Highly Conducting One-Dimensional Solids PDF

Highly Conducting
One-Dimensional Solids
PHYSICS OF SOLIDS AND LIQUIDS
Editorial Board: Jozef T. Devreese. University of Antwerp, Belgium

Roger P. Evrard • University of Liege, Belgium
Stig Lundqvist • Chalmers University of Technology, Sweden
Gerald D. Mahan. Indiana University, Bloomington, Indiana
Nonnan H. March. University of Oxford, England
SUPERIONIC CONDUCTORS
Edited by Gerald D. Mahan and Walter L. Roth
HIGHLY CONDUCTING ONE-DIMENSIONAL SOliDS

Edited by Iozef T. Devreese, Roger P. Evrard, and Victor E. van Doren
ELECTRON SPECTROSCOPY OF CRYSTALS

V. V. Nemoshkalenko and V. G. Aleshin
Highly Conducting
Edited by
Joze! T Devreese
State University of Antwerp (RUCA and UIA)
Antwerp, Belgium
Roger P. Evrard
University of Liege
Liege, Belgium
and
Victor E. van Doren
State University of Antwerp (RUCA)
Antwerp, Belgium
PLENUM PRESS • NEW YORK AND LONDON

Library of Congress Cataloging in Publication Data
Main entry under title:

Highly conducting one-dimensional solids.
(Physics of solids and liquids)
Includes bibliographical references and index.
1. One-dimensional conductors. I. Devreese, Jozef T. II. Evrard, Roger P.
III. van Doren, Victor E.
QCI76.8.E4H53 537.6'2 78-11396
ISBN-13: 978-1-4613-2897-1 e-ISBN-13: 978-1-4613-2895-7
001: 10.1007/978-1-4613-2895-7
© 1979 Plenum Press, New York

Softcover reprint ofthe hardcover 1st edition 1979
A Division of Plenum Publishing Corporation
227 West 17th Street, New York, N.Y. 10011
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted,

in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
Contributors
John Bardeen, Department of Physics, University of Illinois at Urbana-

Champaign, Urbana, Illinois
A. 1. Berlinsky, Department of Physics, University of British Columbia,
Vancouver, British Columbia
R. Comes, Laboratoire de Physique des Solides, Associe au Centre
National de la Recherche Scientifique, Universite Paris-Sud, Orsay,
France
R. A. Craven, I.B.M. Watson Research Center, Yorktown Heights, New
York; present address: Monsanto Company, St. Louis, Missouri
V. J. Emery, Department of Physics, Brookhaven National Laboratory,
Upton, New York
H. Gutfreund, Racah Institute of Physics, Hebrew University, Jerusalem,
Israel
A. J. Heeger, Department of Physics and Laboratory for Research on the
Structure of Matter, University of Pennsylvania, Philadelphia,
Pennsylvania
W. A. Little, Department of Physics, Stanford University, Stanford, Cali-
fornia
T. D. Schultz, I.B.M. Watson Research Center, Yorktown Heights, New
York
L. 1. Sham, Department of Physics, University of California at San Diego,
La Jolla, California
G. Shirane, Brookhaven National Laboratory, Upton, New York
Preface
Although the problem of a metal in one dimension has long been known to
solid-state physicists, it was not until the synthesis of real one-dimensional
or quasi-one-dimensional systems that this subject began to attract
considerable attention. This has been due in part to the search for high-
temperature superconductivity and the possibility of reaching this goal with
quasi-one-dimensional substances. A period of intense activity began in
1973 with the report of a measurement of an apparently divergent conduc-
tivity peak in TfF-TCNQ. Since then a great deal has been learned about
quasi-one-dimensional conductors. The emphasis now has shifted from
trying to find materials of very high conductivity to the many interesting
problems of physics and chemistry involved. But many questions remain
open and are still under active investigation.
This book gives a review of the experimental as well as theoretical
progress made in this field over the last years. All the chapters have been
written by scientists who have established themselves as experts in theoreti-
cal and experimental solid-state physics. The book is intended to be of use
both to students and researchers entering the field as well as to more
advanced physicists. The wealth of ideas and information it contains ought
to be useful to anyone interested in quasi-one-dimensional systems, organic
solids, or the search for novel conduction and superconduction mechanisms.
The editors are very grateful to the authors for their collaboration in
this book. In particular, the editors want to thank Professor Bardeen for his
suggestions and ideas from the very beginning of this enterprise.
The editors would like to thank Miss H. Evans and Mr. M. De Moor for
their technical assistance.
faze! T Devreese
Roger P. Evrard
Victor E. van Doren
VII
Contents
Chapter 1
IntroductIOn to HIghly ConductIng One-DImensIOnal SolIds
A J Berltnsky
IntroductIOn
2 Some Preliminary Thought,
3 ExcItonIC SuperconductiVIty 3
4 TCNQ Salts and KCP 4
4 I NMP-TCNQ 5
42 ITF-TCNO 5
43 KCP 6
5 ITF-TCNO and TSeF-TCNO 7
5 1 Structural Transilions In TTF- TCNO 7
52 Electromagnetic Properties of TTF-TCNO 8
51 FSR and Alloy, of ITF TCNO and TSeF-TCNO 9
6 Theory 10
7 Some Concluding Thought, 12
References 13
Chapter 2
X-Ray and Neutron Scattenng from One-DImensIOnal Conductors
R Comes and G Shlrane
IntroductIOn 17
1 1 Lattice Instabilltle, dnd Phonon Anomalies 17
1 2 X-Ray Diffuse Scattenng 22
1 1 Neutron Scattenng 27
2 Structural Studies of KCP and Related Platinum Chain Complexes 28
2 1 Structure and One Dlmen"onal Electncal Properlies of KCP 29
2 2 X-Ray Diffuse Scattenng from KCP 31
2 3 Neutron Scattenng Studies of KCP 33
2 4 Study of Other Platinum Complexes 42
3 Structural Studies of Organic One-DimensIOnal Conductors 43
3 1 Structure and ITF TCNO Crystals 44
3 2 High-Temperature Precuf';or Scattenng In ITF- TCNO 46
31 The Modulated Phd,e, of ITF-TCNO 54
IX
x Contents
3 4 SPin Waves In ITF-TCNQ? 58

35 The Interpretation of the Sequence of Modulated Phases In ITF-TCNQ 59
36 Study of Other Orgamc One-DimensIOnal Conductors 61
4 Condudlng Remarks 62
References and Notes 64
Chapter 3
Charge-Density Wave Phenomena In One-Dimensional Metals:
ITF-TCNQ and Related Orgamc Conductors
A J Heeger
1 IntroductIOn 69
2 Strength of InteractIOns, BandWidth, Electron-Electron and Electron-Phonon
InteractIOns 77
2 lOne-Electron Energies, Band Structure 77
2 2 Electron-Electron Interactions Nuclear Magnetic Resonance and Magnetic
SusceptibilIty 80
2 3 Electron-Phonon InteractIOn 87
3 The Peleris InstabilIty In ITF-TCNQ Structural Aspects and Phonon Softening 93
4 The Pseudo gap Optical Properties 97
5 ElectTlcal ConductiVity 105
5 1 DC Measurement, 105
52 Microwave Medsurements 120
6 The TranSItion Region 38°K < T < 54°K 122
7 The Pinned Regime at Low Temperatures 127
8 NonlInear Transport In ITF TCNQ at Low Temperatures 131
9 ConclUSIOn 138
References and NO{fI 140
Chapter 4
The Orgamc Metals (TSeF)ATTF)l-x-TCNQ-A SystematIc Study
T D Schultz and R A Craven
1 In trod uctlOn 147
2 PreparatIOn CharactenzatlOn and Lattice Structure 149
3 Phase Diagram 154
4 MetallIc Phase T Tc J
> 156
4 1 Transport Properties 156
4 2 MagnetIC Propertle'i 166
4 3 Phonon AnomalIe, 176
5 Metal-Semiconductor Phase TranSitIOn 183
~ 1 VanatlOn of TranSItIon Temperature Tel WIth AllOYing 184
5 2 ThermodynamIC, and Cntlcal BehaVIOr of the Metal-SemIconductor Phase
TranSItIon 189
6 Semiconductlllg Phase T < Tc I 196
6 1 Transport PropertIes III the SemIconductIng Phase 197
6 2 MagnetK PropertIe, 202
(, 3 Super/attIce, dnd Phonon Anomahe, 206
Contents Xl
7 Summary 217
Chapter 5
Perturbation Approach to LattIce Instablhtles In
Quasl-One-DImensIOnal Conductors
L J Sham
Introduction 227
2 The One DimensIOnal Electron-Phonon System 228
3 Fluctualtons In the One Dimensional System 232
4 Effects of InterchaIn CouplIng 236
5 Effects of ImpurIlte<; 239
Chapter 6
Theory of the One-DImensIOnal Electron Gas
V J Emery
BasIc PhYSICS 247
1 1 IntroductIon 247
1 2 Phase Transitions and Long Range Order 248
1 3 Mathematical Model 250
1 4 Strong CouplIng 252
2 SPIn less Fermion, 255
2 1 Defintlton of the Conltnuum Limit 255
2 2 Boson Representations and the Free Energy 258
2 3 Boson Representattons of Fermion Fields 263
24 CorrelatIOn Functions of the Interacting System 267
3 Large On Site Interaction 271
3 1 Attractive Interactl(ln 271
3 2 RepulSive Interact"ln 274
3 3 Correlation Function, 274
4 ContInuum Limit-Energy Gaps 276
4 1 Separation of Charge and SPIn Degrees of Freedom 277
4 2 Reduction to SpInkss Fermlons 278
4 3 Solution of the <;pInless Fermion Problems 280
44 CorrelatIOn Funcltons 283
4 5 Relattonshlp to Other Problems 286
5 RenormalIzalton Group Method 289
5 1 ScalIng Equatl!ln' 290
5 2 TrajectorIes and Energy Scales 291
5 1 Low Temperature Properttes 294
5 4 Four Parltcle Functions 297
AppendiX A Some Results That Are Useful for WorkIng With Boson
Representations 299
AppendiX B Antlcommutatlon of Different Fermion Fields 300
AppendiX C Charge DenSity Wave Gap and CDW CorrelatIOns 301
References 302
Xll Contents
Chapter 7
The Prospects of EXCltomc SuperconductIvity
H Gutfreund and W A Little
1 IntroductIOn 305
2 The Nature of SuperconductIVIty 309
2 1 Background 309
2 2 Phonon Mechanism 310
2 3 LImItatIOn on T, 312
2 4 Isotope Effect 315
2 5 ExcIton Mechanism 317
3 Problems of SuperconductIVIty Unique to the ExcIton Mechanism 319
3 1 Exchange 320
1 2 Apparent LImitatIOn on A - JJ- * 320
1 1 Vertex Correltlon, 322
1 4 EquatIOn for T, 323
1 5 The Kernel Ulp k) 326
3 6 Effects of the Phonons on the EXCIton Mechanism 328
4 Effects of LImIted DImensIonalIty 331
4 1 Effects of FluctuatIons 333
4:2 Types of Order In a One-DImensIonal Electron Gas 338
4 3 Relevance of . g-ology 344
4 4 Intercham ( ouplmg 345
4 5 LocalIzatIOn and ImpUrItIes 349
4 6 Effects of ScreenIng 352
5 Real Models 356
5 1 Model of a FIlamentary Excltonlc Superconductor 357
5 2 DISCUSSIon 365
6 Summary 366
References 367
Chapter 8
Recent Developments and Comments
John Bardeen
1 IntroductIOn 373
2 Nature of the Phase TransItIOns 378
2 1 TranSItIOn m KCP 379
2 2 TranSItIons m TCNQ Compounds 380
2 3 TranSItIons In TaS" NbSe, 382
3 Coulomb InteractIons and Magnetic SusceptibIlIty 382
1 1 LImItatIons of the Hubbard Model 382
1 2 MagnetIc SusceptIbIlIty 384
11 Problem of TTF-TCNQ 385
3 4 Propertle, of NMP TCNQ 386
4 CollectIve Transport 387
4 1 SuperconductIvIty dnd One-DImensIOnal FluctuatIOns 388
4:2 DIelectrIC PropertIes 390
Contents xiii
4.3. Transport above Tc 392

4.4. Nonlinear Field Dependence of Conductivity 398
5. Concluding Comments 401
References 402
Author Index 405

Subject Index 413
1
Introduction to Highly Conducting
A. 1. Berlinsky
1 IntroductIOn
In the past several years a number of materIals, such as

K2Pt(CN)4XO 3 x H 2 0, X = Cl, Br (KCP), the tetracyanoqumodlmethane
(TCNQ) salts, and the sulfur-nttrogen polymer [(SN)x]' have been dis-
covered whose microscopIc structure consists, to a greater or lesser extent,
of a collection of weakly coupled, highly conducttng chatns Such materIals
are called quasl-one-dlmenslOnal metals and they are of tnterest because
nearly one-dimensional metals are expected to exhibit unusual collective
behaVIOr <;uch a<; the Peleris instabilIty and varIOUS ktnds of superconduc-
tivity The purpose of this chapter IS to provide an entry po tnt tnto the
subject of quasl-one-dlmenslOnal conductors, by defintng the terms,
tntroductng the most Important arguments, and ralstng some of the ques-
tIOns that Will be discussed 10 the followtng chapters
2 Some Prellmmary Thoughts
The problem of a metal 10 one dimensIOn has long been known to

theoretical sohd-state physIcists as pathological for reasons first discussed
(I)
by Pelerls Beyond this fact however, there was httle tnterest 10 the
problem for qUite some time because It did not seem to relate to any real
A ] Berlmsky • Department of PhYSICS University of Bntlsh Columbia Vancouver B C

V6T lW5
1
2 A. 1. Berlinsky
physical system. All this has changed in the past few years with the
synthesis of TCNQ saits,(2) the planar platinum compounds such as KCP,(3)
and most recently of the metal trichalcogenides such as NbSe3.(4) In addi-
tion, the suggestion that quasi-one-dimensional systems might favor exci-
tonic superconductivity(5) has stimulated several important theoretical
developments. :j:
However, the immediate stimulus for much of the work in this volume
was not a new material or a new theory but rather a measurement that was
first reported at an American Physical Society meeting in 1973 by Alan
Heeger and later published under the title "Superconducting Fluctuations
and the PeierIs Instability in an Organic Solid.,,(7) In this work, Heeger and
his co-workers at the University of Pennsylvania reported the observation
of an apparently divergent conductivity peak in tetrathiafulvalene-(TTF-)
TCNQ at 60°K, below which the conductivity fell exponentially, with
decreasing temperature, to zero.
Subsequently the field has developed rapidly in several directions.
Detailed studies of the conductivity peak in TTF-TCNQ and related
materials have shown that although this peak can be quite large, it certainly
remains finite(8.9) (see Figure 1). Extensive experimental work, employing
virtually every probe available to solid-state physicists, has provided much
useful information about the microscopic behavior of these substances,
although certain interesting bits in the puzzle remain elusive.
On the theoretical side, considerable progress has been made in the
development of semiphenomenological models, similar to the Landau-
Ginzberg theory of superconductivity, to describe the physical behavior of
real quasi-one-dimensional conductors. In addition, there have been
significant advances in the theory of one-dimensional systems with the help
of modern renormalization group techniques and new exact solutions to
model problems.
The prehistory, mentioned above, of the subject of one-dimensional
conductors dates to the early 1950s and the work of PeierIs(1) and Froh-
IichYO) The conclusions of Peierls are coveniently summarized in the
chapter on "Cohesive Forces in Metals" in his book Quantum Theory of
Solids.(!) There he argues that a one-dimensional metal is inherently
unstable against a distortion of the lattice whose wavelength is such that
scattering of electrons by the distortion will cause a gap to appear at the
Fermi energy. This happens when the wave vector of the distortion is 2k F ,
where kF is the Fermi wave vector. The energetics of this instability are
discussed in detail in Chapter 5 of this volume. "It is therefore likely,"
PeierIs concluded, "that a one-dimensional model could never have metal-
t Work on one-dimensional conductors has been the subject of several published conference
proceedings. Some of these are given in Reference 6.
Introduction to Highly Conducting One-Dimensional Solids 3
...
·._TTF-TCNQ
...
.........
Figure 1. DC conductivity of a typical
TIF-TCNQ crystal (from Reference 11) NMP-TCNQ
.. ... ...
and of a high-purity NMP-TCNQ crys-
100 200 300
tal (from Reference 9). T IK)
lic properties." However, he immediately proceeded to qualify this remark

by noting its dependence on " ... the adiabatic approximation [which] is
not valid in the case of a metal, so that the energies calculated with the
nuclei at rest cannot be used in physical arguments without great care."
This caveat was particularly appropriate in view of the work of Froh-
lich(IO) on the possibility of superconductivity in a one-dimensional
"lellium" model. Here, as in the case of a crystal, the positive ions, which
are represented by an elastic continuum, distort periodically with wave
vector 2 kF' and the electronic spectrum develops a gap at the Fermi
energy. However, in this continuum model, it is clear that there is no
preferred spatial location for the distortion, and hence the phase of the
distortion is free to move through the solid. Since the motion of the
electrons and the distortion are coupled, this collective motion, which is
now called a sliding charge-density wave (CDW), can carry a current. For
small velocities, Frohlich showed that this current is essentially undamped.
3. Excitonic Superconductivity
Very little progress was made on the subject of one-dimensional

conductors for the rest of the 1950s and the early 1960s, probably because
no real materials were available for study. In 1964, at about the time when
4 A. 1. Berlinsky
organic semiconductors were beginning to appear, Little(5) proposed a

model for a high-temperature organic superconductor which involved
conduction along linear chains surrounded by polarizable dye molecules.
Little's suggestion follows from the Bardeen-Cooper-Schrieffer (BCS)
formula for the transition temperature of a superconductor
Tc = 1.148D e- 1/ Ap
where 8D is the Debye temperature (k8 D is a typical phonon energy), and

Ap is the dimensionless electron-phonon coupling constant. If the attrac-
tive interaction between electrons on a linear chain were instead mediated
by the exchange of excitons along polarizable side chains, then Tc would
become
where Ee and Ae are, respectively, the exciton energy and the exciton-
electron coupling constant. Little argued that the coupling constant for the
phonon and exciton mechanisms would be comparable and that the much
larger exciton energy would lead to transition temperatures well above
room temperature.
Details of this theory of excitonic superconductivity in one-dimen-
sional materials are presented in Chapter 7. For the purpose of this
introduction, however, two facts are worth emphasizing. First, no one has
yet been successful in producing an excitonic superconductor, high Tc or
otherwise. Second, much of the recent progress in one-dimensional
conductors, both organic and inorganic, was stimulated, directly or
indirectly, by Little's proposal. Even TTF-TCNQ, with its highly conduc-
ting TCNQ stacks and polarizable TTF side chains, bears a strong resem-
blance to the Little model.
4. TCNQ Salts and KCP
An important breakthrough in the manufacture of highly conducting

one-dimensional solids was made at Dupont in 1962 with the synthesis of
the organic molecule TCNQ and a large class of charge transfer salts
involving TCNQ.(2) The most highly conducting of these salts, quino-
linium-TCNQ, was at that time the best organic conductor known. These
materials are one-dimensional or more precisely "quasi-one-dimensional,"
because the TCNQ molecules, which are large and flat, form stacks in the
solid, much like a stack of poker chips. Electrical conduction tends to be
high along these stacks and rather small between neighboring stacks. More
information on the structure of TCNQ salts may be found in Chapters 2-4.
4.1. NMP-TCNQ
The first of the highly conducting quasi-one-dimensional organic
materials to be studied in detail by solid-state physicists was N -methyl-
phenazinium-(NMP-) TCNQ, which has a room-temperature conductivity
of about 200 (0 cmr\ which increases by about a factor of 2 in a broad
peak around 200 o K(1l) (see Figure 1). Below 200 0 K its conductivity falls in
a manner that can be described equally well by a temperature-dependent
activation energy or by an e- a / TI12 law, characteristic of a variable-range
hopping(12) model for a disordered one-dimensional system. The former
possibility, an activated behavior, was attributed(11) to the occurrence of a
Mott-Hubbard metal-insulator transition to a low-temperature ground
state with one electron per site and an energy gap equal to U -4/, where U
is the Coulomb energy for two electrons on a site and 4t is the electronic
bandwidth in a tight-binding model. The question of variable-range hop-
ping in a disordered system versus a Mott-Hubbard metal-insulator tran-
sition was hotly argued at the time and remains to this day a subject of
controversy. Nevertheless it was clear to all that the two properties of
NMP-TCNQ on which the controversy hinged, namely, the intrinsic dis-
order in the NMP stack and the apparently quite large on-site Coulomb
interaction U, both acted to reduce the magnitude of the conductivity. In
this sense, the synthesis of TTF-TCNQ represented a significant step in the
direction of high conductivity. The higher symmetry of the TTF molecule
eliminated the problem of intrinsic disorder, and the closer stacking of the
TCNQ molecules increased the bandwidth while the larger polarizability of
the TTF molecules apparently led to a smaller effective value of U. All of
this is reflected in the much higher conductivity of TIF-TCNQ(13) when
compared to that of NMP-TCNQ (Figure I).
4.2. TTF- TCNQ

The discovery of large conductivity peaks in TTF-TCNQ and the very
apparent metal-insulator transition below 60 K raised a number of inter-
0
esting questions. Was the conductivity above 60 0 K enhanced by some

collective mechanism such as superconducting fluctuations? What was the
nature of the metal-insulator transition; and, if the two were linked, why
would fluctuations toward an insulating state enhance the conductivity?
The original proposal of Coleman et al.,(7) that conductivity enhancement
resulted from fluctuations toward a BCS superconducting state, was
quickly abandoned in favor of the proposal of Bardeen(!4) that the fluctua-
tions were toward Frohlich superconductivity. Later experiments by Tiedje
et aIY S ) showed that the magnetoresistance at the conductivity peak was
negligible in fields of 50 kG, and thus that conventional superconducting
fluctuations were probably not involved.
6 A. 1. Berlinsky
Bardeen's proposal, and later detailed calculations by Allender et

al., (16) attempted to explain the conductivity enhancement in terms of
sliding CDW fluctuations. The transition to the low-temperature insulating
state, which they described in terms of mean field theory, resulted from the
fact that the fully formed CDW was expected always to be pinned to the
lattice. Their theory was unable to account, however, for the observed
magnitude of the conductivity enhancement in the fluctuation regime.
Anderson et alY 7) remarked that the observed magnitude and large
apparent temperature range of the conductivity fluctuations required a
mean field transition temperature of several hundred degrees rather than
the 60 K used by Allender et alYh) However, they also favored the sliding
0
CDW mechanism. Soon afterwards, Lee, Rice, and Anderson (LRA)(18)

proposed a detailed model for the metal-insulator transition, attributing it
to a three-dimensional ordering of CDWs on neighboring chains. Using
the results of Scalapino et al.,(19) for the statistical mechanics of one-
dimensional systems that can be described by Landau-Ginzberg theory,
they showed that three-dimensional ordering would occur at roughly one
quarter of the mean field transition temperature. Later LRA extended
their theory by calculating the low-temperature dielectric properties of the
pinned CDW.(20) Three mechanisms were considered by which pinning
could occur: interchain coupling, pinning by impurities, and pinning caused
by commensurateness of the CDW with the underlying lattice. With regard
to this third mechanism they showed that unless the order of the com-
mensurateness was very low, i.e., unless the wavelength of the CDW was a
small integral multiple of the lattice constant, then the pinning effect was
negligible. Thus except for certain special cases (the nearly half-filled band
being the most important exception), CDWs are not seriously affected by
umklapp scattering.
Regardless of the exact nature of the pinning mechanism, LRA
showed that the low-frequency dielectric response of the system is of the
form
W~ 47Tne 2 /m*
e(w)=1+-2+ 2 2
6~ WT-W
where Wp is the plasma frequency for the conduction band, ~ is the Peierls
energy gap in the electronic spectrum, m * is the effective mass of the
CDW, and WT is the pinning frequency. Thus the dc dielectric constant is
greatly enhanced by the presence of a pinned CDW, particularly when WT
is small.
4.3. Kep
Despite the fact that most of the theoretical activity had resulted from
efforts to explain the conductivity peaks in TTF-TCNQ, the first appli-
cations of these theories were to the quasi-one-dimensional square planar

platinum salts KCP,(3) K2Pt(CN)4X03·xH20, X = Cl, Br. In these salts the
platinum atoms lie in stacks, each atom surrounded by four CN molecules.
The Pt-Pt spacing within the stack is quite small, being comparable to that
in platinum metal. The crystals are gold-colored and shiny with room-
temperature conductivities on the order of a few hundred (0 cmr! along
the chain direction and an anisotropy of longitudinal to transverse conduc-
tivities of about 10 5 Y 1) The conductivity of KCP increases very slightly
below room temperature and then decreases in an activated manner at
lower temperatures. Early explanations for the conductivity were based on
disorder theory,(22) but the observation, by neutron(23) and inelastic x-ray
scattering,(24) of a "giant Kohn anomaly" at 2 kF in the phonon spectrum
strongly suggested that the Peierls instability was playing an important
role. Detailed studies of the optical absorption spectrum by Briiesch et
at.(2S) in terms of the LRA model suggest a mean field transition tempera-
ture of about 1000oK. At low temperatures they observe a well-defined

Peierls gap of roughly 0.2 eV and a pinned CDW mode centered at about
15 cm -I. As the temperature is increased, the pinned mode broadens and
decreases in peak intensity until, at room temperature, the low-frequency
conductivity spectrum is fairly flat, even down to w = O. Thus it appears
that the CDW never really becomes unpinned and that there is no
significant collective contribution to the dc conductivity from sliding CDWs
in KCP. This is not too surprising in view of LRA's calculation of a
divergent impurity scattering rate for the CDW at w = O.
5. TTF- TCNQ and TSeF- TCNQ
5.1. Structural Transitions in TTF - TCNQ
In 1975, the long-awaited observation of the Peierls distortion in

TTF -TCNQ was reported by Denoyer et at.(26) (see Chapter 2 ). Using
inelastic x-ray scattering Denoyer et al. found scattering, which cor-
responded to a three-dimensional superlattice, below about 55°K. The
b-axis (which is the stacking axis) superlattice periodicity was found to be
(3.66 ± 0.1 )b, which corresponds to a 0.28 ± 0.01 filled band according to
Peierls's model, where a half-filled band (one electron per atom) leads to a
doubling of the periodicity. An important difference between the situations
in KCP and TTF- TCNQ, with regard to the implications of the wavelength
of the superlattice distortion, is that in KCP one could infer the filling of
the bands from chemical arguments. That is, the halogens remove 0.3
electrons per Pt from the conduction band leading to an 0.15 filled hole
band. Then the observation of the phonon anomaly at q = 0.31T/ c' (where
8 A. f. Berlinsky
c is the Pt-Pt separation) fit neatly into place. On the other hand, in
TTF-TCNQ, no compelling evidence existed a priori for a charge transfer
of slightly more than half an electron per TTF-TCNQ molecule. Thus our
present knowledge of the fraction of charge transfer, and hence of the
value of k F , relies completely on the interpretation of the superlattice
structure as a Peierls distortion.
Since the original measurements of Denoyer et al.(26) there has been a
great deal of activity among neutron and x-ray scattering groups studying
the structural transitions in TTF-TCNQ. This work is reviewed in detail in
Chapter 2. Two particularly interesting new results have emerged. One is
the discovery of a transition at 49°K,(27 .28) where the periodicity, transverse
to the chains, of the CDWs begins to readjust, going from 2a to 4a at the
third transition at 37°K, which had previously been observed in conduc-
tivity measurements.(29) The other new result is the observation of 4 kF
scattering,(30) which has been interpreted in terms of two-electron cor-
relations due to their repulsive Coulomb interaction.(31)
The value 4kF arises because, if repulsive Coulomb interactions are
dominant, the electrons will stay as far apart from each other as they can.
For Ne electrons on N sites, the largest separation possible is Ac = aN/ N.,
which corresponds to a wave vector kc = 2rrNe/ Na. The Fermi wave vector
is kF = rrNe/2Na, and hence kc = 4k F. Possibly as interesting as the studies
of the three-dimensionally ordered states at low temperatures are the
diffuse x-ray and inelastic neutron scattering studies of the giant Kohn
anomalies at 2kF and 4kF in the phonon spectra above 60°K,(32-34) The
x-ray studies show that diffuse 4kF scattering is evident all the way up to
room temperature, while the 2kF scattering only appears below about
150 K.
0
5.2. Electromagnetic Properties of TTF- TeNQ
In addition to the neutron and x-ray scattering studies, much work has
also been done, particularly by the Pennsylvania group, on measuring the
optical properties of TTF_TCNQ.(35) A reflectivity minimum in the
infrared at 1.2 eV has been interpreted as the plasma frequency. Further
out in the infrared, the absorption coefficient is found to peak at about
1000 cm -) and to decrease with lower frequency. The far-infrared behavior
must, ot course, depend on the temperature since the conductivity, for
example, is fairly large at dc and microwave frequencies above 400K and
drops to very low values at liquid-helium temperatures. No detailed study
has yet been performed that is comparable to that of Briiesch et al. (25) for
the pinned Frohlich mode in KCP. The reason is, as suggested by the work
of Coleman et al.,(36) that if such a mode exists, it must be pinned at a
frequency below 10 cm -), where experiments using conventional far-
infrared spectrometers and Michelson interferometers are extremely

difficult. Microwave experiments below 3 cm -I are possible, but they are
subject to problems of interpretation.
Nevertheless, in spite of the fact that direct observations of the pinned
Frohlich mode have not been made, the apparently quite large value of the
low-temperature microwave dielectric constant(37) does suggest the
presence of such a mode, and the Pennsylvania group has used this model
to interpret essentially all of the experimental data on TTF-TCNQ (see
Chapter 3). In addition they suggest that, at temperatures above 54°K, up
to nearly room temperature, the dc conductivity is dominated by the
collective contribution due to sliding CDWs. This is in contrast to the
situation with KCP, and (as is the case for the pinned Frohlich mode in
TTF-TCNQ) the direct evidence for such behavior is rather sparse. The
Pennsylvania theory implies that the high-temperature conductivity is
sharply peaked at dc and that it drops by an order of magnitude between 0
and 1 cm -I at lOooK. This would mean that coaxial resonator measure-
ments at 30 GHz (30 GHz = I cm -I) of Hardy et al.(38) are mainly measur-
ing the imaginary part of the complex conductivity. While it is difficult to
rule out this possibility on the basis of published data, we note that these
measurements are also consistent with a frequency-independent conduc-
tivity and u'» lut especially since the temperature dependences of the
microwave and dc conductivities are nearly identical.
5.3. ESR and Alloys of TTF- TCNQ and TSeF- TCNQ
In charge transfer compounds such as TTF-TCNQ where two distinct,

and to some extent decoup\ed, systems of conducting chains occur, it is
particularly useful to develop methods for separating phenomena that
occur on the two different kinds of chains. In the case of TTF-TCNQ, the
most fruitful of these methods have been electron spin resonance (ESR)(39)
and the study of the isostructural compounds, TSeF-TCNQ and alloys of
the type (TTF)x(TSeF)I_<-TCNQ.(40) The interpretation of the ESR
measurements depends on the fact that the g tensors of TTF and TCNQ
molecules are slightly different. The observed resonance is an average of
the two contributions, and thus the temperature dependence of the posi-
tion of the ESR lines give a measure of the relative contributions of the two
kinds of stacks to the magnetic susceptibility of the conduction electrons.
This and analogous NMR techniques involving the substitution of J3 C
atoms on one of the two types of molecules, (41) together with independent
measurements of the total susceptibility,(42) have allowed a detailed
separation of TTF and TCNQ susceptibilities. Such a separation is par-
ticularly interesting in the vicinity of the phase transitions. A more
complete discussion of these techniques is given in Chapter 4.
10 A. 1. Berlinsky
The synthesis of the TSeF molecule by Engler and Patel at IBM(43)

provided an additional microscopic probe, which can differentiate the
behavior on the two chains. TTF-TCNQ and TSeF-TCNQ are iso-
structural, but the phase transitions that occur are slightly different for the
two materials. TSeF-TCNQ apparently has only a single transition at about
27°K, just below the conductivity peak. By studying TTF-TCNQ crystals
doped with small amounts of TSeF, Etemad(29) was able to infer a fairly
complete picture of the transition region. This work is also discussed in
Chapter 4.
6. Theory
Much progress has been made, in the past few years, on the problem
of the interacting electron gas in one dimension (see Chapter 6), although
the connection between theory and experiment has remained rather tenu-
ous. An important stimulus for this work was the Little proposal concern-
ing superconductivity on one-dimensional chains.(5) There is a well-known
theorem that states that long-range order cannot exist in a one-dimen-
sional system where the interactions are of finite range.(44) Some doubt
existed, however, about whether this theorem applied to superconduc-
tivity, since in the reduced BCS Hamiltonian the forces are of infinite
range. Calculations by Rice,(45) using Ginzburg-Landau theory, and by
Ferrell,(46) using hydrodynamic arguments. led to the conclusion that the
superconducting transition could only occur at T = 0 in one dimension.
However, perturbation-theory calculations by Bychkov, Gor'kov, and
Ozyaoshinsky (BGO)(47) suggested a transition for T> O. This result was
subsequently contradicted by Hohenberg,(48) who derived a rigorous
inequality that implies that the type of long-range order (called "off-
diagonal long-range order") required for the superconducting and
superftuid states cannot exist in one or two dimensions. Meynhard and
S6Iyom(49) then reexamined the BGO solution and showed that it was
equivalent to a solution of the first-order renormalization group equations.
Extending the calculation to second-order theory, they found that the
spurious transition at finite temperature was renormalized to T = O.
The new results that emerged from the work of Meynhard and S6lyom
concerned the a~ymptotic behavior of the one-dimensional electron gas at
low temperature~. This behavior is determined by the nature of the elec-
tron-electron interaction. Two kinds of interaction were considered. The
first causes two electrons on opposite sides of the Fermi surface to scatter
across the Fermi surface. The interaction that leads to BCS superconduc-
tivity is of this type since electrons near ± kF exchange a phonon of wave
vector - 2kF The second type of interaction involves electrons on opposite
sides of the Fermi surface, each scattering into nearby states, with small
momentum transfer. Meynhard and S61yom defined coupling constants gl
and g2, respectively, for the strength of these large and small momentum
transfer interactions, and this nomenclature is now widely used.t Since the
constants gl and g2 determine the low-temperature behavior of the various
response functions of the system [e.g., whether the uniform magnetic
susceptibility X(O) remains finite or goes exponentially to zero as T ~ 0],
one can construct a kind of phase diagram, with respect to.these variables,
for the low-temperature state of the system. Exact results exist for gl = 0
(the Tomonaga model('iO 'il), for gl = 2g2 (the Hubbard modeJ(52.53), and
for gl = -(61T'/5)VF (the Luther and Emer/54 ) solution), where VF is the
Fermi velocity. The Luther-Emery result for negative gl was particularly
important since the renormalization group calculations are least reliable in
this region.
Scalapino et al. ("') and Imry et al, (56) have recently discussed the
importance of obtaining accurate solutions to the one-dimensional prob-
lem if one is to understand the behavior of a weakly interacting collection
of conducting chains such as occur in quasi-one-dimensional conductors. If
interchain interactions are weak, then the individual chains behave
independently until the temperature is low enough and the correlation
lengths on the chains become long enough that interchain interactions take
over. If the intrachain problem is well understood then the interchain
interaction can be treated perturbatively using mean field theory. Thus the
problem of three-dImensional ordering of weakly interacting chains
depends crucially on understanding the single-chain problem.
Depending on the signs and size of the coupling constants, several
different types of in!>tabilitie~ arc predicted to occur in one dimension at
low temperature, involvmg singlet and/or triplet superconducting states or
spin and/or chargc-den-.ity waves. When interchain couplings are included
the effect is to build up correlations between fluctuations on neighboring
chains. When such three-dimensional correlations become important, then
a three-dimensional ordering transition occurs as in the theory of LRA.(18)
Thus the relative strength of the different one-dimensional instabilities
depends on the single-chain interaction parameters gl and g2, but the
nature of the three-dimensional low-temperature ordering also depends on
the ability of the one-dimensional fluctuations on different chains to inter-
act. Lee et al.('>?) have argued that, in quasi-one-dimensional systems where
interchain hopping is negligible, the long-range interchain Coulomb inter-
action will always stabilize the CDW ground state. Thus even though the
short-range repulsive Coulomb interaction (the Hubbard U) may drive a
:j: In Chapter 6 Emery w,cs a ,lIghtly dIfferent set of couplIng constants taken from the theory
of the Kondo problem The relatIonshIp between the dIfferent sets of coupling constants IS
gIven In Table 1 of Chapter (,
12 A. f. Berlinsky
4kF instability at high temperature, as it apparently does in TTF-TCNQ, it

is not surprising that the 2kF CDW ultimately wins out at low tempera-
tures.
7. Some Concluding Thoughts
In the chapters that follow, most of the ideas that have been discussed
in this introduction will be developed in some detail. One fact that will
certainly strike the reader is that a great deal of work has been done in a
very short time. In order to confront such a large amount of information in
a reasonable way, it is important to keep in mind what is special about
these quasi-one-dimensional solids and why they are being studied. Several
particular points come to mind.
One of the most important questions that remains to be settled for
quasi-one-dimensional conductors in general and in particular for TTF-
TCNQ is whether or not sliding CDWs can actually enhance the dc
conductivity. In Chapter 3 Heeger attempts to demonstrate conclusively
that the large conductivity in TTF-TCNQ can only be explained by a
collective mechanism. Schultz in Chapter 4 takes the point of view that
none of the models that have been proposed can account for the detailed
behavior of TTF-TCNQ in a convincing way. It is this author's feeling that
the question is still wide open and that nothing will be settled until some-
one develops a reliable method to measure the complex conductivity of
TTF-TCNQ between 1 and 10 cm- I in both the conducting and insulating
regimes.
Another question of particular interest concerns the nature of the
electron-electron interactions. It has been argued that, on large organic
molecules such as TCNQ, electrons can correlate in such a way as to reduce
their effective Coulomb repulsion. Further, Little claims that the exciton
exchange mechanism can lead to a large attractive electron-electron inter-
action and hence to high-temperature superconductivity. Thus the ques-
tion of the nature of the electron-electron interaction within the chains is
of central importance.
Similarly, the size and effect of interchain interactions must be con-
sidered. In the case of TTF-TCNQ it seems clear that the high-temperature
phase is characteristically one-dimensional. On the other hand, in the
sulfur-nitrogen polymer (SN)x, this is apparently not the case. (58) An
important question, then, is what is the criterion for quasi-one-dimen-
sionality, and how sensitively does it depend on the details of the inter-
action parameters?
Finally, one might wonder how much we have learned about modify-
ing the structure of these solids and designing them to our specification.
IntroductIOn to Highly Conductzng One-DimensIOnal Solzds 13
This IS particularly relevant in the case of the orgamc conductors, where

small modlficatlOns of the individual molecules are routinely performed by
orgamc chemists On the one hand, one feels a sense of frustratlOn because
a large class of TCNQ-based matenals have been found to behave in very
simIlar ways In particular, a room-temperature conductiVIty of about
2 x 10 3 (0 em) 1 appears to be the hmlt for these salts On the other hand,
It does seem that It may be possible to develop classes of related matenals
m whIch the differences between members of the class can be related to
differences m the smgle-cham couplmg constants gl and g2 or m the
mtercham hoppmg or Coulomb mteractlOns and thus to provide more
stnngent expenmental tests of the kinds of theones that are presented in
Chapters 5-7
ACKNOWLEDGMENT
The author gratefully acknowledges the hospltahty of the Natuur-

kmdlg Laboratonum of the Umverslty of Amsterdam, where most of thiS
mtroductlOn was wntten
References
1 R E Pelerls Quantum Theory of Solzds, Oxford Umverslty Press, London (1955)

