
Mott law, 248
Muon spin polarization, 409
Muon spin rotation relaxation, 409
Nanodomains, 329
Nanofilaments, 332
Network, 65
Network connectivity, 71, 73, 143
Network constraints, 193
Network glasses, 1, 4, 43, 146
Network stiffening, 123
Network stress, 73
Neutron diffraction experiments, 315
Neutron scattering, 321, 390, 426
Neutron-transmutation doping, 292
Non-Arrhenius behavior, 104
Non-crystalline solids, 65
Non-Debye relaxations, 106
Non-ergodic diffusion, 115
Non-ergodic dynamics, 111
Nuclear magnetic resonance, 124, 414
Nucleation, 232
Orbital shift, 414
Oxygen, 344
Oxygen ordering, 332
Oxygen vacancies, 433
Pair-breakers, 403
Pairing mechanism, 413
Pauling electronegativity, 190
Percolating charged strings, 390
Percolation, 39, 358, 390, 398
Percolation processes, 39
Percolation theory, 123
Percolation threshold, 399
Percolation topology, 40
Perovskite, 311
Perovskite oxides, 311
Phase diagram, 293, 380,
Phase separation, 33
Phase stability, 189
Phase transitions, 19, 235
Phonons, 398
Photomelting, 79
Polaronic charge density wave, 384
Polarons, 389
Polymeric networks, 65, 135
Polypeptide chains, 441
Porosity, 225
Protein structures, 441
Pseudogap, 431
Pseudoground states, 254
Quadrupole interaction, 414
Quantum computers, 1, 3
Quantum critical point, 385
Quantum mesoscopic stripes, 384
Quantum percolation, 12, 357
Quantum phase transition, 291
Quantum wires, 375
Quartz, 237
Quasi-particles, 319
Raman scattering, 75, 8586
Raman shifts, 314
Raman spectroscopy, 156
Random close packed ionic structures, 192
Random networks, 66, 174
Rayleigh-Ritz, 25
Re-assembly, 243
Reduced density matrix, 23
Relaxation processes, 111
Relaxation times, 103
Reversibility, 99, 238
Rigidity, 85, 168, 181, 225
Rigid units, 88
Rigidity percolation, 44, 51
Rigidity threshold, 67, 243
Rigidity transition, 66, 71, 163, 171
Ring distribution, 181
Scaling exponent, 268
Scaling relationship, 115
Scaling theory, 8
Scattering measurements, 219
Screening, 247
Second quantization, 24
Self-organization, 43, 51, 157, 375, 436
Self-organize, 65
Self-organized networks, 55, 71
Self-trapped excitons, 389
Self-trapped state, 389
Shear modulus, 233
Short-range order, 125
Silicate alloys, 197
Silicate glasses, 172
Slow relaxation, 254
Solidity, 101
Solidity length, 104
Specific heat, 9
Spectral diffusion, 255
Spin correlations, 363
Spin pseudogap, 18
Spin shifts, 414
453
Spontaneous vortex phase, 353
Stability, 62
Stiffness transition, 85, 90
Stressed rigid phase, 76
Stress-free, 65
Stress-free oxide network, 66
Strings in manganites, 400
Stripe formation, 381
Stripes, 414
Structural relaxation, 87
Superconducting gap, 432
Superconducting networks, 61
Superconducting phase, 382
Superconductive transition temperatures, 2
Superconductivity, 1, 33, 209, 331, 341, 362, 369, 375,
403, 408, 413, 436
Supercooled liquids, 111
Superstripes, 375
Swiss cheese, 40
Target structure, 444
Ternary systems, 135
The Pebble Game, 48
Thermal relaxation, 85
Thermodynamic entropy, 115
Thermodynamic fluctuations, 254
Thermodynamic stability, 443
Threshold, 389
Time-reversal symmetry, 292
Transition metal silicates and aluminates, 190
Transition region, 85
Transition temperature, 161
Transport properties, 14, 263, 359
Twins, 433
Two-dimensional lattices, 38
Universality class, 292
Vacancy-ordering, 362
van der Waals interactions, 444
Variable range hopping, 248, 253, 292
Vibrational properties, 88
Viscosity, 233
Viscous liquids, 101, 103
Weak ferromagnetism, 341
Wigner localization, 382
Window glass, 171
Wrong bond, 99
X-ray absorption fine structure, 216
X-ray diffraction, 243, 333
Youngs modulus, 7, 65
Yukawa potential, 265
Zeeman splitting, 414
Zeolite collapse, 242
Zeolites, 225
Zero-frequency modes, 48
454