2 L R Melby R J Harder W R Hertler W Mahler R E Benson, and W E Mochel,
Substituted qumodlmethans II Amon-radial derIvatives and complexes of 7,7,8,8-
tetracyanoqumodlmethan ] Am Chem Soc 84,3374-3387 (1962)
3 K Krogmann and H D Hansen, "VIOlettes" KalIumtetracyanoplatmat,
K2Pt(CN)4Xo 3 2 5H 20 (X = Cl, Br), Z Anorg AUg Chem 358,67-81 (1968)
4 N POng and P Monceau Anomalous transport propertIes of a lInear-cham metal
NbSe3, Phys Rev B 16 3443 (1977)
5 W A Little POSSibilIty of synthesIZIng an orgamc superconductor, Phys Rev 134,
A1416-A1424 (1964)
6 Lecture Notes In PhYSICS Vol 14, One DimensIOnal Conductors, SprInger, New York
(1975), Low DimensIOnal CooperatIVe Phenomena, H J Keller (ed ), Plenum Press, New
York (1975), Chemistry and PhYSICS of One-DimensIOnal Metals, H J Keller (ed),
Plenum Press New York (1977), Lecture Notes In PhYSICS, Vol 65, Organic Conductors
and Semiconductors SprInger New York (1977)
7 L B Coleman, M J Cohen D J Sandman, F G Yamaglshl, A F GarIto, and A J
Heeger Superconductmg fluctuatIOns and the Pelerls mstabllIty In an orgamc solId, Solzd
State Commun 12 \12~ \132 (1973)
8 G A Thomas, D E Schafter F Wudl, P M Horn, D Rlmal, J W Cook, D A
Glocker, M J Skove, C W Chu R P Groff, J L Gillson, R C Wheland, L R Melby,
M B Salamon, R A Craven G De PasqualI, A N Bloch, D 0 Cowan, V V
Walatka R E Pyle R Gemmer T 0 Poehler G R Johnson, M G Miles, J D
Wilson J P Ferram T F Fmnegan, R J Warmack, V F Raaen, and D Jerome,
ElectrIcal conductiVity of tetrathIOfulvalemum-tetracyanoqumodlmethamde (TTF-
TCNQ) Phys Rev B 13 51O~-~110 (1976)
14 A. 1. Berlmsky
9 Marshall1 Cohen, L B Coleman, A F Ganto, and A 1 Heeger, Electromc propertIes

of tetrathlOfulvahmum-tetracyanoqulnodlmethamde (TTF-TCNQ), Phys Rev B 13,
5111-5116 (1976)
10 H Frohhch, On the theory of superconductIvity the one-dimensIOnal case, Proc R Soc
London Ser A 233, 296-305 (1954)
11 L B Coleman, 1 A Cohen, A F Ganto, and A J Heeger, ConductlVlty studies on
hlgh-punty n-methylphenazlnlUm tetracyanoqulnodlmethane, Phys Rev B 7, 2122-
2128 (1973)
12 A N Bloch, R B Weissman, and C M Varma, IdentIficatIOn of a class of disordered
one-dimensIOnal conductors, Phys Rev Lett 28,753-756 (1972)
13 1 T Tledje, "ConductlVlty and Magnetoreslstance of TTF-TCNQ," Masters ThesIs,
UmvefSlty of BntIsh Columbia (1975)
14 J Bardeen, Superconductlng fluctuatIOns In one-dimensIOnal orgamc sohds, Solid State
Commun 13,357-359 (1973)
15 1 T Tledje, J F Carolan, A 1 Berhnsky, and L Weller, Magnetoreslstance of TTF-
TCNQ, Can] Phys 53, 1593-1605 (1975)
16 D Allender, J W Bray, and J Bardeen, Theory of fluctuatIOn superconductlVlty from
electron-phonon Interactions In pseudo-one-dlmenslOnal systems, Phys Rev B 9, 119-
129 (1974)
17 P W Anderson, P A Lee, and M Saltoh, Remarks on giant conductivity In TTF-
TCNQ, Solzd State Commun 13,595-598 (1973)
18 P A Lee, T M Rice, and P W Anderson, FluctuatIOn effects at a Peleris tranSition,
Phys Rev Lett 31,462-465 (1973)
19 D J Scalaplno, M Sears, and R A Ferrell, Statistical mechamcs of one-dimensIOnal
Ginzburg-Landau fields, Phys Rev B 6, 3409-3416 (1972)
20 P A Lee, T M Rice, and P W Anderson, ConductlVlty from charge or Spin denSity
waves, Solzd State Commun 14,703-709 (1974)
21 H R Zeller and A Beck, Amsotropy of the electncal conductlVlty In the one-dimen-
sIOnal conductor Kz[Pt(CN)4]BrO 03 3(H zO), ] Phys Chem Solzds 35, 77-80 (1974)
22 M 1 Rice and J Bernascom, Interrupted strand model for quasl-one-dlmenslOnal
metals, ] Phys F 3, 55-66 (1973)
23 B Renker, H Rletschel, L PlntschOVIUS, W Glaser, P Breusch, and M J Rice,
ObservatIOn of giant Kohn anomaly In the one-dimensIOnal conductor
KzPt(CN)4Br03 3H 2 0, Phys Rev Lett 30, 1144-1147 (1973)
24 R Comes, M Lambert, and H R Zeller, A low temperature phase transition In the
one-dimensIOnal conductor KzPt(CN)4Br030 xHzO, Phys Status Solzdl B 58,587-592
(1973)
25 P Bruesch, S Strassler, and H R Zeller, FluctuatIOns and order In a one-dimensIOnal
system A spectroscopical study of the Peleris transition In KzPt(CN)4Bro 3 3(H2 0), Phys
Rev B 12, 219-225 (1975)
26 F Denoyer, R Comes, A F Ganto, and A J Heeger, X-ray-dlffuse-scattenng eVidence
for a phase tranSitIOn In tetrathlOfulvalene tetracyanoqUinodlmethane (TTF-TCNQ),
Phys Rev Lett 35,445-449 (1975)
27 Per Bak and V J Emery, Theory of the structural phase transformations In tetra-
thlOfulvalene-tetracyanoqulnodlmethane (TTF-TCNQ), Phys Rev Lett 36, 978-982
(1976)
28 W D Ellensen, R Comes, S M Shapiro, G Shlrane, A F Ganto, and A J Heeger,
Neutron scattenng study of 49 K phase transition In TTF-TCNQ, Solzd State Commun
20 53-55 (1976)
29 S Etemad, Systematic study of the transitions In tetrathlOfulvalene-tetra-
cyanoqulnodlmethane (TTF-TCNQ) and Its selenIUm analogs, Phys Rev B 13, 2254-
2261 (1976)
IntroductIOn to HIghly Conductmg One-DImensIOnal Solrds 15
30 J P Pouget, S K Khanna, F Denoyer, R Comes, A F Ganto, and A J Heeger, X-ray

observation of 2kF and 4kF scattenngs m tetrathlOfulvalene-tetracyanoqumodlmethane
(TTF-TCNO), Phys Rev Lett 37,437,440 (1976)
31 V J Emery, New mechanism for a phonon anomaly and lattice distortIOn m quasl-one-
dimensional conductors Phys Rev Lett 37 107-110 (1976)
32 S Kagoshlma, T Ishlgura and H Auryal J Phys Soc Japan 41,2061 (1976)
33 S K Khanna, J P Pouget R Come, A F Ganto, and A J Heeger, X-ray studies of 2kF
and 4kF anomahes In tetrdthlOfulvalenc-tetracyanoqumodlmethane (TTF-TCNO), Phys
Rev B 16, 146R-1479 (1977)
34 S M Shapiro, G Shlrane A F GdrItO, and A J Heeger, Phys Rev B 15,2413 (1977)
35 D B Tanner, C S Jacobsen A F GarIto, and A J Heeger, Infrared studies of the
energy gap m tetrathlOfulvalene-tetracyanoqumodlmethane (TTF-TCNO), Phys Rev B
13,3381-3404 (1976) and references cited therem
36 L B Coleman, C R Fincher Jr A F GarIto, and A J Heeger, Far-mfrared smgle
crystal studies of TTF-TC NO Phys Status SOltdl B 75, 239-246 (1976)
37 S K Khanna, E Ehrenfrcund A F GarIto, and A J Heeger, Electncal conductivity of
tetrathlOfulvalene-tetrac\anoqumodlmethane (TTF-TCNO), Phys Rev B 10, 1298-
1307 (1974)
38 W N Hardy A J Berhnsky and L Weller Microwave conductivity measurements by
the coaXial resondtor technique I DescnptlOn of the method and apphcatlOns to TTF-
reNO and related compounds Phys Rev B 14, 3356-3370 (1976)
39 Y TomkieWICZ and A Taranko, Roles of the donor and acceptor chams m the metal
msulator transItIon m TTF TCNO (tetrathlOfulvalene-tetracyanoqumodlmethane), Phys
Rev Lett 36 751-7';4 (1976)
40 S Etemad, T Penny E M Engler B A Scott, and P E Selden, dc conductiVity m an
Isostructural family of organiC metals Phys Rev Lett 34,741-744 (1975)
41 E F Rybaczewskl L S Smith A F Ganto A J Heeger, and B Silbernagel, 13 C
Knight shift m TTF TCNQ (' 'C) determmatIon of the local susceptlblhty, Phys Rev B
14,2746-2756(1976)
42 J C Scott, A F Ganto and A J Heeger MagnetIc susceptlblhty studies of tetra-
thlOfulvalene-tetracyanoqumodlmethane (TTF) (TCNO) and related organiC metals,
Phys Rev 810 1Hl 1119(1974)
43 E M Engler and V V Patel Structure control m organiC metals SyntheSIS of tetra-
selenofulvalene and Its charge transfer salt With tetracyano-p-qumodlmethane, J Arn
Chern Soc 96 7176-7178 (1974) and SyntheSIS of ClS- and trans-dlselenadlthIaful-
valene and ItS highly conductmg charge-transfer salt With tetracyano-p-qulnodlmethane,
J Chern Soc Chern Commun 9 671-672 (1975)
44 L Van Hove Sur I mtegrale de configuratIOn pour les systemes de partIcules Ii une
dimenSion, Physlca 16 117-143 (1950)
45 T M Rice, SuperconductiVity In one and two dimenSIOns, Phys Rev 140, A1889-
A1891 (1965)
46 R A Ferrell, Posslblhty of one-dimensIOnal superconductivity, Phys Rev Lett 13,
330-132 (1964)
47 Yu A Bychkov 1 P Gor'kov and I E Dzyaloshmsky, POSSlblhty of superconductIVIty
type phenomena m a one-dimensIOnal system, Sov Phys -JETP 23, 489-501 (1966)
48 P C Hohenberg, EXistence of long-range order m one and two dimenSIOns, Phys Rev
158,383-386 (1967)
49 N Meynhard and J Sol yom ApphcatlOn of the renormahzatlOn group technique to the
problem of phase trdnsltlon 111 one dimensIOnal metalliC systems I Invanant couplIngs,
vertex and one partIcle Green s functIOns J Low Temp Phys 12,529-545 (1973)
50 D C MattiS and E H Lleb Exact solutIOn of a many-fermIOn system and ItS associated
boson field J Math PhVI' 6 104-112 (196'i)
16 A. 1. Berlinsky
51. A. Luther and I. Peschel, Single-particle states, Kohn anomaly, and pairing fluctuations
in one dimension, Phys. Rev. B 9, 2911-2919 (1974).
52. E. H. Lieb and F. Y. Wu, Absence of Mott transition in an exact solution of the
short-range, one band model in one dimension, Phys. Rev. Lett. 20, 1445-1448 (1968).
53. A. A. Ovchinnikow, Excitation spectrum in the one-dimensional Hubbard model, Sov.
Phys.-JETP 30,1160-1163 (1970).
54. A. Luther and V. 1. Emery, Backward scattering in the one-dimensional electron gas,
Phys. Rev. Lett. 33, 589-592 (1974).
55. D. 1. Scalapino, Y. Imry, and P. Pincus, Generalized Ginzburg-Landau theory of pseudo-
one-dimensional systems, Phys. Rev. B 11,2042-2048 (1975).
56. Y. Imry, P. Pincus, and D. 1. Scalapino, Phase transitions in quasi-one-dimensional
magnetic structures: Quantum effects, Phys. Rev. B 12,1978-1980 (1975).
57. P. A. Lee, T. M. Rice, and R. A. Klemm, Role of interchain coupling in linear conduc-
tors, Phys. Rev. B 15, 2984 (1977).
58. W. I. Friesen, A. 1. Berlinsky, B. Bergersen, L. Weiler, and T. M. Rice, The three-
dimensional band structure of polysulphur-nitride, J. Phys. C 8, 3549-3557 (1975).
2
X -Ray and Neutron Scattering from
One-Dimensional Conductors
R. Comes and G. Shirane
1. Introduction
Early in 1972, just outside the dark room, we and our colleagues were
examining very closely a puzzling and still wet x-ray pattern (reproduced in
Figure 1(I» from a fascinating compound, K2Pt(CN)4Bro.30· xH 2 0 (KCP).
Besides the usually observed reflections of a rotating crystal pattern, it
showed continuous diffuse satellite lines at about one-third the reciprocal
lattice spacing from the layer lines of Bragg spots. It was the beginning of a
long succession of exciting experimental results obtained either from x-ray
or neutron scattering studies of one-dimensional conductors in relation to
their particular metal-to-insulator phase transitions. After a brief intro-
ductory section, this chapter describes in some detail the results so far
obtained with two main families of such compounds: the Krogman salts and
the organic conductors. (2,1)
1.1. Lattice Instabilities and Phonon Anomalies
1.1.1. The General Case of Displacive Phase Transitions
Since the original formulation in the early sixties(4) of a dynamical

approach to a certain type of phase transition, a great deal of structural
work has been devoted to the study of displacive phase transitions and the
R. Comes • Laboratoire de Physique des Solides, Associe au Centre National de la

Recherche Scientifique, Universite Paris-Sud, 91405 Orsay, France. G.
Shirane • Brookhaven NatIOnal Laboratory, Upton, New York 11973.
17
18 R Comes and G Shlrane
• •
•
Figure 1 X-ray diffuse scattenng pattern from K2Pt(CN)4BrO30 xH 20 (KCP) With Mo K",
radiation The film holder IS a flat camera, and the crystal was onented with ItS c aXIs rotated
15° from the vertical around a honzontal aXIs (from Reference 1)
assocIated precursor phonon anomahes (soft mode) Hlstoncally, the soft

mode Idea was conceIved from the temperature dependence of the dlelec-
tnc constant at para-ferroelectnc phase transItions, where m many cases the
stattc dlelectnc constant eo follows a Cune-Welss law m the hlgh-
temperature phase
1
The relatIOn to a polar opttc phonon mode IS then gIven by the

Lyddane-Sachs-Teller relatIOn
2
eoo WTO
-=-2-
eo WLO
where eoo IS the hIgh-frequency dielectnc constant. ThIS leads to a Bnlloum-

zone-center phonon anomaly schemattcally shown m FIgure 2a, the lowest
frequency of whIch IS the soft mode
w~oOCT-Tc
X-Ray and Neutron Scattering 19
At the phase transition, the eigenvectors of this mode condense into a

pattern of static displacements that give rise to the spontaneous permanent
polarization of the low-temperature ferroelectric phase.
This dynamical approach to displacive phase transitions was later
generalized to all structural phase transitions, provided they were of
second order, or nearly soY) In other words, the origin of the pattern of
static displacements of the less symmetric low-temperature phase was
assigned to a corresponding wave vector phonon anomaly or soft mode in
the high-temperature phase acting as a precursor. This led to the prediction
and subsequent observations of Brillouin-zane-center soft modes other
than ferroelectric,(6) but also to phonon anomalies mostly located some-
where on the Brillouin zane boundary (Figure 2b) and giving rise to
relatively simple low-temperature superstructures.(6) The detailed
experimental results are somewhat complicated by the observation of very
Q
E(k )
r .IT.
E(/i
ta) (d)
h
~
: I I /J. :
I ,
c
tb)
w
tel
Figure 2. Schematic phonon anomalies. (a) For ferrodistortive precursors: zone center soft
mode; (b) for antiferrodistortive precursors: zone boundary soft mode; (c) for modulated
lattices: arbitrary wave vector soft mode. Schematic modification of the electronic bands in
the case of a Peierles transition: (d) The conduction band of the high-temperature metal; (e)
the opening of a gap at the wave vector kF lowers the energy of the occupied states and raises
the energy of empty states: kF becomes a zone boundary of a distorted lattice with a
modulation period of 27T/2k F .
20 R. Comes and G. Shirane
strong anisotropies,(7·8) highly damped or even overdamped modes,(8) and

the existence of the additional central peak response. (9) This initial picture,
however, remains the most fruitful simple approach to structural tran-
sitions.
So far this concerns cases in which the unit cell below the transition
temperature is an integral multiple of the original high-temperature unit
cell. In reciprocal space, this yields phonon anomalies at wave vectors that
can be written as simple fractions of the reciprocal lattice vectors of the
high-temperature phase. We may regard the low-temperature phase as an
average structure essentially identical to the high-temperature one, with a
condensed phonon mode in addition. One can imagine cases in which the
condensed phonon has a wave vector that cannot be written as a simple
fraction of the reciprocal lattice vectors. The unit cell below the transition
point is then not an integral multiple of the high-temperature one. The
precursor phonon anomaly corresponding to such a case is shown in Figure
2c. Such structures indeed exist and are referred to as modulated struc-
tures.
Modulated structures where the modulation period is incommensurate
relative to the main lattice were observed long ago, for instance in those
alloys in which there are concentration modulations. Closer to the problem
of displacive phase transitions, one finds for instance ferroelectric
compounds such as NaN0 2,(lO) thiourea,(l1) and the more recent example
of K2Se04.12 However, the low-dimensional metals of either one-dimen-
sional or two-dimensional nature(J3) have been the most extensively
studied from a structural point of view because in their case the lattice
instability is understood to arise from the coupling of the conduction
electrons to the 2kF wave vector phonons and therefore has a direct
relation to the electrical properties.
1.1.2. The Case of One-Dimensional Metals
In some analogy with the historical origin of the soft mode idea for
ferroelectric compounds, the lattice instability in low-dimensional metals
originates from an anomaly in the electronic dielectric constant of metals at
the wave vector 2k F . Very schematically this can be shown as follows.
(i) The static electron gas susceptibility X(q) has the form(l4)
x(q) = L f(E k )- f(E k+ q )

k E k+ q - Ek
where Ek is the energy of the electronic state of momentum k and f(Ed is

the Fermi distribution function.
(ii) The motions of ions in metals are screened by the conduction

electrons and this leads to renormalized frequencies w (q) given by
W2(q) = w~(q)[l-Ax(q)]
where wo(q) is the unrenormalized phonon frequency and A the strength of
the electron-phonon coupling.
It is clearly seen that anomalies in X(q), and consequently in the
dispersion spectrum, will arise from scattering processes conserving
energy. These are scattering processes between opposite sides of the Fermi
surface and wave vector 2k F . In the three-dimensional case, X(q) has only a
smooth anomaly, giving rise to the generally shallow Kohn anomaly of ordi-
nary metals. (1 'i) In the one-dimensional case, on the contrary, since the Fermi
surface is restricted to a pair of singular points such as Ek - E k + 2kF = 0
over the entire range of k in the sum, x(q) diverges at the wave vector
2kF , leading to a pronounced phonon anomaly. The condensation of this
phonon anomaly into a pattern of static displacements produces a modu-
lated structure of period 27r/2k F along the chain direction; this period can
be incommensurate with the lattice period, depending on the precise filling
of the electron band. and as a consequence a gap will appear in the
conduction band.( I,,)
Another view of the same lattice instability was given by Peierls(l7)
and may be more evocative of the relation between electrical and structural
properties. If we consider a partly filled band (Figure 2d), the energy of the
electrons can always be lowered by opening a gap around the Fermi surface
(Figure 2e); all occupied states are then in the lower-energy band, while
the higher-energy band contains only empty states. The opening of such a
gap at the wave vector kF requires a structural distortion so that the
Brillouin zone boundary of the distorted insulating phase coincides with
the wave vector kF of the undistorted metallic phase, that is, it requires a
new lattice periodicity of 27T/2k K . This brings us back to the wave vector of
the Kahn anomaly. Whether this can happen or not depends on the
balance between the gain in electron kinetic energy and the loss in elastic
energy due to the lattice distortion. In strictly one-dimensional systems,
such a gap opening is expected to occur only at OOK(17); at finite tempera-
tures, enough electrons are thermally excited to the upper band in order to
cancel the gain in electron kinetic energy. The tendency of the lattice to
distort then reveals itself dynamically at sufficiently low temperatures by a
giant Kohn anomaly in the phonon dispersionY 6 ) In real systems, which
are only quasi-one-dimensional, because of possible charge compensations
between adjacent chains,(IX) or slight nonplanarity of the Fermi surface,(J9)
a real phase transition is expected at finite temperature between a high-
temperature conducting state and a low-temperature insulating state. One
reaches then a picture very similar to that of a structural phase transition,
the 2kF "giant Kohn anomaly" providing a high-temperature soft mode (in
the metallic state), which condenses to form a low-temperature modulated
structure (in the insulating state).
Here again, detailed theoretical studies(18~2Q) have predicted more
complicated spectra than shown in the schematic illustration of Figure 2c,
but the qualitative approach, as given above, provides a useful simple
picture.
The next two sections will briefly outline how x-ray and neutron
scattering techniques can be used to study this type of phase transition.
1.2. X-Ray Diffuse Scattering
1.2.1. The Origin of Diffuse Scattering
Any deviation from the periodic crystalline structure, which gives rise
to Bragg diffraction, produces an extra diffuse scattering at general scatter-
ing angles corresponding to general points in reciprocal space. Here we
shall deal only with displacive deviations from the mean structure, and this
brief introduction on x-ray scattering will therefore exclude other types of
disorder. X rays have energies on the order of 106 time that of the thermal
vibrations in crystals, and the energy of x rays scattered by phonons is
consequently essentially identical to the energy of the incoming beam; this
means that x rays will ignore the dynamical aspect and in the case of
phonons will only measure mean squared amplitudes (u), which are pro-
portional to the inverse of the squared phonon frequency w (for low-
frequency phonons and sufficiently high temperatures):
Ix rays OC (U)2 OC 1/ w 2
For ordinary crystals, phonons therefore only produce a relatively slowly
varying background that is superimposed on the much more intense Bragg
reflections.
In the presence of a phonon anomaly of one of the types shown in
Figure 2, however, to the minimum in the frequency at a wave vector qQ of
a dispersion branch there corresponds a maximum phonon amplitude u,
and consequently a maximum of the scattered intensity can be observed
within the weak phonon background at the same wave vector qQ. This is
why x rays can give fruitful preliminary information on dynamical anom-
alies, with the possibility of characterizing their wave vector. Measure-
ments performed in different Brillouin zones also permit the determination
of the corresponding phonon eigenvectors (provided one deals with a
phonon anomaly exclusively taking place on one dispersion branch).
We shall briefly illustrate below, using a simple one-dimensional
modulation example, how the scattered intensity arises.
X-Ray and Neutron Scattenng 23
If k. and kF are, respectively, the wave vectors of the incoming and

scattered beam, the momentum transfer Q is given by
(1 )
For x rays Ik.1 = IkF I and Equation (1) defines the Ewald sphere (Figure 3a).
If we consider two atoms separated by a distance r, the difference of phase
between beams scattered by these atoms is simply
tl</J = Q . r
The amplitude of the beam scattered from an assembly of atoms with
scattering factor II located at the extremity of vectors rl measured from an
Bragg diffrac lion
•
•
incoming
------->-
beam
.... L-..._ _...........
• •
• • •
I a)
·------..--
l' scattered beams
• • ,,;,;1
...
I "
i~c.?!:"lnJt _> _ *-
beam
---- ...... - -_ . - -- - ---
I~':'"''
c---- e • • •
c·L ___ ....
• •
I b)
Figure 3 The Ewald construction (a) DiffractIon IS only observed for reciprocal lattIce nodes
located on the Ewald sphere G hkl = ha* + kb* + lc* = k, - kF (b) For a modulated lattIce, the
reciprocal satellIte plane~ always Intersect the sphere Scattenng IS observed for any onen-
tatlon of the crystal
arbitrary origin is then

A = L II exp(iQ . rl)
I
In the case of a perfect three-dimensional lattice, with one atom per

unit cell, lattice constants a, b, c, and dimensions N\a X N 2 b X N3C, the
scattered amplitude becomes
Nl,N2,N~
A= I L exp[ i Q . (n Ia + n2b + n3c)] (2)
or
sin[!NI(Q· a)] sin[!N2 (Q· b)] sin[!N3 (Q· c)]
A = I sin !(Q . a) sin !(Q . b) sin !(Q . c) (3)
If we use the reciprocal lattice vectors a*, b*, c*, in a real crystal
(N\, N 2 , N3 ~ ex) the scattered amplitude becomes
with
G hkl = ha* + kb* + lc* (4)
These conditions are equivalent to the Bragg relation and lead to the
well-known Ewald construction for x-ray diffraction (Figure 3a).
Now let us consider a very simple example of a one-dimensional
modulation of this originally perfect lattice. Along the c direction, the
atoms are displaced sinusoidally from their average position n3C, with an
amplitude ~ and a wave vector qo. The phase between the different chains
parallel to c is arbitrary. The atoms of the distorted crystal are then located
at the extremity of vectors rl given by
rl = n la + n2b + n3C + ~ exp{i[ ±qo . n3C + <!>(n\n2)]}
where <!> (n \, n 2) IS the arbitrary phase on the different chains. So far this
modulation is static; in the case of x rays a dynamical modulation due for
example to a single phonon mode leads to identical results-it would just
introduce a temperature-dependent amplitude reflecting the Boltzmann
population factor. The scattered amplitude becomes
A = L II exp{iQ· d exp[±iqo' n3c+<!>(n\, n2)]} exp[iQ(n\a+n2b+n3c)]
For small-amplitude modulations the first exponential factor can be

expanded. To first order this gives the following two termst:
(i)
+ If the expan~lon I~ made to second order. It Just adds the corresponding Debye-Waller factor.
which is the Bragg diffraction of the average lattice [identical to Equation

(2) above] and
(ii) A D =t(LiQ.aex P[i(Q±qo)·n 3

n,
c)
which gives rise to an additional diffuse scattering.

Because of the arbitrary phase <fJ(nln2) of the modulation on different
chains, the summation of the amplitudes can be performed only on N 3 , and
the total diffuse scattering is just the sum of the intensities (squared
moduli) scattered by independent chains:
2 2 sin2[~N3(Q±qo)' c]
[dIffuse = t (Q . a) NIN2 . 2[1(Q
SIn 2
)]
±qo' c
(5)
---~v---~
D
This expression contains two main factors:
(i) The diffraction term D, which gives the localization of the diffuse
scattering with a maximum intensity for momentum transfers Q such as
(l an integer) (6)
This relation shows that Os is independent of the coordinates in the
a* and b* directions, thus defining sheets in reciprocal space perpendicular
to the c direction of the one-dimensional modulation, and located at ±qo
from the reciprocal lattice layers perpendicular to c. Figure 3b shows the
corresponding Ewald construction.
(ii) The factor Fd = lea . Q)2, which determines the relative intensity
of each such sheet: the diffuse structure factor.
This example is of course oversimplified; however, it already gives a
good feeling of how a pattern such as that shown in Figure 1 can arise. In
the case of a dynamical modulation of wave vector qo, frequency w, and
amplitude Ll, one simply has to take a time average (for x rays) that
replaces the squared amplitude Ll 2 by the mean squared amplitude. In the
case of real crystals and phonon anomalies such as shown in Figure 2,
Equation (5) has to be modified:
(i) There are generally several atoms per unit cell.
(ii) In the case of dynamical modulation, the mean squared dis-
placement depends on the Boltzmann population factor.
(iii) One has to consider more than one mode-all the different modes
that form the phonon anomaly; instead of a simple coherence
length N3C. this introduces correlation functions.
For a phonon mode and several atoms j per unit cell, the structure
factor Fd = l(J1 . Q)2 of Equation (5) is replaced by
(7)
where iJ, Wj' ej, rj are, respectively, the scattering factor, the Debye-Waller
factor, the phonon polarization, and the position in the unit cell of atom j;
w is the phonon frequency; E = hw(n +!) and n is the Boltzmann factor.
In the context of phase transitions one conventionally uses the
Ornstein-Zernicke correlation function; the diffraction term in relation (5)
is then replaced by a Lorentzian,
(8)
and the correlation length g is given by the inverse half-width at half-

maximum of the diffuse scattering.
1.2.2. The Measurement of X-Ray Diffuse Scattering
Relation (5) shows that the diffuse scattering is proportional to

f\Q . .1)2 compared to the intensity of Bragg reflections from the main
average lattice, which is proportional to l. This gives an order of magni-
tude of the expected intensity from a phonon modulation. Phonon ampli-
tudes are generally smaller than 1% of the lattice spacing, showing that the
intensity arising from a one-dimensional anomaly will be at least four
orders of magnitude smaller than the Bragg reflections.
The measurement of x-ray diffuse scattering therefore requires the use
of monochromators in order to eliminate the continuous spectrum
delivered by x-ray tubes (of the order of 10- 3 of the characteristic K lines).
Bragg reflections from the continuum would give more intensity than the
expected diffuse scattering. Figure 4a shows a schematic setup for such
measurement. In the case of counter detection, sample and counter are on
a single crystal goniometer but, in order to reach sufficient intensity, with a
very relaxed resolution compared to the instruments used for structure
analysis. In the case of photographic measurements, one can use the very
simple fixed-crystal, fixed-film technique, sometimes also called the
monochromatic Laue technique.(l) Diffuse scattering is indeed less local-
ized in reciprocal space than the usual Bragg diffraction; for instance in the
case of one-dimensional modulations, it is clear from Figure 3b that the
satellite sheets will intersect the Ewald sphere for any orientation of the
crystal and therefore give rise to diffuse scattering. In principle, one can
also use Weissenberg or precession techniques, but in most cases this will
X-Ray Source
s
\ retectlon
10 )
Reactor
"" ~ detec tlon
~S .A
_____ )
,,
1b)
Figure 4. Schemal1c experimental setup for the study of diffuse scattering: M = mono-
chromator, S = sample. A 0' analyzer. (a) For x-ray scattering; (b) for elastic or inelastic
neutron scattering.
lead to prohibitive exposure times (several weeks) and make temperature-

dependent measurements almost impossible.
1.3. Neutron Scattering
For the same range of wavelengths as x rays, thermal neutrons have

energies comparable to the phonon energies. An incoming neutron can
therefore lose (or gain) an appreciable amount of its initial energy by
creation (or annihilation) of a phonon. The wavelength A and the magni-
tude of the wave vector of the scattered neutron can then be different from
those of the incoming neutrons:
As a consequence, an energy analysis of the scattered beam can determine

whether the process that gives rise to the diffuse intensity is static (no
change in wavelength) or dynamic, and, in this latter case, can measure its
frequency.
It is clear that for static effects and for coherent neutron scattering,
one obtains expressions that are identical to those worked out for x rays. In
the case of phonons, the frequency analysis introduces an additional
dimension. Starting directly from the more general Equation (7) above, the
cross section for neutron scattering for phonons of frequency Wo, wave
vector qo, and measured in the Brillouin zone G=ha*+kb*+Ic* can be
written
(9)
where the scattering length bj replaces the x-ray scattering factor fj.
Even for ideal crystals, the 8 functions of Equation (9) would be
broadened by the experimental resolution function.(21) In real crystals, and
particularly in the case of soft phonon modes related to phase transitions
that are soft because of anharmonic interactions between atoms, the wave
trains of the lattice vibrations are limited in space and time. Using an
exponential decay for both space (as done above for x rays) and time, the 8
functions of Equation (9) are replaced by Lorentzian functions:
This yields the correlation length ~ as derived above for x-ray scatter-
ing and the damping constants r, respectively, obtained from the line
profiles in space and frequency.
For the measurement of inelastic neutron scattering, one uses the now
classical three-axis spectrometers(22); Figure 4b schematically illustrates
the experimental setup. comparison with Figure 4a for the x-ray case shows
that the main difference in the neutron apparatus is the existence of the
analyzing crystal.
2. Structural Studies of KCP and Related Platinum Chain

Complexes
KCP was first synthetized by Knop(23) in 1842, and is the oldest

compound known to have one-dimensional metallic properties. In fact,
these properties were first suggested by Krogmann from the structure
analysis of a series of platinum chain complexes.(2) Among these, KCP
[K2Pt(CN)4Bro3o· xH 2 0] has been the most extensively studied, both
because of its tetragonal symmetry and the possibility of growing large and
good-quality single crystals; it is therefore the prototype salt of this family.
In the next sections we shall describe in some detail the x-ray and neutron
scattering studies performed on KCP, and in the final section we shall more
briefly mention studies on other platinum chain complexes.
2.1. Structure and One-Dimensional Electrical Properties of Kep
The main feature of KCP (and most other members of this class of
materials) is the square planar arrangement of four ligands around a
central platinum ion, so that in the solid state the planar squares can be
stacked one on the other producing direct interaction between the metal
ions perpendicular to the plane of the complex. The structure of KCP
indeed exhibits such parallel stacks centered on the platinum metal ion as
shown in Figure 5. The stacks are separated from one another by 9.87 A,
and within each chain the platinum atoms are equally spaced with a spacing
of 2.88 A,(2) which is only slightly larger than the 2.77-A spacing in pure
platinum.
From this very general feature of the structure, it is already possible to
understand the origin of the one-dimensional electrical properties.
K 2 Pt(CNkxH 2 0 is an insulator, with a similar type of stacking but a
spacing between platinum atoms of 3.35 A. In this compound, the last-
filled electronic band is produced by the overlap of the platinum d; orbitals
parallel to the chain direction. The addition of a small amount of halogen
acceptors, which introduce energy levels well below the d; band of
platinum, removes the corresponding number of electrons from the pla-
tinum, giving rise to a metallic band structure.
Each platinum atom provides two electrons to the d; band; with
0.30 Br per platinum a~ in KCP. the metallic band is 85% filled. The Fermi
CN
CN
CN CN
C'N
N- C CN
CN CN
CN
(a) (b)
Figure 5. (a) d~ orbitals of Pt(CN)~; (b) overlap of d~ orbitals between stacked complex ions.
The ions are staggered to reduce the Coulomb repulsion between ligands (from Reference 1).
wave vector is then equal to O.857T/e ', with e ' = 2.88 A, or using the true
lattice constant e = 2e ',
kF(KCP)= O.85(27T/e)= O.85c*
The unit cell of KCP contains two units of K2Pt(CN)4BrO.30 . xH 20,
which implies disordered Br and only partially occupied Br sites. The initial
structure suggested by Krogmann was the centrosymmetric space group
P4/mmm with an additional disorder for the potassium ions involving
eight possible sites for the four available potassium ions per unit cell.(2)
More recent work performed on both deuterated and protonated KCP,
and on its chlorine analog, has shown that the structure is not the
centrosymmetric space group P4/mmm and ordered layers of K+
ions.(24.2S) However, one possibly important discrepancy remains from
these structural studies: An additional density found close to the Br site has
been assigned either to a small halogen occupancy(2S) or to a third water
site(26); the latter assignment is more consistent with recent NMR studies
•
Pt 0 C N K 0 Br
a,
z=o
• EB®
x.
e
eo
z=
z= Y:t
• e0
z == % 0
Figure 6. Projection on an ac plane of the structure of KCP. The shaded area close to the
center of the unit cell has been assigned either to an additional Br occupancy (Reference 25)
or to an additional 0 2 0 (or H 2 0) site (Reference 26).
on deuterium in KCP.(27) Figure 6 gives the projection of the crystal

structure of KCP.
2.2. X-Ray Diffuse Scattering from KeF
Figure 1, as already mentioned above, shows a typical x-ray diffuse

scattering pattern from KCP at room temperature as obtained with the
fixed-film, fixed-crystal technique by Comes et a/.(l) Perpendicular to the
chain direction (which is vertical for Figure 1), the layer lines of Bragg
spots that correspond to the Pt spacing (/ even) are accompanied by
continuous satellite diffuse lines. These diffuse lines correspond to planes
in reciprocal space perpendicular to the chain direction.
Analyzing the scattering more carefully, the following is found:
(i) The scattering vector component along the chain direction of
these reciprocal planes is
011 = (l ± 0.30)c* I even
(ii) The intensity from one sheet to the next increases as the scatter-
ing-vector component 011 increases along the chain direction:
where fo varies as an x-ray scattering factor.

(iii) The scattering is uniform in each diffuse plane, decreasing only
slowly with increasing scattering-vector component 01. perpendicular to
the chain direction as expected for an x-ray scattering factor fo:
These features are exactly as predicted by Equation (5) for the

simplest example of a sinusoidal modulation with displacements ac parallel
to chain direction and a wave vector qo = 0.3c*. With the more realistic
Ornstein-Zernicke correlation function and the diffraction term given by
Equation (8), the diffuse scattering satellite sheets of KCP at room
temperature correspond to an intensity distribution 1(0) given by
2 1
I(O)ocfo(Q· ~c) lle+[Q-(/±O.30)c*] I even
In principle the width of the diffuse lines along the chain direction can
give an estimation of the correlation length g; however, in the case of KCP,
all attempts so far have yielded widths limited to the instrumental resolu-
tion, giving only lower limits for the correlations. We shall come back to this
point in the section on neutron scattering.
The x-ray diffuse scattering from KCP thus corresponds to the scat-
tering expected from the giant Kohn anomaly of one-dimensional conduc-
tors; it is observed at the wave vectort qo:
qo = ±O.30c*
and
2k F (KCP) = 2 x O.85c* = 1. 70c*
which is equivalent in the reduced Brillouin zone to 2c* -1.70c* = O.30c*.

As already mentioned above, it is obvious that x rays cannot be used
to determine whether the observed scattering is due to a static modulation
or to a dynamical phonon anomaly. Another feature of the x-ray results
arises from the fact that the x-ray scattering factors are proportional to the
atomic numbers of the atoms involved; as a consequence, compared to the
contribution of platinum, the other atoms eventually taking part in the
modulation make a negligible contribution to the scattering. Therefore the
scattering only reveals the modulation of the platinum chains.
As shown by Comes et al.,(2R) low-temperature x-ray measurements
on KCP showed no appreciable modification of the room-temperature
results down to about 120oK. For lower temperatures, the diffuse satellites
lose progressively the aspect of continuous diffuse lines and the intensity
progressively concentrates at the wave vector (rr/a, rr/a, 2kF)' At 77°K,
the lowest temperature of the x-ray study, the diffuse sheet satellite scat-
tering has taken the aspect of dotted diffuse lines (Figure 7). This
modification of the diffuse scattering reveals a progressive coupling of the
phase of the modulation on successive parallel chains. Since the conden-
sation of the intensity occurs on the transverse zone boundary (rr/ a, rr/ a),
this coupling doubles the unit cell in the a and b directions and cor-
responds to modulations which are in opposite phase in successive chains
along the a and b directions.
This particularly simple type of three-dimensional correlation indeed
produces charge compensation between the adjacent sites on parallel
chains, an argument that was earlier used to explain the martensitic tran-
sitions of V3Si or Nb,Sn. (lR) As can be seen on the pattern of Figure 7b, at
7rK the maxima of diffuse intensity at the wave vectors (rr/ a, rr/ a, 2k F )
are still part of a diffuse line: only short-range coupling between chains is
achieved at this temperature.
t In the initial x-ray work, that satellite sheet wave vector was reported to be 0.33c*; this was
later corrected by the more precise neutron counter measurements(29-31) and higher-
resolution photographic x-ray patterns.
a b
Figure 7 Enlarged sectIOns from x-ray diffuse scatterIng patterns of KCP, showmg the
bUildup of three-dimensIOnal order at low temperature (a) Room temperature the diffuse
scatterIng IS a contmuous satellIte diffuse lIne (one-dimensIOnal scatterIng) (b) 77°K the
mtenslty on the diffuse Ime has condensed half-way between two successive Bragg spots
(transverse zone boundary) (after Reference 28)
23 Neutron Scattenng StudIes of Kep
As already mentioned, Kep can be grown relatIvely eaSily m large,

good-qualIty smgle crystals on the order of 1 cm 3 or more, which are well
SUIted for neutron studies The earlIer melastIc neutron scattenng
measurements performed by Renker et at (29-31) unambiguously showed a
contmuous gIant Kohn anomaly over the entIre 2kF sheet as shown m
Figure 8 In additIon a quaslelastlc cross sectIon was revealed m measure-
ments on deuterated crystals The three-dimensIOnal ordenng at low
temperature was found to Involve only the quasIelastIc component, while
the dynamical part, although altered, In first approximatIon retams Its
one-dimensIOnal character at all temperatures We shall therefore descnbe
the elastiC and the melastlc scatterIng results m two dlstmct sectIOns
JO
75
70
75
7.0
Figure 8 Room temperature dispersIOn surface 05
of KCP for acoustIc phonons propagatmg along
the cham directIOn The giant 2kF anomaly IS o.z 03
clearly VISible for the longltudmal modes (from
Reference 10)
(jlt/
2.3.1. Elastic Neutron Scattering from Kep
The buildup of three-dimensional correlations in KCP was studied in

great detail using neutron scattering. Figure 9 shows elastic scattering
measurements performed by Renker et al. (30) within the 2kF reciprocal
plane around the (7T / a, 7T/ a, 2k F ) position where the x rays showed the
intensity condensation below 120°K. These results confirm the x-ray
observation of a buildup of three-dimensional correlations below 120oK,
but also add several other features of interest.
(i) Long-range three-dimensional ordering is not achieved at the
lowest temperature of 6°K. There is a rapid increase of the transverse
correlation between 120 and SooK, but a saturation value of about 50 A is
reached at lower temperatures with no further evolution (Figure 10). The
6000
7000 6K~
T=87K
5000
78
6000
w 87K 4000
...J
«
~ 5000
['
r:i
..... 3000
gj 4000 1000
« 1l0K 2000
>-
..... 900
in 3000
z
....w
z 800
4.25 05
- 2000
[ 0.0.~ 1
Ol+---~--~--~--~--~--~--~--~--~
o .1 .2 .3 .4 .5 .6 .7 .8 .9
Figure 9. Temperature dependence of the elastic scattering from KCP in the 2kF plane and a
(~, ~, 2k F ) direction. The inserts show scans through the 2kF plane and along the (0, 0, ~) and
(Tr/ a, Tr/ a, ~) directions. The width at half-maximum in these latter scans is given by the
resolution of the spectrometer and remains constant over the whole temperature range (from
Reference 30).
UJ
...J 7000
«
u
~
U'l
a:
.....
6000
.1.
iii N
0::
5000 -,
\.
~
71.000 .3 ~
:I:
~
.2 ;:::.,
Figure 10. Temperature dependence of the

~2 ":'
.....:;
!;j; .1
scattering intensity at (TTl a, TTl a, 2k F ) (full >
circles) and of the linewidth (full width at .....
iii
half-maximum) of the intensity distribution z O L-~~~~~~==~
....zw o i,0 80 120 160 200 2L,() 280 320
in the (~,~, 2k F ) direction (open circles) from T(K)
Reference 30). ....
width in chain direction, that is, through the 2kF reciprocal plane, is limited
at all temperatures by the best available instrumental resolution, yielding a
lower limit for a longitudinal correlation length of about 100 Pt spacings
(=300 A). The final low-temperature state of KCP can therefore be
described in terms of elongated areas of = 300 x 50 x 50 A3 in which the
three-dimensional distortion giving rise to the insulating properties are
correlated, but there is no long-range order. Although such results clearly
show that there is no real phase transition in KCP, they are reminiscent of a
phase transition and we shall refer to it as the three-dimensional ordering
with TJD = 100 c K.
(ii) The scans illustrated in Figure 9 have a nonsymmetrical line
shape. This was shown to arise from a structure factor effect. Indeed if the
modulation only involves displacements of Pt atoms, the structure factor
would be a constant for neutron scattering; if, on the contrary, more atoms
respond to the modulation displacements, the chains acquire spatial extent
perpendicular to the chains, resulting in a variation of the structure factor
perpendicular to the c axis. In a perfectly three-dimensional ordered
modulated structure, this would only be reflected in the relative intensity of
the different ('TTl a, 'TTl a, 2k F ) reflections. Here, because of the incomplete
transverse order which broadens these reflections upon Q, it is also
reflected in the line shape of the scans of Figure 9.
It was shown by Lynn et al.(32) that the structure factor of rigid Pt(CN)4
units was required in order to account for the asymmetrical line shapes.
The purely one-dimensional cross sections shown in Figure lla are
modified by the three-dimensional correlations to produce the solid lines
shown in Figure 11 b as fitted to the experimental line shapes. Further work
by Eagens et al. m ) using the intensivity of eight different satellite
reflections, completely confirmed that the modulation in KCP involved
rigid Pt(CN)4 complexes, and determined the modulation amplitude to be
36
I-D STRUCTURE FACTOR

'"<.> OF Pt (C )4 CHAIN
(~, LO]
'2
"
"-
.c
a
'"
v
lL
b
0::
g
z 20.0.0.
0 T:6.5°K
~
0
0
0
.....
(f)
f-- 10.0.0.
z
=>
0
u
)..-LLI L I I
Ct:
~ 20.0.0.
z
o T=12QoK
:::;:-
c
8E
00
",""/10.0.0.
;n-
f--
Z
=>
o
u
L I I
•••
. •
T =29Qo K
50.0. -------- ~~
'. . ~
.
_ _ _e_ _ _ -.::. •
T= 20.0.° K
~~~~~~~~~~~ /
- y -
'. .
0..0. 0..2 0..4 0..6 0..8 1.0. 0.2 0.4 0..6 0..8 .10.
~-
FIgure 11 (a) Square of the one-dImensIOnal structure factor perpendIcular to the cham aXIs
for a cham of Pt(CN)4 complexes The dependence on lIS due to the consecutIve 45° rotatIOns
of the Pt(CN)4 groups (see FIgure 5b) (From Reference 32 ) (b) Temperature dependence of
the satellIte scatterIng perpendIcular to c at a serIes of temperatures The solId curves are fits
of the data to the model of correlated Pt(CN)4 chams Note that the fits reproduce well the
pronounced d,ymmetrv and shIfts of the maxImum mtenslty (from Reference 32)
(0.0047 ± 0.0005)c at 7°K. This provides an excellent example of

complementarity of the x-ray and neutron scattering measurements: Using
x rays alone, a very detailed and perhaps unfeasible series of measurements
would have been required to find out that atoms other than platinum are
displaced in the modulation waves, whereas with neutron scattering,
because the neutron scattering lengths are more comparable for the
different atoms, this result was attained much more directly. From the
fitting procedure shown in Figure 11 b, Lynn et al. also obtained the
temperature dependence of the modulation amplitude, which provides
another characterization of the three-dimensional ordering around 100oK.
The amplitude is found to increase continuously with decreasing tempera-
ture, and no distinct feature is observed around lOOoK (Figure 12).
There have been many speculations about the origin of the incomplete
ordering in KCP. The simplest approach is to invoke pinning on impurities,
.;;x 0.025 KCP

z
0
i= 0.020
a::
0
I-
(f) 0.015
Ci
u.
0
w 0.010
0
:::>
I-
:J 0.005
0-
::;:
«
0
0.20
-I 0.15
j j
J
.~
'- 0. 10
0.05
0
0 100 300
T( K)
Figure 12. Temperature dependence of the amplItude of the sInusoidal charge-density wave.s
in each chain, and the Inverse correlatIOn range perpendicular to c as deduced from fits as
~hown In Figure 11 b (from Reference 32).
lattice defects, or disorder which could be responsible for the dephasing of

static atomic displacement waves in the chains(34); this is of course suppor-
ted by the existing Br disorder. A partial answer is given by the study of the
well-ordered organic one-dimensional conductor TTF-TCNQ (see Section
3), where the low-temperature state appears long-range ordered. This
seems indeed to favor an explanation in terms of defects due to the intrinsic
disorder of KCP. In the same direction are the interpretations of other
physical properties of KCP, the interrupted strand model for the conduc-
tivity,(35) and the NMR data.(35)
So far we have assumed, as in most models for KCP that have been
worked out in some detail, that the distortion is uniform throughout the
crystal. However, another approach was proposed by Philips(36) in which
the chains are divided into distinct conducting and distorted segments. The
temperature dependence can then be obtained by a progressive growth of
the distorted sections at the expense of the metallic sections. A similar
description was also proposed for TTF-TCNQ.(37)
2.3.2. Inelastic Neutron Scattering from Kep
We have already mentioned the initial experiments by Renker et

aIY9,30) which established the existence of a 2kF phonon anomaly in KCP.
Further investigations of the dynamical properties were performed by
Carneiro et at. (38) The study of the inelastic scattering within the 2kF
phonon anomaly has required special consideration. Since the scattering
near 2kF is very steep in energy, the constant-Q method of scanning (i.e.,
fixing the momentum transfer and varying the energy transfer(22») is not
satisfactory. In such circumstances, the usual technique is to employ
constant-E scans (fixing the energy and varying the momentum transfer).
Here, however, even this technique does not lead to an unambiguous
representation of the dispersion near 2k F • In order to provide data free
of interpretation, the experimental results were presented in the form of
w, q intensity contours, as shown in Figures 13 and 14. Besides the
low-temperature three-dimensional correlations revealed by the elastic
scattering maximum at (1T/a, 1T/a, 2k F), the inelastic parts of these
intensity contours around 2kF in first approximation clearly retain a
one-dimensional character, that is to say only a little dependence on Q.L'
Moreover, instead of the very broad 2kF plateau observed down to 240 0 K
(Figure 13), the low-temperature measurements reveal two intensity
maxima located at 2.5 and 5.5 meV.t
+A third peak found at 3.3 meV (Figure 14a) apparently comes from an interaction with the
T A mode. (38) This maximum was also reported to have a relatively large extension parallel to
the chain direction (11)
r A Z
8
?:
5
EB~I
I 0.40 I
!. 4 -
3
..t: INTENSITIES:
5 ••••••
3 10 --
15 ---
20 --
30 ••••••
2 40 - - COUNTS
50 - - Tm;;l
100 .,. ••••
200 - -
300 - - -
0 0.1 0.2 0.3 0.4 05

~ [C*j
Figure 13. Normalized (w, q) intensity contours of the neutron scattering of KCP from the
excitations of wave vector q = (c* at 240 o K. The elastic incoherent scattering and the inelastic
background have been subtracted. The normalized intensity unit " counts/(2 min)" is not to be
taken as representa tive of the actual counting time (from Reference 38).
2.3.3. The Interpretation of the Neutron Scattering Study of Kep
The results described above clearly show that there is a quasielastic as

well as an inelastic component to the one-dimensional 2kF scattering in
KCP, and that limited three-dimensional ordering involving the quasielas-
tic part of this scattering occurs around lOO°K. There is, however, no clear
understanding of the relation between these two components.
Clearly one does not observe a soft mode whose frequency decreases
with temperature and condenses at the phase transition, thus giving rise to
a low-temperature insulating three-dimensional modulated structure. The
comparison of Figures 13 and 14 seems to indicate an overall hardening of
the broad spectra below 100oK, but there is no clear vanishing of the 2kF
anomaly. In contrast with structural phase transitions of more usual three-
dimensional compounds, the problem is not so much here to account for
the existence of the quasielastic component (central peak), which seems
r A z
e
7 :K2 P!(CNl4 B'o.33.2 O2

: T'SOK
6 RESOLUTION(FWHM)
5
>"
E INTENSITIES
'34 5 .. .. .. .
~
10--
15 - - -
20--
3 30·······
4 0 - - COUNTS
50 - - ~
100 . ... . . .
2
200 - -
300 - - -
400 - -
500 . ..... .
o 0.1 0.2 0.5

a t [C*j
0.4
err-_________.--------~r-----A----_.--_._r--",_--------,
/.
z
7
Kl'(CN~ B'ru 3.20 p
T ' I60 K
6
RESOLUTION (FWHM)
5
0------<
0.040
INTENSITIES
5 ...... .
10 - -
15 - - -
20--
30 ...... .
2 40 - - COUNTS
50 ---- 2fOOl
100 •••••••
200 --
300 -- -
400 - -
500 ..•.. ..
o QJ 0.2 0.3 0.4 0.5

c [c* J
Figure 14. Normalized (w, q) intensity contours of the neutron scattering of KCP. (a) From
the excitations of wave vector (0, 0, (c*) at 80 and 160o K. (b) From the excitations of wave
vector (a* /2. a* /2. (c*) at 80 and 160°K. The elastic incoherent scattering and the inelastic
M v A
8 ,---------.---------.---------r---~--_,,_------_.
6 T'80K
RESOLUTION(FWHM)
INTENSIT IE S
50
1 ........
--
15 - - -
20 - -
30 . . .... . .
40 - -
jCOUNTS
2min
2
Igg~
200- -
300---
500· ...... .
0.4 0.5
~ [C· J
v A
3
of' INTENSITIES
5 .. .... .
10 - -
15 - ' -
~~
COUNTS
188 ~
200--
2m,n
~ .. '~
0 ~----~O~.I----~O~
.2-~~-<~~~~~_-OL
.4-----0
~.5
C[Co)
background have been subtracted The normalIzed intensity Unit counts/{2 min)" IS not to be
taken as representative of the actual counting time (from Reference 38)
predicted by theory in the case of one-dimensional metals, (20) but to

understand the inelastic spectrum itself and its temperature dependence.
A first explanation is to assign the two low-temperature peaks ob-
served at 5.5 and 2.5 meV to, respectively, the amplitude and phase
mode.(39·40) This assignment was originally given by Bruesch et al.(41) from
infrared experiments and by Steigmeier et al. (42) from Raman measure-
ments.
An alternate suggestion was made by Carneiro et al.(38) Using the
response of damped harmonic oscillators, they showed that the low-
temperature intensity contours could be explained with a continuous dis-
persion, as shown schematically in Figure 2c. In both models increasing
dampmg WIth temperature is sufficient to account for the temperature
dependence.
The important question, before trying to compare these results quan-
titatively with the theoretical predictions concerning Peierls transitions, is
to wonder whether KCP with its intrinsic disorder and its incomplete
ordering can really be considered as an appropriate model compound. It
has provided material for the first obervations of the qualitative features
predicted along ago for one-dimensional conductors, but other well-
ordered systems may be better suited for further understanding of the
genuine properties of one-dimensional conductors.
2.4. Study of Other Platinum Complexes

A few other platinum chain complexes, as listed in Table 1, have been
studied, for the most part briefly by x-ray diffuse scattering. Most of these
compounds have lower symmetry than KCP, yielding generally highly
twinned crystals, but they display the same type of one-dimensional stack-
ing. (2) They can also often exist in several forms according to very easy
dehydration, making their study in general more difficult than for KCP.
Table 1. Platinum Chain Complexes
Complex Reference
Cyamde~
- K2Pt(CN)4Clo 32' xH 20
-Na2Pt(CN)4Bro 23' xH20 1
- KI 7sPt(CNk 1.5 H 20 43,44
-Rbi 7sPt(CN)4'xH20 44
-CSI7SPt(CN)4·xH20 44
-[C(NH2hhPt(CN)4BrO 23' xH20 44
Oxalates
-KI dPt(C 20 4hl'xH 20 45
-Mgo dPt(C 20 4hl'xH20 46
, , \ I
FIgure 15 DIffuse x-ray pattern from Mgo 82Pt(C204h xH 2 0 obtaIned wIth Cu Ka radIatIon
at room temperature and a cylIndncal camera (from Reference 46).
In all cases one-dimensional diffuse x-ray scattering was observed

at a wave vector along the chain direction, in close agreement with
the calculated 2kF value from the chemical formula. The beautiful
pattern of Figure 15 corresponds to the last-listed compound
{Mgo dPt(C 20 4h]· xH 20}, which has been studied in some detail by
Dubois(46l; he found in particular no three-dimensional ordering of the 2kp
waves down to nOK, extended two-dimensional scattering in addition to
the one-dimensional 2k p scattering, and another phase transition toward a
modulated structure taking place near 11°C and with no apparent relation
with the 2k p scattering.
Only KI 7sPt(CN)4·1.5D20 (K Def CP) has been briefly investigated
by neutron scattering.(43) A well-defined Kohn anomaly was observed at
the commensurate 2kF wave vector predicted by the chemical formula.
Surprisingly, the phonon anomaly is well observed in a transverse acoustic
branch, while the longitudinal branch also seems to soften at 2k F • No phase
transition or three-dimensIOnal ordering was reported down to 80o K.
3. Structural Studies of Organic One-Dimensional Conductors
KCP has the appealing feature of possessing only one type of conduc-
ting stack, and a conduction-band filling unambiguously known from the
chemical formula. It has therefore made it possible to establish the exis-
tence of "giant Kohn anomalies" in one-dimensional metals and to reveal a
three-dimensional ordering of the charge density waves at low temperature
related to a Peierls instability. As in the case of all other presently known

single-stack conductors, KCP is not well ordered, which complicates a full
understanding of its properties. The only known well-ordered one-dimen-
sional conductors are the organic charge transfer salts, which by definition
possess two different types of conducting stacks. Moreover, the charge
transfer and consequently the band filling of these compounds can only be
indirectly estimated. The organic charge transfer metals have therefore
richer properties, but the two-stack systems bring another type of
complexity, and a detailed study is required in order to find out the
respective contributions from each type of stack.
The neutron and x-ray scattering studies of tetrathiafulvalene tetra-
cyanoquinodimethane (TTF-TCNQ), which will be dealt with in the next
sections, are particularly relevant in this respect. In a last section we briefly
mention studies of other one-dimensional conductors.
3.1. Structure and TTF- TCNQ Crystals
It is not overrated to say that in recent years the crystalline structure of

ITF-TCNQ has been one of the most frequently and most carefully
investigated structures,(47,48) with structure analysis performed not only at
room temperature but also at 100, 60, 53, and 45°K. As established by
Kistenmacher et al.,(47) TTF-TCNQ crystallizes in the monoclinic system
(space group P 2dc) with unit cell constants a = 12.298, b=3.819, c=
18.468 A, and {3 = 104.46°.
The unit cell contains two formula units, as shown in projection
perpendicular to the b axis in Figure 16. The almost-planar TTF radical
cations and the TCNQ radical anions form segregated columnar stacks in
the b direction (which is here the one-dimensional direction) with inter-
planar spacings of 3.47 and 3.17 A, respectively. The dihedral angle
between the least-squares planes of the cations and the anions is approxi-
mately bisected by the [010] direction. The stacking of TTF-TCNQ is
schematically shown in Figure 17.
Figure 16. Projection perpen-

dicular to the chain axis (a, c)
plane of the molecules in the unit
cell of TIF-TCNQ (from
TCNQ Reference 47).
[10 0J
TC NO
TTF
Figure 17. Schematic herringbone stacking of TTF-TCNQ.
The one-dimensional metallic state of TTF-TCNQ is achieved by the

overlap along the stacks of the highly directional 1T molecular orbitals and
the unpaired electrons on the acceptor (TCNQ) and on the donor (TTF),
which result from a simple transfer:
(TIF)o + (TCNQ)o ~ (TTFtx (TCNQr x
The precise amount of charge transfer is generally unknown; usually it is
evaluated by interpolation from the bond lengths in the charge transfer
compound and those of neutral and singly charged molecules [for instance
(TCNQ)o and (TCNQr l
Given the error bars on the bond lengths, this procedure yielded
charge transfers of 1(47) or 0.5 electrons.(48) A direct evaluation of the
charge transfer by numerical integration of charge-density distribution
performed by Coppens(49) indicated a charge transfer of 0.48-0.60 elec-
trons from a TTF to a TCNQ. As will be explained below, this is in close
agreement with the measured value of the 2kF wave vector of diffuse
scattering experiments, which determined this charge transfer to be 0.59 ±
0.01 electron.(50)
These ambiguities concerning the amount of transferred charge, and
consequently the value of the 2kF wave vector, may partly explain why, in
spite of several attempts, conventional structure investigations could not
detect the Peierls transition predicted from the earliest transport
measurements. (51)
The existence of such a phase transition was established by the obser-
vation in two independent diffuse x-ray studies by Denoyer et at.(52) and
Kagoshima et at.,(52) respectively, of new satellite Bragg peaks observed
below 54°K. A further characterization of the structural instabilities and
modulated phases of TTF-TCNQ was achieved by a combination of neu-
tron and x-ray studies. (<;0.53-55) In both cases, the experimental conditions
were unusually difficult. In the x-ray case special equipment had to be
constructed in order to reach sufficiently low temperatures, which had to be
maintained for weeks. Other types of difficulties were encountered for the
case of neutron scattering.
Neutron triple-axis spectrometry, as briefly described in the intro-

ductory section, and as used for the study of KCP, has now become a
well-established tool, and the study of elastic or inelastic scattering from
"large and simple" crystals is almost a matter of routine work. The case of
TTF-TCNQ offered, however, several complications. Available crystals
are extremely small ab plane platelets, with typical dimensions for un-
twinned crystals of 10 x 2 x 0.05 mm 3 (to be compared with a size of the
order of 1 cm 3 for the usual crystals studied by neutron scattering). Further
difficulties originate in the relatively complicated crystal structure: there is
a total of 68 atoms per unit cell in TTF-TCNQ. In contrast with simpler
structures, where many of the Bragg reflections for which the scattering by
all the atoms are in phase produce high intensities, TTF-TCNQ has only a
few strong reflections, moreover not necessarily at the appropriate location
in reciprocal space for a given investigation. The best reflection (013) has
only about ~ of the potential in phase scattering, and most of the other
reflections are considerably weaker. These intensity conditions also affect
the diffuse scattering, and even the well-defined satellites from the low-
temperature modulated phases were hardly measurable and then only in a
few Brillouin zones.
There have been three independent diffuse scattering investigations of
TTF-TCNQ. Below, we shall describe the combined x-ray and neutron
studies by the groups from Orsay<54) and Brookhaven(50,53,55) and the x-ray
investigation by Kagoshima et at. (56) which led to coherent results. The
neutron measurements by Mook et al.(S7) revealed important discrepancies
with the other studies; we shall therefore discuss these disagreements in a
separate section. In the Brookhaven study, most of the elastic scattering
measurements (with analyzer set for zero energy transfer) were performed
on a single deuterated untwinned crystal. Weaker effects and inelastic
phonon measurements were carried out with an assembly of 17 carefully
selected, mostly untwinned, deuterated samples, which were aligned one
by one with respect to their c* direction and mounted on an aluminium
plate.(50,53) This resulted in an aggregate sample of about 15 mm 3 with an
effective mosaic spread of 0.5° in the b* (chain) direction and 2° perpen-
dicular to b*, and less than 5% twinning.
As the x-ray and neutron work on TTF-TCNQ have alternatively
added new features, here we shall not separate the two techniques, but
rather distinguish the high-temperature precursor scattering from the low-
temperature modulated phases.
3.2. High- Temperature Precursor Scattering in TTF- TCNQ
The unmodulated room-temperature structure is stable above the

Peierls transition, which takes place at 54°K.(50) Above this temperature,
one-dimensional precursor effects attributed to the expected giant Kohn

anomaly were barely visible and only briefly described in the earlier pre-
liminary x-ray studies.(52) More detailed inelastic neutron measurements
by Shirane et at.(S3) revealed several features at the wave vector O.295b*.
This wave vector corresponds to the components along the chain direction
of the low-temperature satellite reflections of the modulated phases and is
therefore assigned to 2k F • At room temperature the acoustic branches of
deuterated TTF-TCNQ along the chain direction b* (Figure 18a) do not
exhibit any pronounced anomaly in contrast to the case of KCP. There is
just a shallow but clearly existing anomaly on the longitudinal branch
around the wave vector O.295b*. At lower temperatures this shallow
longitudinal anomaly does not develop. Surprisingly, however, another
anomaly at the same vector was found to grow below 1500 K in the
transverse branch mainly polarized along c *:j: (Figure 18b). This progressive
e
f--ee
6 6
~
E
...,>- 5 5
<r
w
Z
w
z 4 4
0
Z
D
:I:
0..
3 3
2 L 0.2 0.3
21:F
0 0.1 0.2 0.3 0 .4 0.5

~ (b· units)
Figure 18. Dispersion curves from TTF-TCNQ(D) for acoustic modes propagating along
[010], the chain direction, at 295°K. The insert shows the transverse branch around 2kF at
84°K. Note that the Kohn anomaly occurs in the branch mainly polarized along c*.
6 120'K
!>O 6
................. 30
~ ... :/40
------
/ ·' 30
, 60 C: / ....
,
5}1.--</
... ....
2'F
t
2
0.2 0.3 0.1. 0.1 0.2 0.3 0,4
( 0.1+ ~ , 3 ) (O,h S.3)
z6
o
z
'.. .""
84'K •.. 30
...................... ~.~., 40
30·······
1.0 __ _
l COUNTS
fJf!
~O -
,/, // /... ·· 30
o 60 ...... 10 m,n
::t: ;,..--- ...... ,
r2j / . ..
0. 80 - --
//
100 -
50
,.:.::-:..~.:~~~-\ RESOLUTl0(g)
4 60 " / ......... I. :
:.~.. . ... ........
. J
Figure 19. Normalized w, q intensity
contours of the neutron scattering of
TTF-TCNQ from the acoustic excita-
2 L-____L-__~~__~
0.1 0.2 0.3 0.4 tions of wave vector (0, g,O) mainly
(0.1+e,.31 polarized in the c * direction and at
200, 120, and 84°K.
development of the transverse anomaly is clearly visible on the intensity

contours of Figure 19. At lower temperatures the typical scans of Figure 20
show that the intensity has become too low to follow in detail the expected
condensation. The limited but very clear data of this difficult neutron
scattering investigation were fully confirmed by subsequent x-ray
measurements.(54,56) The improved conditions of the x-ray investigation by
Pouget et al.(54) further established unambiguously the one-dimensional
character of the 2kF scattering, and revealed the existence of another
:j: For phonons propagating along the b* direction, a symmetry analysis shows that one can
separate the spectrum into pure longitudinal and transverse modes. But this analysis cannot
fix a particular direction for the polarization of the transverse modes. The experimental
results show only that the 2k p transverse anomaly has a polarization quite close to the c* or
c direction. In the one-dimensional, a c* direction, which is situated in the plane of
symmetry of the molecular chains, seems more likely for the transverse acoustic modulation
of the intermolecular spacing than a c direction (see Figure 16).
one-dimensional precursor at the wave vector 4kF also studied in the x-ray
• .. (56)
counter mvestlgahon.
Figure 21 shows a series of diffuse x-ray patterns from TTF- TCNQ at
different temperatures. As was the case for KCP (Figure 1), diffuse
intensity maxima running along the layer lines perpendicular to b*, and
forming the so-called "satellite sheets," are clearly visible. Yet there are
noticeable differences between the precursor effects observed in KCP and
in TTF-TCNQ.
(i) Two Types of One-Dimensional Scattering are Observed in 7TF-
TCNQ. The first type assigned to 2kF is found at the wave O.295b*; this
scattering mainly develops below IS00K and corresponds to the phonon
LA DEUTERATED TTF - TCNQ

200 OK
600 l
TA (o,{ ,3) 6E '2.5 meV
~
A
:J TA
c:
'E ..c : SCAN
0 400
C\J
-... t
<11 q
I- 2kF
Z
::>
8
200
................. ./
60 OK (40 min )
O L--I.LO-------IL.I-------I.L2-------I,L3-------IL.4------~
(O,{ ,3)
Figure 20, Constant energy scans for de ute rated lTF-TCNQ crystals at 200 and 60°K, Note
that a small maximum is observed at 60 0 K for the 2kF wave vector, consistent with the
contours of Figure 19, but the experimental conditions are too severe to follow the conden-
sation at even lower temperatures, The clear separation of the LA and T A branches at small
wave vectors reflects th e good resolution conditions achieved with the aggregate sample
described in the text.
Figure 21. Diffuse x-ray patterns of TIF-TCNQ in the undistorted high-temperature phase
(T> 54°K). The sample is oriented with b* and c* directions in the equatorial plane, the angle
between the incident x-ray beam and b* is 124°. The rings observed around the incident beam
are powder parasites from the sample holder. Bragg spots with O.5b* components are due to
the A/2 contamination from the continuous spectrum of the x-ray source which is also
reflected by the monochromator. 600 K: The satellite diffuse sheets are clearly visible at the
wave vectors O.295b* identified as 2kF and O.59b* identified as 4k F. 1100 K: The 2kF
scattering has decreased in intensity and is broadened whereas the 4kF scattering remains
sharp. 1500 K- The 2kF scattering is undetectable by eye and only 4kF scattering is visible
(from Reference 54).
anomaly measured by inelastic neutron scattering (Figure 19) which

sharpens in the same temperature range.
The second type of scattering is observed at the wave vector O.59b*
(or 0.41b* in the reduced zone), i.e., at twice the value of the wave vector
of the former scattering, and therefore assigned to 4kF • This second type of
scattering clearly is a genuine precursor, since above 1500 K it is the
predominant scattering, as can be seen from the patterns of Figure 21. The
temperature dependence of the intensity of both types of scattering
deduced from microdensitometer readings by Khanna et ai.(54) is shown in
Figure 22. Comparable x-ray counter measurements(56) are illustrated by
the typical scans of Figure 23.
The excellent agreement between these independent sets of data
clearly establishes the different temperature dependence of the two types
TTF-TCNQ
~1O 1\,
If)
<{
a::
I-
00
a::
<{ 5
o
-0-----£-0
o
-4~
60 100 150 200 250 300
Figure 22. Temperature dependence of the peak intensity of the 2kF and 4kF scatterings
recorded by microdensitometer at the point (-3 , 6, 2.705 , 2.05) and (-3.6,2.41,1.25). The
ordinate is plotted as l i T to eliminate the temperature dependence of the phonon population
factor, which is proportional to kT for low-frequency phonons (from Reference 54).
of one-dimensional scattering. The subsistence of a very weak anomaly at

2kF up to room temperature, visible in Figure 23, very likely corresponds
to the shallow longitudinal anomaly observed by inelastic neutron scatter-
ing. In Figure 23 we can also notice a shift in wave vector from 0.41b*±
0.02 at 600K to about 0.45b* ± 0.02 above 2000K for the 4kF scattering.
4000 5000
lO.2 +?, \)
...;
.... 3500 4500
'"
g
~
'".....z
:::>
..., 3000
0
4000
>-
I-
V"l
Z
W
I-
z 2500
Figure 23 . X-ray counter measurements of

the 2kF and 4kF scattering in the 021 zone 2000~~~-~~-~~
and as a function of temperature (from 0.1 Q2 0.3 0.4 05
Reference 56). '7
(ii) The Diffuse Scattering Appears as Interrupted Lines. This occurs

because in TTF-TCNQ the charge-density waves modulate the distances
between relatively large molecules consisting of atoms that have compar-
able scattering factors; therefore the intensity strongly depends on 01-' In
first approximation it was found that translations of rigid molecules
account for the general features of the intensity dependence upon
01-.(54)
(iii) The Polarization Has Several Components. (54,56) The 2kF scatter-
ing has both a longitudinal and a transverse (mainly directed along c*)
component. The 4kF scattering, not yet investigated by neutron scattering,
is found to have only a longitudinal component. It is at first puzzling that
charge-density waves have a transverse component. Charge-density waves
are produced by the modulation of the intermolecular distances, In KCP,
the modulation involves the Pt(CN)4 groups which are perpendicular to the
chain direction; in this case, as for strings of atoms, only a longitudinal
polarization can modulate the intermolecular spacing and give rise to
charge-density waves. In TTF-TCNQ, the molecular stacking is different
(Figure 17), i.e., the molecules are tilted around the a axis of the unit cell;
therefore there exist two polarization components that can modulate the
intermolecular spacing: the longitudinal component along b, and the
transverse component perpendicular to the axis of rotation a, which coin-
cides with the reciprocal c* axis. The comparison between the two cases is
shown schematically in Figure 24. More precisely, these two polarizations
are mainly observed experimentally in TTF-TCNQ.
The question still remains why only the anomaly in the transverse
branch is observed to grow at low temperatures in the neutron experi-
ments. According to the symmetry operations of the P2d c space group,
the transverse acoustic mode can couple with a longitudinal optic mode.(58)
~K~ I I I , I I I I I lal
Figure 24. Possible polarization of
--t--. I III
"'
c* charge-density waves according to the
Ibl
orientation of extended molecules. In
KCP, the Pt(CN)4 planes are perpen-
dicular to the chain direction (a), so only
longitudinal modulation can give rise to
charge-density waves (b). In lTF-TCNQ,
the almost planar molecules are tilted with
respect to the stacking direction (c); both
longitudinal (d) and c* polarized modula-
tions (e) can given rise to charge-density
waves (from Reference 54).
ITTF - TeNa I
ONSET OF lD ORDER
AT 54K
Figure 25. Microdensltometer scans In the 2kF plane

along the a * direction , showing the onset of three-
dimensional order close to 54°K, and the corresponding
formation of the (3 5, 3.295,3) satellite. Two suc-
cessive satellites of the 2kF plane in the a* direction a*
have been recorded just below the phase transition In
order to provide a scale (from Reference 54)
This could naturally account for the occurrence of the two well-defined
components found by x-ray scattering.:j:
Down to about 60 o K, the diffuse scattering sheets of TTF-TCNQ at
2kF and 4kF wave vectors have a one-dimensional character. Between 60
and 54°K the 2kF diffuse x-ray scattering qualitatively changes, and a
progressive build up of the diffuse sheets into satellite spots corresponding
to a wave vector (0.5a*, 0.295b*, Oc*) can be observed (Figure 25), until
the phase transition temperature of 54°K is reached. Within this tempera-
ture range the 4kF scattering keeps its one-dimensional character.
t Here we discard the very weak 2kF scattering observed at room temperature and which can
correspond to the shallow phonon anomaly found in the longitudinal branch, and we only
deal with the well-defin ed longitudinal component of the 2kF scattering observed at low
temperatures for which no corresponding sharp longitudinal acoustic phonon anomaly has
been measured to date
3.3. The Modulated Phases of TTF- TCNQ
With the condensation of the 2kF scattering at 54°K described above,

there begins a sequence of modulated phases first characterized from
elastic neutron scattering measurements by Comes et al. (50) and
subsequently fully confirmed by x-ray investigations. (54.56) Three successive
modulated phases are observed in TTF-TCNQ, with phase transitions at
54°K (the Peierls transition), 49°K, and 38°K as shown in the temperature
dependence of the 2kF satellite intensity of Figure 26. An additional phase
transition around 46°K, suggested to be related to the condensation of the
4kF scattering, has also recently been c1aimed(59\ structural results are,
however, still inconclusive in this respect.
3.3.1. Modulated Phase J, 49°K < T < 54°K
Modulated phase I is stable between 54 and 49°K with satellite

reflections observed at the wave vectors (0 .5a *, 0.295b*, Oc*), which
TTF - TCNQ NO.7
COUNTING TIME
° «(,1.295,3) 2min
• (1+!,2.705,O) 20min
1500
0.25 s:! :S0.50
o
if) °o
f-
3
o
1000
u
500
o 70
Figure 26. Temperature dependence of the intensity of two different satellites measured by
inelastic neutron scattering. Note that the (~, 1.295,3) satellite, which has an important c*
component , persists up to 54°K, while the (1 + ~, 2.705, 0) satellite, which has a b* component
but no c* component, extrapolates to zero intensity around 49°K (from Reference 50).
corresponds to a modulated lattice with a superstructure 2a x 3.40b xc.

Satellites of this phase are extremely weak, so only one well-defined
satellite could be followed up to 54°K quantitatively by neutron measure-
ments: namely, the satellite with components (0.5a*, 1.295b*, 3c*), which
is close to the strongest (013) reflection of the unmodulated lattice; its
temperature-dependent intensity is shown in Figure 26. Photographic x-
ray patterns by Khanna et al.(54) exhibit a series of such weak but well-
defined satellites. An estimation of the intensity of these superlattice
reflections relative to the closest Bragg peaks gives a figure of the order of
10- 5 at SocK. No satellites are observed in this temperature range with a
4kF wave vector component.
Noticeable is the fact that all satellites between 54 and 49°K have a
nonzero scattering vector component along c*; no satellite reflection was
found in the (hkO) scattering plane. since the intensity of this type of
reflection (condensed phonons) is proportional to (Q . e), where e is the
polarization of the atomic displacements [see relation (7) above], this
observation, as suggested by Comes,(60) shows that only the transverse
component along c* from the higher-temperature one-dimensional pre-
cursor scattering (T> S4°K) is condensed in this phase. This is again
consistent with the observation of a transverse (c*) phonon anomaly and
with the subsistence below S4°K of relatively intense one-dimensional
scattering with the 2kF wave vector component, possibly due to the
uncondensed longitudinal component.
3.3.2. Modulated Phase II, 38°K < T < 49°K
(i) The 2kF Satellites. As clearly seen in Figure 26 and confirmed by

additional measurements,(61) below 49°K, 2kF satellites with zero scatter-
ing-vector component along c* become observable, implying that the
longitudinal component of the 2kF scattering condenses around 49°K.
There are no available inelastic neutron data close to this temperature, but
this suggests that a longitudinal Kohn anomaly has developed and finally
condensed at 49°K.
The 49°K phase transition is in fact more subtle and unique than just
described. At 49°K the wave vector component along a* of the condensed
2kF satellites starts to decrease continuously from 0.5a* towards about
0.3a* and then locks discontinuously at the 0.25a* value at 38°K. In terms
of modulations in real space, the modulation changes from its simply
incommensurable (along b) superstructure 2a x 3.40b x c above 49°K, to a
temperature-dependent doubly incommensurate (along a and b) super-
structure x(T)a x 3.40b x c, before it locks in a simply commensurate
superstructure again at 38°K with the modulation 4a x 3.40b x c. The 49°K
phase transition was in fact only discovered after a reanalysis by Bak and
0.06
TTF-TeNQ
q = ( 1/2 :!:. 8)
a
0.05 o SAMPLE E9
• SAMPLE N6
0.04 0.30
::s
N
<0 0.03~ ..:
0.35 *
0
0.02~ I
001 t- 0040
0.0f- 0.50
36 40 44 48 52 56
T (K)
Figure 27 The occurrence of three phase transitions in TTF-TCNQ is best visualized from
the temperature dependence of the satellite Bragg peak position in reciprocal space when
plotted as a function of 8 2 = d-qa)2, where qa is the satellite wave vector component along
a* (from Reference 55). This was first suggested by Bak and Emery (Reference 62).
Emery(62) of the initial data and fully determined by the high-resolution

measurements of Ellenson et al.(55) (Figure 27).
(ii) The 4kF Satellites. Satellites with 4kF wave vector component
along b* were observed on x-ray patterns as well as by neutron measure-
ments, (54.55) but the temperature dependence of the intensity and position
of these satellites was only investigated in detail in the x-ray counter work
by Kagoshima et al. (56) This work concludes that the 4kF satellites are
clearly visible at low temperature at wave vectors that are always twice
those of the 2kF satellites. Their intensity extrapolates to zero somewhere
above 45°K (Figure 28). Their weaker intensity as well as their position in
this temperature range are consistent with a higher-order effect. The
relative and absolute intensities of the 2kF and 4kF satellites suggest a
(/)
w !:::
«> :>
z (~, 2.41,1) a
:i:
w *
!::: z'" 0 .5
1,· 1
0.4
~,~
...J
0 .3
w 0a::
...J
t- t- 0.2
« u 0.1
'\
Figure 28. Temperature dependences of (a) (/)
w 0
the 4kF satellite positions along the a* axis > b
5000 ·
and (b) intensities of the corresponding -:
satellites found at 2.41 b* in the (021) zone.
u
W
The index {, represents the wave vector of >-
t-
(/)
o
the lattice modulation along the a * axis in (/)
~ 2500 ·
2
units of a* (after Reference 56). The open w (/)
t- t-
circles correspond to satellites observed in 2
~ :>
the other investigations; the full circles to the 0
g 020
satellite only revealed by the x-ray counter 30 40 50
study (Reference :;6). TEMPERATURE (K)
harmonic of 2kF indicating that below 45°K the modulation along the
chain direction is probably not strictly sinusoidal.
According to this behavior the 4kF satellites do not modify the modu-
lated lattices deduced from the 2kF ones. What is to date not clearly
established is whether these satellites (which correspond to a longitudinal
polarization) condense simultaneously with the longitudinal component of
the 2kF satellites at 49°K, or whether they condense at a lower tempera-
ture (46°K?) as suggested from recent specific heat measurements(S9) and
theoretical work by Barisic. (63) This would of course add a new phase
transition to this already rich sequence. Above 45°K the 4kF satellites
become so weak that no clear answer can be given at present. A second
type of 4kF satellite with a wave vector of Oa*, O,59b*, Oc* was also
reported from the x-ray counter investigation (Figure 28); these satellites
were observed neither in the photographic x-ray studies nor in the neutron
measurements.
3,3.3. Modulated Phase III, T< 38°K
At 38°K, a first-order phase transition with a hysteresis of about 1°

takes place and the superlattice locks in its 4a x 3.40b x c modulation and
no further structural change is observed below this temperature.(SO,64) The
intensity of the 2kF satellite reflections relative to the closest Bragg peaks
of the main lattice has reached the order of 10- 4 in this phase: from this
value one can estimate the distortion amplitude to be about 1% of the
lattice spacing. This amplitude is comparable with the amplitude of the
charge-density waves in KCP.
3.4. Spin Waves in ITF-TCNQ?
There have been several suggestions to account for the simultaneous

occurrence of both 2kF and 4kF scattering in TTF-TCNQ. Some of these
models do not imply particular interactions between electrons, such as for
instance the model developed by Weger et al. (65) Their description assumes
libration of molecules rather than simple translations with, however, a
large polarization component along the a direction. This seems inconsis-
tent with the presently available structural data, which determined the
main components of the polarization of the charge-density waves to be
directed along band C*.(54) As initially suggested by Torrance,(66) the first
model worked out by Emery(67) after the experimental observation, and
most other attempts, involve repUlsive Coulomb interactions between
electrons and strong coupling on either one or both species of molecular
stacks. (67)
Easiest to describe is the schematic case of the strong coupling
limit. (68) Let us consider already localized electrons along the chains as in
the low-temperature modulated phases. For a ~-fiIled band, the normal
situation with two electrons of opposite spins per site gives rise to a charge-
density wave period of 4b = 2tr/2k F (Figure 29a). In the strong coupling
limit, only one electron can be accommodated on each site, producing then
a charge-density wave period of 2b = 2tr/4k F (Figure 29b) as required to
explain the 4kF scattering. Figure 29b also shows the existence of spin
waves with a period 4b = 2tr/2k F identical to the period of the normal 2kF
. (69)
scattenng.
This brings the discussion on the inelastic neutron scattering investi-
gation carried out by Mook et al. (57) These authors were the first to report a
giant Kohn anomaly extending down to 1.2 meV and observable at room
temperature in the longitudinal acoustic branch along the chain direction.
As already described above, the subsequent inelastic neutron measure-
211'
- 2kF--+
0) 11 I I I 11 I
1l
b) t r
---
J
I
21T
4kF
Figure 29. Schematic representation of the localized electrons in the low-temperature

insulating phase for a i-filled conduction band in the high-temperature conducting phase. (a)
Each site can accommodate two electrons of opposite spins: the lattice distortion due to the
charge-density wave is 21T/2k F ; (b) each site can only accommodate one electron; the lattice
distortion is 21T/4k F• but there is also a spin-density wave of 21T/2k F (from Torrance, in
Reference 67).
ments by Shirane et al.(53) on de ute rated ITF-TCNQ revealed no strong

anomaly in the longitudinal acoustic branch but only a gentle curvature
near 2k F . To the contrary, in this later study, the Kohn anomaly was found
to develop only at low temperatures, on the transverse (c*) acoustic
branch. Because of these results, it has been conjectured that the samples
used in the two experiments were fundamentally different. It was also
suggested(70) that the soft mode was not in an acoustic branch but was a
particular C- H (or C- D) stretching vibration and that the differences in
the two experiments arose from the different structure factor due to the
opposite signs of the scattering lengths of Hand D. The validity of these
arguments has been recently questioned.(7J) Torrance et al.(69) have also
suggested that the giant anomaly reported by Mook et al. at room
temperature was associated with a spin-density wave rather than with a
charge-density wave (see the discussion above and Figure 29).
Repeated neutron inelastic scattering measurements, carried out
jointly at Brookhaven and at Oak Ridge on both protonated and deu-
terated samples,(72) have unambiguously shown that no distinct anomaly at
q = 2kF and room temperature can be observed in the longitudinal acoustic
branch, thus confirming the dispersion shown in Figure 18. The possibility
of spin-density waves remains an open question requiring further study;
such experiments for small spins distributed over large molecules are
extremely difficult.
3.5. The Interpretation of the Sequence of Modulated Phases in TTF-

TCNQ
Based on some experiment work,(73) the models developed in order to

understand the sequence of successive phases in TTF-TCNQ generally
assume distinct and successive distortion of the two types of molecular
stacks. Bak and Emery(62) showed that a Ginzburg-Landau theory using
the proper order parameters (see Figure 27) and Coulomb interaction
arguments between parallel chains then accounts for the successive phases.
The Peierls transition on one type of molecular stack (generally thought to
be the highly conducting TCNQ stack) occurs at 54°K, whereas the tran-
sition on the second molecular species (assumed to be the TTF stack) takes
place at 49°K and then drives the transverse modulation along a from 2a to
larger values until it locks at 4a at 38°K.
The charge-density waves in the successive phases of ITF-TCNQ can
be visualized by the schematic diagrams shown in Figure 30. In the models
developed, respectively, by Bak and by Bjelis et al.,(74) at 49°K the charge-
density waves start sliding in the b direction in a uniform way relative to
each other. It is this displacement that shows up as a continuous change of
the a * component of the satellite reflections. As the temperature
y,b 54K> T>49K T - 48 K
Figure 30. Schematic diagram of phase transition in TTF-TCNQ. The full curves indicate the
qb modulation along the TCNQ chains. The dotted curves indicate the 2qb (4k F ) modulation
below 38°K (from Reference 74).
0.06
I TTF-TCNQ
I
I q
a
= (112 !8)
I
0.05 ,+. o SAMPLE E9
I
I • SAMPLE N6 ~
I /T3 c:
0.04 0.30
.,u"
0
":.0 0.03
0
u
0.35 0.
0.02 ·u
~
TI
*
0
0.01 0.40
+
0.00 0.50
36 40 44 48 52 56
T(K)
Figure 31. The unique kind of hysteresis observed in TTF-TCNQ between 38 and 49°K
(from Reference 'is).
approaches 38°K, neighboring planes are approximately 90° out of phase.

Again the charge-density waves slide with respect to each other along b,
this time discontinuously, so that below 38°K the chains are pairwise in
phase and neighboring pairs 180° out of phase.
While this gives a relatively complete picture for the sequence of
modulated phases, further theoretical treatment is needed in order to
account for the unusual kind of continuous hysteresis reported by Ellenson
et al.(SS) and shown in Figure 31.
In spite of the good agreement of such models with the available data,
they are still lacking full support from a structure determination of the
modulated phases, and other theories can be worked out with simultaneous
condensations on both molecular speciesYS) The successive condensation
of a transverse (c*) anomaly at 54°K, and another longitudinal anomaly at
49°K strongly suggested by the structural data, is not in itself in contradic-
tion with the two chain models; but it also provides other possibilities.
3.6. Study of Other Organic One-Dimensional Conductors
Besides conventional structure determinations, only limited diffuse

x-ray work has been reported on other organic one-dimensional conduc-
tors. It is, however, important regarding the observation of both 2kF and
4kF scattering in TTF-TCNQ to find out if other systems display such
double one-dimensional precursor scattering.
The compounds in Table 2 revealed only a single kind of one-dimen-
sional scattering.
Only TSeF-TCNQ was studied at low temperatures, with three-
dimensional ordering of the charge-density waves reported below 29°K
toward a modulated lattice 2a x 3.15b x nc, where n was not deter-
mined.(76)
Figure 32 shows a diffuse scattering pattern of TSeF-TCNQ obtained
by Megtert et al.(77) compared to TTF-TCNQ, the heavier selenium atoms
(high x-ray scattering factor) produce extremely well-contrasted one-
dimensional scattering.
Table 2. Compounds Showing Only a Single Kind of One-

Dimensional Scattering
Charge
Compound Measured 2kF transfer Reference
TseF-TCNQ (lOOOK) 0.315 0.63 76, 77

HMTSeF-TCNQ (300 0 K) 0.37 0.74 76
HMTTF-TCNQ (lOOOK) 0.36±O.01 0.72 78
--- -~
Figure 32 X-ray diffuse scattenng pattern from TseF-TCNQ at 300 K showmg the well-
defined 2kF one-dimensIOnal satellite sheet scattering at the wave vector O.31Sb* (from
Reference 77).
Other investigations by x-ray scattering on the alloy TTF (So 97SeO 03)-
TCNQ showed that both 2kF and 4kF scattering were still present, with a
sequence of phase transitions similar to that of TTF-TCNQ but at some-
what lower temperatures. (79)
4. Concluding Remarks
In the preceding sections we have only considered real one-dimen-

sional conductor~, in which furthermore the structural investigations have
revealed charge-density waves. There are, however, also three-dimensional
crystals with distinct metallic chains in several crystallographic directions
that have already been approached in a one-dimensional approximation.
This was done, for instance, for the well-known A15 compounds such as
V 3 Si or Nb 3 Sn. (80.18)
A recently discovered mercury complex displaying one-dimensional-
electrical properties has a structure bearing some resemblance to the A15
compounds.
Hg 288 AsF 6 , as shown by Cutford et al.,<81) consists of metallic bonded
polymercury cations in a tetragonal lattice consisting of an array of
r
(AsF 6 l anions. The mercury chains are situated in nonintersecting chan-
nels along a and b axes of the tetragonal lattice and there are no chains
along the c axis. These authors also reported one-dimensional diffuse x-ray
scattering. This one-dimensional scattering (Figure 33) is very different
from the "satellite sheet" type of scattering described above for KCP or
X-Ray and Neutron Scattertng 63
TTF-TCNQ Its very hIgh mtensIty alone rules out a dIsplacIve orIgm of
the usual type, and rather suggests that the mercury chams behave as
entItIes that are mdependent of the host AsF6 crystallIne network, except
for the locatIOn of the mercury channels
The repeat dIstance along the mercury chams deduced by Brown et
at (82) from the one-dImensIOnal dIffuse scatterIng (aHg = 2 64 A) IS mdeed
mcommensurate WIth the AsF 6 lattIce constant (a = b = 7 54 A) At room
temperature there IS no coherence between the dIfferent mercury chams
The streaks of FIgure 33 are therefore one-dImensIOnal Bragg peaks
ThIs compound has been studIed 10 great detaIl, (83) mcludmg neutron
scatterIng by Hastmgs et at , (84) as a functIon of temperature Very mteres-
tmg features mvolvmg dIfferent and succeSSIve couplIng of the mercury
chams are observed at low temperatures, but so far no clear eVIdence of a
gIant Kohn anomaly was found ThIS IS somewhat surprIsmg, smce such
mercury chams are, at least at room temperature, the closest real example
of a one-dImensIOnal metal, unless the three-dImensIOnal couplIng
between the mercury chams occurs at a temperature that IS too hIgh and
prevents the development of a 2k p phonon anomaly
The clear conclusIOn from these studIes IS that further work IS reqUIred
for a full understandmg of the many new structural aspects revealed by the
one-dImenSIOnal conductors, and the relatIOn to the overall propertIes of
such materIals
Figure 33 X-ray diffuse scatterIng pattern from Hg 286 AsF6 shoWIng the one-dimensIOnal
Bragg reflectIOns of the Independent mercury chaIns of this compound (from J P Pouget and
S Khanna unpUblished)
64 R. Comes and G. Sh"ane
ACKNOWLEDGMENTS
We are grateful to all our colleagues from Brookhaven National

Laboratory (1. D. Axe, P. Bak, W. D. Ellenson, V. J. Emery,
S. M. ShapIro), from the UnIversity of Paris-Orsay (F. Denoyer, M.
Lambert, J. P Pouget), and the University of Pennsylvania (A. F. Garito,
A J. Heeger, S. K. Khanna), who participated in many discussions and
mvestigatlOns on the structural aspects of the one-dimensional conductors.
Work at Brookhaven was performed under the auspices of the U.S. Energy
Research and Development Admmistration.
References and Notes
R Comes, M Lambert, H LaunOls, and H R Zeller, Phys Rev B 8, 571 (1973)

2 K Krogmann, Angew Chem Int Ed Engl 8,35 (1969)
3 A F Ganto and A J Heeger, Acc Chem Res 7,232 (1974)
4 P W Ander~on FlZlka D,elecmkov, G I Skanavi (ed ), Acad Nauk SSSR, Moscow
(1960), W Cochran, Phys Rev Lett 3,412 (1959), Adv Phys 9,387 (1960), Adv Phys
10,401 (1961)
5 W Cochran and A Zla, Phys Status Solidi 25, 273 (1968)
6 For reviews see G Shlrane, Rev Mod Phys 46, 437 (1974), J D Axe, Trans ACA,
Feb 1 (1971) Proc Conf Neutron Scattenng, Gatlmburg (1976) (ORNL CONF
760601 PI) B Dorner and R Comes, m DynamiCs of Solids and LiqUIds by Neutron
Scattering, T Spnnger (ed ), (Spnnger, BerlIn (1977»
7 J Harada and G Hon]o J Phys Soc Jpn 22,45 (1967), R Comes, M Lambert, and A
GUlmer, C R Acad SCI Pans 266, 959 (1968), F Denoyer, R Comes, and M Lambert,
Sohd State Commun 8, 1979 (1970), R Currat, R Comes, B Dorner, and E Wlesen-
danger, J Phys C 7 2~2 (1974)
8 Y Yamada G '>hlrane and A Lmz, Phys Rev 177,848 (1969), K Gesl, J DAve, G
Shlrane, and A Lmz Phys Rev B 5 1933 (1972)
9 T Rlste, E J 'lamuelsen, and K Otnes, Solid State Commun 9, 1455 (1971), S M
Shapiro, J D Axe, (J Shnane, and T Rlste, Phys Rev B 6, 4332 (1972)
10 Y Yamada, I Shlbuya dnd S Hoshlno, J Phys Soc Jpn 18,1594 (1963), S Hoshmo
and H Motegl Jpn J App/ Phys 6, 708 (1967), G DollIng, ] Sakurai, and R A
Cowley, Proc 2nd IMF 1969, J Phys Soc Jpn, Suppl 28,258 (1970)
11 Y ShlOzakl Ferroelectncs 2, 245 (1971), D R McKenZie, J Phys C 8,1607 (1975)
12 M Izumi] D Axe G Shlrane, and K Shlmaoka, Phys Rev B 15, 4392 (1977)
13 For reVlew~, see Lecture Notes In PhYSICS, Vol 34, One DimensIOnal Conductors,
Spnnger, New York (1975), Low DimensIOnal Cooperaflve Phenomena, H J Keller (ed ),
Plenum Pre", New York (1975), ChemIStry and PhYSICS of One-DimenSIOnal Metals, H J
Keller (ed) Plenum Press, New York (1977), Proc Conf Orgamc Conductors and
Semiconductors SlOfok Hungary (1976), m Lecture Notes In PhYSICS, Vol 65, Orgamc
Conductors and Semiconductors, Spnnger, New York (1977), L N Bulaevskll, Usp Flz
Nauk 115 261 (197~) A ] BerlInsky, Contemp Phys 17,331 (1976)
14 See other chapter~ of thl' book
15 W Kohn Phys Rev Lett 2, 193 (1959)
16 A M Afanase\ and Y Kagan, Sov Phys JETP 16,1030 (1963)
X-Ray and Neutron Scattermg 65
17 R E Peleris, Quantum Theory of Solids, Clarendon, Oxford (1964), H Frohch, Proc R

Soc London A 223, 296 (1954)
18 S Banslc, Phys Rev B 5, 941 (1972), Ann Phys (Pans) 7,23 (1972), L Gorkov, m
Col/ectlVe Properttes of Physical Systems, Band S Lundquist (eds ), Academic Press, New
York (1973)
19 B HOrowitz, H Gutfreund, and M Weger, Phys Rev B 9,1246 (1974)
20 S Banslc, A BJehs, and K Saub, Solid State Commun 13,1119 (1973), W Dletench,
Solid State Commun 17,445 (1975), W Dletench, Advan Phys 25,615 (1976)
21 M J Cooper and R Nathans, Acta Crystallogr 23, 357 (1967)
22 B N Brockhouse, Inelastic Scattering of Neutrons In Solids and LIqUids, p 113, Inter-
natIOnal AtomiC Energy Agency, Vienna (1961)
23 W Knop, Ann Phys (LeIpZIg) 43,111 (1842)
24 H J Delseroth and H Schultz Phys Rev Lett 33,963 (1974)
25 J M Wilhams, L Jeffrey, and J L Petersen, Phys Rev Lett 33, 1079 (1974)
26 G Heger, B Renker, H J Delseroth, and H Schultz, Mat Res Bull 10,217 (1975)
27 H Nledoba (to be published)
28 R Comes, M Lambert, and H R Zeller, Phys Status SolidI 58,587 (1973), R Comes,
In Lecture Notes In PhYSICS, Vol 34, P 32, Spnnger, New York (1975)
29 B Renker, H Rletschel L PmtschovlUs, W Glaaser, P Bruesch, D Kuse, and M J
Rlce,Phys Rev Lett 30 1144(1973)
30 B Renker, L PmtschovlUs, W Glaser, H Rletschel, R Comes, L Liebert, and W
Drexel, Phys Rev Lett 32, 836 (1974), B Renker, L PmtschovlUs, W Glaser, H
Rletschel, and R Come~ Lecture Notes In PhYSICS, Vol 34, P 53, Spnnger, New York
(1975), B Renker and R Comes, m Low DimensIOnal CooperatIVe Phenomena, H J
Keller (ed), Plenum Pres, New York (1975)
31 R Comes, B Renker I PmtschovlUs, R Currat, W Glaser, and G Scheiber, Phys
Status Solidi 71, 171 (1975), L PmtschovlUs, B Renker, and R Comes, Proc Conf
Neutron Scattenng, GatlInburg Oak Ridge NatIOnal Laboratory, Conf 760601-Pl
(1976)
32 J W Lynn, M Izumi G Shlrane S A Werner, and R B Saillant, Phys Rev B 12,
1154 (1975)
33 C F Eagens S A Werner and R B Saillant Phys Rev B 12, 2036 (1975)
34 See L N Bulaevskll m Reference 13, L J Sham and B R Patton, Phys Rev Lett 36,733
(1976), D Hone, P A Montano, T Tonewaga, and Y Imry, Phys Rev, B 12, 5141
(1975), P N Sen dnd C M Varma, Soild State Commun 15,1905 (1974)
35 D Kuse and H R Zeller Phys Rev Lett 27,1060 (1971), M J Rice and J Bernascom,
] Phys F 3 55 (1973), H LaunOls and H Nledoba, m Low DimensIOnal CooperatIVe
Phenomena, H J Keller (ed), Plenum, New York (1975)
36 J C Phillips, Phys Status Solidi (b) 78, 259 (1976)
37 J C PhilliPS, Phys Status Solidi (b) 78, 371, 471 (1976),79,111 (1977)
38 K Carneiro, G Shlrane S A Werner, and S Kaiser, Phys Rev B 13,4258 (1958)
39 P A Lee, T M Rice, and P W Anderson, Phys Rev Lett 31,462 (1973)
40 P A Lee, T M Rice, and P W Anderson, Solid State Commun 14,703 (1974)
41 P Bruesch and H R Zeller, Solid State Commun 14, 1037 (1974)
42 E F Stelgmeyer R Loudon G Harbeke H Auderset, and G Scheiber, Solid State
Commun 17,1447 (197")
43 K Carneiro, J Eckert, and G Shlrane, Soild State Commun 20,333 (1976)
44 A J Schultz, G D Stucky J M Williams, T R Koch, and R L Mafly, Soild State
Commun 21, 197 (1977)
45 A Bertmottl, C Bertmottl, and G Jehanno, C R Acad SCI Pans 278B, 45 (1974)
46 J Y DubOIS The~e ,erne Cycle Umverslte Pans-Sud 91405, Orsay, France (1975)
66 R Comes and G Shlrane
47 T 1 Klstenmacher T E Phillips, and D 0 Cowan, Acta Crystallogr Sect B 30, 763

(1974)
48 R H Blessing and P Coppens, Solid State Commun 15,215 (1974), A 1 Schultz, G D
Stucky, R Craven, M 1 Schaffman, and M B Salamon, J Am Chern Soc 98,5191
(1976), A 1 Schultz and G D Stucky,J Am Chern Soc 98,3194(1976)
49 P Coppens, Phys Rev Lett 35,98 (1975)
50 R Comes, S M Shapiro, G Shlrane, A F Ganto, and A 1 Heeger, Phys Rev Lett 35,
1518 (1975), Phys Rev B 14, 2376 (1976)
51 1 P Ferrans, D 0 Cowan, V V Walatka, and 1 H Perlstein, J Am Chern Soc 95,
948 (1973), L B Coleman, M 1 Cohen, D 1 Sandman, F 1 Yamaglshl, A F Ganto,
and A 1 Heeger, Solid State Commun 12, 1125 (1973)
52 F Denoyer, R Comes, A F Ganto, and A 1 Heeger, Phys Rev Lett 35,445 (1975), S
Kagoshlma, H Anzal, K Kajlmura, and T !shlguro, J Phys Soc Jpn 39, 1143 (1975)
53 G Shlrane, S M Shapiro, R Comes, A F Ganto, and A J Heeger, Phys Rev B 14,
2325 (1976), S M Shapiro, G Shlrane, A F Ganto, and A 1 Heeger, Phys Rev B 15,
2413 (1977)
54 1 P Pouget, S K Khanna, F Denoyer, R Comes, A F Ganto, and A 1 Heeger, Phys
Rev Lett 37 437 (1976), S K Khanna, 1 P Pouget, R Comes, A F Ganto, and A 1
Heeger, Phys Rev B 16, 1468 (1977)
55 W D Ellenson, R Comes, S M Shapiro, G Shlrane, A F Ganto, and A 1 Heeger,
Solid State Commun 20,53 (1976), W D Ellenson, S M Shapiro, F Shlrane, and A F
Ganto, Phys Ret B 16, 3244 (1977)
56 S Kagoshlma, T !shlguro, and H Anzal, J Phys Soc Jpn 41,2061 (1976)
57 H A Mook and C R Watson, Phys Rev Lett 36, 801 (1976)
58 P Bdk and V 1 Emery (unpublished)
59 D Djurek, K Franulovlc, M Prestner, S TomlC, L Glral, and 1 M Fabre, Phys Rev
Lett 38, 715 (1977)
60 R Comes, In ChemlSlry and PhYSIC.1 of One-DimensIOnal Metals, H 1 Keller (ed ),
Plenum, New York (1977)
61 1 P Pouget, ~ M Shdplro, G Shlrane, A F Ganto, and A J Heeger (to be published)
62 P Bak and V J Emery, Phys Rev Lell 36,978 (1976), T D Schultz and S Etemad,
Phys Rev B 13 4928 (1976) A sImIlar suggestIOn was bnefly mentIOned In earlier
papers K Saub, <; Banslc and 1 Fnedel, Phys Lett 56A, 302 (1976), B HorOWitz and
D Mukamel Solid State Commun 23 285 (1977)
63 S Banslc (to be published)
64 The first expliCit ,uggestlOn of a phase transition at 38 K was made by D Jerome, W
Muller, and M Weger, J Phys (Pam) Lett 35, L77 (1974).
65 M Weger and J Fnedel J Phys (Pans) 38,241 (1977)
66 1 B Torrance, pnvate commumcatlOn to the authors (1an 1976)
67 V J Emery, Phys Rev Lett 37, 1227 (1976), J B Torrance, In ChemIStry and PhYSICS of
One-DimensIOnal Metals, H J Keller (ed), Plenum, New York (1977), P A Lee, T M
Rice, and R A Klemm Phys Rev B 15,2984 (1977), 1 Kondo and K Yama]l, J Phys
Soc Jpn 43,424 (1977), J Hubbard, Phys Rev B 17,494 (1978)
68 A A OvchInmkov, Soc Phys JETP 37 176 (1973), J Bernascom, M J Rice, W R
Schneider and <; Strassler Phys Rev B 12 1090 (1975)
69 J B Torrance, Phys Rev B 17, 1099 (1978)
70 K Carneiro, Phys Rev Lett 37, 1227 (1976)
71 M 1 Rice dnd N 0 Lipan, Phys Rev Lett 38,437 (1977)
72 H A Mook, G Shlrane and S M Shapiro (to be published)
73 Y TomkieWICZ A R Taranko, and 1 B Torrance, Phys Rev Lett 36,751 (1976), E
Rybaczewskl S Smith A F Ganto A J Heeger, and B Silbernagel, Phys Rev B 14,
2746 (1976)
74 P Bak, Phys Rev Lett 37, 1071 (1976), A BJehs and S Banslc, Phys Rev Lett 37,
1517 (1976), T D Schultz and S Etemad, Phys Rev B 13,4928 (1976)
75 E Abrahams, J Solyom and F Woynarovlch (to be pubhshed)
76 C Weyl, E H Engler K Bechgard, G lehanno, and S Etemad, Solid State Cornrnun
19,925 (1976)
77 S Megtert, A F Ganto and R Comes (to be pubhshed)
78 S Megtert, R Comes, A F Ganto, K Bechgard,l M Fabre, and L Glral, ] Phys Lett
(Pam) 39, L118 (1978)
79 S Kagoshlma, T Ishlguro E M Engler, T D Schultz, Y TomkiewIcz (to be published),
K Smger, T Wei, A F Ganto, and 1 P Pouget (to be pubhshed)
80 J Labbeandl Fnedel ] Phys (Pans) 27, 153,303(1966)
81 B D Cutford, W R Datar, R 1 Gillespie, and A Van Schyndel, Adv Chern Ser 150
(1976)
82 I D Brown, B D Cutford, C G DaVies R 1 Gillespie, P R Ireland, and 1 E Vekns,
Can] Chern 52,791 (1974)
83 C K Chiang, R Spal A Denenstem, A 1 Heeger, N D Mlro, and A G MacDiarmid,
Solid State Cornrnun 22 293 (1977), see also Phys Rev B 15,4607 (1977), T Wei, A
F Ganto, C K Chiang N D Mlro and A G MacDlarOid (to be published)
84 J M Hastmgs,l P Pouge! G Shlrane A 1 Heeger N D Mira, and A G MacDiarmid,
Phys Rev Lett 39 14R4 (1977)
3
Charge-Density Wave Phenomena in
One-Dimensional Metals: TTF-TCNQ
and Related Organic Conductors
A. f. Heeger
1. Introduction
Considerable effort has been devoted toward understanding the one-

dimensional metallic behavior of tetrathiafulvalene-tetracyano-
quinodimethane (TTF-TCNQ).(1) There are two overlapping aspects of this
interest: the study of TTF- TCNQ as a prototype one-dimensional metallic
system, and the study of TTF-TCNQ as an example of a metallic system
made up of organic molecules with molecular properties that might be
changed and improved by design and synthesis. (2) In this chapter we will
focus primarily on TTF- TCNQ as a one-dimensional system; the molecular
properties will be brought in as a means of understanding, on a microscopic
basis, the origin of the competing intermolecular and intramolecular inter-
actions that lead ultimately to the metallic solid.
Although it has long been known that a one-dimensional metal is
intrinsically unstable to the formation of charge- or spin-density waves, (3-5)
the study of real physical metallic systems that are, in effect, truly one-
dimensional in electronic structure has emerged only in the past few
years.o) The chainlike stacking Of(6,7) planar molecules in the TCNQ salts
with the implied 1T-electron overlap was suggestive of highly anisotropic
properties and, in a rough sense, of quasi-one-dimensionality. However, it
is important to recognize the difference between "anisotropy" and "one-
dimensionality" since clearly all experimental chain like systems form in
A 1. Heeger • Department of Physics and Laboratory for Research on the Structure of

Matter, University of Pennsylvama, Philadelphia. Pennsylvania 19104.
69
70 A. 1. Heeger
three-dimensional structures with, perhaps, weak coupling between

individual chains. Understanding of the reality of the one-dimensional
limit came with quantitative measurements of the anisotropy of the opti-
cal(8-19) and transport(15-17) properties of one-dimensional conductors. The
observation in polarized optical reflectance of a plasma edge for E parallel
to the conducting axis and a flat frequency-independent reflectivity
through the far ir for E perpendicular to the conducting axis implies that
even though the conductivity perpendicular to the chain is finite, this
transverse transport does not arise from metallic behavior; i.e., w ~7"1I > 1
and w ~7" ~ < 1, where Wp and 7" are the plasma frequency and scattering
time. Moreover, the measurements of the magnitude of the transverse
components(15-17) of the conductivity tensor yield an estimated mean free
path I~ of order 10- 3 of a lattice constant, whereas in the parallel direction
the mean free path as estimated from the dc or the optical conductivity
near the plasma edge implies metallic propagation. Thus III> all and I~ « a~,
where all and a~ are the lattice constants parallel to and perpendicular to
the principal conducting axis. The implication is that the interaction of the
electrons with the lattice can be sufficiently strong to localize electrons on
individual chains with diffusive hopping(16.17) between chains. This local-
ization converts a metal with an anisotropic band structure into an array of
one-dimensional metallic chainsY S )
In this sense TTF-TCNQ, KCP, and a growing class of related
compounds are one-dimensional metals and can be expected to exhibit the
unique phenomena associated with the mathematical one-dimensional
instabilities. On the other hand, poly(sulfur-nitride), (SN)x, although
chainlike in structure,(J9,20) has large enough interchain coupling to prevent
the one-dimensional localization. As a result, the optical properties,(21-25)
electron energy loss results,(26) and ultimately the absence of a Peierls
transition(27-29) indicate that (SN)x is not a one-dimensional metal, but an
anisotropic three-dimensional metal. Reaction with halogens appears to
further separate the (SN)x chains reducing the interchain coupling leading
to more nearly one-dimensional behavior of (SNBro.4)xYO)
We will focus in this chapter on the charge-density wave phenomena
associated with the Peierls instability in the TTF-TCNQ system bringing in
experimental results from related compounds where available and appro-
priate. TTF-TCNQ is a one-dimensional metal at high temperatures that
undergoes a metal-insulator transition(,l) at 54°K to a high-dielectric-
constant(17,32,33) nonmagnetic(31,34) ground state. The crystal and molecular
structures are shown in Figure 1.(6) Recent x-ray scattering(35,36) and elastic
neutron scattering(37) studies demonstrated the existence below 54°K of a
low-temperature incommensurate superlattice having a periodicity of 3.4b
in the chain direction. In the conducting state above 54°K the x-ray results
showed one-dimensional diffuse scattering(35,36,38-40) consistent with a one-
Charge-Density Wave Phenomena 71
~ H\ l I;~
c\ I
c =c\ I
C
c=c C= C
0'
/ t =c!
I \
'c\~
N H H N
TCNQ TTF
Crystal data:
a = 12.298 A
b =3.819 A
~i-o
c = 18.468 A
{3 = 104.46"
V = 839.9A 3
Z= 2
!~
d = 1.62 g/cm 3
Figure 1. Molecular constituents and parallel cham structure of TIF-TCNQ (Reference 7).
x corresponds to a. y to b. and z to c. The c* direction is normal to a and b.
dimensional lattice distortion or a phonon anomaly. Inelastic neutron

scattering studies(41) of the phonon spectrum revealed a Kohn anomaly at
O.295b at room temperature, which becomes stronger(41) with decreasing
temperature. These structural studies, therefore, established the existence
of the charge-density-wave ground state arising from the Peierls instability
and the associated giant Kohn anomaly in the fluctuation regime above
54°K, and provided detailed information on the temperature dependence.
The Peierls instability is a soft-mode structural metal-insulator tran-
sition driven by the one-dimensional electronic system. It is useful to view
72 A. f. Heeger
the instability in terms of the electronic dielectric response function which

relates the induced charge density to an external perturbing potential:
[E 1(q,w)-1)]=(47Te 2/q2)[p(q,w)/Vext (q,w)] (1)
The dielectric response function can be written in the following form:

2
~1 47Te
E (q, w )-1 = -2 F(q, w, T) (2)
q
where F(q, w, T) is the temperature-dependent Lindhard function,
F(q, w, T) = I n p+l - np (3)

p Ep - E p +q + liw
where np is the Fermi distribution function for state p. The Lindhard
function is the fundamental response function of the metallic electron sea
and is shown in Figure 2a for one-, two-, and three-dimensional systems. In
three dimensions the Lindhard function varies smoothly with an infinite
slope at q = 2k F , where the perturbing wave vector spans the Fermi surface
such that Ep - Ep+2kF = O. In two dimensions the response function has a
cusp and a discontinuous derivative at q = 2k F . In one dimension, since the
Fermi surface is a plane, the nesting condition Ep - Ep+2kF = 0 is satisfied
over the entire Fermi surface, resulting in a logarithmic singularity. With
increasing temperature, the smearing of the distribution function in the
numerator of Equation (3) smooths out the singularity resulting in a sharp
peak at 2kF which progressively weakens with increasing temperature
(Figure 2b).
Equation (2) is a free-electron result consistent with the simple Froh-
lich Hamiltonian neglecting long-range Coulomb interactions. To obtain
the proper long-wavelength, high-frequency behavior (plasmons), and
metallic screening, one must include the long-range electron-electron
Coulomb repulSIOn. However, for our purposes here Equation (2) is
adequate, since we will be interested in local response to lattice distortions
for which charge neutrality is maintained within the unit cell. Combining
Equations (2) and (3) we obtain
p(q, w) = F(q, w )Vext(q, w) (4)
where F(q, w) is the Lindhard function shown in Figure 1b for w = 0 and

assuming a tight-binding band structure.
The Peierls instability arises from the divergent response of the one-
dimensional metal to an infinitesimal driving potential at q = 2kF as a
result of the logarithmic singularity in F(q, T) as T ~ O. Such an external
potential arises from a distortion in the crystal. Consider, for example, the
half-filled band case, and a distortion 8 where the lattice constant goes
2.0r---y----m--,----,
Figure 2. (a) Lindhard function F(q.o.o) for 1.5

one-, two-, and three-dimensional electron 1 dimension
gas (normalized to the density of states at the
Fermi level). (b) Temperature-dependent
Lindhard function for the tight-binding F (q) 1.0 f-,:=",---'--:":"::":"
band. The calculations were carried out for N(O)
the half-filled band. The logarithmic depen-
dence of F(2k F ) as a function of (EFI k T) is
shown on the left-hand side. 0.5
(b)
•
4 4
50
0
0
•
.I
3 :3
,
0-
"0-
X ,
><
"...
-'"
) ..><
"
0-
.I
N
2 2
. 0-
j
>< •
•
/
OL-~_-L __-L_~_L--J~~__J -_ _L-~O

1.2 1.4 1.6 1.8 2.0
74 A. 1. Heeger
from a to a + 8 or a - 8 on alternating sites. Such a distortion doubles the

unit cell, making the Brillouin zone of width 2TT/2a = TT/ a = 2kF . The
existence, then, of the doubled zone implies
(5)
As a result, the electronic energy is lowered by an amount
(6)
More generally, we can write
ilEe1(q) = -iF(q, w)V~ (7)
where Vq is an external potential, for example, imposed by a phonon in the
crystal. In this context, Equation (4) is written in the spirit of the Born-
Oppenheimer approximation, i.e., the electronic system is assumed to
adiabatically follow the much slower nuclear motion.
We therefore consider a system described by the Hamiltonian
H = Ho+ H e - e + H e - p (8)
where Ho represents the unperturbed electron and phonon subsystems
Ho = L Eku C :aCka + L hw~(nq + 4) (9)

k. a q
and Eku is the tight-binding one-electron band structure. H e - e and H e - p

represent electron-electron and electron-phonon coupling terms, respec-
tively. We assume initially (see Section 2.2) that the effective electron-
electron interaction term, He-e. is not important. In this limit, one has the
simple Frohlich Hamiltonian
'(J(= L EkuC:uCku+ L g(q)(aq+a~q)cLq+Lhw~(nq+~) (10)

k.a k.q q
where g(q) is the coupling constant, C lu(cku) is the electron creation

(annihilation) operator, and a~(aq) is the phonon creation (annihilation)
operator.
To calculate the phonon spectrum including the effect of the electron-
phonon coupling we utilize Equation (7) and write
Hphonon = 4K L (UI +! - U I )2 -41el L F(q)V~+iM L IOd 2 (11)

I q I
The first term is the usual "spring constant" potential energy contribution
with UI being the displacement of the lth molecule from equilibrium, and
the second term is essentially a phonon self-energy contribution arising
from the electronic polarizability. The final term is the kinetic energy
contribution.
Imagine imposing a periodic distortion (wave number q) on the lattice
as in a phonon lattice wave. The resulting Vq can be obtained from the
form of the electron-phonon coupling. If the nearby cores are displaced,
the electrons at the lth site see a potential
(12)
The resulting Vq from a periodic wave is
(13)
2h 1/2
. 21 t
= 'Y ( - -) SIn '2q(a- q +aq ) (14 )
NMw q
The phonon spectrum is readily calculated from Equation (11):
Hphonon = L hwq(nq +~) (15)

q
where
2 = (O)2[
W q Wq 1- -llel.
KN SIn 2(1'2qa )F (q, T,wq
)] (16)
and
W ()
q = 2( K / m )1J?- SIn
. '2qa
1
Because of the peak in F(q, T) at q = 2kF (see Figure 2b) the calculated
phonon spectrum shows a relatively sharp anomaly near q = 2kF • This giant
Kohn anomaly in the phonon spectrum ultimately arises from the perfect
nesting of the Fermi surface in the one-dimensional metal. (42)
The mean field transition occurs when a solution for the screened
phonon frequency appears at zero frequency, for in this case one has a
condensation of 2kF phonons, i.e., a true distortion in the crystal. Thus, the
equation for the mean-field Peierls transition temperature, Tp , is given for
the simple acoustic mode case (hw « E F ) by<43.4 4)
(17)
giving
(18)
where A is the dimensionless electron-phonon coupling constant,
76 A. 1. Heeger
A = (2g2 /hw q )N(o)' where g~q) = "/2h/ MW q and N(o) is the density of states
at the Fermi energy.
The mean field temperature defined in Equation (18) should not be
viewed as a phase transition temperature, since the one-dimensional
fluctuations shift the actual phase transition down to T = OOK in the
absence of interchain coupling. Rather, kBT~F defines the characteristic
energy scale below which the electrons and 2kF phonons are strongly
coupled; there exists short-range order and a dynamical distortion with
finite coherence length, geT), but no static distortion. A static distortion
will appear in the structure as a result of pinning(45-47) the dynamical
distortion by a phase-dependent potential which may arise, for example,
from interchain Coulomb coupling, commensurability of the distortion
with the underlying lattice, or impurities.
A soft-mode transition similar to that described above would be
expected in any metal whose Fermi surface contains macroscopic, nested
sections. Given such nesting, a periodic lattice distortion (PLD) will arise
whose periodicity results in the opening of a small (compared to the
bandwidth) gap in the electronic energy spectrum at the Fermi surface. The
PLD is accompanied by a periodic distortion in the conduction electron
charge density: a charge-density wave (CDW). In a one-dimensional metal,
the Fermi surface is perfectly nested, implying a PLD with wave number
qo = 2k F , and an energy gap that spans the entire Fermi surface.
In a single-particle semiconductor, the gap results from the periodic
perturbing effects of the equilibrium ionic lattice and is thus necessarily
tied to the lattice. The Peierls gap, however, results from the PLD and is
thus tied only to the distortion and not necessarily to the lattice, itself. (4) If
the PLD is commensurate with the lattice, it corresponds to a standing
wave, and the Peierls state is semiconducting. If the PLD is incommen-
surate with the lattice, it is possible to visualize excited states where the
phase of the PLD moves through the crystal carrying the CDW. In this
case, the Peierls state is conducting in a collective many-body sense as first
described by Frohlich.(4)
The plan of the chapter is as follows: Section 2 reviews the experi-
mental data relevant to the strength of the interactions (bandwidth, elec-
tron-electron, and electron-phonon). In Section 3, we discuss briefly the
structural aspects of the Peierls instability in TTF-TCNQ (see the Chapter
by Comes and Shirane for a more complete discussion). Section 4 discusses
the optical properties in the range from far-infrared through the visible
with emphasis on the pseudogap. The results on electrical conductivity are
presented in Section 5, with emphasis on the metallic state. The transition
region (38°K < T < 54°K) is discussed briefly in Section 6. Section 7 focuses
on the pinned CDW regime at low temperatures with emphasis on the
optical properties and the unusually large microwave dielectric constant.
The nonlinear transport recently discovered in TTF-TCNQ and related

CDW systems is discussed in Section 8. Conclusions and remarks are
presented in Section 9.
2. Strength of Interactions; Bandwidth, Electron-Electron and

Electron-Phonon Interactions
A variety of experiments have been performed to clarify the dominant

interactions and obtain a measure of the overall scale of energies involved
in TTF-TCNQ. In the following we discuss in turn one-electron energies
(i.e., band structure), the electron-electron Coulomb interaction, and the
electron-phonon interaction.
2.1. One-Electron Energies; Band Structure
Estimates of the overall bandwidth were obtained from the plasma

frequency, (II .12) t h
ermoei ' power, (4849)
ectnc . an d magnetic . susceptI'b'l'
llty. (34)
In each case, a value of order 0.1 eV was found for the intrachain (b-axis)
transfer integral, indicating a bandwidth of order 0.5 eV. It is difficult to
place realistic error bars on the bandwidth. The various estimates,
however, are consistent with one another.
Electron energy-loss experiments(50) on thin films of TTF-TCNQ have
resulted in the direct observation of the plasmon peak at 0.75 eV (q = 0).
The results are in agreement with the earlier optical studies.(13.14)
Measurements(50) of the plasmon dispersion relation and lifetime provide
more direct information on the conduction bandwidth and yield a value of
0.5 eV ± 0.1 eV, in excellent agreement with the above estimates. It seems
clear that a relatively narrow band picture is appropriate, and that tight-
binding theory is a good starting point.
Neglecting, initially, the interchain electronic transfer, two possible
inequivalent band structures can be envisioned as shown in Figures 3a and
3b. In Figure 3a, the two bands associated with the TTF and TCNQ chains,
respectively, are "parallel"; i.e., both are either normal (minimum at k = 0)
or inverted (minimum at k = 11'/ a). The two bands could in general be of
different widths, and might even cross. In this structure kFl and kF2 (k F1 =
11'/ a - kF2 ) are the Fermi momenta of the electron and hole systems,
respectively. The alternative structure is shown in Figure 3b. In this case
the bands cross at k F .
Since the molecular orbitals of the unpaired electrons on TTF and
TCNQ are made up primarily of linear combinations of the 1T-pz wave
functions, the normal or inverted structure is determined by the detailed
stacking of one molecule above the other in the chain. The situation for the
78 A. f. Heeger
a b
,.
,I
TT k TT
b b
Figure 3. (a) Schematic parallel band structure for two-chain donor-acceptor system. The
bands do not cross. (b) Schematic crossed band structure for two-chain donor-acceptor
system. Interchain tunneling causes the hybridization (dashed curve).
TCNQ stack is such that the molecules slip-stack in such a way that the
phases always match across the length of the molecule. This slip-stacking
basically changes the symmetry from odd to even, resulting in a normal
(s-like) band. The symmetry argument for the TTF stacking is less clear
since the phases do not match uniquely and there is a partial cancellation.
Direct calculations of the overlap and transfer integrals suggest that the
TTF band is inverted.(51) In the absence of interchain coupling, the normal
or inverted structures do not lead to different properties; in either case the
band structure is that of two sets of independent one-dimensional chains.
Even within the narrow-band tight-binding approximation interchain
coupling must be considered since the segregated donor and acceptor
chains comprise a three-dimensional structure. The effect of small inter-
chain coupling energies(SI) (ta/tb _10- 2 , tc/tb _10- 1) on the parallel band
structure (Figure 3a) would be minimal. The Fermi surfaces would be
slightly curved, but these effects would not be of major qualitative
importance. In the case of Figure 3b, however, the finite fa would remove
the band crossing by admixture leading to hybridization energy gap (32) as
shown by the dashed curves in the figure. Going out along (~, 2kF , 0) the
gap goes to zero at the zone boundary so that the system would be a
zero-gap semiconductor. (52)
The bandwidths associated with wave vectors along the a axis or c*
axis are in any case much less than that associated with the b axis. Berlin-
sky and co-workers(5!) have made estimates of the transverse integral ta
(between TTF and TCNQ chains) and fc (between like chains) with the
result that fc -0.01 eV and fa =0.003 eV. As a rough experimental
measure of this anisotropy one can use the anisotropy in the conductivity,
i.e., of order 10 3 at room temperature. Although a crude estimate, this
suggests that the bandwidth associated with ka(Wa ) is of the order of 30

times narrower than that associated with the principal conducting (b) axis
(Wb). Using the results obtained from the plasma frequency,(1l·12) the
susceptibility,(34) and the thermoelectric power,(48.49) we have Wb = 0.5 eV,
so that Wa -10 2 K. Transport in such a narrow band (i.e., Wa < k8D ,
where 8D is the Debye temperature) is not metallic but diffusive(17) and
severely limited by the phonon thermal disorder. That a diffusion-theory
approach to the transverse conductivity is necessary can be seen from an
estimate of the mean free path along the a or c* axis. In the tight-binding
approximation one can estimate the mean free path from the measured
electrical conductivity:
u1T1i
1= VFT" = --2- (19)
2Ne a
where N is the carrier density, a is the lattice constant in the direction in
question, and U is the measured conductivity. Using u1-1 (0 cmrl (16.17)
leads to an estimate of 1< 0.01 A, i.e., much less than a lattice constant.
Thus, in the transverse directions the carriers do not propagate as in a
metal, but remain on a given chain for relatively long times, then hop from
chain to chain in a diffusive manner. This conclusion is further strength-
ened by ~tudies of the a-axis optical properties(II,I1,14) (see Section 4
below). There is no observable a -axis plasma frequency and no sign of
metallic reflectivity. The reflectivity (for E polarized along a) is approxi-
mately 10% from the visible through the far ir, characteristic of an insula-
tor with dielectric constant of about 2.5. This is as expected for such an
anisotropic system where w ~ « w ~ since the transverse effective mass is
large. Consequently w;r < 1 and the transverse plasma oscillations are
overdamped. In effect, band theory is not appropriate for the transverse
directions.
Experimental information on the interchain hopping times has been
obtained by Soda et al.(S1) from their studies of the nuclear magnetic
resonance. Basically, they find clear evidence of one-dimensional diffusion
cutoff at long times by the interchain hopping. From analysis of the
frequency dependence of the nuclear spin-lattice relaxation rate, Tl\ they
obtain values in agreement with the calculations of Berlinsky.(SI)
The experimental data thus provide strong evidence that the transport
in the transverse directions is diffusive, indicating that, even in the presence
of the weak interchain coupling, the electron wave functions are localized on
individual chains with diffusive hopping from chain to chain. In this sense
the one-electron band structure should be considered pseudo-one-dimen-
sional; the weak three-dimensional coupling is localized into one-dimen-
sional chains by the thermal disorder. Cohen(18) has observed that this is in
effect an Anderson disorder localization from a three-dimensional band
80 A. 1. Heeger
structure into a collection of one-dimensional band structures driven by the

thermal disorder.
2.2. Electron-Electron Interactions: Nuclear Magnetic Resonance and

Magnetic Susceptibility
Information on the strength of the electron-electron interaction is
available from nuclear relaxation rate (T~I) and magnetic susceptibility
studies. The proton spin-lattice rela,xation rates (T~l) were measured in
(TTF-TCNQ) and its de ute rated analogs TTF(D4)-TCNQ and TTF-
TCNQ(D 4) as a function of temperature from 4.2 to 300o K. Since the
protons are directly coupled to the electronic system through the hyperfine
interaction, by selectively deuterating the proton sites on either the TTF or
the TCNQ chains, one can use NMR to probe the local electronic suscep-
tibility as seen by the remaining proton spins.
The early studies of Rybaczewski et al.(S8) demonstrated that X2Tl T
exhibits only a weak temperature dependence, where Tl is the proton
spin-lattice relaxation time in TTF(D4)-TCNQ, TTF-TCNQ(D 4), and
TTF-TCNQ and X is the total measured spin susceptibility. These results
were analyzed(s>-) in terms of the enhanced one-dimensional response
function(5S) at q = 2k F , with the conclusion that the electron-electron
interaction was not dominant.
Subsequent measurements of the frequency dependence of TI in
TTF-TCNQ by Soda et al.(S3) demonstrated the existence of the w I/2
dependence characteristic of one-dimensional diffusion cutoff at times
greater than Tc hy interchain hopping [Tc = (2rr/h)tlN(o), where t.l is the
transverse interchain hopping integral and N(o) is the density of states].
Soda et al.(">3) generalized the earlier theory to include both the q = 0 and
q = 2kF contributions to the relaxation rate(56):
T~l =2'Y?~T A 2 I x1 (q,WN) (20)

g /.LB q WN
where A is the hyperfine constant, WN is the nuclear Larmor frequency, 'Y
is the nuclear gyromagnetic ratio, and X 1(q, WN) is the imaginary part of
the response function. For a quasi-one-dimensional system with kB T« E F ,
only Iql = 0 and Iql = 2kF contribute. Soda et al.(S3) argue that at long
wavelengths, X"(q = 0, WN) should be viewed as having diffusive charac-
ter(S?) since the electron mean free path is in the range of a few lattice
constants as a result of the short electron scattering time, Tv. For Iql = 2kF ,
the features of X 1(q, WN) are described by the coherent tight-binding
picture used earlier. (')4,55) They obtain for the Korringa product(S3)
A ~
[T· Tx;r 1= (g/.LBr (2S r l{ C:) 1/2
g(we)K'(a)+ K(a)
}
(21)
where
s = gJ-LBh (22)
41TkB
_ f1 + (1 + W2T~ )1 / 2 112
g(W)- L 2(1 +W2T~) J (23)
K(a) = (I - a )2([1- aF(q, T)r 2 )q_2k

F (24a)
K'(a) = (1- a /([ 1- aF(q, T)r 3/2 )q_O (24b)
and a = UN(o) IS the product of the effective Coulomb interaction times the
density of states at the Fermi energy, and F(q, T) is the temperature-
dependent Lindhard function
Equation (21) for the Kornnga product can be compared directly with
experiment. The low-field values for Tl TX;(T) have been calculated from
the TJ data of Reference 54 takmg the temperature-dependent local
susceptibility appropnate to the TTF and TCNQ chains, respectively, as
inferred from the 13 C Kmght-shlft studies (58) The results are plotted in
Figure 4 as [TJ TX;(A/ gJ-LB)2 S 1] for direct companson with Equation (21).
I
• TTF CHAIN PROTON DATA (i . F)
x TCNOCHAIN PROTON DATA ( i- Ol
0 . 12 I- -
• •
...... ....---. .. .......... ....

• •• ••••
•
•,• •••
(f)
NO)
N~
01
'- 0 .08
l-
r- ~
-
I--
N
N
X
._
._
• x
x
•
•••
<l x •
••
x ., x x x x
0 .04 r- •• -
'.
x
XX x •
••
•• x
o~~~Y~Y?----~I----------~I----------_~
I --~
o 100 200 300
T (K)
FIgure 4 Nuclear magnetic relaxation data for TTF-TCNQ(D 4) (000) and TTF(D4)-TCNQ
(x x x) A 2 X2 T, T(g2 f.L ~) I IS only weakly temperature dependent above 60"1< and the two
chams YIeld Similar magnitudes Implymg little difference m Coulomb mteractlOn strength on
the TTF dnd TCNQ chams (see Reference 56)
82 A. 1. Heeger
Using the solution hyperfine constants(59) for TTF(AF) and TCNQ(A Q ),

the rates for TTF(D4)-TCNQ and ITF-TCNQ(D4) are enhanced over the
bare Korringa product by factors of 20 and 12 respectively.
The frequency dependence observed by Soda et al. (53) implies that the
first term is dominant at low frequencies (or applied fields). Thus, from the
absolute magnitudes, we obtain (1- a )1/2[ 'TJTv] 1/2 = 20 for the TCNQ
chain with the corresponding factor being approximately 12 for the TTF
chain. The quantity ('Tc/'TS/2 can be estimated from the experimental
values, 'Tc = 4 X 10~12 sec as obtained from the frequency dependence of
T j ,(53) and 'Tv = 3 X 1O~15 sec as obtained from analysis of the optical
reftectance(1l) near the plasma frequency giving ('TJTv)I/2 - 30. The
agreement is satisfactory given the uncertainty in using 'Tv from the optical
data in the TIl analysis appropriate for w = O. Thus the magnitude and
frequency dependence are given correctly by Equation (21) in the weak
Coulomb limit.
To obtain an upper limit on the magnitude of a we use the results of
Jerome and Weger/ 53 ) which show that at high fields (H< 100 kOe)TI
deviates from the HI/2 dependence and appears to saturate at a value
approximately three times larger than the low-field value (see Figure 5). If
we identify the high-field value as an upper limit for the field-dependent
second term of Equation (21), we obtain
(1- a)2
---'------'-----""""'2~4 (25)
[1-aF(2k F , T)]
Numerical estimates using EF = 0.2 eV and T = 3000 K yield
F(2k F , 30(fK) = 2. The magnitude and weak temperature dependence
observed for X2Tl T as discussed above imply the weak Coulomb limit. We
therefore take the positive root of Equation (25) so that (1- a)/ (1- 2a) ~
2 and a ~ ~ in approximate agreement with earlier estimates. (58) Thus,
Equation (21) accounts for the absolute magnitude, the frequency depen-
dence, and the temperature dependence of the proton relaxation rate
E
1--
150
100
50 t-(
r--
<xH
112
,
,, Figure 5. Magnetic field dependence of Tl in TTF-

°0~-1~0----1~
0-0 --
20~0---- TCNQ(D 4 ) showing the Hb 12 (wb 12 ) dependence
characteristic of one-dimensional diffusion and the
Ho (KOe) saturation at high fields (see References 53~55).
provided that the effective electron-electron interaction is relatively weak

(a::::; 0.3). Moreover, the weak temperature dependence of X;T 1 T at low
magnetic fields and the absolute magnitude of T~l appear to be inconsis-
tent with the strong Coulomb limit. (01-6,) At high magnetic fields, the second
(2k F ) term dominates the relaxation rate. Since K(a.T) increases with
decreasing temperature, Equation (3) predicts an increase in [TI TX;r l as
the temperature is lowered, as was observed by Soda et at.(S3) at high fields.
We note particularly the similar results for TTF-TCNQ(D 4) and
TTF(D4)-TCNQ (see Figure 4 and References 53, 54, and 58). The
conclusion that U / W < 1 is therefore appropriate for both the TTF and
TCNQ chains.
The weak temperature dependence of X;T 1 T and the absolute
magnitude of T~l appear to be inconsistent with the strong Coulomb limit.
For U 2: W, the expression (21) for X(q, w) is inadequate, since it treats the
exchange enhancement in RP A. In the strong coupling limit, we expect a
weak temperature dependence of TIl. Coll(60) has shown that for the
quarter-filled band Hubbard model, the orbital and spin decoupling leads
to a well-defined spin-wave spectrum (E max - J) in addition to the single-
particle band excitations (E max - W). Based on our understanding of the
half-filled band case, where experiment(61) and theory(h~) are in good
agreement, the spin-wave excitations should give a contribution to T~1 of
order A 2/ J, possibly reduced by the lifetime of a given electron on a given
site. A contribution from the single-particle excitations may add an addi-
tional linear term. The actual magnitudes of the two terms would be
determined by the spectral weight function associated with each part of the
spectrum. These spectral weight functions are presently unknown. It is
likely, however, that for U» W the spin-wave process will dominate,
whereas for U« W the spin-wave contribution will be negligible and the
above approach [Equation (21)] should be a good approximation. In any
case we expect the spin-wave contribution to be typical for the S = t 1 - d,
antiferromagnet, i.e., dominated by diffusion processes at high tempera-
tures and approaching a finite value (of order A 2 / J, where J is the inter-
chain exchange) for kB T < 1.
The local susceptibilities on the individual TTF and TCNQ chains
were determined(58) through a study of the Knight shift associated with the
13C-labeled CN group located on the TCNQ molecule. The Knight shift in
metals is given hy K -- - f-L; I HhfX" where Hhf is the average hyperfine field
per electron and X, is the Pauli (local) susceptibility per molecule. The
Knight shift thus provides a direct measure of the time-averaged local
magnetic field at the TCNQ molecule and thereby determines the
temperature dependence of the TCNQ chain local magnetic susceptibility.
When combined with the total susceptibility, XT, the data can be analyzed
to obtain the individual chain contributions (XO for the TCNQ and XF for
84 A. 1. Heeger
1.0
0 .9
>-
I- 0 .8
-'
co 0 .7
f=
n.
w 0 .6
u
Vl
::> 0 .5
Vl
a
w 0.4-
N
:i (TCNOIX o
<! 0 .3
~ . ~....-.-
a: ~.----r--
0 0.2
2 --- • .!!-- . - - --
0.1 -~.
50 100 150 200 250 300

TEMPERATURE (K)
Figure 6. Normalized local susceptibilities on the ITF(XF) and TCNQ(XQ) chains (Reference
54). The results were obtained from the DC Knight-shift studies together with the static
magnetic susceptibility (Reference 58).
the TTF chains). The result of this decomposition is shown in Figure 6.

Similar, although somewhat different, results have been obtained from
deconvolution of the ESR g-value measurements.(lOO.126.127)
Torrance et al. (64) have argued that the room-temperature static
susceptibility is enhanced over that expected from the Pauli susceptibility
and have used such arguments as evidence for strong Coulomb coupling.
The experimental ratio(58) XF/ Xo = 7/3 was determined from the 13 C
Knight-shift studies (see Figure 6), where XF and XO refer to the room-
temperature local susceptibilities on the TTF and TCNQ chains, respec-
tively. Using XO + XF = Xtot and the experimental value Xtot = 6 X 10- 4
emu/mole,(34) one finds Xo = 1.8 X 10- 4 emu/mole and XF = 4.2 X 10- 4
emu/mole. For a cosine tight-binding band, the enhanced q = 0 suscep-
tibility is given by X "-- (1 - a rlNop, M7Tt sin(7Tp/2) r\ where p = 0.59 is the
charge transfer. The above values for XO and XF together with a = 0.3 as
suggested by the Tl data yield 4to = 0.46 eV and 4tF = 0.20 eV, where to
and tF are the transfer integrals associated with the TCNQ and TTF chains,
respectively. These values are well within the uncertainty in the theoretical
estimatesYiI ) Moreover, the absence of Landau damping in the plasmon
dispersion('iO) sets an upper limit on the bandwidth of W < 0.55 eV. We
conclude that the enhancement of the static susceptibility is relatively small
and is consistent with the results obtained from the nuclear relaxation
studies.
The principal arguments(64) presented in favor of the large U limit
come from studies of the optical properties of the TCNQ salts. Torrance(64)
has argued that the existence of additional structure in the optical spectrum
should be assigned to charge-transfer transitions associated with excita-
tions across the Hubbard gap. He asserts that the "B" peak near 1 eV is a
measure of the strong Coulomb interaction. However, the same "B" peak
is evident for HMTSeF-TCNQ,(65) where magnetic, optical, nmr, and
galvanomagnetic studies(66) all indicate weak Coulomb interactions. Thus,
the attempts to identify features in the optical data indicative of the
strength of electron-electron interactions are less than convincing.
The nonmagnetic ground state, absence of localized moment forma-
tion, and particularly the non-enhanced Korringa relaxation provide evi-
dence that electron-electron Coulomb interactions are not dominant in
TTF-TCNQ. Evidently the interaction term, He-e. has been reduced by
intramolecular correlation, metallic screening and correlation, and general
polarizability effects to the point where Coulomb effects are relatively
weak (at frequencies well below both the plasma frequency of the resulting
metal and the dominant singlet molecular TT-TT* transitions responsible for
the polarizability). Although it is somewhat surprising that such screening
effects would be so efficient, the residual Coulomb interaction appears to
be relatively weak and not the dominant interaction responsible for the
overall solid-state properties.
The origin of the temperature-dependent susceptibilities in the metal-
lic regime (see Figure 6) has been the subject of considerable discussion
and speculation. Two principal ideas have been proposed:
(1) The temperature dependence is a band-structure effed52 ) resul-
ting from a deep minimum in the density of states near EF as a result of
TTF and TCNQ wave-function hybridization in the crossed band structure
(see Figure 3b).
(2) The temperature dependence results(34.58) from a pseudogap in the
density of states, resulting from the 2kF scattering from the dynamical
distortion expected in the one-dimensional fluctuation regime below the
mean field transition temperature, T~F. Theoretical studies(67) have shown
that
(26)
where No is the unperturbed band density of states, VF is the Fermi

velocity, g(T) is the longitudinal coherence length, and ~;;;;;(~2>1/2 is the
energy gap (21l = 3.5 kBTr: F ).
86 A. 1. Heeger
The band-crossing hybridization explanation seems unlikely in view of

the clear localization of the electronic wave functions on individual chains
due to thermal disorder. In this case the "hydridization gap" would be
removed by the localization at least at temperatures above the 54°K phase
transition. More convincing experimental evidence comes from systematic
studies of the susceptibility in this class of related systems.(34) The suscep-
tibilities of many of the compounds in this class have now been studied,
including TTF-TCNQ, ATTF-TCNQ, and TMTTF-TCNQ. The effects of
methyl substitution on the rings or Se substitution in the molecule is to
change both the size and electronic properties of the resulting molecules.
As a result, major differences in the interchain transfer integrals are
expected on general grounds since the intermolecular overlap and detailed
crystal structure and packing are not precisely the same. These differences
would be reflected in the magnitude of the spin susceptibility of the salts,
and indeed, Xs at room temperature is noticeably different for the
different systems (TTF, 6.0 x 10- 4 emu/mole; ATTF, 3.4 x 10-4 emu/
mole; and TMTTF, 4.8 x 10- 4 emu/mole). However, the normalized
spin susceptibilities are nearly identical over the entire temperature
range.(34)
This argument is further strengthened by the observation that the class
of materials with the characteristic temperature dependence of X first
observed in TTF-TCNQ is not restrictive. The complex salt (TTT)
(TCNQh (tetrathiotetracene) has been reported by Shchegolev and co-
workers(6H) to be metallic. with conductivity similar to the early obser-
vations on poorly formed crystals of TTF-TCNQ. Although the crystal
structure and molecular arrangements are quite different, the normalized
temperature dependence of X is nearly identical to that of TTF-TCNQ.
The Se substituted salt, TSeF-TCNQ, again yields a X vs. T curve of nearly
identical form.(69) This apparent generality for systems with different crys-
tal structures indicates that the characteristic temperature dependence
does not arise from structure in the electronic density of states (i.e., from
band structure). A more general mechanism is responsible for the major
features of this temperature dependence.
The interpretation of X(T) in terms of a pseudogap in the density of
states has been shown to be in agreement with the predictions of Ginz-
burg-Landau theory for a complex order parameter system (amplitude and
phase of the distortion).(SH) The resulting expression for the coherence
length is given h/58 )
(27)
where b is the lattice constant, f is the fractional charge transfer, and kF

the Fermi wave number. Using this and Equation (26) leads to(58)
(28)
where XP = 2ILBNo is the bare density of states. Although Equation (28) is

in agreement with the TCNQ chain susceptibility (see Figure 6), the diffuse
x-ray data(40) indicate a considerably stronger temperature dependence for
fd). However, a more detailed comparison requires thermal averaging
over the full energy-dependent density of states using the experimental ft.])
as input. Such an analysis has not been carried out. Moreover, the potential
role of the 4kF scattering(38.39.4 0 ) has not been included in the above
arguments.
2.3. Electron-Phonon Interaction
There are two dominant sources of the electron-phonon interaction in

a tight-binding organic metal:
(a) Intermolecular (acoustic and optical mode) lattice phonon modu-

lation of the electronic energies
(b) Intramolecular vibrational mode modulation of electronic ener-
gies.
Phonon modulation of the intermolecular transfer integral is a neces-

sary consequence of a tight-binding metal. The transfer integral scales with
the overlap integral, and since the latter is expected to vary exponentially
with distance between molecules, a strong electron-phonon contribution is
implied. Note, however, that as the bandwidth narrows, this contribution
becomes less important.
Phonon modulation of the solid-state ionization potential is to be
expected in such ionic metals in which the Madelung energy and polaron
polarization bond make important contributions to the cohesive energy.
The polaron energy, E B , is a strong function of distance, (r- 4 ), indicating a
special sensitivity to lattice vibrations. Note that although an electronic
polarizability is involved in the excitonic polaron, this polarizability enters
in a quasistatic manner as the molecules vibrate in the presence of a
phonon. However, such phonon modulation of the cohesive energy neces-
sarily comes in only in second order since the first-order terms are identi-
cally zero in a stable crystal. Nevertheless, these basically large interactions
[the scale is on the order of the (l - A) charge transfer energy, where I is
the donor ionization potential and A is the acceptor electron affinity,
- 4 e V rather than the bandwidth, - 0.4 e V] which have their origin in
interchain coupling must be kept in mind. Such strong second-order terms
88 A. 1. Heeger
may play an important role, for example, in the origin of the 4kF scatter-
. (38-40)
mg.
On adding an electron to a molecule, it is experimentally observed
that the various molecular bond lengths change,(6.7) i.e., there is an inter-
action between the electron and the intramolecular optical phonons. The
details of the molecular distortion depend on the wave function of the
added electron; qualitatively a given bond will expand if the electron wave
function has a node in the bond region or contract if the wave function has
an antinode. This is the source of the coupling of the electrons to the
intramolecular vibrations.
We consider the two leading contributions in turn.
2.3.1. Intermolecular
H~~~ = (k) L (Uj+l - Uj )(c ;+luCjO' + c;O' Cj+IO')

j
2h
(~t') L (--0)
1/2
= (sin ~q) e -q{j+1/2)
j.q Nmw q
x (a ~q + aq )(c i+lcj + C iCj+d (29)
Transforming to Bloch states and summing over j, we obtain
. I 1)(a-q
H (aJ
e_p = t L.
1 '\ (
-2h
N0 )1/2 sm 2q cos (k +2q t t
+aq) Ck+qCk (30)
q.k mWq
= L g(k, q )(a ~q + aq)c t+qCk (31)
q.k
where
2h 1/2
g(k,q)= -tl(Nmw~) sin~qcos(k+~q) (32)
Since the most important interactions are for electrons near E F ,

2h ) 1/2
g(k, q) = g(kF' q) = tl(Nmw ~ sin 2 ~q (33)
Note that
(t ' )2 . 2 I
NK sm 2q (34)
where w~ = 2(K/m)1/2 sin ~q and K is the effective "spring constant"

describing the bare ionic restoring forces. Recalling the usual definition of
the dimensionless electron-phonon coupling constant, Ao,
Ao=2 g2 (d)1 ,
N(0)=(t )2_1_ (35)
liw q F.S. K 21Tt
where N(O) = N/21Tt is the density of states at the Fermi energy (one kind
of spin). Equation (35) is written for the special case of the half-filled band
(kF = 1T/2a).
2.3.2. Intramolecular
(36)
(37)
with
Ii ) 1/2
g(q) = -~ (2NMn~ r" (38)
and n; is the frequency associated with a particular intramolecular mode

with effective force constant K" and reduced mass M, and r" represents the
coupling of the electron to that mode. The effective indirect attractive
interaction is given by
(39)
and
opt
Aph =
("
:-
r;')N
K" (0) =:-""
Aph
(40)
A relatively strong coupling of the electrons to the lattice and molecu-

lar phonons may be inferred directly from the mean free path, which is
estimated to be on the order of one to two lattice constants at room
temperature. Thus the electrons are very strongly coupled to the lattice or
molecular modes, and the system is near the limit of applicability of band
theory.
Strong electron-lattice coupling has been inferred from infrared
reflectivity studies of the scattering rate, 'T ~~. Using Hopfield's expres-
sion(70) for the dimensionless coupling constant A = N(O) Vph ,
A= (_li_)
21T'T
(kTf1
ph
(41)
(valid for T> ()D), a strong coupling value of A-I was obtained. Analy-
SiS(ll)of the temperature dependence of the scattering rate indicates that
90 A. I. Heeger
T(T) is consistent with single-phonon scattering, with a contribution from

temperature-independent defect scattering which varies from sample to
sample and which increases as the number of defects increases. However,
Seiden and Cabib(7l) have attributed T(T) to electron-electron scattering
(see Section 5.1 below).
Independent evidence of acoustic mode electron-phonon coupling in
ITF-TCNQ was found in studies of the Young's modulus.(72,73) Compared
with other TCNQ salts, the magnitude of the Young's modulus is small and
shows a remarkable decrease with increasing temperature. The small re-
storing force to shear strain has been interpreted as evidence of elastic
one-dimensionality indicative of weak interchain elastic coupling. This
shear "softness" appears to be an important distinguishing feature between
ITF-TCNQ and other pseudo-one-dimensional systems such as KCP.
Strong coupling of the conduction electrons to the intramolecular
lattice modes has also been suggested. The transferred electron density on
the TCNQ anion is concentrated on the terminal di-cyanomethylene
groups with maximum amplitude on the C N as evidenced by direct NQR
studies.(74) As a result, fluctuations in charge density affect the various
bond lengths and strongly couple to the intramolecular vibrational modes.
Rice et al. (75) have carried out estimates of the symmetry-allowed electron-
phonon coupling constants for the TCNQ molecule. Their results are
summarized in Table 1. The characterization of the ten symmetric Ag
modes as obtained by Takenaka(76) is given in columns 1 and 3, while their
respective frequencies, as refined by Girlando and Pecile,(77) are shown in
column 2. The results of Table 1 imply A~~t/N(0)=0.1geV, which-
combined with the tight-binding density of states [N(0)=4/7TW] for a
bandwidth W =~ 0.5 eV-leads to A ~ht = 0.5.
An independent estimate of the coupling of the conduction electrons
to intramolecular modes can be obtained from measurements of the
changes in bond lengths on going from the neutral molecule to the ionic
salt.(6,7,78,79) On adding an electron to a molecule, it is experimentally
observed that the various molecular bond lengths change; i.e., there is an
interaction between the electron and the intramolecular optical phonons.
The total polarization energy lowering for one electron added may be
written
(42)
where Kv is the effective force constant (obtainable from ir measurements

of bond vibration frequencies) and IlXv is the change in length of the vth
band. The bond length data exist for both neutral TCNQ and for the
negative anion in several TCNQ salts. As shown in Table 2, the estimated
polarization energy is 0.11 eV in NMP-TCNQ, and 0.18 in Cs 2TCNQ3.
These are cases where there is little doubt that complete charge transfer
Table 1. Properties of a. Modes in TCNQ and Their Coupling with the b2g(Tr)
Electron State a
Analogous benzene
Estimated modes (em-I)
Frequency g and el g ( 7T) coupling
Modes (em-I) Characterization (meY) constants (meY)
VI 3048 CH stretching (100%) =0
"2 2229 C::=N stretching (86%); =0

C-C wing stretching
(14"1" )
"3 1602 C=C ring stretching (35%); =50 "16(1595)
CH bending (23%); g16 = 51
CC ring stretching
(17"10 )
1454 C=C wing stretching (58%); =50
"- C=C
0')%)
ring stretching
"s 1207 CH bending (64%); =50 "17(1178)

C=-C ring stretching g17 = 45
(31%)
"6 948 C-C ring stretching (38'1'0); =50
c- C wing stretching
(19",,)
C(CNh scissoring (22%)
"7 711 C-C ring stretching (40%); =50
C- C wing stretching
(29"" )
ring deformation (19%)
"s 602 C(CN) scissoring (43%); =50
('--C wing deformation
(29°/" )
"9 334 Rmg deformation (31 %); =50
C. -('=:oN bending (16°/,,);
ring deformation (16%);
c==c wing stretching
(lb";,,)
"10 144 C -('=:oN bending (61 %); =0
ring deformation (17%)
a See Reference 75_
has occurred, so we may consider the resulting numbers as representative.

Thus, for TCNQ we estimate V';,hl = 0.2-0.3 eV. The resulting value of A~~I
is in the range 0.5-0.8, in good agreement with the more detailed
theoretical results of Rice et ai.(75) described above.
In summary, TTF-TCNQ is a narrow-band, one-dimensional metal
with relatively strong electron-phonon coupling. Such a system is expected
~
Table 2. Polaron Binding Energy and Induced Orthogonality Band Narrowing from Molecular Vibration Modes. Calculated from
the Bond l"ength Changes between Neutral and Negatively Charged TCNQ Molecules"
k Wo Xu XI Ebl X2 Em
Bond No (10' dyne em) (1 () 12 erg) (A) (A.) (10 I' erg) (A) .lX2 (1 () I' erg)
.lXl YO l Y02
-C:=N 4 17.73 0.42 1.140 1.156 0.016 91 0.87 1.152 0.012 51 0.486
C=C 2 9.6 0.33 1.346 1.341 0.005 2.4 0.015 1.355 0.009 7.8 0.047
C=C 2 9.6 0.33 1.374 1.388 0.014 18.8 0.115 1.411 0.037 132 0.797
=C-C- 4 5.18 0.18 1.441 1.420 0.021 45.6 Ull 1.419 0.022 50 loll
-C-C- 4 4.50 0.18 1.448 1.434 0.014 17.7 0.39 1.426 0.022 43.7 0.97
Totals 175.5 2.40 284.8 3.41
(O.lleY) (0.18 eY)
a See A. A. Bnght. P. M Chaikin. and A. R. McGhie. Phys. Rev. B 10. 3650 (1974).
h Subscnpt 1. data from NMP-TCNQ (Reference 78).
c Subscript 2: data from CS 2 TCNQ3 (Reference 79).
~
~
~
'"
~
...
to be unstable toward a soft-mode structural transitIOn driven by the

divergent response of the electron gas at q = 2k F , the Peierls instability.
Given the experimental strong coupling of the electrons to lattice acoustic
modes and intramolecular vibrational modes, the Peierls distortion with
q = 2kF could result either in a crystalline distortion characterized by
molecular motion (e.g., dimerization in the case 2kF = 'TT/a or longer
wavelength analogs for 2kF < 'TT/ a) or by small bond length distortions or,
perhaps more likely, by a combination of the two.
The renormalized phonon frequency for the case of coupling to an
acoustic mode was derived in Section 1 [see Equation (16)]. For the
(unrealistic) case of a single intramolecular optical mode, the correspond-
ing equation is
(43)
In the absence of the electron-phonon coupling each of the inter-

molecular acoustic modes and intramolecular optical modes are true
normal modes of the lattice. The electron-phonon interaction, via the large
response of the one-dimensional electron gas near 2k F , provides strong
indirect mode-mode coupling. Because of the conduction electrons, any
inter- or intramolecular displacement on one molecule is correlated with
any inter- or intramolecular displacement on any other molecule. As a
result strong coupling to intramolecular modes in zeroth order leads to an
enhanced giant Kohn anomaly in the new acousticlike mode in higher
order. In a model containing a single representative intramolecular mode
!1 q and a single acoustic mode W q , the dispersion relation for the new
acousticlike mode is(RO)
W
2 =2:l{ Wq+H
2 r.2q - [( wq+u
2 r.2)2
q -
4 WqU
or.o(
q l-AF(q,w)
)]1/2} (44)
where w~ and !1~ are the bare frequencies, Wq and !1 q are given by
Equations (16) and (43), and A =Al +Ao. The result is clear; if Al>Ao the
intramolecular coupling can dominate the acoustic mode lattice dynamics.
Since in a real system ,\ 1 arises from a great many separate intramolecular
modes, each of which is relatively weakly coupled, the resultant indirect
effect on the acoustic mode is likely to be more important than the
corresponding direct effect on any of the high-frequency intramolecular
modes where the response function F(q,fl remains small.
q)
3. The Peierls Instability in TTF- TeNQ: Structural Aspects and

Phonon Softening
The analysis of Section 2 defines the structural aspects of the
Peierls instability. At low temperatures a well-defined three-dimensional
94 A. f. Heeger
superlattice is expected with a superperiod along the one-dimensional chain

direction of As, where
21T/ As = 2kF (45)
For a simple tight-binding band, 2kF = 1T(f/ b), where O:s f:S 1 is the
fractional band filling, i.e., the number of electrons per molecule. Thus the
period of the superlattice determines the charge transfer, f = 2b/ As. The
position of the superlattice satellite peaks in the a * and c 1< directions will
depend on the detailed interchain coupling mechanism.
At higher temperatures, one expects a loss of transverse order leaving
one-dimensional scattering with little or no correlation between adjacent
chains of the phase of the modulation in the chain direction. This is the
temperature regime of the giant Kohn anomaly, the softening of the
phonon spectrum at q = 2kF . At still higher temperatures, the Kohn
anomaly will gradually weaken as the peak in the Lindhard function
becomes progressively less pronounced.
Diffuse x-ray scattering studies(35,36.38-40) above and below 54°K and
detailed neutron elastic scattering(37) measurements below 54°K have
provided the crucial structural evidence of the Peierls instability in TTF-
TCNQ. These data are described in detail in the accompanying article by
Comes and Shirane (Chapter 2 in this book). We briefly summarize the
principal results here in the context of the related electronic properties.
The x-ray and neutron scattering results have established unam-
biguously the appearance of new satellite Bragg peaks below 54°K, imply-
ing a modulated structure. The satellites appear at incommensurate posi-
tions along the chain or b* direction with a periodicity of 0.295b*. From
Equation (45), we obtain 2kF = O.295b* with the superlattice period
As = 3.39b. The charge transfer is therefore f = 0.59 electrons.(Sl)
Further characterization of these 2kF satellites was carried out in the
work of Comes et at.(m These experiments established that the two pre-
viously observed electronic transitions at 38°K(82.83) and 54°K have clear
structural counterparts. Below 3g u K, satellites appear at the points
(±O.25a*, ±O,295b*c*), in agreement with the x-ray data of Kagoshima et
at.(36) At 38°K, the modulation in the a* direction becomes incommen-
surate and displays an abrupt increase gradually reaching the value 0.5a*
at 49°K and remaining at this value until the 54°K transition.(37) Below
38°K all satellite widths are resolution limited, indicative of long-range
order. However, the satellite peaks do show progressive broadening in the
transition region as T approaches 54°K.
Above 54°K, one-dimensional precursor effects attributed to the giant
Kohn anomaly were observed in the x-ray studies. Well-defined diffuse
sheets occur at O.295b*, indicating that 2kF , and hence the charge transfer,
remains constant through the ordering transition. From an estimate of the
width of the 2kF lines, the coherence length of the fluctuating distortion is
about 50 lattice constants at 58°K.(40) With increasing temperature the 2kF
diffuse lines weaken in intensity with considerable broadening. (38-40) The
residual 2kF scattering at room temperature(39) is broad and barely
observable and suggests a coherence length of a few lattice constants.
The half-width at half-maximum (HWHM) and intensity (normalized
by 1/ T to take the temperature dependence out of the Bose factor) are
plotted in Figures 7a and 7b. The data were obtained from Khanna et al.(40)
using microdensitometer tracings of the 2kF sheet. Figure 7a provides a
measure of the growth of the 2kF coherence length. The diffuse scattering
intensity is proportional to the number of phonons, 12kF oc kT/hw2kF. Thus
Figure 7b is a measure of the phonon softening in the vicinity of 2kF (i.e.,
within experimental resolution of 2k F ). Figure 7c shows the product
(I2kF/T)(HWHMbF' Note that the product in Figure 7c is approximately
constant, suggesting that the dominant effect in the range from 3000 K
down to 55°K is the growth of coherence with the associated temperature-
dependent softening being a smaller effect.
This conclusion is reminiscent of the results obtained for the CDW
instability in the layered compounds. McMillan(84) has analyzed the latter
in terms of a strong-coupling CDW theory in which a local (short
coherence length) distortion remains well above the three-dimensional
ordering temperature. Such a local distortion (in effect a collection of
incoherent small polarons) may be the source of the temperature-depen-
dent susceptibility and optical pseudogap. As the temperature lowers, the
local distortions would begin to order into a long coherence length 2kF
mode as observed in the scattering data.
Inelastic neutron scattering investigations have resulted in observation
of a giant Kohn anomaly at 2kF = O.295b*. Shirane et al.(41) observed an
anomaly which develops clearly only below 150o K. The temperature
dependence of the 2kF scattering as observed in the diffuse x-ray stud-
ies(38-40) described above is in general agreement with the results of
Shirane et al.(41) The strong temperature dependence (see Figure 7) rules
out the possibility of a sharp phonon anomaly at room temperature,(85)
which would require enhanced phonon amplitude at 0.295b*.
Thus the giant Kohn anomaly predicted for the one-dimensional
Peierls' instability has been observed in TTF-TCNQ. The results of
Shirane et al.(41) show a broad and weak anomaly at room temperature
developing into a sharp structure below 150 0 K and becoming more pro-
nounced as the transition at 54°K is approached. Below 150 0 K the anomaly
develops near the transverse branch at 2k F .
In addition to the 2kF scattering, x-ray studies(38-40) have revealed
diffuse scattering at O.59b*, which is equivalent to 4k F. This indication of
two anomalies has been interpreted(38) as arising directly from the two
96 A. 1. Heeger
IS~-------------------------------r~~~~
I TTF - lCNa
.]s.-------------------~I------,
A- I TTF - lCNa
.] HWHM
b
.2
.15
.1
.05
60 ISO 200 250 300

7"K
c
0 .6 ~ 0 2kF
::::;: 0.5 ~ 0 0
r
::: 0.4 x_ _ _ _ _ x
r
x 0 .3
I- 0.2
----
.....
H
0.1
I I I I I I I I
0
60 80 100 120 140 160 180 200 220 240 260 280 300
TEMPERATURE (OK)
FIgure 7 (a) ScatterIng mtenslty (II T) as a functIOn of temperature for the 2kF and 4kF
anomalIes as obtamed from dIffuse x-ray studIes (see Reference 40) (b) Half-wIdth at
half-maxImum (HWHM) as a functIOn of temperature for the 2kF and 3k F anomalIes as
obtamed from dIffuse x-ray studIes (see Reference 40) (c) The product, (II T) x (HWHM) as a
functIon of temperature for the 2kF and 4kF anomalIes
chain structure of TTF-TCNQ with independent TIF and TCNQ chains.

In this picture, one type of chain behaves as a normal one-dimensional
metal and gives rise to the expected 2kF Kohn anomaly at O.295b*, and the
other type of stack, because of electron interactions, gives rise to the
second anomaly. Repulsive interactions between electrons in the strong
coupling limit on one molecular species would require single occupancy of
each momentum state and thus would double the Fermi wave vector giving
rise to the 4kF anomaly. Using somewhat weaker intrachain Coulomb
coupling Emery(86) has shown that a 4k p anomaly can arise from phonon
modulation of the electron-electron interaction, giving rise to simul-
taneous scattering of two electrons across the Fermi surface with a
momentum change of 4k F . Finally, it has been suggested(87) that the scat-
tering at 0.59b* is the 2kF anomaly with the O.295b* scattering arising
from spin-wave-phonon interaction. Each of these mechanisms requires
relatively strong intrachain Coulomb interactions. However, as described
in Section 2.2 above, such interactions do not appear to be dominant.
Moreover, the nuclear relaxation data indicate little or no difference
between the strength of intrachain interactions on the TTF and TCNQ
chains. The origin of the 4kF scattering may well be in the interchain
coupling. Lee et at. (88) have shown that the inclusion of interchain coupling
leads to a 4kF anomaly even for zero on-site Coulomb repulsion.
Moreover, the strong interchain attractive forces which ultimately lead to
the stabilization of the solid are necessarily the largest energy in the
problem. The effect of phonon modulation of the interchain coupling in
second order (see Section 2.3) should be studied as a possible source of the
4kF scattering. This second-order effect (with energy scale of order 4 e V)
might be more important at high temperatures than the first-order effect
from the modulation of the transfer integrals (energy scale -0.4 eV).
4. The Pseudogap: Optical Properties
The optical properties and frequency-dependent dielectric function

expected for the Peierls-Frohlich charge-density wave conductor have
been described in detail in earlier publications. 0,13,14) The 2k p scattering of
the electron Bloch waves in the vicinity of kp leads to a pseudo gap in the
density of states, which becomes a real gap when the CDW is pinned at low
temperature.
In the fluctuation regime, as shown in the previous section, the finite
coherence length leads to states within the gap. However, because the
coherence length is considerably greater than a lattice constant, these
states are strongly scattered and hence effectively localized for hw < 2d.
The conductivity in this region would therefore be diffusive with very low
98 A. 1. Heeger
mobility and should go to zero as the coherence length grows long. For
frequencies hw > 2~, the optical conductivity would be unaffected by the
2kF scattering and hence should show the high-frequency Drude-like
behavior expected for a metal, in the ideal case. The expected increase in
conductivity at hw ~ 2~, in semiconductor terms, is due to the interband
transition across the Peierls gap. Finally, at very low frequencies, oscillator
strength associated with the Frohlich CDW collective mode should
dominate.
A schematic diagram of O'I(W) vs. w showing the characteristic fea-
tures of the collective mode and single-particle excitations is shown in
Figure 8. In the conducting state the collective mode is centered near zero
frequency with a width determined by the collective mode lifetime, 'Te(T).
The pseudo gap leads to a broad minimum in O'l(W) with the single particle
(semiconductor interband) transitions showing up at higher frequencies.
Pinning of the collective mode will shift the oscillator strength into the
infrared.
The fundamental signature of the CDW collective mode Frohlich state
is therefore relatively high electrical conductivity at dc and microwave
frequencies in the presence of a pseudogap in the electronic excitation
spectrum. The low-frequency conductivity is expected to be extremely
sensitive to pinning by defects and impurities. These are precisely the
features observed in TTF-TCNQ for T> 54°K (see Section 5 below).
Polarized reflectance data obtained from single crystals are shown in
Figures 9a(13) and 9b.(X9) Figure 9a shows the room-temperature results for
Ellb and Ella in the range 50-10,000 cm -I. The data below 300 cm- I were
obtained from unpolarized reflectance from a carefully aligned mosaic of
single crystals mounted on thin gold wires. In this long-wavelength region,
- - E -~
9 I
0", (w) I
w
Figure 8. Schematic representation of the frequency-dependent conductivity of the Peierls-
Frohlich conductor showing the collective-mode peak and the single-particle oscillator
strength at higher frequencies.
300K
w
u
Z
<t
f-
u 0.4
w
-'
u.
w
a::
0.2
(a)
0'
0 2000 SCXXl 10,CXXl
FREQUENCY (em')
to.
e.g
0..8 Ellb
/
,r\ 15K
0..7 ,..}'v'~.~, ISOK
300K
\ ~
\\ 1'''--- . . . . . . .
o.S J "-
w
u "-
Z "-
~ 0.5 "-
u "-
~
u.
"-
UJ
\
0.4 \
'" \
\
0.3 \
\
(b)
a 1000 2000 3000 <000 5000 6000 7000 8000.
FREQUENCY (em-I)
Figure 9. (a) Polarized reflectance of TIF-TCNQ single crystals over the whole region,
50-10,000 cm -1 The dashed lines below 350 cm -1 indicate data derived from the unpolarized
reflectance of an oriented mosaic (Reference 13). (b) b-Axis reflectance of TIF-TCNQ single
crystals at selected temperatures (Reference 89).
100 A. J. Heeger
~(Ellb) was obtained by assuming £ll(Ella) is constant at the 300 cm- I value
so that ~(Ellb)= 2~meas-~(Ella) (polarized data in the far ir are sum-
marized below).
The single-crystal measurements were repeated(89) in independent
experiments and extended to lower temperatures as given in Figure 9b,
where the data for Ellb from 1000 to 8000 cm - I are shown. The plasma
edge sharpens up with decreasing temperature, consistent with the earlier
results on single crystals(ll) and thin films.(13·14) In addition, the absolute
value of the long-wavelength reflectance increases significantly from about
60%-65% at room temperature to about 80% at 15°K with intermediate
values at 160 oK. This temperature-dependent increase in reflectance
indicates that the relatively low reflectance is intrinsic and not the result,
for example, of imperfect surface scattering.
Single-crystal polarized reflectance measurements have been exten-
ded into the far ir.(90) Results for two representative temperatures, 4.2°K in
the insulating, high dielectric constant (pinned) regime and 1000K in the
conducting (unpinned) regime are shown in Figure 10. Data obtained from
different runs and on completely different mosaic samples were in good
agreement.
At low frequencies (v < 20 cm -1) the polarized 1000K reflectance
turns rapidly upward, heading toward 100% as required for a system with
large dc conductivity. This observation of nearly complete reflectance at
the lowest frequency is of critical importance for it demonstrates that the
65%-70% value in the region above 20 cm -I is intrinsic and not the result
of experimental difficulties such as diffuse scattering, small particle effects,
interrupted strands, misaligned mosaic, etc. The corresponding structure
(decrease) in the 4.2°K reflectance below 20 em-I will be discussed below.
Figure 10. Polarized (Ellb) reflectance of

single-crystal mosaic of TIF-TCNQ at
lOooK and 4.2°K (Reference 90).
The relatively low b-axis reflectance at long wavelengths indepen-

dently implies the existence of an energy gap at both temperatures. The
reflectance expected for a simple Drude metal with a plasma edge near
7000 cm -I and a dc conductivity of order 50000 cm -I would be in excess
of 90% for all frequencies below 100 cm -I. Therefore, the low reflectance
( -70%) at lOooK above 20 cm -I implies a very small FIR conductivity in
the gap region. An estimate of the dielectric constant and conductivity in
the gap can be obtained by comparison of the low- and high-temperature
data. At 4.2°K TTF-TCNQ is an insulator; therefore, the low-frequency
reflectance is determined completely by the real part of the dielectric
constant e I(W), Using the standard relation
(e t!2 _1)2
.cJl(W) = (e:!2 +1)2 (46)
gives the value e 1(20 cm- 1< jj < 80 cm -I) "'" 90, in excellent agreement with
the thin-film results and the room-temperature single-crystal results
obtained from a full Kramers-Kronig analysis of .cJl(w). The small overall
increase in reflectance above 30 cm -I at lOooK compared to that at 4.2°K
can be used to set an upper limit on the far-infrared conductivity in the
gap region. Expanding the full expression for reflectance in the limit
e21 e 1 < lone obtains
.cJl(W)=(:l;::~r (1+:;/~) (47)
which leads to an estimate of the conductivity in the gap region at lOOoK of

(Tb (50cm- I )=50(Ocmr l , two orders of magnitude below the dc and
microwave values.
Using this result, the frequency-dependent conductivity at long
wavelengths is plotted in Figure 11. The far-infrared data together with the
typical dc(16) and microwave data(17) at lOooK are shown. The solid curve is
from an approximate analysis of the E//b reflectance using a single Lorentz-
ian oscillator centered at zero frequency and a residual conductivity in the
far infrared of 50 (0 cmrl. These results provide clear evidence of the
pseudogap present in the conducting regime above 54°K.
The value obtained for e 1 (w) in the far infrared can be used to obtain
an estimate of the Peierls gap using the standard expression for the single-
partIc utlOn (176791)
. Ie contn'b' ' ,
(48)
With hW p =1.2eV and e~=2.5, we obtain hw g =O.leV, in good

agreement with the results of the full Kramers-Kronig analysis(l3,14) of the
single-crystal and thin-film reflectance data as sketched in the inset to
102 A. I. Heeger
9000r 8000
T
§ 6000
8000
8CXXl
7000
I~ 6000, 1,.:.
E _ 60
~ 5000 ~ 3 40-
.;- 20 -
3
I -20
b 4000
1 i7 (em')
I
3000
1
2000 jl
1000
o
o
~,=.-.=--.-.-.-.-.-.-.-,=.-e-e-t-e-t-e-.
4 8 12 16 20 24 28 32 36 40 44 48
FREQUENCY (emt)
FIgure 11. Frequency dependence of <TI(W) below 50 cm- J at lOooK. Inset: Sketch of full
frequency dependence of <TJ(w) and Ej(W) at lOooK (Reference 102).
Figure 11. The full frequency dependence of crt

(w) shows a maximum at
1000 cm -1 with a smooth Drude-like frequency dependence observed at
higher frequencies. The corresponding behavior of E t (w) is like that of a
semiconductor with transitions across the energy gap sufficiently strong to
give negative values between 1000 and 6000 cm -1. At lower frequencies d
is positive with magnitude of approximately 100 in the gap region. From
the maximum in crt t
(w) and the zero crossing in E (w), we infer a pseudo-
gap with magnitude hW g =0.1 eV (10 3 cm-\
The results in the conducting regime are therefore qualitatively
consistent with the x-ray observations of strong 2k p scattering from the
dynamical distortion with relative long coherence length.
An alternative explanation of the optical infrared properties has

recently been proposed by Gogolin and Mel'nikov(92) based on the well-
known effects of localization of wave functions by scattering in one-
dimension. In their theory the zero crossing in E(w) and the peak in u(w)
occur at the characteristic frequency w = 1/ T, where T is the scattering
time. Although they obtain good quantitative agreement with the general
features of the thin-film data above 50 cm -1, their theory is unable to
explain the large dc and microwave conductivity in the presence of the
assumed localization. However, the dependence of the spectral features on
the magnitude of T (e.g., by comparing film and single-crystal data, or by
purposely adding defects) should be studied to determine if, indeed, such
scattering localization is the dominant feature.
The dielectric constant, c 1, of a Drude metal would be negative at
frequencies less than the plasma frequency; near zero frequency the Drude
dielectric constant approaches
With typical values for u ~ 10 3 0- 1 cm -1 and hwp ~ 1 eV (coo is the high-

frequency core dielectric constant, Coo = 2-3), c1(0) should be negative with
magnitude of order 10 2 • However, as shown above, the Kramers-Kronig
(KK) analysis of the room-temperature reflectivity data for TTF-TCNQ
and (TSeT)zBr(92a) shows c 1 to be positive at low frequencies in the far
infrared ( ~ 10 2 cm -1).
Taking into account the difficulty of performing the optical reflectivity
measurements on small crystals, the problem of surface quality control,
and the indirect nature of the data obtained by using the KK analysis, it is
important to have direct measurements of c 1 at microwave frequencies on
a series of organic metals in the high-temperature conducting regime.
However, the cavity perturbation method commonly used for this purpose
at low temperatures is insensitive to c 1 when C2 = 47ru/ W is large.
The results of a microwave (~1 cm- 1 ) standing-wave experiment have
recently<92b) provided a measurement of c 1 at room temperature. The
results for the series of compounds (see Table 3) indicate c 1 to be large and
positive, which is inconsistent with simple metallic behavior. The positive
room-temperature values imply the existence of a zero crossing in c1(W)
below the plasma frequency, consistent with the results of the above-
mentioned infrared (ir) studies on TTF-TCNQ (see above) and
(TSeT)zBr.(92a) The large magnitudes suggest a strong resonance in u(w) at
relatively low frequencies.
The results summarized in Table 3(92b) are incompatible with a picture
of these compounds as simple Drude metals. However, these data are in
qualitative agreement with results obtained from infrared optical studies.
104 A. 1. Heeger
Table 3. Room-Temperature Values for EI and (J' Obtained with the Microwave
(~I cm ') Standing- Wave Technique and the Dielectric Constants at 4.2°K
(Where Available) Determined from Cavity Perturbation Measurements at
0.3 cm- I
Number of
Material crystals O'(flcm)-I £1 (RT) £1 (4.2)
Q-TCNQ 3 130±25 900±200 300

TfF-TCNQ I 200 2500 3500
TSeF-TCNQ :\ 650± 100 6000± 1500 15,000
TMTIF-TCNQ 4 260±60 2200±700 2600
HMTFF-TCNQ I 405 4500 700
(nTh-I, 4 7S0± 100 14,500±600 Not measured
(TSeThBr 3 IOOO±SO 22,600 ± 1500 Not measured
In the case of TTF-TCNQ, reflectance and electron energy loss measure-

ments indicate a plasma edge with the dielectric function crossing zero to
negative values below 7500 cm- I . Extension of these results into the far ir
on films and single crystals indicates a second zero crossing to positive
values below 1000 cm -I reaching a value of about 80 at 300 cm -I. The
data of Table 3 then imply that E J continues to increase to its value of
several thousand at 1 cm -1. This would necessarily verify the presence of a
"gap" in the single-particle excitation spectrum. Similar results(92a) have
been obtained for (TSeThBr where E I increases from E I = 80 in the far
infrared to E J = 22,600 at 1 cm -I. Far-infrared data are not available for
the other systems. However, the results of Table 3 indicate that the second
zero crossing is a general feature.
The low-frequency dielectric constant can be written in terms of the
frequency-dependent conductivity using the Kramers-Kronig relation
£1(0)=1+8 1o
00 O'(w)- 0'(0)
w
2 dw
Thus, independent of detailed model, the large magnitude of EJ at

microwave frequencies implies additional oscillator strength at very low
frequencies, suggesting a strong low-frequency resonance in O'(w) centered
at a finite frequency (> 1 cm- I ), but with a zero-frequency tail responsible
for the dc conductivity.
In the context of the schematic representation of O'(w) presented in
Figure 8, these data for E I in the "metallic" regime imply that the collective
mode is centered above w = 0, thereby giving E 1 large and positive at
microwave frequencies. We note that the dielectric constant measured at
room temperature is of the same magnitude as that measured at 4.2°K;

and, moreover, the 3000 K and 4.2°K results appear to scale in the various
systems. Thus the dramatic change in dc conductivity below the metal-
insulator transition involves only the strength of the zero-frequency tail of
the collective mode with little or no change in the center frequency.
5. Electrical Conductivity
5.1. DC Measurements
Standard four-probe electrical conductivity measurements along the

crystallographic b axis on all samples showed that, with decreasing
temperature, the conductivity rapidly increases to a maximum value which
varies typically from crystal to crystal. The temperature at which the
maximum occurs, TM, is in the range 58-60o K. Below the maximum, the
conductivity suddenly decreases, characteristic of a metal-insulator tran-
sition. The variation of the conductivity maximum from crystal to crystal is
due to the extreme sensitivity of one-dimensional metals such as TTF-
TCNQ to crystalline defects, twinning, and impurities.
DC conductivity measurements on TTF-TCNQ have been further
examined using the techniques(93) of Montgomery and co-workers on
highly anisotropic conductors. Montgomery's results can be applied(16) to
the question of inhomogeneous currents wherein a nearly zero current
density can result at the surface where the voltage leads contact the
anisotropic sample. In this method, one purposely uses a lead configuration
designed to give a well-defined inhomogeneous current distribution. When
the current is applied with the leads aligned parallel to the principal
conducting axis, the voltage drop between the opposite two leads is
considerably reduced by the anisotropy. The apparent conductivity in this
configuration can be approximated by(16)
(49)
where CTII(CT.L) is the intrinsic conductivity parallel (perpendicular) to the

principal axis, 11l(l.L) the corresponding distance between current and
voltage leads, A the anisotropy CTIII CT.L, and G a simple geometric factor.
The apparent conductivity would then be determined by the magnitude
and temperature dependence of the anisotropy CTIII CT.L. Consequently,
detailed measurements of the temperature dependence of the electrical
anisotropy have been carried out. (16)
The b-a anisotropy (CTit / CT~) is shown in Figure 12 as a function of
temperature. The magnitude of the anisotropy implies pseudo-one-
dimensional behavior as described above. The experimental observation of
106 A. f. Heeger
I I
6000 -
j\
.1 I\
5500 -
-
jl · \
5000
If·
-
4500
-
:,: ::::' f. \ .~
4000
-
-
2500 f- \.J
Y -
-
.-.-.-
2000 f--
1500
1000 f- -
500 f- -
I I I I I I I I
030 40 50 60 70 80 90 100 110 120
T (OK)
Figure 12. The electrical anisotropy of TTF-TCNQ[(Tlr/(T~l as a function of temperature

(Reference 16).
double maxima in the anisotropies has been used as a convenient internal

check(16) on the validity of O'~ (T) four-probe measurements. False giant
conductivity maxima were generated using a completely misaligned lead
configuration similar to that used by Schafer et al.,(94) which is nothing
more than a slightly modified Montgomery configuration. The fact that the
anisotropy dominates the measurement is clearly signified by the double
maximum in the false conductivity. The sensitivity of the double anisotropy
maxima as a self-consistent check on the validity of O'it data has been
shown experimentally. Experiments show(16) that even a slight enhance-
ment due to inhomogeneous currents is accompanied by the clear signature
of the second peak in the data near 40°K.
The two peaks in the anisotropy correspond to the two phase tran-
sitions observed in the conductivity studies. The first peak corresponds to
the onset of long-range order near 54°K. At that temperature O'b drops
rapidly (a large minimum in din O'b/ dT) while O'a varies smoothly.
Similarly, the lower-temperature peak (Figure 12) arises from the 38°K
transition where again d In O'h/ dT has a sharp minimum. The 38°K tran-
sition was first observed by Weger et al.(82) and later demonstrated

conclusively by Etemad.(83)
The value of 0'11 (TM )1 0'11 (300 0 K), the maximum normalized conduc-
tivity, varies from crystal to crystal and is associated with the extreme
sensitivity of one-dimensional metals to crystalline defects, twinning, and
impurities. The intrinsic anisotropy altO': > 10 4 near 58°K makes TTF-
TCNQ extremely sensitive, and crystal perfection at the level of parts per
million is required, for any defects will either remove a given chain alto-
gether, force carriers to tunnel through the defect, or force transverse
current flow between chains.
In order to quantify the role of defects on the conductivity peak ratio
(CPR), and the magnitude and overall temperature dependence of the
conductivity in the range T> 54°K, a study of the effect of radiation-
induced defects was undertaken.(95)
Fast-particle bombardment is known as one of the means of controlled
introduction of lattice defects into crystalline solids, and hence is an
excellent tool for studying in a general way the influence of such defects on
the physical properties. In particular for organic one-dimensional solids,
radiation-induced defects lead to cause-effect relationships which are
direct and easy to establish from studies on a given set of crystals, and do
not rely on subsequent synthesis and crystal growth. The radiation-induced
defect studies therefore serve to explicitly quantify the importance of
defects in nominally "pure" crystals.
Studies of the low-temperature transport(96) after irradiation at 5 x
10 cm- 2 indicate that the ohmic conductance at 4.2°K was increased by
14
more than two orders of magnitude, and reproducible from sample to

sample. This defect contribution is described(96) as O'd = (T~ exp( - Edl kBT)
with (Edl k B ) = 20 and O'~ = 0.5 (n cmr 1. The single characteristic activa-
tion energy implies a single dominant defect. Spin resonance studies(97) of
the most heavily irradiated samples (flux 5 x 10 14 cm- 2 ) indicate an induced
concentration of spin-l /2 impurities of approximately 0.1 % (1000 ppm).
The measured value of 0'~=O.5 (ncmr 1 is consistent with 0.1%
donor I acceptor states with a carrier mobility of order 2 cm 2 IV sec, consis-
tent with that inferred from the room-temperature conductivity of TTF-
TCNQ. Detailed studies of the resistivity in the temperature range 35-
60 0 K show that for samples exposed to 5 x 10 14 cm- 2 of 8-MeV deuterons,
the phase transitions, observed in pure samples as anomalies in the resis-
tivity, (98) are suppressed. Similar experiments on alloys(99) indicate a
comparable suppression of the phase transition at a level of about 3%
TSeF in TTF-TCNQ. Because of the structural and electronic similarity of
TSeF to TTF and the fact that the transitions are driven by the TCNQ
chains, (58, 100) we expect that for a given concentration the effect on the
phase transitions of alloying will be somewhat less than that due to the
108 A. J. Heeger
18 .---.---~--~--~----
Figure J3(a). The electrical conductivity of

TTF-TCNQ after irradiation as a function of
::.:: temperature near the conductivity maximum
o (Reference 95). e e e, nonirradiated;
o,...,
+.+ +, 1.1 x 1013 cm- 2 ; t-, t-, t-" 2.4x 1013
b cm- 2 ; XXX,5.2x10 13 cm- 2 ;OOO,9.3X
"- IO!3 cm- 2 ; 0'0 0,51.5 X 10 13 cm- 2
I-
3 ~ __ ~ __- L__ ~ ____L -___
50 60 70 80 90 100
TE MPERATURE (K)
E
<.)
Figure l3(b). Resistivity In the

transition region from several
samples with progressively higher
defect concentration (Reference 95).
(a) nonirradiated; (b) 1.1 x
10 13 cm- 2 ; (c) 2.4x 10 13 em- 2 ; (d)
5.2 x 10 13 cm -2; (e) 9.3 x 10 11 em -2; TE MPERATURE (K)
(051.5 x 10 13 em 2
i'rradiation-induced defects. Thus, we conclude that the value of 0.1 %

defects from an incident flux of 5 x 10 14 cm -2 is approximately correct. We
shall take this value to set the scale of the concentration of induced defects.
Therefore, the experimental results cover the range from approximately
20 ppm to 1000 ppm.
Figure 13 shows the electrical conductivity of TIF-TCNQ after
irradiation(95) as a function of temperature near the conductivity maxi-
mum. The series of nested curves (each one normalized to the room-
temperature value measured for that crystal) result from progressively
higher flux or equivalently progressively higher defect concentration
(Figure 13a). The typical behavior of TTF-TCNQ has not changed due to
the irradiation; the electrical conductivity increases with decreasing
temperature, goes through a relatively sharp maximum at a temperature
TM, and becomes activated (a semiconductor) at low temperatures.
The phase transitions, observed in pure ITF-TCNQ in the tempera-
ture range 35-54°K, remain observable for the irradiated samples.(95)
However, the transitions broaden and shift to lower temperature with
increasing defect concentration. Figure 13b shows the resistivity in the
transition region from several samples with progressively higher defect
concentrations.
The results for the three characteristic temperatures are shown in
Figure 14(95):
110 A. 1. Heeger
70 r---.---~----~---r---.----.
Figure 14. Characteristic tempera-

40 tures of TTF-TCNQ as a function of
incident flux or defect concentration;
TM is the temperature of the conduc-
tivity maximum, T J is the first tran-
sition associated with the onset of
long-range order in pure TTF-TCNQ,
30 ~--~--~----~--~----~--~ and T2 is the 38°K transition asso-
o 10 20 30 40 50 60 ciated with the locking of CDWs
13 2
FL U X ( 10 lem ) (Reference 95).
TM: The temperature of the conductivity maximum occurs at 58°K in

pure TTF-TCNQ and increases to 68°K at a defect level of 1000 ppm.
Note that samples with conductivity maximum at 60 0 K have on the order
of 25 ppm added defects.
T 1 : The first transition, associated with the onset of long-range order
in pure TTF-TCNQ, shifts to lower temperatures. The effect of disorder is
to suppress the phase transition from 52.3°K in our pure samples down-
ward with an initial slope dT2 / de = 150 K per percent defects.
0
T 2 : The 38°K transition, associated with the final locking of the

CDWs on the two chains in pure TTF-TCNQ also shifts to lower tempera-
tures; dT3/ de = 200 0 K per percent defects.
There has been considerable controversy(16.94.95.101.102) concerning the
dc electrical conductivity of TTF-TCNQ centering on the absolute magni-
tude of the room-temperature conductivity, the size of the conductivity
peak ratio [CPR = a(Tja(3000K)] and the temperature at which the peak
occurs (Tl)' The results for TM are shown in Figure 14,(95) the results on
the absolute magnitude of a (300 0 K) and for the CPR are shown in Figures
15 and 16.(95)
J
~~ -T~---'
TTF-TCNQ
SINGLE CRYSTAL
>
t:; 60 o
o
J
:::J
o o
Z o
o o
u 40
Z
w
(j)
<l 20
w
cr
u
w
o L _ _I
o --
o 10 20 30 40 50 60
FLUX (ld 3/cm 2 )
FIgure 1') Room-temperature conductIVIty, CT(3000K), as a functIOn of particle flux and

defect concentration The solId curve represents on-lIne measurements of a smgle sample, the
expenmental pomts are from separate samples measured subsequently (Reference 95)
Figure 15 shows the room-temperature conductivity, o-(300 0 K), of

TIF-TCNQ as a function of particle flux and defect concentration. The
solid line represents the experimentally determined variation of the
conductivity obtained by on-line measurements. A sample was mounted
for four-probe conductivity measurements and placed in the scattering
chamber; the voltage and current were remotely monitored at short inter-
vals (beam off) dunng the Irradiation. The experimental points were
obtamed from samples IrradIated and subsequently measured. The
---
extreme sensitivity of the room-temperature conductivity to defects is
16~ .
i\
,
8
r<l
14
b 12 - •
~
\.
10
~.--
o
E
--
8
FIgure 16 The conductIvIty
peak ratIO as d functIOn of 6
partIcle flux and defect con-
o 10 20 30 40 50 60
centrdtlon (Reference 9')

112 A. 1. Heeger
clearly evident. The room-temperature conductivity values reported by

other laboratories(1OI) typically fall in the region from 300 to 500 (0 cmf!.
Figure 15 demonstrates that such values are characteristic of defect levels
in excess of about 100-200 ppm. Thus the correlation suggested earlier(102)
between the room-temperature conductivity and increased sample quality
is verified.
The conductivity peak ratio, O'(Tj 0'(300 0 K), is plotted as a function of
particle flux and defect concentration in Figure 16.(95) As long as the
CPR» 2, this quantity will be an approximate indicator of sample quality,
in effect a measure of the residual resistivity if interpreted in the con-
ventional metallic sense. With the starting samples having only moderate
quality (CPR -16), it is not possible to extrapolate back to predict the
CPR of the purest defect-free TTF-TCNQ. However, by observing the
defect concentration needed to reduce the CPR by a factor of 2 we can
evaluate the approximate defect and/or impurity level in nominally high-
purity TTF-TCNQ, i.e., characterized by 0'(300 0 K) = 700 (0 cmf! and
CPR = 16-20 (typically). Figure 16 quantitatively characterizes such
samples as having of order 100 ppm defects and/or impurities.
The temperature dependence of the electrical resistivity of TTF-
TCNQ in the high-temperature conducting regime was discussed by Groff
et alyo3) Taking the point of view of simple metallic behavior, the resis-
tivity is separated into a temperature-dependent intrinsic term and a
temperature-independent residual term, i.e.,
(50)
Equation (50) assumes a simple power law for the intrinsic temperature
dependence. Underlying Equation (50) is the assumption that Po will be
proportional to the defect concentration whereas the power, A, and the
coefficient A = PI Tr/ are assumed to be intrinsic and therefore indepen-
dent of defects at least in the low-concentration limit.
In an attempt to evaluate the validity of Equation (50), least-squares
fits to the data were carried OUt.(95) To avoid a preference for either
conductivity or resistivity, the following was minimized:
X2 == I (Y! - Yeal)2 (51a)

, y,
where y, denotes the data point at a temperature T and Yeal is the cal-
culated value from Equation (50). For each A, x == TA was defined so that
Peal = Po + Ax and the matrix equation aX2 / apo = 0, aX2 / aA = 0 was
inverted to find the best po and A. The power A was incremented to find
the absolute minimum in X2. For a given set of data (over a given tempera-
ture range) there is no fitting error; i.e., A, Po, and A are the best
parameters that describe the data. For a given temperature range the only
3.00 ,..---- - - - - , - - - - - - - r - - - - - - - - ,
2.8 0
,"
2.60 ....." ,,'
, ' ."41 .,...,,1······ ,"
Figure 17. Power law parameter A

2 . 20 ~
as a function of TF where the fitting
parameters were determined from 2.00 50 100 - - -I...J.S-0- - - -2=-'00
data from successive intervals TF <
T < 3000 K (Reference 95)
error comes from the scatter in the data. The variance in A was calculated
using the formula
LlA = [ ~ (::1 Llyrf /2

(SIb)
where Lly. is the error in an individual data point. For an (overestimated)

scatter of Lly/y =2%, we find LlA =0.035.
Figure 17(95) represents a test of the validity of Equation (50). For a
typical nondamaged sample the fitting procedure(95) described above was
carried out over a temperature interval TF < T < 300o K. The resulting AF
is plotted as a function of TF . If Equation (50) were a true description of
the data, the same AF would be found for any temperature interval, i.e., AF
would be independent of TF . The results show AF varying significantly; the
error bar indicates the variance, ~A, as calculated above assuming 2%
scatter in the data. The value of AF approaches A = 2.3 when the fit
includes the entire temperature interval 800 K < T < 3000 K in agreement
with the results of Groff et aZ.o°3 ) who found A = 2.33 ± 0.14. However, the
strong dependence of AF on TF indicates that Equation (50) is only a rough
approximation and should not be taken to have fundamental physical
significance. The true temperature dependence is more complex, involving
at least higher terms and possibly an altogether different functional form.
Nevertheless, with the caveat in mind, it is useful to categorize and quantify
the effects of induced defects in terms of the parameters of Equation (50).
Least-squares fits of the data to Equation (50) were carried out for
irradiated samples over the full range of induced defect concentrations.
The results(95) are shown in Figures 18-20. Figure 17 shows the best fit to
two samples, one not irradiated and the second exposed to a total flux of
5.2xl013cm~2. The data are plotted as Inpvs.ln T to show clearly the
entire range. Absolute values of p are given for the two samples. As
indicated above, the resistivity increases on irradiation. The overall
114 A. f. Heeger
4.0
3.0
2.0
1.5
10- 3
8.0
6.0
E
~
Cl 4.0
Q... 3.0
2.0
1.5 +
10- 4
8.0
6.0
4.0
40 60 80 100 150 200 300
T (K)
Figure 18. Examples of best fit (solid line) of Equation (50) to two samples; the lower-
resistivity sample was not irradiated. the higher-resistivity sample was exposed to a total flux
of 2.8 x 10 14 cm -2 (Reference 95).
temperature dependence for both samples is in agreement with Equation

(50) as indicated by the solid curves which represent best fits (input data for
fitting restricted to the region above lOOOK). The parameters for the two
curves are given in the figure caption. As expected, Po increases with
induced defects. However, analysis shows that over the defect range
studied, no significant change in A could be detected (i.e., any small change
that might be present is within the uncertainty as described above; see
Figure 17).
One qualitative feature of the data shows up in plots such as given in
Figure 18. In high-purity samples, the data points between about 58°K and
68°K typically fall below the fitting curves, suggesting the possibility of an
additional contribution(104) to the conductivity near the maximum.
800 - ~ ,------,----,-------.----,
600 -
E
o,
•
c:; 400 -
=l o~ -
/
0 1 •
Q;: 200 L~ o~
~-
O~--L---~--~---L--~--~
o /0 60
FLUX (l OI3/cm2)
Figure 19. Residual resistivity, Po, as a function of particle flux and defect concentration. The
open circles result from fitting the data to Equation (50); the solid points correspond to the
measured minimum resistivity (at TM) (Reference 95).
Figure 19 shows the residual resistivity, Po, as a function of flux and

defect concentration.(95) The results indicate that Po = ac, where c is the
defect concentration and a = 5 X 10- 3 (0 cm)/%. The extreme sensitivity
of the quasi-one-dimensional system to defects is evident. Figure 20 shows
the coefficient A = PI To" [see Equation (50)] as a function of flux and
defect concentration.(95) Note that both the values for Po (Figure 19) and A
(Figure 20) are dependent on the careful determination of sample dimen-
sions and require stringent reproducibility for different samples. The data
40~
3.0
't'
0
0
0
0
«
"-
«
1.0,-
60
FLUX (l 0 13/ cm 2)
Figure 20. The prefactor A = PI To' from Equation (50) as a function of particle flux and
defect concentration (the initial value is Ao = 1.1 x 10-8 ). The solid curve is the normalized
on-line room-temperature data from Figure 15 (Reference 95).
116 A. f. Heeger
meet this standard as evidence by the excellent agreement between the

on-line data and the room-temperature results from different samples
shown in Figure 15. The strong dependence of the coefficient A on defect
concentration is unexpected and represents an unusual breakdown of
Matthiessen's rule. The sensitivity of the "intrinsic" temperature-depen-
dent term to defects is not an artifact of the data analysis; the results
plotted in Figure 19 can be anticipated by direct examination of the
room-temperature data in Figure 15. The initial decrease in a(3000K)
followed by saturation at higher flux levels is the direct result of the change
in the temperature-dependent term. Note that the smooth variation of
CPR (Figure 16) and Po (Figure 19) as a function of particle flux provides
additional evidence that the defect concentration is proportional to the
particle flux over the entire range under study.
Any discussion of the mechanism and temperature dependence of the
electrical conductivity in the "metallic" regime must take into account the
known structural aspects of the problem as described in Section 3 above. In
the conducting regime, above 58°K the dynamical distortion has a well-
defined periodicity (As = 3.4b) and a relatively long, temperature-depen-
dent coherence length. o5 --41) The x-ray observations(39.40) indicate stronger
scattering as the temperature is lowered with an apparent divergence as
T --'» 54°K (see Figure 7).
An instantaneous snapshot of the distortion in the fluctuation regime
would look schematically as shown in Figure 21. The lifetime of the
dynamical distortion is relatively long on the scale of electronic times, and
certainly greater than 1/ Wph, so that the electrons will experience a
periodic superlattice potential (As = 3.4b) over a distance A 2':: VF/ Wph 2'::
100 A. Therefore, just as the x rays scatter from the finite coherence length
dynamical distortion, so, too, would the electronic Bloch waves scatter as a
result of the electron-phonon interaction. Thus the increasing 2kF x-ray
scattering with decreasing temperature would directly imply a correspond-
ing increase in electron 2kF back-scattering across the Fermi surface with a
reduction in the electron transport mobility and the formation of a pseu-
dogap in the electronic density of states. The x-ray data for 58°K < T <
150 K therefore lead to the conclusion that the single-particle contribution
0
to the electrical conductivity should decrease as the temperature is

lowered.
The observed increase in TM with defect concentration can be under-
stood in terms of a fluctuation CDW collective transport mechanism. The
relatively long coherence length CDW fluctuations would be expected to
exhibit local random pinning of the phase in the presence of defects and
impurities as shown by Lee et al.,(46) Fukuyama et ai.,(1oS) and Gorkov.(47)
Qualitatively, we would also anticipate a rounding of the peak since the
defect pinning is random and unrelated in detail to the buildup of trans-
Figure 21 SchematIc IOstantaneous "snapshot" of the distortIOn 10 the fluctuatIOn regime.

The modulated distortIOn (wavelength As = 3 4b) IS plotted along the b axis over a length
comparable to the coherence length, g(T) (Reference 95)
verse phase coherence. Moreover, the effect of defects would be to

suppress the onset of long-range order,(l06) as observed experimentally.
The opposite dependence of TM and TI on defect concentration is
difficult to understand on the basis of single-particle theorY, in which the
decrease in conductivity just above the phase transition is usually attri-
buted to the onset of transverse coherence and the associated formation of
an energy gap. The experimental results indicate a suppression of long-
range order (and by implication the onset of a mean-field gap) to lower
temperatures while TM increases.
Alternatively, the decrease in CPR and the shift in TM toward higher
temperatures would be expected from one-dimensional localization of the
single-particle electronic wave functions due to the induced disorder.
Byc hkov, (107) Berezmsky,
. (l08)
and Gogohn. 1(109)
et a . have shown that for
independent electrons in a one-dimensional system containing defects or
impurities, the conductivity should decrease as the temperature is lowered
approaching zero with a temperature dependence determined by the ratio
of the impurity and phonon scattering rates, 7,mp/ 7ph(T). Within this
theory, the conductivity will increase with decreasing temperature until
7Ph(T) > 7,mp at whIch pomt the defect localization will dominate. The
experimental result~ for (J(T) in the irradiated samples are in qualitative
agreement with such predictions. However, a critical test of the appli-
cability to TTF-TCNQ comes from the temperature dependence of the
118 A. 1. Heeger
microwave dielectric constant in irradiated samples. In the localization

regime, co(T) > 0, as observed experimentally in QnTCNQ2 and related
systemsY 10) If s(T)~ 0 as T ~ T3 from below and remains less than zero in
the conducting regime (as is the case in pure TTF-TCNQ), then defect
localization of electronic wave functions would not be the dominant effect.
Microwave transport experiments(111) indicate that c 1 continues to go
through zero in the vicinity of the 38°K transition. Thus the defect local-
ization arguments do not appear applicable.
The effect of induced defects on the temperature dependence of the
conductivity can be discussed in terms of the parameters Po, A, and A of
Equation (50). The approximately linear dependence of po(e) is expected
and simply provides some indication of the importance of defects as scat-
tering centers. For example Po = ae with a = 4 X 10- 3 (0 cmr 11% is in
magnitude consistent with defects blocking a given chain causing interchain
(P.L) series resistance to play an important role. Since P.L is only weakly
temperature dependent above 58°K, such an effect would appear as an
apparent residual resistivity.
The TA dependence has been interpreted as resulting from single-
particle transport dominated by electron-electron scattering. (71) However,
the dependence as observed experimentally appears to be stronger than
T2. Moreover, the traditional T2 power law dependence associated with
electron-electron scattering in three-dimensional systems is not expected
in a one-dimensional systemY 12) For one-dimensional metals electron-
electron scattering resistivity should be linear in T; the Fermi surface is a
point so that angular scattering around the Fermi surface is forbidden, and
the traditional second factor of kBTI EF does not appear.
The theoretical understanding of the sliding mode ac conductivity,
(TF(T), within the Peierls-Frohlich model has developed slowly since the
original suggestion by BardeenY 13) Patton and Sham argued initially(114)
that the Frohlich state was pararesistive and later obtained results(llS)
indicating (TF = 00. Allender et ai.(116) developed a phenomenological
theory for the initial fluctuation regime at high temperature. Their results
were derived microscopically by Strassler and Toombs.(1l7) M. 1. Rice(1l8)
has calculated the dc electron-hole relaxation rate, 1/7, leading to an
estimate of the Frohlich conductivity through the relation(118)
(52)
where M* is the Frohlich charge-density wave effective mass. Since the
theory neglects anharmonic elastic effects, the resulting value of (TF
represents an apparent upper limit to the sliding mode conductivity. The
resulting electron-hole relaxation rate is given by
!=~AW6(N(wI2)) tan(~) (53)

7 2 No 4kBT
where A is the electron-phonon coupling constant, Wo is the bare phonon

frequency, No is the unperturbed electronic density of states at the Fermi
energy, and N(w/2) is the actual electronic density of states in the one-
dimensional fluctuation regime at frequency w/2. Using the Lee et al.
theory(46) for the density of states in the fluctuation regime, one obtains
(S4a)
where m* is the band mass, 2~ = Eg is the energy gap, geT) is the order
parameter coherence length, VF is the unperturbed Fermi velocity, and Ns
is the number of electrons in the condensed state. For temperatures well
below the mean-field Peierls temperature, T~F, the order parameter and
energy gap are well formed and N s / N = 1. The existence of the optical
energy gap at temperatures in the "metallic" regime provides evidence that
this is the case in TTF-TCNQ at least below IS0oK. In this case
(S4b)
where (To = (Ne 2 /m*)(b/vF)= SOO-1000 (0. cmr! using values appropriate
to TTF-TCNQ. The implied temperature dependence reflects the growth
of the coherence length geT) for which experimental data are available
from the diffuse x-ray scattering studies of Khanna et at. (40) Comparison of
Equation (S4) with Equation (SO) and the TA dependence used to describe
the conductivity data implies g(T)/ b - T-(Hl). We replot the results of
Khanna et al.(40) (see Figure 7) in Figure 22. The observed temperature
03
I
I
I
I
o
1
02 01
I
0/
o //
all
/
/ "
Sv/
o~~
0:: _ -~ RESOLUTION
~- --- - -- - - - - - - - - -- - - - --
T (K)
Figure 22. The half-width at half-maximum (HWHM) of the 2kF streak as a function of
temperature (from Khanna et at.. Reference 40). The dashed curve represents a TA+I
dependence as suggested by Equations (50) and (54b) (Reference 95).
120 A. 1. Heeger
dependence of the coherence length is consistent with a power law depen-

dence with A = 2.3 in satisfactory agreement with the transport results.
Figure 22 implies that experimentally (J' - Tg(T) with a numerical
coefficient consistent with Equation (54b).
The breakdown of Matthiessen's rule reflected in the rapid change in
room-temperature resistivity (Figure 15) and in the coefficient A = PI To)"
with irradiation is a surprising result and may simply reinforce the
conclusion that Equation (50) does not have fundamental significance.
Such an effect is difficult to understand from single-particle theory, where
the different scattering mechanisms would not interfere. From the point of
view of Equation (54) the change in A is attributed either to a change in
the effective oscillator strength (Ne 2 / M*) or to a reduction in mean-field
scale temperature and an associated decrease in the coherence length.
Although both the Frohlich effective mass and the scale temperature might
be sensitive to impurity or defect concentration, there is currently no
detailed theoretical understanding of this result.
One final comment on the electrical conductivity is in order. It is by
now well known that in one-dimension impurities cause localization of
wave functions rather than just scattering. Indeed, as indicated above,
Bychkov,(107) Berezinsky,(1OS) and Gogolin et aly09) have shown rigorously
that for free particles in a one-dimensional system containing impurities
the conductivity decreases as the temperature is lowered, approaching zero
with a temperature dependence determined by the ratio of the impurity
and phonon scattering rates, T,mp/Tph(T). Thus the existence of high
conductivities in pseudo-one-dimensional systems is necessarily a many-
body effect resulting from attractive electron-electron interactions.
Our conclusion is that dc conductivity at room temperature is approx-
imately 10 3 !r 1 cm -1 and that there is strong evidence of peak conduc-
tivities (near 58°K) of order 10 5 n- 1 cm -1. Peak values exceeding
104 n-1 cm -1 have been observed in many laboratories. The sensitivity of
the dc conductivity in a one-dimensional system to defects and impurities is
not surprising, and should in fact be anticipated as a feature of the problem
in future studies of anisotropic solids.
5.2. Microwave Measurements

.
E xpenmenta I stu d'les 0 f t he mIcrowave
. . (173233119120)
propertIes . . .. 0
f
pure crystals of the organic salt TTF-TCNQ were carried out, including
electron spin resonance and a complete study of the dielectric function,
EJ - iE2, from room temperature to 4.2°K as measured along the principal
conducting b axis and the transverse a axis using the highest-purity
material. The spin resonance line is asymmetric, (17) characteristic of a
metal with skin depth less than the sample dimension, when the rf magnetic
field is perpendicular to the b axis, and symmetric (Lorentzian) when the rf

magnetic field is parallel to the b axis. These results are understood in
terms of the Dyson-Bloembergen theory of resonance line shape as ap-
plied to the pseudo-one-dimensional metal.
The cavity perturbation technique of Buravov and Shchegoiev(l21) has
been extended(119) into the surface impedance regime. This allowed, for
the first time, direct analysis of the cavity Q data in the high-conductivity
regime where the skin depth is less than the sample dimensions.
In the surface impedance regime.(119) the loss in the cavity due to
inserting the metallic sample is given by
L= A/(J'{) (55)
where {) is the classical skin depth {) = C/(21r(J'W )1/2. The constant A is a
geometrical factor determined by the shape and filling factor of the sample.
For an ellipsoid of revolution (semiaxes a and b)
(56)
where Eo is the unperturbed microwave electric field at the anti node of the
TE101 cavity. The resulting expression for the microwave conductivity is
(57)
where a = 2 V s/ Vc is the filling factor, Vs is the sample volume, f is the

operating frequency, c is the velocity of light, and A is the change in the
cavity width (at half-power) divided by the frequency.
Buravov and Shchegolev(12!) derived the expression 8 = (f - [0)/[0 =
a/ T/ for the fractional shift in resonant frequency of the cavity in the
presence of an ellipsoidal sample of sufficiently high conductivity, but with
skin depth large compared with the sample dimensions. This relation
remains valid even in the skin depth regime, since it has to do only with the
boundary value problem in the space outside the sample. The boundary
condition E~:~ = 0 at the surface holds (with very small corrections) in the
skin-depth regime, and {) = a/ T/ continues to be true. Hence, (57) can be
recast in the simple form
(J'1/2 = C;2)( ~2)(l~2b)(±) (58)
where all quantities are directly measurable.

Before analyzing the data according to Equation (58), it is necessary to
demonstrate that the samples are in the skin depth regime. Earlier work on
electron spin resonance has established that the skin depth is smaller than
the sample dimensions. This was first shown(17) by the observation of the
122 A. 1. Heeger
asymmetric Dyson-Bloembergen line shape in ESR characteristic of a

metal whose skin depth is much less than the thickness. The estimated skin
depth at 10 GHz for TTF-TCNQ using a room-temperature conductivity,
alr(R.T.)~ 500 (0 cmr!, is 8::5 20 f.Lm. Thus, for samples with transverse
dimensions greater than 50 J-I,m, the surface impedance regime should be
appropriate throughout the temperature interval 45°K < T < 300°K.
As an independent experimental check on the validity of the classical
skin effect for anisotropic TTF-TCNQ, the coaxial resonator technique
developed by Richards(122) and by Hardy and Berlinsky(123) has been
utilized. In this experiment the sample forms the center conductor of a
coaxial transmission cavity with Q given by
2b R
Q=-ln- (59)
811 b
where R is the radius of the resonator and b is the radius of the sample.
The resulting Q is directly determined by the skin depth; if the microwave
fields penetrate uniformly the Q approaches a value of order unity. These
experiments give values of Q = 30 (x band) at room temperature for
typical samples of TTF-TCNQ implying 8« b. Analysis of eight samples
gives values of a~(R.T.)= 500± 250 (0 cmr! for TTF-TCNQ.(!!9) The
results indicate that for samples with thickness in excess of 50 J-I, m the
surface impedance regime is appropriate for TTF-TCNQ even at room
temperature where the longitudinal conductivity is relatively low and the
anisotropy (a~/a:)= 10 3 .
Figure 23 shows the microwave conductivity(19) of five samples of
TTF-TCNQ as measured with the cavity perturbation method. Each of the
samples has transverse dimensions in excess of 80 J-I,m. The resulting values
for aIT(R.T.) fall in the range 650±350 (Ocmr 1 typically observed in dc
experiments. Figure 23 shows a series of nested curves reminiscent of the
quantitatively similar results found in dc experiments(!6) and confirmed in
more detail through the induced defect studies.(95) These relatively high
conductivity results are to be contrasted with the work of Bloch and
co-workers,(33) who reported modest conductivities from measurements on
material of relatively poor quality.
6. The Transition Region 38°K < T < 54 oK
Two transitions, at 54°K and 38°K, in TTF-TcNQ have been widely

studied with direct structural evidence coming from x-ray and neutron
studies. Principally by analyzing transport measurements, Etemad(83)
suggested the 54°K transformation results from ordering on TCNQ chains,
and the 38°K transformation from ordering on TTF chains. From elastic
100
80
60
b
O"II(T}
b
O"II(RT)
40
20
OOL-2~0--~~6LO--L-~10-0~==14LO~~138~0~~22~0~~2~6~O~~3~OO
T (K)
Figure 23. Microwave conductivity of a series of samples of TIF-TCNQ. The results were ob-
tained using the cavity perturbation technique in the surface impedance regime (Reference 119).
neutron studies, Comes et ai.(37) showed that the region 38°K < T < 54°K is
characterized by an incommensurate superlattice where the a axis modu-
lation changes continuously from 2a near 54°K to 4a with a discontinuous
step at 38°K. On the basis of their Ginzburg-Landau treatment of the two
sets of coupled chains, Bak and Emery(124) have shown the neutron results
for TTF-TCNQ are consistent with three transitions. They suggested long-
range order sets in on one of the chains at 54°K and on the second chain
near 49°K with the a-axis modulation remaining at 2a between 54 and
49°K. From 49°K to 38°K both chains continuously order with respect to
124 A. 1. Heeger
one another, finally locking discontinuously to 4a at 38°K. The 38°K

transition is first order with hysteresis observed in both the structural(37,39)
and transport studiesY25)
The Bak-Emery theory assumes that the TTF and TCNQ chains have
scale temperatures that are large compared to the transition temperature
so that their coherence lengths are long, and the nearly static lattice
distortions may be regarded as complex order parameters in terms of which
the free energy may be expanded. These assumptions appear to be valid, as
indicated by the structural, optical, and magnetic data summarized above.
Although the Bak-Emery analysis of the structural data provided the
first evidence of the 49°K transition, the independent ordering of the
TCNQ and TTF chains had been inferred from transport(83) and spin
resonance data.(126,127) The 49°K transition is clearly observable in the
local susceptibility data as obtained from the 13C Knight-shift studies.(58)
These Knight-shift data as shown in Figure 24 directly identify the tran-
sitions; the TCNQ chains order at 54°K and the TTF chains near 49°K.
In the region 49°K < T < 54°K, where only the TCNQ chains are
ordered, the superlattice periodicity is ~a *, (37) implying that the two TCNQ
chains within the unit cell are antiphase ordered. With the onset of TTF
0.4
1 11
38K 49K 54K ~
•______ .
0.3
'-I~
•
----.---
0.2
0.1 ....... +~XQ . ----

..
o.o~
•
.~//. ./
o 10 20 30 40 50 60 70 80 90 100 I/O 120
TEMPERATURE (K)
Figure 24, Temperature dependence of the TTF and TCNQ chain local susceptibilities shown
in detail (Reference 58),
chain order at 49°K the periodicity becomes incommensurate,(37) q +1a*,

with q varying continuously in the region 38°K < T < 54°K and jumping
discontinuously to q = -~a* at the first-order 38°K transition.
Physically, these features can be understood in terms of interchain
Coulomb coupling between the charge-density waves. The Coulomb inter-
action between two chains with charge-density waves Pi = Po cos ({!i on each
.
may be wntten
(128129)
'
1 2
Umtercham = -Po[2Ko(2kFli)] COS«({!i - ((!j) (60)
4£_
where e ~ is the perpendicular static dielectric constant, Ko(x) is the
complete elliptic integral of the first kind, and d is the interchain spacing
{for x » 1, 2Ko(x) = [(n-j2x) e -xf12}. For 49°K < T < 54°K, where only the
TCNQ chains are ordered, this Coulomb energy is minimized if «({!i-({!j)=
7T; i.e., the chains are antiphase ordered. In this case, the net interaction on
a TTF chain located symmetrically between is zero. However, if order sets
in on the TTF chain the total energy is lowered by breaking the symmetry
and allowing «({!i - ((!;) = 7T + e so that the total energy is minimized.(130)
Bak and Emery(L'4) have included these features in their two-chain
coupled order parameter theory. They write the free energy as
(61)
where I/Ilq represents the Fourier transform of the order parameter on the
TCNQ chain and 1/12q that on the TTF chain. Minimizing in terms of the
phase difference as described physically above and taking advantage of the
symmetry of the structure
where I/Ilq is assumed to take on a finite average value, a, b, and care

functions of I/Ilq and T but not of q, and a vanishes when I/Ilq = O. Near
54°K, the coefficients band c are positive as evidenced by the lack of order
on the TTF chains and the experimental fact that q = O. Minimizing with
respect to q, one finds
(63)
(64)
to fourth order in 1/12q- In absence of the interchain coupling, order on the

TTF chain would occur when b < 0; interchain coupling increases the
ordering temperature while at the same time causing q ~ O. The induced
transition occurs at the temperature T2 when (b - a 2 /4c) = O. Assuming
mean-field behavior at temperatures just below T 2, i.e., I/I~q = a(T2 - T),
126 A. 1. Heeger
Bak and Emery find(l24)
q2 = {O'a2 a (65)
-2(Tz -T), T<T2
4c
The experimental results of Ellenson et aZY31) are plotted in Figure 25 as
q2 vs. T and are in complete agreement with the Bak-Emery theory.
Thus the observed phase transitions are consistent with the accumu-
lating evidence of a complex order parameter coupled chain system with a
scale temperature well above the actual transition temperature. Some
evidence of one-dimensional to three-dimensional crossover has also been
obtained from detailed analysis of the temperature-dependent conductivity
near 54°K.(112) More recent studies(lll) of the effects of the dilute induced
defects on the phase transitions have demonstrated the destruction of
long-range order as expected for a quasi-one-dimensional system in which
the three-dimensional phase arises from a one-dimensional to three-
dimensional crossover.
006
TTF- TCNQ (E9)
q , ( 1/2 ~ 8)
005~-
a
004- 0.30
N
003- "
'" T3
-~
0.35 *c
002-
001- 0.40
00- 0.50
36 40 44 48 52 56
T (KI
Figure 25. Plot of ohserved peak positions along a* (right scale) and derived values of 8 2 (left
scale). The three transition temperatures are also indicated (Reference 131).
Early evidence of the 54°K transition came from the observation of a

specific heat anomaly(133) associated with the ordering. Recent high-sensi-
tivity differential specific heat studies have successfully detected the heat
capacity anomalies associated with the other transitions. (134)
7. The Pinned Regime at Low Temperatures
The experimental features that most clearly characterize the tempera-

ture range below 38°K in TTF-TCNQ are the charge-density wave ground
state observed in the structural studies and the unusually large low-
frequency dielectric constant. (17 ,32) Lee et ai, (46) have calculated the dielec-
tric constant associated with the complex order parameter CDW state in
which the phase is pinned, They find
2 Wp2 lLp{"\2
1'1-1=--2+-2 (66)
3 WG WF
where w; = 41TNe 2 /m* describes the single-particle oscillator strength
(hw p = 1.2 eV) in the presence of an energy gap, WG. The single-particle
oscillator strength is determined by the total electron density, N, and the
electron effective band mass, m *, In addition to this usual contribution to
el expected for a simple semiconductor, the second term represents the
collective mode oscillator strength due to optically active phase oscillations
of pinned CDW where WF is the characteristic pinning frequency of the
Frohlich mode. The phase mode oscillator strength is given by =n;
41TNse / M*, where Ns is the density of condensed electrons (Ns = N in the
2
pure case where the CDW state is not gapless) and M* is the Frohlich
mass(4,113,67):
M* 1 w7:x
- = 1 + - -2 (67)
m* A Wo
where A is the dimensionless electron-phonon coupling constant intro-

duced in Equation (35), WG is the gap frequency, and Wo is the bare phonon
frequency at 2k F ,
Extensive microwave measurements have been carried out on TTF-
TCNQ to determine the low-temperature dielectric constant along the
ch am ' (b)
' b aXIs elan d t h e transverse a aXIs
. (a)(17323312°)Th
e 1. ' " .. I
e pnnclpa
results include the unusually large value for I' ~ = 3500 and an anisotropy,
e ~/ e~, of several hundred consistent with the one-dimensional nature of
the electronic system, The large value for I' f implies the presence of
oscillator strength at relatively low frequencies.
Although it is natural to associate the large b-axis dielectric constant
with the pinned CDW, let us examine the alternative possibility, If d were
entirely the result of single-particle effects, the observed values for e rand
128 A. 1. Heeger
ENERGY (eV)
0.04 0.1
~~ T, " , - - - "---,---.--
I
0.61
i
( T TF ) ( TeNQ)
I-
I
300 K
z'" 0.4 :--

!
*E
"-
E
"~
02
ol
y)() 40,000
FREQUENCY (cm- I )
Figure 26. Number of effective electrons, obtained from the oscillator strength sum rule, for
TIF-TCNQ for two polarizations, vs. frequency from 300 to 37,000cm- 1 • Notice the log-
arithmic frequency scale. The gentle plateau for Ellb implies the exhaustion of the single tight-
binding band oscillator strength just below the onset of interband transitions (Reference 13).
ht/lp would require that the single-particle oscillator strength be exhausted

at frequencies of order 0.02 eV. The number of effective electrons
obtained from the oscillator strength sum rule
(68)
is shown (13) in Figure 26. The single-particle oscillator strength does not
begin to take on appreciable values until well above 1000 cm -1. Figure 26
is obtained from analysis of the room-temperature reflectance. However,
the data at low temperatures(l4.90) lead to the same conclusion. Direct
analysis of the 4.2°K far-infrared reflectance data(90) yields £ = 90 in the r
frequency range 20 cm -1 < jj < 80 cm -1. This value is consistent with the
single-particle contribution expected from the first term of Equation (66)
using the known plasma frequency (hw p = 1.2 eV) and the Peierls gap(!3)
(liwG = 0.14 eV)"
10 3 ----------------e
38 K
10'
---- _-------e
E
<.>
c:
b
Figure 27. Dc and lO-GHz conductiv,ty

compared at several temperatures indica-
ting a strong frequency dependence. The
inset shows IT(w) inferred from the dc and
microwave data, microwave dielectric
1/
constant, and far-infrared results.
The Kramers-Kronig relations relate the dielectric constant at

frequency w to the frequency distribution of oscillator strength:
FI(WO)= 1+8 Io
x [u(w)-u(O)]
2
w -Wo
2 dw (69)
Using Equation (69) and the measured dc and microwave conductivity

values, we can attempt to synthesize a picture of the frequency dependence
of the conductivity at low frequencies. Figure 27 shows the dc and
microwave conductivity values at a series of fixed temperatures. At 4.2°K,
there is a dramatic increase in u(w) in going from dc to 10 GHz. The large
positive value for FI(l7) (lOGHz) at 4.2°K implies through Equation (69)
that u(w) continues to increase with substantial oscillator strength at higher
frequency. In the far infrared(l3.14) (f > 15 cm -Ihl is only 100. This implies
that the oscillator strength giving rise to the large microwave C I is localized
at f:S 15 cm - I . This may be the pinned collective mode anticipated in the
Peierls-Fri:ihlich collective transport theory. At higher temperatures, Udc
and U(lOGHz) approach one another and are essentially indistinguishable
above 38°K, where FI crosses zero. However, from Equation (69), CJ <0
implies that u(w) decreases with increasing frequency in the conducting
regime for temperatures near the phase transition. This frequency depen-
dence has been observed(lIl) at microwave frequencies (10 GHz and
130 A. 1. Heeger
30 GHz). However, as described above(92b) the microwave dielectric

constant at 300 K is again positive and about equal to its value at 4.2°K. In
0
terms of the distribution of oscillator strength, Equation (69) requires that

<T(w) increase above 10 GHz at 3000 K once again. At 300o K, the lO-GHz
and 30-GHz results are in agreement; the increase in <T(w), implied by
E\ > 0, must occur at frequencies above 30 GHz. These results are sum-
marized in the inset to Figure 27, where we plot schematically <T(w) based
on the combined dc and frequency-dependent microwave data, the micro-
wave dielectric constant [through Equation (69)], and the far-infrared
results.
Systematic studies of the microwave dielectric constant(120) have
included measurements of the isostructural compounds (TTF)(TCNQ),
TMTTF-TCNQ, DSeDTF-TCNQ, and TSeF-TCNQ. The results suggest
that a common mechanism is involved. The temperature dependence of E ~
is shown for all of these systems in Figure 28. For TTF-TCNQ, E~
18
16
14
12
r<l 10
o
)( 8
W
2
C>
\ x \
o
°0~--------~I~
O--------~2~0~--------~
3~0----------4~0~~
T (K)
Figure 28. Temperature dependence of the dielectric constant for TIF-TCNQ and related
derivative salts (Reference 80). TIF-TCNQ, 000; TSeF-TCNQ, 000; TMTIF-TCNQ,
666; DSeDTF-TCNQ, • • • ; TIFo 97 TSeFo 03- TCNQ, X x x (Reference 120).
increases slightly from the 3.5 x 10 3 value at 4.2°K, leveling off at about
30oK, then drops toward zero at 39°K. Above 38°K there is strong indica-
tion that E ~ goes negative; however, the experiment is insensitive to a
negative E 1 in the presence of a large 0"1. The derivative systems exhibit a
qualitatively similar temperature dependence. The TSeF-TCNQ system is
exceptional; the 4.2°K dielectric constant is 16,000 and drops monotonic-
ally to zero at about 20oK.
Infrared data are not yet available for the derivative systems, so that a
decomposition into single-particle and collective phase mode contributions
is not possible. However, the similar crystal structures, the similar values
for W p ,(l1.136) the similar temperature dependence of the magnetic suscep-
tibilities, (34,69) and the observation of one-dimensional diffuse scatter-
ing(137) above the transition temperature all suggest that the energy gap is a
common feature and that f' f is dominated by the collective phase mode at
low temperatures.
The temperature dependence of E? has not yet been analyzed in
detail. The success of mean-field theory(l24) in explaining the details of the
transition region in TTF-TCNQ suggests a simple renormalization of the
pinning frequency with a corresponding increase in the microwave dielec-
tric constant and conductivity. When the pinning frequency approaches
zero, I' ~ is driven negative by the low-frequency oscillator strength as can
be seen from the Kramers-Kronig transform, Equation (69). El(W)<O
requires that (dO",/dw)<O in the vicinity of w. Alternatively, fluctuation
effects may be important even in the low-temperature regime leading to
thermally de pinned regions which increase in fractional volume with
increasing temperature. Such an effective medium approach (1 38) has been
applied to TTF- TeNQ with results qualitatively similar to the experimental
temperature dependence.
8. Nonlinear Transport in TTF- TCNQ at Low Temperatures
The unusually large low-temperature dielectric constant and the

related low resonance frequency for the coherent pinned phase mode
imply relatively weak pinning forces and suggest the possibility that an
applied electric field at low temperatures can cause the weakly pinned
COW condensate to partially depin and thereby becoming conducting. The
recent experimental observation(96) of nonlinear J- V characteristics for
ITF-TCNQ at low temperatures provides initial evidence of such COW
phenomena.
Figure 29a shows the current-voltage characteristics of a single crystal
of TTF-TCNQ measured along the principal conducting b direction over
the temperature range between 1.5 and 4.2°K. The data shown were
132 A. 1. Heeger
obtained using dc techniques, and the upper limit of 1 p,A was chosen to
avoid heating the sample. At the lowest fields, corresponding to voltages
less than about 0.2 V, the data are ohmic and the conductivity decreases
exponentially as the temperature is lowered. As the electric field is
increased, the data become nonohmic with the dynamic conductance,
dI/ dV increasing with increasing field. The degree of nonlinearity as
indicated in the full logarithmic plot increases dramatically as the
temperature is lowered, with the result that the curves appear to be
converging at some characteristic electric field. The degree of nonlinearity
can be clearly seen in the linear plot in the inset.
To examine this effect more carefully, the measurements were extend-
ed to higher current densities using high-impedance pulsed techniques. (96)
Figure 29b shows the results at three temperatures. The data are plotted as
10- 6
1.0 8 00 00>
0 00 cA>
000 <DO
.8 000000
000000
000000
.6 000000
IV-< AJ
10- 7 o 00 000
.4
o 00 000
.2 0 00 00 0
HAl o 00 000
5 10 15 o 0 000
o
V(voltsl
-9
10
o 0
o
o
o
o (a)
-IO~~~__~~~~__-L~~_ _~_ _~_ _~~
10 .01 .02 .05 .10 .20 ,50 1,0 2,0 5.0 10.0
V (volts)
Figure 29, (a) The dc current-voltage characteristic of single crystals of TTF-TCNQ measured
along the b axis at temperatures between 1,6°K and 4.2°K, (b) i vs. E for TTF-TCNQ covering
current density versus electric field; the various temperature curves con-
verge near a field of 400 V / cm. The inset displays the same data plotted
linearly. The I-V curves qualitatively appear to be approaching an off-on
situation at absolute zero where current is not generated until a critical
electric field is reached.
It is natural to question whether the observed nonlinearities in the
I - V characteristics of TTF-TCNQ are related to its one-dimensionality.
To this end, the I-V characteristics were measured(96) for single crystals of
TTF-TCNQ whose needle axis was oriented along the crystallographic a
axis, the direction of alternating stacks of TTF and TCNQ molecules.
Figure 30 shows the normalized field-dependent conductivity, (T(E) = J/ E,
plotted vs. electric field for both b and a directions. The conductivity along
b increases by more than 10 4 over the range of electric field where the
10
8
o
B
o
1.0 o
6 Q)
q,
Q)
10- 1 <tIE 00
at>
~<'>4
, <to
6'0~
OOQ>
'0 00
2 oe08
-
B (}
0 00
qp
00
....~.".,
0 0
o
o 0
100 200 300 400 0
0 o
E (V/cm) 0 o 0
0 0 o
0
0
0
0 0 o
0 0 o
0
0
0 0
0
0 0 0
T=4 .2K 0 0 0
00 0
0 3.00
00 1.6 0
000 0
(b)
I07~________~__- L________~____~________~
1.0 5.0 10.0 50.0 100 500

E (V/cm)
the range 10- 6 A/ cm 2 < I < 10 A/ cm 2 ; dc techniques were used for j < 5 X 10- 2 A/cm 2 ; pulsed
measurements covered the range above 10- 2 A/cm 2
134 A. 1. Heeger
104
o
o
10 3
0'" (E)
0'" (0)
10 2
o
o
o
o
10 o
o
o
o
o
o
o
o a Figure 30. The normalized field-
1, 0 _~OILO~O-OO_o..o....o_o_.2'b _ _ ~ _ _ ~ 2~~_
dependent conductivity a(E)=
j(E)/ E of TTF-TCNQ measured
1000 along the principal conductivity b
1.0 10.0 10 0
axis and the transverse a axis at
E (Vlcm)
4.2°K.
a-axis conductivity is ohmic. The conductivity along c was too small to

provide measurable currents at applied fields of less than 103 V /cm so we
were unable to repeat the measurements in that direction.
Similar measurements(96) have been carried out on TSeF-TCNQ using
a combination of dc and pulsed I-V measurements. The ohmic conduc-
tivity at 1.soK is more than seven orders of magnitude greater than that of
ITF-TCNQ. Nevertheless, the TSeF-TCNQ conductivity becomes
nonlinear at approximately the same electric field.
Detailed experimental studies(96) lead to the conclusion that the
nonohmic conductivity data are intrinsic to the material and do not arise
from rectifying contacts, sample heating, or impurity / defect related impact
or field ionization of carriers.
The nonlinear transport can be qualitatively described in terms of the
pinned charge-density wave ground state of TTF-TCNQ. The pinning
potential is necessarily periodic and the simplest form(139) is
V = Vo(1 - cos 'P ) (70)
If the pinning were to arise from the Coulomb interactions between charge-
density waves on neighboring chains, the periodicity of the potential would
be the same as that of the charge-density wave, i.e., 'P = (21T/A s )X, where
As = 3.4b.(35-41) If the pinning were to arise from the commensurability of
the charge-density wave with the equilibrium lattice, the periodicity would
be that of the lattice, i.e., 'P = (21T/ b)X. The magnitude of the pinning
potential can be estimated by expanding the potential and comparing the

quadratic term with the pinning frequency determined by far-infrared
reflectance measurements:
21T)2 * 2
Vo ( T =MFwF (71 )
A = b (commensurability pinning) or A = 3.4b (Coulomb pinning). A Froh-

lich effective mass(90) of M~ ::.= 1400m* (m* = single-particle band mass)
and a pinning frequency(90) of WF::'= 2 cm -I corresponds to a pinning
potential energy of e Vo::'= 17°K (commensurability pinning) or 1SK (Cou-
lomb pinning).
The current carried by a charge density wave system depends on the
coherent motion of its phase with respect to the stationary latticeY40) The
full Lagrangian density of the COW system must include a potential energy
due to its compressibility. This has been found by Rice et a1Y39) to be
(72)
where Ns is the one-dimensional density of condensed electrons, q* =

(21T/ As), and Co is the characteristic phase velocity given by Co =
vF(m */ n ~ )1/2. The equation of motion(139) with the phase as the general-
ized coordinate is
(73)
If the pinning potential of Equation (70) is used, this is the sine-Gordon

equation for which there exists a class of exact nonlinear solitary wave
·
so 1utlOns .
given b y(119141)
-
(p±{X-vt)=4tan
-I{ exp± [(WF) (1-v2/c6)I/2
Co
X-vt ]} (74)
These correspond to phase "kinks" of ±21T localized over a length d-

co/wF(1-v 2/c6)I/2 and propagating with a velocity v; for TTF-TCNQ
d(v = O)~ 20b - 6As. The phase kinks correspond to compressions and
rarefactions of the CDW which are, in essence, mobile localized charges.
The Lagrangian density can be converted into a Hamiltonian density
and the solutions of Equation (73) result in an energy spectrum given
by(139)
(75)
136 A. 1. Heeger
where the energy has been expressed in terms of a "rest mass":
(76)
The phase kinks can thus be thought of as charged quasiparticles with

mass M<p and rest energy M<pd. In the absence of a driving force, these "q;
particles" are thermally excited with an activation energy of the order of
the rest energy.
For TTF -TCNQ, the q;-particle rest mass as estimated from Equation
(76) is about 12°K (commensurability pinning) or 1400 K (Coulomb pin-
ning). The measured(96) zero-field activation energy of 14°K is in close
agreement with the former. Similar measurements(96) on TSeF-TCNQ
found an activation energy of about 5°K. The conductivity due to q;
. l
partIe b
es can '
e wntten as (13996)
. .
(77)
We have assumed that the density of q; particles scales with the density
of condensed conduction electrons and that the mobility was thermally
diffusive. The diffusion constant is a thermal velocity given by ~mv~H = kT
times a characteristic length which we assume to be a lattice constant. At
1 SK, Equation (77) gives ~<p - 10 (cm 2 IV sec). Using the expression
above for the density of particles, the conductivity prefactor of TTF-TCNQ
and TSeF-TCNQ can be used to imply a mobility. The conductivity pre-
factor of TSeF-TCNQ is of order 10 2 0.- 1 em -I, which implies a mobility of
about 10 cm 2 IV sec, in agreement with the calculated value. On the other
hand, the conductivity prefactor of TTF-TCNQ is six orders of magnitude
lower, implying a mobility of only 10.- 5 cm 2 IV sec.
The phase kinks described earlier are solutions of the equation of
motion [Equation (73)] only in the absence of a driving force. Qualitatively
as the electric field is increased the effective pinning potential is decreased;
thus an electric field would have the effect of lowering the q;-particle
activation energy. The low-temperature CDW conductivity can thus be
pictured as being characterized by two values of the electric field. At the
lowest fields the nonlinearity is determined by the field (Eo) required to
reduce the q; particle activation energy to zero. This is of order 150 V I cm
as discussed above. The upper characteristic electric field (E I ) is that which
is sufficient to reduce the magnitude of the potential to zero; (AsI2rr)EI =
Vo. For interchain Coulomb pinning, the upper (E I ) field is of order
10 3 V / cm. Above this upper field, the conductivity would again be ohmic
as the entire CDW system is depinned.
The order of magnitude of the conductivity of the depinned CDW
system can be estimated from the oscillator strength and lifetime as
determined from the far-infrared experiments to be of order 10 2 (0 cmfl.
For TSeF-TCNQ, (Tmax = 30 (0 cmfl at ISK, implying that the CDW
system is nearly completely depinned at E -10 V /cm. For TTF-TCNQ,
(Tmax=2xl0~2 (Ocmr l at 400V/cm. In both cases, sample heating
restricts the measurements to fields below E I.
The rest mass energies and characteristic electric fields inferred from
the data are in reasonable agreement with theoretical estimates. The
absolute magnitude of the high-field conductivity and the low-field mobil-
ity present a more serious puzzle. For TTF-TCNQ, the values are approx-
imately four to five orders of magnitude too small (even though j/ E has
increased by more than six orders of magnitude in the experimental range
studied). For TSeF-TCNQ, both the low-field mobility and the high-field
conductivity are in good agreement with the magnitudes expected for
cp-particle diffusive transport (at low fields) and the nearly complete
depinned CDW transport (at high fields). The problem may in part arise
from the experimental limitations; the TTF-TCNQ curve at 1.5°K is so
steep that a 10% increase in applied electric field would bring the absolute
magnitude up to the fully depinned value. Unfortunately, this higher-
current regime brings the power level up to such high values that heating
effects are unavoidable even with pulsed techniques.
Larkin and Lee (INa) have shown that impurities or defects lead to a
localization of solitons. Since solitons are required to tunnel through the
impurities to contribute to the dc conductivity, the dc conductivity prefactor
is reduced by the tunneling probability e -A. In the strong pinning case in
which the soliton energy E, is much less than the barrier height V, we have
A=2 -(M*)
m
1/2 V (M*)
In-=-
E, m
1/2
(78)
where M* is the Frohlich CDW effective mass. Whereas the dc conductivity

would be drastically reduced by soliton localization and tunneling, at
microwave frequencies the conductivity would approach the intrinsic soliton
value since motion of the soliton over macroscopic dimensions is not
required. Experimental studies on TrF- TCNQ show that the dc con-
ductivity is more than five orders of magnitude smaller than predicted
by the simple soliton picture. In contrast, the microwave conductivity shows
the same activation energy but is 105 times larger in magnitude. Moreover,
the fact that the low-temperature microwave conductivity is unaffected by
irradiation while at 1.5°K the nonlinear dc conductivity is reduced by a
138 A. 1. Heeger
factor of 4 (96) may be further evidence for the soliton mechanism. Thus, the
Larkin-Lee(139a) tunneling prefactor offers a possible explanation of the
experimental results, provided e ~ A:s 10-°_10-6. Using Equation (78), we
find (M*/m)I/L?II.S-14 or M*/m"PIOO-200. Although somewhat
smaller than other estimates, the ratio has the correct order of mag-
nitude.
The applicability of the phase-kink soliton idea to real COW systems
must be viewed critically. Although structural studies have provided
detailed evidence of the pinned COW ground state, the excitations will
depend on the nature and origin of the pinning potential V «(,0 ). We have
assumed V «(,0 ) to be of the simple cosinusoidal form, thus leading naturally
to the sine-Gordon equation and (,O-particle excitations. A general property
of this kind of system, characterized by an anharmonic potential plus
dispersion, is the existence of two classes of solutions: small-amplitude
harmonic solutions (in the COW case there are the phason modes of Lee et
ai.,(46) and large-amplitude nonlinear kink solutions (the (,0 particles of
Rice et ai.(139»). There is ample experimental evidence in the pinned COW
systems for the small~amplitude solutions. In particular, the unusually large
dielectric constants (f 1 = 3500 for TTF-TCNQ(I7·33, 120) and E 1= 16,000 for
TSeF_TCNQ(120») in the presence of a relatively large energy gap indicate
low-frequency oscillator strength consistent with the small-amplitude
excitations. The only data available relevant to the question of the exis-
tence of the nonlinear kink excitations is the nonlinear transport described
above. However, more direct experimental detection of the nonlinear
excitations is clearly required.
9, Conclusion
There has been remarkable progress in the study of organic one-

dimensional conductors. The Peierls instability and COW ground tate c
have been established as experimental fact as a result of the structural

measurements. The detailed role played by the TCNQ and TTF chains has
been clarified through measurements of the local susceptibility using NMR
techniques(58) and g-value labeling.(99,100,126,127) The important theoretical
contributions to the developing understanding of the phase transitions
have provided confidence in an approach based on a complex order
parameter description of the coupled chain system,
Perhaps the most important question remaining to be settled is
whether the Frohlich collective mode or any collective behavior is the
dominant transport mechanism in the conducting state. The large conduc-
tivity in the presence of well-defined 2kF scattering which apparently gives
rise to the pseudogap as observed in the optical and magnetic properties

provides the primary evidence of collective behavior. The presence of
oscillator strength at very low frequencies in the dielectric regime as
indicated by the microwave and far-infrared studies provides similar evi-
dence in the low-temperature regime.
On the other hand it has been argued that the thermal conduc-
tivity,(144) optical scattering time,(7J) and temperature dependence(103) of
the conductivity are all consistent with simple metallic behavtor with elec-
tron single-particle transport limited by mean-free-path scattering.
However, two recent experimental studies appear to argue against such
simple single-particle approaches to the transport. First, the large positive
dielectric constants,(92ol discovered for the entire class of one-dimensional
conductors in the "metallic" regime, are incompatible with a simple Drude
metal. Thus the analysis of the thermal conductivity by Salamon et at. (144) is
oversimplified and cannot be viewed as meaningful evidence of single-
particle transport. The studies of the effect of induced defects on the
transport,(9S) structure.(l42) and ir properties(l45) of TTF-TCNQ have pro-
vided additional insight into these questions. The resulting changes in
transport properties can be qualitatively understood in terms of collective
transport in the "metallic" state which is quenched upon formation of a
static superiattice through interaction with defects. Further confirmation
comes from infrared studies,(14» which show no observable change in the
single-particle scattering time at the highest defect levels. These latter
results indicate that the optical scattering time and the dc conductivity are
unrelated in contrast to the arguments of Seiden and Cabib.(71)
The existence of high metaIIike conductivities in quasi-one-dimen-
sional systems as evidence for many-body effects remains under continued
active study. The experiments involving controlled amounts of impurities
and/or defects(95) are of clear importance, for through such experiments
one should be able to sort out the various aspects of disorder localization
expected for one-dimensional systems.
As a result of the lack of long-range order and the presence of intrinsic
dissipative processes, the Peierls-Frohlich condensate is not a supercon-
ductor. However, a broken symmetry conduction mechanism arising from
a (nearly) phase-independent charge-density wave would represent the
only example yet found in nature of collective electron transport that is
different from paring superconductivity. The experimental evidence for
collective transport in these one-dimensional materials is accumulating and
is a principal reason for their interest.
The possibility of superconductivity at relatively high temperatures in
such quasi-one-dimensional metals continues to be an important goal.
Such systems exhibit the unique features of relatively strong coupling to
140 A. 1. Heeger
the hIgh Debye temperature Intramolecular vibrational modes in the

combined presence of one-dimensional instabilities and a high electronic
density of states. These features may yet give rise to important progress in
superconductIvIty (146)
ACKNOWLEDGMENTS
The work of many people is summarized in this brief review. I par-

ticularly wIsh to acknowledge the contributions of Professor A. F. Garito.
His care and inSIstence on the highest quality materials, together with a
constant attentIOn to detaIl on many phases of the experimental studies,
have been of clear Importance I thank C. K. Chiang, Marshall J. Cohen,
L. B. Coleman, C R Fincher, W. 1. Gunning, S. K. Khanna, P. R. New-
man, P. Nlgrey, E F. Rybaczewskl, 1. C. Scott, D. B. Tanner, and T. S. Wei
for recent cruCial contrIbutIons. I am especially grateful to Dr. R. Comes
and Dr G. Shlrane and theIr colleagues for their beautiful experimental
studies of the structural aspects of the Peierls instability in TTF-TCNQ and
for many valuable dISCUSSIons VarIous aspects of the work at the Uni-
versIty of Pennsylvama descrIbed In this review were supported by the
Advanced Research Projects Agency (DAHC-15-72-C-0174), by the
NatIOnal SCIence FoundatIOn (DMR-21667), by the National Science
FoundatIOn MRL Program (DMR 76-00678), and by the Army Research
Office (DAAG 29-77-G-0089)
Low-Dlmen~lOnal Cooperallve Phenomena, H J Keller (ed ), Plenum Press, New York

(1975), Lecture Notes In PhYSICS, Vol 34, One-DimensIOnal Conductors, Spnnger, New
York (1975), A J Berhnsky, Contemp Phys 17,331 (1976), J J Andre, A BIeber,
and F Gautter Ann Phys (N Y) 1, 145 (1976), Chemmry and PhYSICS of One-
DimensIOnal Metals H J Keller (ed ), Plenum Press, New York (1977)
2 A F GarIto and A J Heeger, Acc Chem Res 7,232 (1974), A F Ganto and A J
Heeger, Nobel Symp 24,129 (1973)
3 R E Peleris Ouantum Theory of Solzds, Oxford UnIversIty Press, p 108, London
(195) )
4 H Frohhch, Proc R Soc London Ser A 223, 296 (1954)
:; A W Overhauser, Phys Rev Lett 4, 62 (1960)
6 R P Shlbaeva and L 0 Atovmyan, ] Struet Chem 13,546 (1972), F H HerbsteIn, In
PerspectIVes In Structural ChemlSlry, J D DUnItz and J Albers (eds ), Vol IV, pp
166-395 WIley New York (1971)
7 T J KIstenmacher T E PhIlhps, and D 0 Cowan, Acta Cryst 830,763 (1974), R H
BleSSIng and P Coppens, Solzd State Commun 15, 215 (1974), A J Schultz, G D
Stucky R H Blessmg, and P Coppens, ] Am Chem Soc 98, 3194 (1976)
8 D Kuse and H R Zeller Phys Rev Lett 27, 1060 (1971)
9 P Bruesch S '>trassler and H R Zeller, Phys Rev B 12, 219 (1975)
10. H. R. Zeller, in Low-Dimensional Cooperative Phenomena, H. J. Keller (ed.), p. 215,

Plenum Press, New York (1975).
11. A. A. Bright, A. F. Garito, and A. 1. Heeger, Solid State Commun. 13, 943 (1973);
Phys. Rev. B 10, 1328 (1974).
12. P. M. Grant, R. L. Greene, G. C. Wrighton, and G. Castro, Phys. Rev. Lett. 31,1311
(1973).
13. D. B. Tanner, C. S. Jacobsen, A. F. Garito, and A. J. Heeger, Phys. Rev. Lett. 32, 1301
(1974); Phys. Rev. B 13,3381 (1976).
14. C. S. 1acobsen, D. B. Tanner, A. F. Garito, and A. 1. Heeger, Phys. Rev. Lett. 33, 1559
(1974).
15. M. 1. Minot and 1. H. Perlstein, Phys. Rev. Lett. 26, 371 (1971); H. R. Zeller and A.
Beck, f. Phys. Chem. Solids 35, 77 (1974); D. Kuse and H. R. Zeller, Solid State
Commun. 11, 355 (1972).
16. M. J. Cohen, L. B. Coleman, A. F. Garito, and A. 1. Heeger, Phys. Rev. B 10, 1298
(1974).
17. S. K. Khanna, E. Ehrenfreund, A. F. Garito, and A. 1. Heeger, Phys. Rev. B 10, 2205
(1974).
18. Morrel Cohen, Lake Arrowhead Conference on One-Dimensional Conductors, Lake
Arrowhead, California (1974).
19. M. Boudeulle, Ph.D. Thesis, U. Claude-Bernard de Lyon (1972); M. Boudeulle, Cryst.
Struct. Commun. 4, 9 (1975).
20. Marshall 1. Cohen, A. F. Garito, A. 1. Heeger, A. G. MacDiarmid, C. M. Mikulski,
M. S. Saran, and 1. Kleppinger, f. Am. Chem. Soc. 98 (13), 3844 (1976).
21. A. A. Bright, M. 1. Cohen, A. F. Garito, A. 1. Heeger, C. M. Mikulski, P. 1. Russo, and
A. G. MacDiarmid, Phys. Rev. Lett. 34, 206 (1975).
22. A. A. Bright, M S. Cohen, A. F. Garito, A. 1. Heeger, C. M. Mikulski, and A. G.
MacDiarmid. Appl. Phys. Lett. 26, 612 (1975); see also Marshall 1. Cohen, Ph.D.
Thesis, University of Pennsylvania (1975).
23. H. Kamimura, A J. Grant. F. Levy, A. D. Yoffe, and G. D. Pitt, Solid State Commun.
17,49 (l975).
24. L. Pinstchovius, H. P. Geserich and W. Moller, Solid State Commun. 17, 477 (1975).
25. J. Bordas, A. J. Grant, H. P. Hughes, A. Jakobson, H. Kamimura, F. A. Levy, K.
Nakao, Y. Natsume, and A. D. Yoffe, f. Phys. C 9, L277 (1976).
26. C. H. Chen, J. Silcox, A. F. Garito. A. J. Heeger, and A. G. MacDiarmid, Phys. Rev.
Lett. 36, 525 (1976).
27. V. V. Walatka, Jr., M. M. Labes, and J. H. Perlstein, Phys. Rev. Lett. 31,1139 (1973);
C. H. Hsu and M. M. Labes, I. Chem. Phys. 61, 4640 (1974).
28. R. L. Greene, P. M. Grant, and G. B. Street, Phys. Rev. Lett. 34, 89 (1975).
29. R. L. Greene, G. B. Street, and L. J. Suter, Phys. Rev. Lett. 34, 577 (1975).
30. C. K. Chiang, M. J. Cohen, D. L. Peebles. A. J. Heeger, M. Akhtar, J. Kleppinger, and
A. G. MacDiarmid, Solid State Commun. 23, 607 «1977); M. Akhtar, J. Kleppinger,
A. G. MacDiarmid, J. Milliken, M. J. Moran, C. K. Chiang, M. 1. Cohen, A. J. Heeger,
and D. L. Peebles, ('hem. Commun. 473 (1977).
31. J. H. Perlstein, J P Ferraris, V. V. Walatka, and D. O. Cowan, AlP Conference
Proceedings 10, 1494 (1972); J. P. Ferraris, D. O. Cowan, V. V. Walatka, and 1. H.
Perlstein, 1. Am. Chem. Soc. 95, 948 (1973).
32. S. K. Khanna, A. F. Garito. A. J. Heeger, and R. C. 1aklevic, Solid State Commun. 16,
667 (1975).
33. A. N. Bloch, 1. P. Ferraris, D. O. Cowan, and T. O. Peohler, Solid State Commun. 13,
753 (1973). The value of OJ (4.2 K) = 50 reported has been demonstrated to result from
a lack of sample purity (Reference 16). With increased sample quality, their result has
142 A. f. Heeger
come into agreement with the value reported in Reference 16 fA. N. Bloch, APS
Meeting, Denver. Colorado, 1975 (unpublished)].
34. 1. C. Scott, A. F. Garito and A. 1. Heeger, Phys. Rev. B 10, 3131 (1974).
35. F. Denoyer, R. Comes, A. F. Garito, and A. 1. Heeger, Phys. Rev. Lett. 35, 445 (1975).
36. S. Kaoshima, H. Anzai, K. Kajimura, and T. Ishiguro, J. Phys. Soc. Japan 39, 1143
(1975).
37. R. Comes, S. M. Shapiro, G. Shirane. A. F. Garito, and A. J. Heeger, Phys. Rev. Lett.
35,1518 (1975).
38. 1. P. Pouget, S. K. Khanna, F. Denoyer, R. Comes, A. F. Garito, and A. 1. Heeger,
Phys. Rev. Lett. 37. 437 (1976).
39. S. Kogoshima. T. Ishiguro, and H. Anzai, J. Phys. Soc. Japan 41, 2061 (1976).
40. S. K. Khanna, J. P. Pouget, R. Comes, A. F. Garito, and A. J. Heeger, Phys. Rev. B 15,
1468 (1977)
41. G. Shirane. S. M. Shapiro, R. Comes, A. F. Garito, and A. J. Heeger, Phys. Rev. B 14,
2325 (1976)
42. A. M. Afana~;ev and Yu. Kagan, Zh. Eksp. Tear. Fiz. 43,1456 (1963) [Sov. Phys.-JETP
16, 1030 (1963)].
43. C. G. Kuper. PrOf. R. Soc. London Ser. A 227, 214 (1955).
44. M. 1. Rice and S. Striissler, Solid Slate Cornrnun. 13,697,1389 (1973).
45. J. R. Schrieffer. in Collective Properties of Physical Systems, B. Lundquist and S.
Lundquist (ed.), p. 142, Nobel Foundation, Stockholm and Academic, New York
(1973).
46. P. A. Lee, T. M. Rice. and P. W. Anderson, Solid State Cornrnun. 31, 462 (1973).
47. L. P. Gorkov, Preprint, Solid State Cornrnun. (to be published).
48. P. M. Chaikin, J. F. Kwak, T. E. lones, A. F. Garito, and A. 1. Heeger, Phys. Rev. Lett.
31,601 (1973)
49. 1. F. Kwak, P. M. Chaikin, A. A. Russel, A. F. Garito, and A. 1. Heeger, Solid State
Cornrnun. 16,7'29 (1975).
50. J. J. Ritsko. D. J. Sandman, A. J. Epstein, P. C. Gibbons, S. E. Schnatterly, and 1.
Fields, Phys. Rev. Lett. 34, 1330 (1975).
51. A. 1. Berlinsky, J. F. Carolan, and L. Weiler, Solid Siale Cornrnun. 15, 795 (1974); D. R.
Salahub, R. P. Messmer, and F. Herman. Phys. Rev. B 13, 4252 (J 976); F. Herman, D. R.
Salahub, and R. P. Messmer. Phys. Rev. B 16, 2453 (1977). A. Karpfen, 1. Ladik, G.
Stollhoff, and P. Fulde. Chern. Phys. 8, 215 (1973). S. Shitzkovsky, M. Weger, and H.
Gutfreund.1. Phys. (Paris)39. 711 (1978).
52. U. Bernstein, P M. Chaikin, and P. Pincus, Phys. Rev. Lett. 34, 271 (1975).
53. G. Soda. D. Jerome, M. Weger, J. M. Fabre, and L. Giral, Solid State Cornrnun. 18,
1417 (1976).
54. G. Soda, D. Jerome, M. Weger, J. M. Fabre, L. Giral, and K. Bechgaard, Proc. Siofok
Conference. Siofok, Hungary, September (1976).
55. G. Soda, D. Jerome, M. Weger. J. Alizon, 1. Gallice, H. Robert, J. M. Fabre and L.
Giral,1. Phys. (Paris) 38,931 (1977).
56. E. Ehrenfreund and A. J. Heeger, Phys. Rev. B 16, 3830 (1977).
57. P. Fulde and A. Luther, Phys. Rev. 170, 570 (1968).
58. E. F. Rybaczewski, L. S. Smith, A. F. Garito, A. 1. Heeger, and B. G. Silbernagel, Phys.
Rev. B 14. 2746 (1976); E. F. Rybaczewski, A. F. Garito, A. 1. Heeger, and E.
Ehrenfreund, Phys. Rev. Lett. 34, 524 (1975).
59. A Q = 1.57g. see P. N. Reiger and G. K. Fraenkel, 1. Chern. Phys. 37, 2795 (1962);
AI' = 1.26K, set' F. Wed!. G. M. Smith, and E. J. Hufnagel. Chern. Cornrnun., 1453
(1970).
60. C. F. Coil. Phys. Rev. B 9, 2150 (1974).
61. L. S. Smith, E. Ehrenfreund, A. J. Heeger, L. V. Interrante, 1. W. Bray, H. R. Hart, Jr.,
and I. S. Jacobs, Solid State Cornrnun. 19, 377 (1976).
62. E. Ehrenfreund. E. F. Rybaczewski, A. F. Garito, A. J. Heeger, and P. Pincus, Phys.

Rev. B 7, 421 (1973).
63. F. Devreux, Phys. Rev. B 13,4651 (1976).
64. J. B. Torrance, B. A. Scott, and F. B. Kaufman, Solid State Commun. 17, 1369 (1975);
J. B. Torrance. B. A. Scott, D. C. Green, P. Chaudhari, and D. F. Nicoli, Solid State
Commun. 14, X7 (1974); J. B. Torrance and B. D. Silverman, Phys. Rev. B15, 798 (1977);
J. B. Torrance, Phys. Rev. B 17, 3099 (1978); J. B. Torrance, in Chemistry and Physics
of One-Dimensional Metals, H. J. Keller (ed.), p. 137, Plenum Press, New York
(1977).
65. M. Weger (private communication).
66. For a review, see D. Jerome and M. Weger, Chemistry and Physics of One-Dimensional
Metals, H. J. Keller (ed.). Plenum Press, New York (1977).
67. P. A. Lee, T. M. Rice, and P. W. Anderson, Phys. Rev. Lett. 31,462 (1973); see also
M. J. Rice and S. Strassler, Solid State Commun. 13, 1389 (1973); and A. Bjelis and S.
Barisic, J. Phys. (Paris) Lett. 36, L169 (1975).
68. L. 1. Buravov, O. N. Eremenko, R. B. Lyobovskii, L. P. Rosenberg, M. L. Khidekel, R.
P. Shibaeva, I. F. Shchegolev. and E. B. Yagubskii, Sov. Phys. JETP Lett. 20, 208
(1974).
69. Studies carried out on samples prepared independently at IBM (S. Etemad, E. M.
Engler, and J. C. Scott, unpublished) and Penn (T. Wei and A. F. Garito, unpublished)
show behavior qualitatively similar to that observed in TIF-TCNQ. See also S. Etemad,
T. Penney, E. M. Engler, B. A. Scott, and P. E. Seiden, Phys. Rev. Lett. 34, 741 (1975).
70. J. J. Hopfield. Comments Solid State PhYs. 3,48 (1970).
71. P. E. Seiden and D. Cabib. Phys. Rev. B 13,1846 (1976).
72. M. Barmatz, L. R. Testardi, A. F. Garito and A. J. Heeger, Solid State Commun. 15,
1299 (1974)
73. T. Ishiguro, S. Kagoshima, and H. Anzer, 1. Phys. Soc. Japan 42,365 (1977); T. Tiedje,
R. R. Haering, M. H. Jericho, W. A. Roger, and A. Simpson, Solid State Commun. 23,
713 (1977).
74. J. Murgich and S. Pissanetzky, Chern. Phys. Lett. 18,420 (1973).
75. M. J. Rice, C. B. Duke, and N. O. Lipari, Solid State Commun.17, 1089 (1975).
76. T. Takenaka. Spectrochim. Acta 27 A, 1735 (1971).
77. A. Girlando and C. Pecile. Spectrochim. Acta 29A, 1859 (1973).
78. C. J. Fritchie, Jr.. Acta Crystallogr. 20, 892 (1966).
79. C. J. Fritchie, Jr.. and P. Arthur, Jr., Acta Crystallogr. 21, 139 (1966).
80. M. 1. Rice (private communication); see Reference 75 and M. J. Rice, Phys. Rev. Lett.
37,36 (1976).
81. P. Nielsen, D. J. Sandman, and A. 1. Epstein, Solid State Commun. 17,1067 (1976); 15,
53 (1974); W. D. Grobman, R. A. Pollak, D. E. Eastman, E. T. Maas, and B. A. Scott,
Phys. Rev. Lett. 32, 534 (1974); P. Coppens, Phys. Rev. Lett. 35, 98 (1975); W. D.
Grobman and B. D. Silverman, Solid State Commun. 19, 319 (1976).
82. D. Jerome, W. Miiller, and M. Weger, 1. Phys. (Paris) 35, L77 (1974); J. R. Cooper, D.
Jerome, M. Weger, and S. Etemad, J. Phys. (Paris) 36, L219 (1975); A. J. Berlinsky, T.
Tiedje, J. F. Carolan. L. Weiler, and W. Friesen, Bull. Am. Phys. Soc. 20, 465 (1975); S.
Etemad, T. Penney, and E. M. Engler, Bull. Am. Phys. Soc. 20, 496 (1975); see also
C. W. Chu, J. M. E. Harper, T. H. Geballe, and R. L. Greene, Phys. Rev. Lett. 31, 1491
(1973), who saw indications of a second transition at a somewhat higher temperature and
first conjectured that the TIF and TCNQ chains might be undergoing separate tran-
sitions. The second peak in the anisotropy (Reference 15) results from the 38°K
transition.
83. S. Etemad, Phys Ret". B 13, 2254 (1976)
84. W. L. McMillan, Phys. Rev. B, 16,643 (1977).
85. H. A. Mook and C. R. Watson, Phys. Rev. Lett. 36, 801 (1976).
144 A. 1. Heeger
86 V J Emery, Phys Rev Lett 37, 107 (1976)

87 J B Torrance (pnvate commuOicatlOn)
88 P A Lee, T M Rice, and R Klemm, Phys Rev B 15, 2984 (1977)
89 C S Jacobsen, Ph D thesIs, The TechOical UOIverslty of Denmark, Lyngby, Denmark
(1975)
90 L B Coleman, C R Fincher, Jr, A F Garito, and A J Heeger, Physlca Status
Solidi (b)75, 239 (1976)
91 David Penn, Phys Rev 128, 2093 (1962)
92 A A Gogohn and V ! Mel'OIkov (to be pubhshed)
92a M G Kaplunov, K I Pokhodnp, Physlca Status SOlidI (In press) For details on
preparatIOn and charactenzatlOn of (TSeT)2Br see the follOWing L I Buravov, A I
Kotor, M L Khldekel, I F Shchegolev, and E B Yagubsku, Bull USSR Acad SCI
Chem Ser 475 (1976)
92b W J GunOlng, A J Heeger, ! F Shchegolev, and S P Zolotukhln, Solid State
Commun 25 981 (1978)
93 H C Montgomery, J Appl Phys 42,2971 (1971), B F Logan, SORice, and R F
Wick, J Appl Phys 42,2975 (1971)
94 D E Schafer, F Wudl G A Thomas, J P Ferrans, and D 0 Cowan, Solid State
Commun 14, 147 (1974)
95 C K Chiang M J Cohen, P R Newman, and A J Heeger, Phys Rev B 16, 5163
{I 977)
96 M J Cohen, P R Newman, and A J Heeger, Phys Rev Lett 37,1500 (1976), M J
Cohen and A J Heeger, Phys Rev B 16, 688 (1977)
97 P R Newman and L S Smuh (pnvate commuOicatlOn) See also Reference 96
98 D Jerome, W Muller, and M Weger, J Phys (Pans) Lett 35, L77 (1974), S Etemad,
Phys Rev B 13 2254 (1976), P Horn and D Rlmal, Phys Rev Lett 36,809 (1976),
T Ishlguro S Kagoshlma, H Anzal, J Phys Soc Japan 41, 351 (1976)
99 Y TomkieWICZ A R Taranko, and E M Engler, Phys Rev Lett 37, 1705 (1976), see
also S Etemad, Phys Rev B 13, 22')4 (1976), P M Chalkm, J F Kwak, R L Greene,
S Etemad, and E M Engler Solid State Commun 19,1201 (1976)
100 Y TonklewKz A Taranko, and J B Torrance, Phys Rev Lett 36, 751 (1976), Y
TomktewlC7 R A Craven T D <;chultz E M Engler, and A R Taranko, Phys Rev
B (m press)
101 G A Thomas 0 E Schafer, F Wudl, P M Horn, D Rlmal, J W Cook, D A
Glocker, M J Skove, C W Chu, R P Groff, J L Gillson, R C Wheland, L R
Melby, M B Salamon, R A Craven, G De Pasquah, A N Bloch, D 0 Cowan, V V
Walatka R E Pyle, R Gemmer, T 0 Poehler, G R Johnson, M G Miles, J D
Wilson, J P Ferrans T F Finnegan, R J Warmack, V F Raaen, and D Jerome,
Phys Rev B 13 'i 1O'i (I976)
102 Marshall J Cohen L B Coleman, A F Ganto, and A J Heeger, Phys Rev B 13,
5111 (1976)
103 R Groff, A Suna, and R Mernfield, Phys Rev Lett 33,418 (1974)
104 M J Cohen Ph D theSIS, UOIverslty of PennsylvaOla (1975)
105 H Fukuyama, T M Rice and C M Varma, Phys Rev Lett 33,305 (1974)
106 Y !mry dnd SMa, Phys Rev Lett 35, 1399 (1975)
107 Y A Bychkov Zh Eksp Teor FIZ 65,427 (1973) [Sov Phys JETP 38, 209 (1974)]
108 V L Berezlnskll Zh Eksp Teor Flz 65, 1251 (1973) [Sov Phys JETP 38, 620
{I 974)]
109 A A Gogohn V ! Mel OIkov, and E ! Rashba, Zh Eksp Teor Flz 69,327 (1975)
110 A A Gogohn S P Zolotukhlm, V I Mel'OIkov, E E Rashba, and I F Shchegolev,
PIS Zh Eksp Teor Flz 22,564 (1975) [Sov Phys JETP Lett 22,278 (1975)]
III W Gunning and A J Heeger, Phys Rev B (in press), W Gunnmg, A J Heeger, and A
J Epstein Ph\'5 Rev B (In press)
112 We thank Dr L P Gorkov for POIntIng this out to us, see L P Gorkov and I E
OZlaloshInsku,Zh Eksp Teor Flz Pls'maRed 18, 686 (l975)[Sov Phys JETPLett 18,
401 (1973)]
113 J Bardeen, Soltd State Commun 13,357 (1973)
114 B R Patton and L 1 Sham, Phys Rev Lett 31,631 (1973)
115 B R Patton and L ] Sham, Phys Rev Lett 33,638 (1974)
116 0 Allender,] W Bray,andJ Bardeen,Phys Rev 89,119(1974)
117 S Strassler and G A Toombs, Phys Lett 46A, 321 (1974)
118 M 1 Rice, Solzd State Commun 16,1285 (1975)
119 M Cohen, S K Khanna, W ] GunnIng, A F Ganto, and A J Heeger, Soltd State
Commun 17, 167 (I97~)
120 W ] GunnIng S K Khanna A F Ganto, and A 1 Heeger, Solid State Commun 21,
765 (1977)
121 L Buravov and I F Shchegolev, P"b Tekh Eksp 2, 171 (1971)
122 P Richards, Advdnced Research Projects Agency Report, 1973 (unpubhshed)
123 W N Hardy and A J Berllnsky, Phys Rev 814,3356 (1976)
124 Per Bak and V 1 Emery Phys Rev Lett 36,978 (1976)
125 T Ishlguro, S Kagoshlma, and H Amzal, J Phys Soc Japan 42, 1365 (1977)
126 Y TomkiewIcz, B A Scott L ] Tao, and R S Title, Phys Rev Lett 32, 1363 (1974)
127 Y TomkiewIcz, A R Taranko, and J B Torrance, Phys Rev Lett 36,751 (1976)
128 P A Lee, Aspen Workshop all I 0 Conductors, Aspen, Colorado (1975), M J Rice,
Aspen Workshop on I 0 Conductors, Aspen, Colorado (1975), 0 J ScalapIno, Y
Imry and P PInCUS Phys Rev 811,2042 (1975)
129 K Saub, S Banslc, and] Fnedel Phy~ Lett 56A, 302 (1976)
130 T 0 Schultz and S Etemad Phys Rev 8 13, 4928 (1976), see also Guy Deutscher,
Phys Rev 813 2714 (1976)
131 W 0 Ellenson R Comes S M Shapiro, G Shlfane, A F Ganto, and A ] Heeger,
Solid State Commun 20 51 (1976)
132 P Horn and 0 Rlmal Phys Rev Lett 36,809 (1976)
133 R A Craven,M B Salomon,G DePasquall,R M Herman,G Stucky, and A Schultz,
Phys Rev Lett 32 769 (1974)
134 D DJurek, K Franulnvlc M Prester S Tomlc, L Glral and J M Fabre, Phys Rev
Lett 38, 715 (1977)
135 See Reference 7 ~, Wo 2 = Ln g~w n 31Ln gn 2 Wn I, where Wn are the vanous mode
frequencies and gn are the respective coupling constants
136 P M Grant, P Mengel E M Engler, G Castro, and G B Street,8ull Am Phys Soc
21 (3), 254 (1976)
137 C Weyl, E M Engler S Etemad, K Bechgaard, and 1 Jehanno, Solid State Commun
19,925 (1976)
138 F P Pan,D Stroud, and 0 B Tanner, Solid State Commun 20,271 (1976)
139 M J Rice, A R BiShop, J A Krumhansl, and S E Trullinger, Phys Rev Lett 36,432
(1976)
139a A I LarkIn and P A Lee, Phys Rev 817,1596 (1978)
140 M J Rice, In Low-DImensIOnal CooperatIVe Phenomena, H J Keller (ed ), p 23,
Plenum Press, New York (1975)
141 G B Whitman, Linear and Nonhnear Waves, John Wiley and Sons, New York (1974)
142 W J Gunmng, C K Chiang, A J Heeger, and A J EpsteIn, Phys Rev 8 (In press)
143 L] Sham and B R Pdtton, Phys Rev Lett 36,733 (1976)
144 M B Salamon,] W Bray G OePasquall, R A Craven, G Stucky and A Schultz,
Phys Rev 811,619(1975)
145 W 1 GunnIng and A J Heeger (to be published)
146 E Abrahams L P Gor kov and G Kharadze, Solzd State Commun 25,521 (1978)
4
The Organic Metals
{TSeF}x{TTF}l-x-TCNQ-A Systematic
Study
T. D. Schultz and R. A. Craven
1. Introduction
The recent interest in quasi-one-dimensional metals, although stimulated

by hopes of finding a high-temperature superconductor, has been sustained
for two other reasons. First, these materials have exhibited a wide variety
of behavior that may be considered anomalous when compared with
normal three-dimensional nonmagnetic materials. Second, while the
strengths of synthetic organic chemistry assure that there will be a wide
range of possible materials with quasi-one-dimensional behavior, the
number of systems that have been studied extensively at this time is very
small. There is therefore the very attractive possibility that the study of
new systems will increase our understanding and uncover even more
interesting patterns of behavior.
Of all these materials, the organic solid tetrathiafulvalene-tetra-
cyanoquinodimethane (TTF-TCNQ) has been the most comprehensively
studied and is among the richest in anomalous properties. To take special
advantage of the organic-molecular nature of TTF-TCNQ, attempts have
been made to make slight modifications in one or the other of the con-
stituent molecules. It was hoped that new compounds having the same
structure as TTF-TCNQ would emerge, and that a comparison of the
physical properties of these compounds (and of their alloys with TTF-
T. D. Schultz • IBM Watson Research Center, Yorktown Heights, New York 10598.
R. A. Craven • Mom,anto Company, St. Louis, Missouri 63166. This work was begun while
the author was at the IBM Watson Re'iearch Center, Yorktown Heights, New York 10598.
147
148 T. D. Schultz and R. A. Craven
TCNQ) with the properties of TTF-TCNQ would increase our under-

standing of TTF-TCNQ specifically and of quasi-one-dimensional metals
generally.
While a number of modifications of both TCNQ and TTF have been
studied, the modification that is sufficiently small to yield isostructural
crystals yet sufficiently large to yield a host of interesting effects has been
the replacement of the four sulfur atoms in TTF by four selenium atoms,
yielding tetraselenafulvalene, which we denote by TSeF.:j: It is the syste-
matic study of TTF-TCNQ, TSeF-TCNQ, and the alloys (TSeF)x(TTF)l-x-
TCNQ, 0 < x < 1, that is the subject of this chapter.
Research on the system (TSeF)ATTF)l-x-TCNQ is well along but far
from "complete." Of all the members of this series, TTF-TCNQ is much
the most thoroughly investigated. A number of studies of TSeF-TCNQ
have also been reported, enough to indicate many similarities and several
remarkable differences with TTF-TCNQ, but neither the extent nor, as yet,
the reliability of the TSeF-TCNQ studies is comparable to those of
TTF-TCNQ. Some experimental work has concentrated on TSeF-doping,
i.e., replacing small percentages of TTF in TTF-TCNQ by TSeF; much less
has focused on TTF-doping of TSeF-TCNQ. There have also been a few
systematic investigations across the whole alloy series.
Because of the incomplete experimental situation at this time, this
chapter is intended as a progress report. We shall, of necessity, devote
considerable space to TTF-TCNQ, thereby overlapping other chapters of
this book. But since it will be our intention to lay the groundwork for
understanding the significance of results in other members of the series, the
emphasis will be on those aspects of TTF-TCNQ where we believe
experiments on the rest of the series have yielded, or can yield, significant
understanding. Where results for other compounds in the series are already
available, we shall report them and discuss their significance. Where
experiments have never been done or are in progress, we shall indicate the
significance that such experiments could have.
In the presentation of the wealth of material already existent, we have
separated the discussions of the metallic phase (Section 4), the semicon-
ducting phase (Section 6), and the transition between these two phases
(Section 5). We recognize that our understanding of either phase (and
especially of the phase transition) is affected by our understanding of the
other, so the division is in a sense arbitrary and we must from time to time
anticipate work reported in later sections.
In the discussion of each phase, we have also grouped the experimen-
tal results into subsections on the transport, magnetic, and structural
:j: In the abbreViations for this and other selenium compounds, selenium has also been
represented in the literature by S (e.g., TSF, HMTSF, etc.). We have adopted TSeF as being
psychologically. if not logically, superior.
The Organic Metals (TSeF)ATTF)I-x-TCNQ 149
properties. This also necessitates occasional anticipation of results in future

sections, but it seems an appropriate course in view of the lack of any single
model that, in our view, convincingly correlates a broad range of known
physical properties. In fact, the lack of a single fundamental model is a
dominant theme of this chapter that emerges from our considerations and
that should be contrasted with the view taken in the chapter by Professor
Heeger.
Threading through all of our considerations are several broad issues:
the roles of the two kinds of stacks, the role of band structure, the
importance of electron-electron interactions, the nature and effects of the
electron-phonon interactions, the nature and significance of correlations in
the metallic phase, and the importance of collective vs. independent-
particle behavior, to name some of the most important. If we appear from
time to time to have assumed answers to certain of these questions, as for
example in our frequent use of the free- (i.e., noninteracting) electron
approximation, it may be only that the assumed answers are more widely
believed and the consequences of alternative assumptions have not been so
thoroughly investigated. We have occasionally tried to remind the reader
of the extent to which these questions are still open.
In the next two sections we set the stage for more detailed discussion.
In Section 2 we discuss the preparation and characterization of the crystal
samples and the structural parameters of their lattices and in Section 3 we
present a general overview of the phase diagram of the system. In Section 4
and Section 6 we give parallel discussions of the metallic and semiconduc-
ting phases and encourage the reader to read in sequence the correspond-
ing subsections of these two sections, if that seems appropriate. Section 5,
devoted to two aspects of the phase transition, can be read independently
of the rest. A brief summary of the high points and some concluding
remarks comprise Section 7.
2. Preparation, Characterization, and Lattice Structure
Since the major thrust of this chapter is to provide a review of what

has been learned and what problems arise when TTF-TCNQ is alloyed
with TSeF-TCNQ (see Figure 1), a discussion of the material properties
and lattice structures of this system is needed to provide an essential
platform from which judgments of physical measurements can be made.
Although the acceptor molecule TCNQ was synthesized and fairly
well characterized by earlier workers(l) it was the more recent synthesis of
the donor molecule TTF(2) and its subsequent combination with TCNQ(3-6)
that produced the material TTF-TCNQ with its wide variety of interesting
properties. Both TTF and TCNQ are now available commercially,
TTF X, - X4 = S
TSeF X, - X4 = Se
CIS-DSeDTF X X2= S
"
X3'~= Se
trans-DSeDTF X, ,X 3 = S
Figure 1. TCNQ and various fulvalene molecules. X2 ,X 4 = Se
although considerable care must be taken in the purification of these

materials before crystals of reasonable quality can be obtained.
The genesis of this chapter's investigations into the unique properties
of the quasi-one-dimensional organic conductors was the successful
synthesis of TSeF by Engler and Patel(7) and the subsequent preparation of
conducting crystals of TSeF-TCNQ.(8) This discovery was particularly
important, because it allowed changes to be made in the electronic struc-
ture of the TTF-TCNQ crystal with minimal complications from the steric
effects associated with alkylated donor molecules.(9.10) Following this first
successful substitution of the sulfur atoms by selenium, Engler and Patel
developed the mixed sulfur-selenium derivatives cis-DSeDTF and trans-
DSeDTF which also formed conducting compounds with TCNQY 1)
Because of the steric similarity between TTF-TCNQ and TSeF-TCNQ
and, very importantly, because these materials form isostructural
solids,(12-15) an isostructural series of alloyed materials (TSeF)x(TTF)l-x-
TCNQ can be made. The synthesis and details of the characterization of
this alloy series have recently been presented by Engler et a1Y2) The
conclusion of this work is that the TTF and TSeF constituent molecules are
uniformly and randomly distributed in the fulvalene stack and there is
complete solubility of both molecules. These conclusions are supported by
the crystallographic x-ray determinations,(12.15) electron-microprobe
. (12) (16 17) d .
ana IySlS, transport measurements, , an spill-resonance measure-
mentsY 7) Until now, such an isostructural series has been a rare occur-
rence. We know of only one other example, the isostructural pair
diethylTTF-TCNQ and diethylTSeF-TCNQ, which has only recently been
discovered(9) and to date has been little investigated. In other systems, like
hexamethylTTF-TCNQ and hexamethyITSeF-TCNQ, there are important
structural differences.
The Organic Metals (TSeF)xCTTF)l_x-TCNQ 151
Table I. Lattice Parameters at Room

Temperature of (TSeFjJTTF)l_x-TCNQ for
Various x a
a b c (3
x (A) (A) (A) (deg)
-~.--
00 12.281 3.823 18.422 104.43

0.03 12.296 3.830 18.413 104.46
018 12.341 3.846 18.433 104.41
042 12.391 2.862 18.432 104.26
068 12.446 3.870 18.448 104.20
095 12.502 3.873 18.494 104.11
098 12.505 3.873 18.504 104.16
100 12.505 3.876 18.514 104.15
-------- .~~- ..
a From Reference 1 '\
The crystal structures of TTF-TCNQ and TSeF-TCNQ were fully

determined by x-ray crystallographic analysis of the scattering from single
crystals of the pure elements.(!3,14) Using these end points as a guide, the
powder x-ray patterns of the alloys were completely indexed for all
concentrations at room temperature,(12,lS) No nonindexable reflections
were observed, an indication of the isostructurality and solubility of TSeF-
TCNQ in TTF-TCNQ,
A compilation of the structural information on the alloy series has
been given by Etemad et aIY S ) We show their data in Table 1 and Figure 2,
1852,-----.---.---.----.------..,
c-LATTICE
CONSTANT (~)
1846
18.40L------'------'---~---'------I
1252,--------,-----,---,----,-------,
1240
12.28't::......--'-----'-----'---~--
3881,----.---.---.---.---
385
~.~~ ,
Figure 2. Lattice parameters vs. x for

(TSeF).(TIFh-x-TCNQ at room
1 ~:~ ~o_f3_(d_e_gr~e -;: -s)_- - ;: ,- -:- -_~~-
0.2 0.4
-_
06
__ =:;:;:;;::=::=dJ
0.8 1.0
temperature. (After Reference IS.) X
152 T D. Schultz and R. A. Craven
The overall effect of the alloying on the structure is an expansion of the

lattice to accommodate the larger orbital of the selenium ions. This struc-
tural variation is quite smooth and does not show any abrupt changes that
could be associated with an instability or tendency for the TIF and TSeF
molecules to cluster or form separate phases.
The effects of this changing structure on key physical parameters and
distances between the crystals are summarized in Table 2. These distances
are indicated in Figure 3 by arrows where possible. Looking at Table 2, we
see that although the changes in lattice parameters with alloying are not
very great, these changes have a significant impact on certain transfer
integrals, thereby influencing the degree of one-dimensionality of the
electronic motion and the strength of the coupling between the electrons
and those acoustic phonons that modulate the important intermolecular
separations. The transfer integrals we have in mind are to and tF along the
TCNQ and fulvalene stacks, tOF between Bloch functions on a TCNQ
stack and a neighboring (along a) fulvalene stack, and too and tFF along c
between neighboring TCNQ and fulvalene stacks. To see this impact of
Table 2. Lattice Parameters, Stacking Distances, Distances a between Closest

Atoms of Neighboring Molecules, and Related Atomic Parameters for TTF-TCNQ
and TSeF- TCNQb
Parameter TIF-TCNQ TSeF-TCNQ
Chalcogen Slater orbital parameter 0.82 (3p-orb.) 0.95 (4p-orb.)

Chalcogen van der Waals radius 1.85 2.00
Nitrogen van der Waals radius 1.50 1.50
b axis
b (powder pattern) 3.823 3.876
TCNQ-TCNQ separation 3.17c 3.21 d
Fulvalene-fulvalene separation 3.48 c 3. 52 d
4tQ (= bandwidth) 0.58' 0.54'
4tF (= bandwidth) 0.48' 0.58'
a axis
a (powder pattern) 12.281 12.505
Shortest chalcogen-nitrogen distance 3.20c 3.16 d
Sum of corresponding van der Waals radii 3.35 3.50
c axis
c (powder pattern) 18.422 18.514
Carbon-nitrogen contact between TCNQs 3.26 c 3.59 d
a After Reference 15.

b All distances in an~stroms, energies in electron volts.
c Reference 14. Reference 13. e Reference 18.
The Organic Metals (TSeF)x (TTF)1 ~ x - TCNQ 153
Figure 3. Nearest-neighbor contacts in a and c

directions in TTF (or TSeF)-TCNQ. Arrows (1) and
(1') indicate the inequivalent C-N pairs that are the
nearest and second-nearest pairs along c, having
almost the same separations; arrows (2) indicate
S-N (or Se-N) pairs that are the nearest pairs along
a, the S (or Se) and N being in molecules displaced
by ±b relative to one another. Hydrogen atoms,
contributing little to 7r orbitals, are ignored. (After
Reference 14.) 0-
replacing TTF by TSeF, let us compare the changes that take place in the
lattice parameters with the changes in the van der Waals radius and the
Slater orbital parameter when going from sulfur to selenium.
One of the clearest consequences of the change in lattice parameters is
the change in the intrastack TCNQ-TCNQ transfer integral to caused by
the increase in b. Recent electronic structure calculations by Herman et
al.(18) yield a change in to from 0.58 eV at x = 0 to 0.54 eV at x = 1 due to
the change in b. Such a decrease in to should not only reduce the TCNQ
bandwidth but should also decrease the coupling of the TCNQ electrons to
the acoustic phonons that modulate b.
Although the intermolecular spacing within the fulvalene stacks also
increases when we go from x = 0 to x = 1, the effects of this are more than
offset by the larger van der Waals radius and Slater orbital parameter of
the selenium atom: while the interfulvalene distance increases by 0.10 A,
the van der Waals radius increases by 0.15 A. The corresponding increase
in the transfer integral tF should result in a greater fulvalene bandwidth, as
was calculated by Herman.(18) This behavior is reflected in measurements
of both the thermoelectric power and the optical plasma frequency, as we
shall discuss in Section 4.1. The increase in tF should also result in a
stronger interaction between the electrons on the fulvalene stacks and the
acoustic phonons that modulate b.
In addition to the changes in b, changes in a also have a noticeable
impact on the electronic properties. The increase in a from 12.28 to
12.50 A is accompanied by an increase rather than a decrease in the
interstack transfer integral tOF for two reasons. First, there is an increase in
the width of the fulvalene molecules that more than offsets the increase in
a and results in a decrease in the chalcogen-nitrogen distance. Second, the

selenium orbitals are more extended than the sulfur, which makes the
effective chalcogen-nitrogen distance even smaller. This increase in tOF is
noticeable not only in the conductivity along a but also in measurements
that are sensitive to departures from electronic one-dimensionality, e.g.,
EPR linewidths (Section 4.2) and critical exponents near the metal-semi-
conductor transition (Section 5.2).
The increase in c with x is, percentagewise, much smaller than that of
a or b, and the chalcogen atoms are not involved in the nearest-neighbor
atom pairs. Consequently, the effect on the relevant transfer integrals too
and tFF is much less.
The change in lattice parameters does, of course, affect interactions
other than the transfer integrals. The change in the Coulomb interactions
between charged stacks may influence the amount of charge transfer from
fulvalene to TCNQ stacks, and the change in the elastic and Coulomb
interactions between charge-density waves that develop on the stacks can
affect many properties associated with soft phonons, with the metal-semi-
conductor phase transition, etc. The systematics of these changes are,
however, more difficult to discuss because of uncertainties about the
degree of Coulomb screening that occurs. Also, the r-1 dependence of the
Coulomb interaction, at least at the distances between neighboring stacks,
may make the sensitivity of these interactions to changes in lattice
parameters less than that of the exponentially dependent transfer integrals.
In later sections, the effect of changing lattice parameters on charge trans-
fer will be taken directly from other experiments, while the effects on the
interactions between charge-density waves will be largely ignored.
3. Phase Diagram
To get a general overview of the alloy system (TSeF)x(TTF)1-x-

TCNQ, we consider its phase diagram as shown in Figure 4. We see that
broadly speaking there are two phases, a high-temperature metallic phase
in which the room-temperature conductivity is high and increases with
decreasing temperature, and a low-temperature semiconducting phase in
which the conductivity is low and decreases with decreasing temperature.
The phases are separated by a transition at the temperature Tel (x), which
we shall call the metal-semiconductor transition.
The metallic phase is also characterized by a magnetic susceptibility
that decreases markedly with decreasing temperature, an EPR line that is
unusually narrow for such a good metal, and sheets of anomalous diffuse
x-ray diffraction intensity corresponding to phonons or other kinds
of distortions having q vectors near the planes qb == Iq· bl! b = const. == Qb
The Organic Metals (TSeF).(1TF)I_x-TCNQ 155
w
a:
::::J
t:ra:
w
a..
~
w
I- SEMICONDUCTING PHASE T.
~ 30 c\'
i=
Vi
Figure 4. Phase diagram of z
<
a:
(TSeF)x (TIF)I-x -TCNQ. Transitions I-
at Tc3 and Tc4 have been seen only for
x =0. x
(the "2kf anomaly")t and near the planes qb = 20b == 01, (the "4kf
anomaly").
The semiconducting phase is also characterized by the development of
a periodic lattice distortion that is incommensurate with the crystal lattice.
Within this phase in the pure and nearly pure systems (x = 0 and x = 1),
additional transitions are seen, a total of three near x = 0, whose tempera-
tures in order of decreasing temperature we denote by Te3 (x), Te4(x) and
Te2 (x), the numbering corresponding to the chronological order of their
discovery, and one near x = 1 at a temperature we also call Te2(x). This
notation has the virtue of adapting to the discovery of ever more tran-
sitions. All of these transitions are very sensitive to fulvalene doping, being
broadened in most cases into unobservability with a few percent alloying
(e.g., x = 0.03 and x = 0.97 for Te2)' In the case of the transition at
Te2 (x =0), the transition temperature is also lowered by about 25%, from
38°K to 29°K, when x = 0.03. Most of the phase transitions, especially the
metal-semiconductor transition at Te 1 (x), have been observed in several
different kinds of experiments including the dc conductivity, spin suscep-
tibility, specific heat, and diffuse x-ray and elastic neutron scattering.
Experimental studies have, of course, concentrated on certain parts of
this phase diagram. The pure systems, especially TTF-TCNQ, have been
the most extensively studied. The lightly doped systems (x 2=: 0 and x:s 1),
especially doped TTF-TCNQ, have also been studied, but to a lesser
extent. In the alloys of intermediate concentrations, only the crystal struc-
ture, EPR, dc conductivity, and thermoelectric power have been investi-
gated so far. Generally speaking, the doping studies have shed light on the
pure systems, while the alloy studies allow one to discern systematic trends
linking the behavior near x = 0 to that near x = 1.
We turn first to a consideration of the metallic phase, with T> Tel,
then to a detailed consideration of the metal-semiconductor transition at
t Because we use F (and Q) in subscripts to denote "fulvalene" (and TCNQ), we use {, rather
than the customary F, to denote Fermi. as in k f and E f .
Tel, and finally to the low-temperature phase(s) including the other tran-
sitions.
4. Metallic Phase, T> Tc I
4.1. Transport Properties
Measurements of the transport properties of TTF-TCNQ, TSeF-

TCNQ, DSeDTF-TCNQ, and the alloy series (TSeF)x(TTF)l-x-TCNQ
have been the source of a great deal of the information as well as much
experimental and theoretical controversy surrounding these materials. In
the high-temperature metallic state, measurements of the dc electrical
conductivity, microwave conductivity, and thermoelectric power have been
made on the pure materials and on selected compositions over the entire
alloy series. Additional measurements of the optical reflectivity and dc
conductivity as a function of temperature and applied pressure have also
been made for TTF-TCNQ and TSeF-TCNQ.
The electrical conductivity measurements at dc and microwave
frequencies demonstrate similar temperature dependences for the pure
materials and across the alloy series, suggesting that a significant part of the
scattering processes is insensitive to the details of the changing molecular
species. There are differences, however, and measurements of the ther-
moelectric power. the optical reflectivity, and the effects of pressure on the
dc and optical conductivity help to clarify these differences. The ther-
moelectric power is a sensitive probe of which molecular stack is respon-
sible for conduction, while measurements of the optical reflectivity and of
changes induced in the optical and dc conductivity with applied pressure
give information about the relative size of the bandwidth in TTF-TCNQ
and TSeF -TCNQ and the sensitivity of the conduction to this banding.
4.1.1. Low-Frequency Electrical Conductivity
The initial experimental results on TTF-TCNQ, which were widely

reported, emphasized the large increases in the dc electrical conductivity
when the temperature was lowered from room temperature.(4,5,19) This
conductivity peaks at some temperature above the metal-semiconductor
transition Tel, but the actual magnitude of the conductivity at this peak is
still the subject of considerable controversy.(20.21) conductivity measure-
ments on TSeF-TCNQ, DSeDTF-TCNQ, and the alloy system
(TSeF)ATTF)J-x-TCNQ indicate that the general shape of the conduc-
tivity curve is the same for all of these isostructural materials,cs, I 5) The
conduction increases with decreasing temperature, reaching a peak value
The Organic Metals (TSeF)x(TTF)I _x-TCNQ 157
1 4r---~--~----~--~---'----'
12
'\
1\
I \
1
10 I
J
TSeF - TCNO
I
I
8 I
J
:.:: I
b J
o
~ 6 J
b I
"- I
f-
I
b
4 I
J
J
J
J
2 I
I
Figure 5. Normalized dc conductIvity I '
I
along b axis vs. T for ITF-TCNQ, 1./
TSeF-TCNQ, and DSeDTF-TCNQ o 50 100 150 200 250 300
TEMPERATURE (OK)
(After Reference 8.)
at some temperature Tp above the phase transition, before undergoing the

metal-semiconductor transition at Tel as illustrated in Figure 5.
Quantitative analysis of the temperature dependence of the conduc-
tivity also suggests that similar relaxation processes are responsible for the
conductivity in all of these isostructural compounds. An intrinsic tempera-
ture dependence of the resistivity can be defined by assuming p (T) =
Po+ eTA. This separates the universal temperature dependence of the
resistivity from effects that can be associated with sample defects, non-
homogeneous current paths, and other variations in crystalline quality.
This method of analysis has been applied to TTF-TCNQ by several
workers. They have found that A = 2.33 ± 0.15 for both the dc(4.22) and
microwave(23.24) conductivity. A similar analysis has been done for TSeF-
TCNQ.(8.23) We illustrate this nearly quadratic temperature dependence by
plotting the measured resistivity versus the square of the temperature in
Figure 6. A similar plot can be made for the alloy series, but the disorder
introduced in the fulvalene stacks causes considerable rounding far above
the transition temperature.
One measure of sampie quality that has been widely used in compar-
ing samples is the ratio of the peak conductivity to the room-temperature
conductivity. This ratio typically reaches values of 25-50 for TTF-TCNQ,
12-15 for TSeF-TCNQ (possibly as high as 22), 6-7 for OSeOTF-TCNQ,
25 TTF - TCNQ -
.' .'
;-
/
;"
20
,,~ DSeDTF - TCNO

15
.c:
E
v
/
,- /
Q 10 ,-
,-
~
I ....-....-"
--- ---
I _-~TSeF-TCNQ
5 1\
\ _.
I
0 2 4 6 8 10
(TEMP ERATURE)2 (10\ (OK)')
Figure 6. DC resIstivity along b axis vs. T2 for one sample each of ITF-TCNQ, TSeF-
TCNQ, and DSeDTF-TCNQ. (After Reference 8.)
and 8-25 for (TSeF)x(TTFh-x-TCNQ. The lowest conductivity ratios in

the alloy series are found for x = 0,5. Although the TTF-TCNQ crystals
have shown larger increases in the conductivity ratio, TSeF-TCNQ crystals
have shown a higher room-temperature conductivity. Both materials reach
conductivity levels of the order of 1-3 x 104 (O-cmr 1 in routine measure-
ments. One of the effects of alloying, the introduction of disorder which
decreases the effective scattering time of the electrons and holes, can be
seen in both the room-temperature conductivity and in the overall shape of
the conductivity curve. As we show in Figure 7, the conductivity drops
rapidly with alloying at both ends of the isostructural series. Analogously,
the width of the transition region as measured by the temperature
difference Tp- T, I increases rapidly with alloying. For ITF-TCNQ,
800
700 20
\:
<.>
600 10
~
S 500
S2 400
0 ~
- 10 S2
0
0
00 300 -20 ~ Figure 7. Average dc conductivity and
'"b 200 ,
a:: thermoelectric power along b axis
, - 30 ~
100
vs. x for several samples of
----_#- -40
(TSeF)x(TTFh_x-TCNQ at room
-50
0 0 .5 1.0 temperature. (After References 15, 48,
x and 49.)
The Organic Metals (TSeF).{ITF)l_x-TCNQ 159
70
..... -.- Tp
-.-. ,
--. --.
60 ....
" \
::.:: \,
~
w
cr
::>
~
cr
w 40
0-
~
W
~
30
Figure 8. Metal-semiconductor
transition temperature Tel and
peak-conductivity temperature Tp 20
vs. x in (TSeF)x(TIF)l_x·TCNQ 0.0 0.2 0.4 0.6 0.8 1.0
(After Reference 15.) x
Tp - Tel = S°K. It increases to IsoK for x = 0.03, is larger than 200K for x =
0.68, and decreases rapidly again in TSeF-TCNQ to about 8°K. This is
illustrated in Figure 8.
Several theories have been proposed to explain the temperature
dependence of the conductivity in the metallic region. Although these
theories are discussed extensively in other chapters in this book, a brief
listing indicates the complexity of the problem, and to some extent the lack
of complete understanding. The temperature dependence has been
ascribed (1) to a collective enhancement of the conductivity due to some
type of Peierls-Frohlich sliding mode,(25-29) (2) to the enhanced effects of
electron-electron scattering due to the reduced dimensionality and nar-
rowness of the bands,(8,30) (3) to electron-two-phonon scattering in a
reduced dimensionality,(31,32) (4) to electron-libron scattering,<33) (5) to
a strongly temperature-dependent mobility caused by scattering from
molecular optical phonons, (34- 17) and (6) to electron transport through an
anisotropic medium that has a mixed structural or electronic phase. (38) (7) In
addition, the possibility of diffusive motion resulting from thermallocaliza-
tion of the electrons, which is particularly easy in one dimension, has been
analyzed. (39,40) (8) Very recently, it has been argued (41,42) that a large part of
the temperature variation IS due to thermal expansion, the conduction at
constant volume varying approximately linearly with T over much of the
metallic range; and in this connection (9) it has been proposed(43) that the
resistive mechanism is the scattering of electrons by low-lying spin-wave-
like excitations. There is also an additional possibility, although somewhat
un l1'ke Iy, t h at t h e (T = T- 7- 11±0 l'i. temperature d epend ence IS
. umque
. to
systems that are strictly isostructural to TTF-TCNQ, If this were the case,
other materials, with different symmetry and crystal coordination would

show different power-law dependences. Transport measurements on other
one-dimensional materials with segregated stacks and nonintegral charge
transfer will be useful in determining the universality of this conduction
behavior. The isostructural pair DETTF-TCNQ and DETSeF-TCNQ,
which was mentioned in Section 2, is one promising example. These form
crystals with a different structure and appear to have much less enhancement
·· WIt
o f t h e con d uctlvlty .hd '
ecreasmg temperature. (94445)
. ,
4.1.2. Thermoelectric Power
When TTF-TCNQ was first studied, it was usually assumed that the
metallic behavior occurred on only the TCNQ stacks. The increase in
conductivity observed when going from TTF-TCNQ to TSeF-TCNQ
suggested, however, that there is appreciable conductivity on the fulvalene
stacks, at least in TSeF-TCNQ, because it seemed unlikely that the
conductivity of the TCNQ stacks would increase markedly as its bandwidth
decreased. This evidence was not completely convincing, however, since
other effects of the replacement of TTF by TSeF could conceivably
enhance the mechanism, as yet undetermined, of the TCNQ conduc-
tivity.
A more informative probe of the roles played by the two kinds of
stacks in electrical conduction has proved to be the thermoelectric power.
Its measurement and analysis by Chaikin and co-workers in TTF-
TCNQ,(46.4 7 ) TSeF-TCNQ,(48) and the alloy series(49) have given important
information about these roles, as we now describe.
The thermoelectric power(50) is simply the energy (measured relative
to the Fermi energy) carried by the electrons or holes, per unit charge,
divided by the temperature of the conducting medium. For a set of iden-
tical, metallic, one-dimensional stacks of very weakly interacting electrons
described by the tight-binding approximation, the thermoelectric power is
(1)
where t is the transfer integral along the stack, II is the number of conduc-
tion electrons per site (0 < II < 2), and r(E) is the mean scattering time for
electrons of energy E. In the square brackets, the first or "band" term and
second or "scattering" term are explicitly temperature-independent for
free-electron and tight-binding bands(51) with fixed charge transfer ll. For a
varying charge transfer, these terms can at best be only weakly tempera-
ture dependent. Thus one expects the thermoelectric power to be pro-
portional to T for a metallic system. For a semiconductor with energy gap
The Organic Metals (TSeF)ATTF)l_x-TCNQ 161
E g , by contrast, the thermoelectric power is

k~(b -} Eg mh) (2)
S=-);I b+1 kBT+lnm e
which goes like T- 1 and can be large. Here, b is the ratio of the electron to
hole mobility and me and mh are the corresponding effective masses. The
sign of the thermoelectric power is negative (positive) if the transport of
energy is dominated by the electrons that are above (below) the Fermi
energy. This is similar, but not quite the same, as saying that the conduc-
tivity is dominated by one or the other group of electrons.
When there are two kinds of stacks that interact only weakly, the total
thermoelectric power is an average of that for the individual stacks,
weighted by their respective conductivities:
(3)
If both kinds of stacks are described by Equation (1), if the scattering term
proportional to T'(E)/ T(E) can be neglected, and if the transfer integrals
are assumed to be of opposite sign as is believed to hold for TTF-TCNQ
and related systems, then the combined thermoelectric power has the form
7T2k~T2 [ (Tl - T2) cot(7Tv/2) ]
S=- 61e1 It l h+lt2IT2sin(7Tv/2) (4)
For the series (TSeF)x(TTFh-x- TCNQ, where it is believed that v < 1, we
see that the sign of S is determined by the stacks with the longer scattering
time. Interestingly, the scattering term, which is proportional to
T'(E)/T(E), cannot be completely neglected for TTF-TCNQ, because
estimates(51) of the strength of the band term give a contribution that is not
quite large enough to account for the room-temperature value of
- 28 IL V;oK. This discrepancy does not seriously detract from the overall
picture of which stack is responsible for the conduction in the alloy series,
however.
The temperature dependence of the thermoelectric power as
measured along the b axis in TTF-TCNQ,(46) TSeF-TCNQ,(48) and the
alloy series(49) is shown in Figure 9. Although the details of these
measurements vary because of the varying contributions from the fulvalene
and TCNQ bands, there is an overall similarity in the temperature depen-
dences. At high temperatures, the thermoelectric power is nearly linear in
temperature, characteristic of the metallic state. For temperatures below
125°K but above the phase transition temperature Tel, an additional
contribution to the thermoelectric power, which is of a positive sign, is
noted. Below Tel the thermoelectric power is essentially proportional to
T- I , characteristic of the semiconducting state.
-60r-------~--------~------__,
-50
-40 x'O.SO / •••••••••••• /
-30
.'
..,/ x.<::o~~. - ·
,.
. ~ .... . "
;
;:
o
-20 , .. ./ ,/
.....
~;;;~:::cc~~;~t~
>
-=--10
0..
UJ
I--
o
10
20 X·I ,O
Figure 9. Thermoelectric power

300L-.l..-..L---- 1..i... 300 along b axis vs. T for several values of
OO---------'200'----------1
x in (TSeF)x(TTFh-x-TCNQ. (After
TEMPERATURE (OK) Reference 49.)
Let us look more closely at the thermoelectric power at room

temperature, where, over the whole alloy series, it is proportional to T, so
that the band term can be presumed to dominate. The experimental results
are shown in Figure 7. We see that in the pure materials S changes from
-28/J.V tK in TTF-TCNQ to +3 /-LV tK for TSeF-TCNQ. Within the
model discussed above, this can be interpreted as arising from an increase
in the fulvalene scattering time that allows the positive (hole conduction)
SF to compete more effectively with the negative SQ.
The behavior in the alloys supports these conclusions and also
emphasizes the role of disorder in shortening the scattering time on the
fulvalene stacks. The rapid drop in 'TF that is expected on fulvalene doping
of pure TTF-TCNQ (or of pure TSeF-TCNQ) increases (or restores) the
dominance of the TCNQ stacks and makes the thermoelectric power more
negative. That the effect is more important at the TSeF-TCNQ end is
expected, because in TTF-TCNQ the scattering time is already relatively
short when compared with 'TQ.
It is interesting to compare the conclusion from the thermoelectric
power measurements that the TCNQ stack is responsible for most of the
conduction at room temperature with conclusions from measurement of
the Hall effect. Based on their dc Hall-effect measurements, Cooper et
al. (52) agree with the interpretation that the main conductivity is due to the
electrons. Based on their recent microwave Hall measurements with a
different orientation of the magnetic field, Ong and Portis(53) disagree,
The Organic Metals (TSeF)x(TTF)I_x-TCNQ 163
however, claiming that the relaxation time for holes is longer than that for
electrons. The microwave measurements are more difficult to interpret, but
this in itself does not make them invalid. A serious problem for both
techniques is the different nature of the conduction perpendicular to the b
axis, i.e., via some sort of diffusive hopping mechanism. The full impli-
cations of the Hall measurements are not yet clear.
4.1.3. Optical Properties

The most significant feature of the optical properties of TTF-
TCNQ(54,55) and TSeF-TCNQ(56) is the metallic-like plasma edge in the
optical reflectivity when the light is polarized parallel to the conducting axis
(Ellb), and the absence of this structure when the light is polarized perpen-
dicular to the b axis (E 1. b). This anisotropy reflects the anisotropic band-
ing of the stacked molecular system. Interestingly, this reflectance spec-
trum does not change significantly in TTF-TCNQ or TSeF_TCNQ(51,57)
when the crystals are lowered to temperatures below Tel .(54) This result is
consistent with TTF-TCNQ and TSeF-TCNQ being metallic in character
above Tel, but also indicates that the gap responsible for the strong
increase in the dc or low-frequency resistivity is small compared with the
bandwidth along the conducting axis.
The optical reflectivity of TTF-TCNQ and TSeF-TCNQ, which is
shown in Figure 10, has been analyzed with a Drude model, using a
0 .50' .....- - - - .- - ---.------.-----,

\
\
\
\
\-TSeF-TcNO
0.40 \ wp=I.96 x 10" (sec- ')
\
\ T =4.20 x 10'''(sec)
\ (",=2 .25
\
\
\
\
:> \
i= \
u \
W
...I \
LL
~ 0 .20 \--.:,-1_ TTF-TCNO
Wp=1.82 x 10'"(sec-')
T = 3.0 x10-'>( sec)
£",=2 .25
0 . 10
Figure 10. Optical reflectivity vs. w for

TIF-TCNQ and TSeF-TCNQ. The elec-
tric field is parallel to the b axis. (After 0 .6 0 .8 1.0 1.2
References 54-56 ) ",(eV)
frequency-dependent dielectric function
(5)
where l' is an energy-dissipative relaxation time, Wp is the plasma

frequency, and EO is the real, constant, background dielectric contribution
at high frequencies. This analysis gives direct information about the
bandwidth of the conduction electrons. Using a tight-binding approxima-
tion for both the fulvalene and TCNQ band structures, the sum of the two
bandwidths becomes
(6)
where nc is equal to the combined density of states and b is the lattice

spacing along the conducting axis. The plasma frequency at atmospheric
pressure, which increases from 1.19 eV to 1.29 eV when TTF is replaced
by TSeF, indicates a corresponding increase in the optical bandwidth.
What is particularly interesting about the optical measurements is that
the temperature dependence of the optical conductivity is the same as that
of the dc and microwave conductivity, despite the fact that these
measurements are above the apparent gap in the far infrared at about
1000 cm -\ which has been interpreted as being related to the high-
conductivity mechanism.(58.59) Seiden and Cabib(30) have shown that in
TTF-TCNQ the optical relaxation time 1'=(po+aT2rl, where O'opt=
w~1'147T. The constant value for po is high when compared to Po for dc
measurements, but this may be an experimental problem associated with
optical measurements on highly anisotropic materials. More recently Blu-
deau et al. (57) have repeated this analysis on TSeF-TCNQ with the same
conclusion. On the basis of this analysis it seems unlikely that the far-
infrared gap (which had been related to the possibility of paraconductive
fluctuations above T, I (';X,';9)) does in fact play a role in the quadratic
temperature dependence.
4.1.4. Pressure Effects on Transport Measurements, T> Tel
The most immediate and obvious effect of applying isotropic pressure

to single crystals of TTF-TCNQ or TSeF-TCNQ is a compression of the
lattice. This compression changes the overlaps of the wave functions
between the molecular constituents, modifying the bandwidths and scat-
tering times. Measurements of the optical reflectivity,(56,60) dc conduc-
tivity,(61,62) and compressibilit/63 ) have all been made as a function of
pressure. The measurements indicate that the bandwidth and the room-
The Organic Metals (TSeF)xCTTF)I-x- TCNQ 165
temperature conductivity both increase with increasing pressure. The

increase in bandwidth with pressure as measured by the plasma frequency
is essentially the same for both TTF-TCNQ and TSeF-TCNQ: dhwp/dP=
8x 10- 3 eV/kbar or din wp/dP= 1 %/kbar. The room-temperature
conductivity shows a much greater sensitivity to pressure. Researchers
have found d In fT(P)/ dP = 23 % /kbar(61,64) and d In fT(P)/ dP =
40% /kbar(62) for TTF-TCNQ, and din fT(P)/ dP = 18% /kbar(62) for
TSeF-TCNQ. Although the fractional change appears to be somewhat less
in TSeF-TCNQ, the total change at ambient pressure is larger because of
the larger conductivity.
The large sensitivity of the dc conductivity to pressure, and presum-
ably therefore to volume, led Cooper(41) to suggest that the strong
temperature dependence of the conductivity might be due in part to the
thermal expansion of the lattice. To estimate the temperature dependence
of fT at constant b, Friend et al.(42) have used measurements of dfT/ dP and
the compressibility along b. They obtain a resistivity that varies nearly
linearly with temperature, a result that could have an important effect on
future attempts to understand the dc-conduction mechanism. For example,
a number of resistivity mechanisms that had been considered attractive
because of their T2 behavior would have to be rejected.
Because the conductivity is much more sensitive to pressure than
would be estimated from the pressure dependence of n (Ef ) (estimated
from din w p / dP) and of the electron-phonon coupling strength (estimated
from dTc d dP(38)), Jerome(43) has proposed that the effective Coulomb
repulsion strength (U in the Hubbard model) is comparable to the
bandwidth and that, in this regime, the resistivity is due to scattering of
single electrons by antiferromagnetic spin-wave excitations. Jerome argues
that such scattering would be very sensitive to U, and therefore to pressure,
and that it would also give a roughly linear temperature dependence.
Conductivity measurements under pressure in TTF-TCNQ and TSeF-
TCNQ also show the remarkable result(62) that the dc conductivities along
all three axes show the same percentage shifts with pressure. The implied
pressure independence of the anisotropy ratios should be contrasted with
their marked temperature dependence.
4.1.5. Conclusions
The transport measurements on TTF-TCNQ, TSeF-TCNQ, and the

alloy series for temperatures above Tc 1 give information as to which stack
is responsible for the conduction and to some extent indicate the complex-
ity of the conduction process. In TTF-TCNQ the conduction is primarily
along the TCNQ stack. In TSeF-TCNQ the conduction is on both stacks.
This is consistent with a larger bandwidth for TSeF.
Optical measurements confirm this bandwidth picture. The conduc-

tion mechanisms in TTF-TCNQ and TSeF-TCNQ are similar, but the
scattering mechanisms, temperature dependence, and pressure depen-
dence are still not understood.
4.2. Magnetic Properties
Although the anomalous peak in the dc conductivity first drew wide

attention to TTF-TCNQ, and other electrical and optical properties
continue to stimulate a host of controversies over which of several
mechanisms best explains the observed phenomena, the magnetic
properties of TTF-TCNQ have also called forth a variety of explanations,
none of which seem to fare well under closer scrutiny. To get some insight
into this challenging situation, we first review the experiments in the
metallic phase of TTF-TCNQ, summarizing two of the most basic contro-
versies about their interpretation. Realizing that the restriction to the
metallic phase in this section is somewhat arbitrary, we shall anticipate
certain magnetic properties of the semiconducting phase where appro-
priate. Second, we consider what has been learned by a systematic study of
the alloy series. Finally we review what is known of the properties of pure
TSeF-TCNQ and how this relates to TTF-TCNQ.
4.2.1. TTF- TCNQ
Both static susceptibility measurements using a Faraday balance and

various EPR measurements were reported by Tomkiewicz et al. (65) and by
1. C. Scott et al.(66) The static measurements are attractive because of their
ease and precision, but they suffer from the uncertainties involved in
subtracting off a large and possibly temperature-dependent diamagnetic
contribution to get the spin susceptibility. The EPR measurements, it was
originally believed, could be missing the entire contribution from a stack
with a very broad line. Consequently, a detailed comparison of the static
and EPR susceptibilities was made by Gulley and Weiher,(67) who showed
that the results agree from 300 K to 3000 K under the assumption of a
temperature-independent diamagnetic susceptibility. Composite results,
including those from the static measurements of Tomkiewicz et al. (65) from
300 K to 420 o K, are shown in Figure 11.
0
Of initial interest was the temperature dependence of Xspm in the

metallic state, which is very unusual: (1) It is not temperature-independent,
as would be expected from the Pauli paramagnetism of a nonmagnetic
metal of free-electron-like quasiparticles, provided only that T« E,; (2) it
is not diverging as 1/ T, as would be expected if the electrons were so
localized as to have negligible interactions between their spins; and (3) it is
The Organic Metals (TSeFMTTF)I-x-TCNQ 167
.2
!:
a.
><
Figure 11. Total spin susceptibili-
ties VS. T for TIF-TCNQ and
TSeF-TCNQ. Composite data from
static and EPR measurements.
(References 65-67.)
not decreasing exponentially in 1/ T, as would be expected for electrons or

spin excitations having an activation energy. Instead, Xspin decreases
roughly linearly with decreasing T from 300 to 60 0 K with
Xspin(60}/ Xspin(300} = 0.4. This behavior was attributed by Scott et al. (66) to
a deepening pseudogap in the density of states created by the growth in
amplitude and, more important, correlation length of Peierls fluctuations
(see Section 4.3) with decreasing temperature, based on the theoretical
work of Lee et al. (68) The fit of this theory to experiment(66) was only rough,
but the theory assumed the Peierls fluctuations to have a wavelength equal
to 2b (i.e., it used a real order parameter). Later use of a theory modified to
reflect the incommensurate nature (Section 4.3) of these fluctuations (i.e.,
use of a complex order parameter) improved the agreement. (69)
This explanation of the T dependence of X,pm was in fact taken as an
important piece of evidence to support the idea that Peierls fluctuations of
large correlation length are present even up to room temperature.
However, three different experiments have provided evidence against this
explanation. First, x-ray scattering experiments to be described in Section
4.3 indicate that the correlation length of such fluctuations is very short
above -lOooK and the fluctuations themselves are unobservable for T>
IS0°K. Second, it was shown by Tomkiewicz et al.(70) that the effects of
TSeF doping on the temperature dependence of the spin susceptibility are
opposite to what would be expected if Peierls fluctuations, whose
development would be inhibited by such TSeF doping, were the case of the
fall in Xspin(T}, as we discuss in Section 4.2.2. Finally, it was argued by
Jerome et al. (71) that if one-dimensional fluctuations were pronounced over
a wide temperature range, this range would shrink under pressure, as the
system became more three dimensional electronically, and the temperature
dependence of Xspm would thereby be affected. EPR measurements of Xspin
under pressure by these authors showed no such effect of any significance.
FIgure 12. E vs. k surfaces for a two-

dImensIOnal array of molecules consis-
tmg of alternating TTF and TCNQ
stacks. Gap due to mterstack mixmg is
maximum (=2~a) for ka = 0 and
vamshes on the zone boundary at ka =
- b"/2 - b"l4 - kf kf b"/4 bO/2 a*/2=Tr/a.
It was also realized by Bernstein et al.(72) that the interstack mixing

along the a direction of electronic wave functions on a pair of neighboring
TTF and TCNQ stacks could produce a gap 2Aa at the Fermi energy
(curves on ka = 0 plane in Figure 12). In a perfect lattice, where this
interstack mixing becomes a banding along a, as shown in Figure 12, the
result is a decrease in the density of states around E f . For kB T:s Aa the
susceptibility would decrease with decreasing T. Furthermore, the suscep-
tibility would be expected to decrease with increasing pressure, as was
observed by Jerome et al.,(7!) because the application of pressure would
increase tOF and thereby enhance the decrease in the density of states.
However, with a temperature-independent band structure, the fit to the
temperature dependence of Xspm was not good.(72) Also, as we have seen,
the conductivity increases with increasing pressure.
It was later appreciated by Torrance et al.(73) that the magnitude itself
of Xspm at and above room temperature might force a radical revision(73.74)
of the widely accepted assumption of essentially free-electron behavior
within the stacks. Even at very high temperatures, the number of spins
capable of contributing to the susceptibility, as measured by TXspm, is lower
if the electrons are free than if they are strongly repelling, because in the
former case some electrons of opposite spins will be in the same spatial
state and therefore unable to flip their spin, whereas in the latter case, all
electrons can contribute. If kB T« E" (as is believed to be the case, at least
on the stack undergoing a Peierls transition), this difference is magnified
because TXspm for free electrons, but not for strongly repelling electrons, is
reduced from its high-temperature value. As a result, it is argued, the
free-electron picture is unable to explain the high value of TXspm observed.
For electron repulsion of intermediate strength, the value of TXspm is
presumably also Intermediate, the so-called Coulomb enhancement. On
the basis of the observed magnitude of TXspm and various estimates of the
bandwidths involved, Torrance(7'») estimates that U/4t - 2-3. It should be
rLmarked that if the Coulomb interactions are strong, then the tempera-
The Organic Metals (TSeF)x(TTF)l-x-TCNQ 169
ture dependence of Xspm will also be modified. In fact, there will then be an
effective antiferromagnetic interaction between the electron spins that will
result in a spin susceptibility that, at sufficiently low temperatures, falls
with decreasing temperature. However, efforts to fit observed Xspm(T) with
Hubbard-model predictions have failed. (71.75) Of course, there are other
implications of such strong electron-electron repulsions, which can be
profound, as Torrance has arguedY'i)
The behavior of X,pm as a function of pressure at constant temperature
has also been adduced as an argument against a free-electron like model.
According to Jerome et al.,(71) Xspm decreases much more rapidly with
increasing pressure than would be expected from estimates of the pressure
dependence of the bandwidths, which are, in turn, derived from the pres-
sure dependence of the plasma frequency. This too has been used to argue
that U - 4to or 4tF.
Since there are two kinds of stacks, there are in fact two magnetic
behaviors to understand, so it is important to distinguish the separate roles
of the two kinds of stacks in contributing to the susceptibility. The key to
this was seen first by Walsh et al., (76) who saw a single Lorentzian EPR line
with an essentially temperature-independent g value over most of the
metallic range, intermediate in magnitude between the g values of the TTF
and TCNQ molecules, rather than distinct lines for the TTF and TCNQ
stacks separately. They interpreted this as evidence that the magnetic
excitations hop rapidly from stack to stack, residing approximately equally
on the two kinds of stacks, over this whole temperature range. A system-
atic study of the roles of the two kinds of stacks (especially near the phase
transition and in the insulating state) was subsequently undertaken by
Tomkiewicz et al .. (74 77) who used the fact that for strongly interacting spin
excitations (indicated by the single Lorentzian line), the observed g value
should average the separate g values go and gF, in proportion to the time
the spin excitations spend on the two kinds of stacks. This is, in turn,
proportional to the respective spin susceptibilities, whenever the inter-
actions between stacks are sufficiently weak for XspmO and Xspm.F to be
meaningful. Thus,
(7)
where Xspm = Xspm,O + Xspm,F is the total spin susceptibility. This equation
can be viewed as a relation between g tensors or between their rotational
averages. Taking for go the g value in NMP-TCNQ or of TCNQ ions in
solution and for gF the g value observed in TTF-TCNQ at -25°K [where
gobs is T independent, implying that Xspm.O = 0, as we shall see, so that
Equation (7) implies gobs = gF] and taking for Xspm either the value deter-
mined by EPR or from static measurements, the total spin susceptibility
was d ecompose d O? 78) mto
.
Xspm.O an d Xspm,F as sown
h 'm F'Igures 13 an d
35
30
., 25
"0
E
"-
::>
E
20
Q)
<t
'0
15
c:
a.
on
)( 10
05
400
TEMPERATURE (OK)
Figure 13 Spm susceptibilIty vs T of TTF and TCNQ stacks, lO-420oK DecomposItion of

total susceptibilIty IS by g-value method (After References 77 and 78 )
14. Two properties first mentioned by Walsh et al.(76) for the metallic state
stand out: the susceptibilities Xspm.O and Xspm.F have similar temperature
dependences, at least for T> 65°K, and have comparable magnitudes
(xspm,O/ Xspm F = 48/52).
In understanding these results there are three possibilities: First, it is
possible that the separate susceptibilities determined by this g-value
method are properties of the two kinds of stacks, considered independently.
In thIs case, the mechanIsms producing the anomalous temperature
16
Qj
"0
E 12
"-
::>
E
Q)
q-
'Q 08
c:
0-
ut
>< 04
°0 70
Figure 14 Spm susceptlblhty vs T of TTF and TCNQ stacks, 1O-60o K DecomposItion by

g-value method (After Reference 77 )
The Organic Metals (TSeF)ATTF)l~x-TCNQ 171
dependence of the two, similar spin susceptibilities should be similar.

Because the roles played by the two kinds of stacks in the metal-semicon-
ductor transition are quite different, as we shall discuss in Sections 6.1 and
6.2, it would seem most unlikely that these mechanisms could be associated
with fluctuations that are the precursor of the phase transition (e.g., Peierls
fluctuations, as proposed by Scott et al. (66») which involve the two kinds of
stacks in entirely different ways. For T < 65°K, where Xspin,O but not Xspin,F,
begins to fall rapidly with decreasing temperature, such fluctuations could
indeed playa role. In fact, this decrease when coupled with the very rapid
drop in Xspm,O below Tel and the smooth variation of Xspin,F through the
transition was the first unambiguous evidence that the TCNQ stacks drive
the 53°K transition.(74.77) If it is Coulomb effects that are responsible for
the anomalous temperature dependence, then, barring radically different
bandwidths on the two kinds of stacks, one would have to conclude that if
the Coulomb enhancement effects are important on one kind of stack, they
are also important on the other.
Second, it is possible that the susceptibilities themselves, or at least the
mechanisms accounting for their anomalous temperature dependence, are
inherently coupled. The only mechanism of this kind to have been pro-
posed is banding along a, which, as we have said, does not give the correct
T dependence.(72) If another such mechanism is proposed that fits the data,
the different behavior of Xspm,O and Xspin,F below 65°K would then be
attributed to the precursor fluctuations, which are predominantly on the
TCNQ stacks, being strong enough to override the coupling mechanism.
A third possibility is that the g shift does not measure the relative
susceptibilities of the separate stacks. After all, when there is significant
magnetic coupling between the stacks, the very notion of susceptibilities on
individual stacks may not be well defined. On the other hand, when the
separate susceptibilities are well defined, and one is in the strong-coupling
regime, indicated by a single Lorentzian line and a field-independent g
shift, Equation (7) is valid.(74)
In addition to the EPR g-shift approach, two other attempts have been
made to determine XSPIO,O and Xspm,F using NMR by Rybaczewski et ai. In
the first(79) the proton NMR TJ was measured in samples in which either
the TTF or TCNQ molecules had been deuterated. Coulomb enhance-
ments were assumed negligible and solution hyperfine constants were used
to relate the respective TJ's to the local susceptibilities, In the second(69)
the Knight shift of DC nuclei introduced into the CN groups of the TCNQ
molecules was measured and the normalization constant of the Knight shift
was assumed to be temperature independent and was chosen to fit the
room-temperature TJ's of the proton NMR experiment. The results are
qualitatively similar to the decomposition by g shifts, i.e., Xspin,O/ Xspin,F is
roughly T independent down to SooK, although the value of this ratio
determined by proton NMR (= 34/66) is quite different from that deter-

mined by the g value (=48/52). Other discrepancies apparent around Tel
and Tc3 and certain internal inconsistencies suggest that the assumption of
a temperature-independent hyperfine coupling may be invalid.(8o) On the
basis of the temperature and frequency dependence of Tl at higher
magnetic fields, Soda et al.(81.82) have also argued that Coulomb effects are
far from negligible.
A final magnetic property of TTF-TCNQ is the EPR linewidth, which
is smaller by two orders of magnitude than predicted by the Elliott relation
for the spin-lattice relaxation time of an isotropic metal. As we shall see
below, it is the alloy series that allows us to identify the broadening as
coming from spin-lattice interactions. It has been argued on the basis of a
weakly-interacting-electron assumption(83) that the very narrow line results
through a combination of three effects: (1) very weak coupling to the
acoustic modes with q = 0 that are involved if k' = k; (2) a selection rule for
spin-flip scattering,(84) following from time-reversal invariance, that makes
the matrix element vanish for k' = -k; and (3) energy conservation, which,
in electronically one-dimensional systems, requires that Ik~ I= Iklll thereby
ensuring that spin-lattice relaxation processes, so dominant in three-
dimensional metals, are approximately forbidden in one-dimensional
metals.
The observed EPR linewidth, besides being so small, has an
anomalous temperature dependence. Whereas EPR linewidths arising
from spin-lattice relaxation normally decrease with decreasing tempera-
ture, the linewidth in TTF-TCNQ increases, by a factor of 2 between 300
and 60 0 K for one orientation. A first attempt was made(85) to explain this
anomalous behavior in terms of an enhancement of on-stack spin-flip
scattering due to a progressive softening with decreasing temperature of
the 2kf phonons (see Section 4.3) that are involved in the process. It was
shown, however, that this mechanism alone is unlikely to explain the
anomalous behavior. In the next section, we shall report on the tempera-
ture dependence of the line width in the alloys and in TSeF-TCNQ, which
should give important additional clues as to the mechanism.
4.2.2. (TSeF)x(TTF)l-x- TCNQ, 0 < x:s 1
Let us now turn to the isostructural compounds (TSeF)x(TTF)l-x-

TCNQ including TSeF-TCNQ. The principal investigations have been the
systematic EPR studies by Tomkiewicz and co-workers of the g value,(7o)
t h e separate susceptl'b'l' .
IltJes Xspm.O an d Xspm.P,·
(8687) an d t h e
linewidth(83,R6,88l for a variety of compositions. Static susceptibility
measurements have also been made for x = 0.03 by Singer et al,(89) and for
pure TSeF-TCNQ by Etemad, Scott, and co_workers(90,91) and by Buravov
The Organic Metals (TSeF)ATTF)l-x- TeNQ 173
et al. (92) Let us first consider Xspin, the g value, and the susceptibility
decomposition and then turn to the linewidth studies.
The first group of EPR results concerns the effect of alloying on the
temperature dependence of the observed g value and its implications for
the mechanism behind the T dependence of Xspin' Referring to the pro-
posal that the temperature dependence of Xspin was due to the developing
pseudogap associated with growing Peierls fluctuations,(66) Tomkiewicz et
al. (70) argued that alloying on the fulvalene stacks should introduce dis-
order there, thereby limiting the growth of Peierls fluctuations and the
development of an associated pseudogap on those stacks. By inhibiting the
fall of Xspin.F, but not of Xspin,Q, the alloying would be expected to shift the g
value toward gF with decreasing temperature, the more so the larger x. The
g values vs. T, as given in Reference 70, are shown in Figure 15 for
x = 0.03,0.18,0.42, and 0.68. It is seen that for x = 0.03 and x = 0.18, for
which the occasional TSeF molecules might be considered impurities, the
effect on the g value is opposite to that predicted. This was taken as
additional evidence against the Peierls-fluctuation theory of Xspin(T) in
TTF-TCNQ,
When coupled with an EPR determination of Xspin(T), the g values
make possible the decomposition of Xspin into Xspin,Q and Xspin,F for a
number of alloy concentrations X,(87) with the qualitative result that the
linear decrease in Xspin,F with decreasing temperature is hastened, while
the falloff in Xspin,Q is inhibited by increasing x. In fact, Xspin,F falls to a very
small value at Tel for x = 0.68, The full implication of these experiments,
which deliberately destroy any similarity that might exist between the two
kinds of stacks and/or which may override any mechanism that couples the
stacks, awaits further analysis,
20.0.9
2.0.0.8
~
__
»
_ .c:;.:--
7' r
-'
I
-----
X=o.18
/" ,,'
2.007
,"
/
I
I
9 I
I
2.0.0.6 / I
, ,,'
,/
,/
Figure 15, g value vs. T in 2005 ,c
,,,,,,,,,
(TSeF)x(TTF)l_x-TCNQ for x =
X=0.68,,'
0.03, 0.18, 0.42, and 0.68. Data 200.4 ---_ ......... '
indicate increasingly rapid drop of
Xspin.F from room-temperature value 60 300
for increasing x. (After Reference 70.)
The EPR studies of pure TSeF-TCNQ are limited to T?:.150oK

because at lower temperatures the EPR line is so broad as to be unobserv-
able. It is also dangerous to extrapolate from values of x for which the line
is still visible, because some physical properties change rapidly with x for
x::s 1, e.g., the transition temperature Tel and, as we shall see, the EPR
linewidth at room temperature. Nevertheless, Xspin,F and Xspin,Q have been
determined at room temperature from the g shift and Xspin with the
remarkable result(86) that Xspin,FI Xspin,Q, which is = 1.1 in TTF-TCNQ, as
we have seen, is reduced to -0.25 in TSeF-TCNQ.
The static susceptibility measurements made on TSeF-TCNQ in the
metallic phase(90-92) extend over a much wider temperature range. The
results of Scott et al. are shown in Figure 11. While the temperature
dependence of Xspin for TSeF-TCNQ is roughly similar to that for TTF-
TCNQ, the magnitude of Xspin is only half as large, only slightly more than
Xspin,Q in TTF-TCNQ. If the only change in Xspin,Q in going from x = 0 to
x = 1 arose from the slight narrowing of the band associated with an
increase in b, we would conclude that Xspin,F in TseF-TCNQ would have to
be extremely small. In actual fact, the g-value decomposition(86) indicates a
slight decrease in Xspm,Q, so Xspm,F is reduced by a factor of -5. Although
the fulvalene bandwidth in TSeF-TCNQ is presumably wider than in
TTF-TCNQ, as we have seen from several points of view, this increase in
bandwidth could hardly reduce Xspm,F by such a large factor. Thus we are
forced to conclude that the reduction in Xspm,F is due to something other
than changing bandwidths.
One possible explanation for the decreased Xspin,F is increased wave-
function mixing between the fulvalene and TCNQ stacks, which would
reduce the density of states at the Fermi level and therefore the suscep-
tibility. Nevertheless, this explanation must be rejected because it would
result in comparable reductions in both Xspin.F and Xspin,Q and would also
affect the temperature dependence of Xspm on both stacks in similar ways.
A second explanation has been proposed by Tomkiewicz et al., (86) viz.,
that the Coulomb enhancement (or Hubbard U) is much less on the TSeF
stacks than on the TTF stacks. How this might be related to the different
temperature dependence of Xspm,F on these two donor stacks has not yet
been analyzed.
The second group of EPR results in the isostructural compounds is
concerned with the linewidth and its dependence on x and T. A plot of the
room-temperature linewidth r vs. x, shown in Figure 16(83) makes it
evident that the very narrow line in TTF-TCNQ (-5 G) broadens by a
factor of -90 in going to TSeF-TCNQ. If it is assumed that the line
broadening is due to spin-phonon interactions, then it should be dominated
by the spin-phonon interactions on the fulvalene stacks, because the g
shifts, which are a measure of the spin-orbit coupling strength, are typically
The Organic Metals (TSeF)x(TTF)l-x- TCNQ 175
en 600 r----.--,--.----.---,-.-----,--,--r-, 30 x 10- 5

en I
~ 25 6 _
400 ~ ~
lJJ "
~ g
I
15 ...J -
b 200 10 [;3"'-
N u.
~
lJJ
~
5 ::J .3'
~,
...J 0:: r...
o
x z
Figure 16. EPR linewidth r and normalized linewidth r /(!!.gF)2 vs. x in (TSeF).(TIF)l_x-
TCNQ. !!.gF is average g shift of fulvalene stacks measured as described in Section 6.2.3.
(After Reference 83.)
30 and 70 times larger on the TTF and TSeF molecules, respectively, than
on the TCNQ molecules. In comparing linewidths for different x, one
should therefore normalize them by dividing by the square of the fulvalene
spin-orbit interaction strength. Taking this to be proportional to the
fulvalene g shift as determined in the insulating state (see Section 6.2), one
obtains the normalized linewidth "YF == r / (~gF )2, also shown in Figure 16,
which now increases by only a factor of -17, lending support to the idea
that the linewidth arises from the spin-orbit interaction.
As we have already remarked, under the assumption of weakly inter-
acting electrons the small linewidth in TTF-TCNQ is due to the one-
dimensionality of the electron states. By the same reasoning, the increase
of "YF with x has been taken(83.87) as a sensitive indication of the increased
mixing of wave functions on different stacks with increasing x. A
comparison of the chaIcogen van der Waals radii with the chaIcogen-
nitrogen contact distances (between the TCNQ and fulvalene molecules in
neighboring stacks along a) for TTF-TCNQ and TSeF-TCNQ has already
indicated a source of this interstack mixing (Section 2). That the interstack
hopping rate is proportional to the intrastack scattering time, as well as to
the square of the interstack transfer integral, has been argued by Soda et
ai.(82) If this is so, the increase in room-temperature linewidth with x could
also be attributed in part to an intrastack scattering time that increases

smoothly with x. Such a behavior is inconsistent, however, with the x
dependence of the room-temperature conductivity along the stacks, which
is sharply decreased with alloying (see Figure 7).
The temperature dependence of the linewidth for a number of alloy
concentrations has been measured. (88.93) The results shown e\"en larger
increases of linewidth with decreasing T for x = 1 than for x = 0, but a
negligible increase in at least the range 0.18 < x < 0.67, and possibly
outside it as well. This suggests that the mechanism producing the increase
in "YF is destroyed by the disorder associated with alloying, one additional
clue in the as yet unsolved problem of the temperature dependence of the

linewidth.
4.2.3. Conclusions
In concluding this section, we note that there ace a number of

important and open questions involving the magnetic properties: the
magnitude and temperature dependence of the susceptibility, the relation-
ships between the two kinds of stacks, the importance (if any) of Peierls
fluctuations for the magnetic properties above 65-70oK, the importance of
electron correlations, the T dependence of the EPR linewidth, the size of
the interstack transfer integrals, and the effects of interstack banding along
a. The systematic study of the magnetic properties of the isostructural
series has added a new dimension to these problems.
On the one hand, the studies have led to two new insights about the
metallic state: (1) The effect of fulvalene-doping on the temperature
dependence of -the g value casts further doubt on fluctuations associated
with the metal-semiconductor transition as the source of the temperature
dependence of XSPIn' (2) The x dependence of the EPR linewidth makes it
likely that the relaxation mechanism is via the fulvalene spin-orbit cou-
pling, and the residual x dependence of the normalized linewidth is evi-
dence that at least one interstack transfer integral, believed to be along a, is
much larger for larger x.
On the other hand, these studies have raised a variety of new ques-
tions:
(1) Why does the spin susceptibility decrease with TSeF doping at
temperatures well below room temperature?
(2) Why does the spin susceptibility on the fulvalene stacks decrease
so much more rapidly with decreasing T for large x than for small x?
(3) Why is XspIn,F at 3000 K so much smaller for TSeF-TCNQ than for
TTF-TCNQ, and if the difference is due to a much smaller Coulomb
enhancement in TSeF-TCNQ, why is the enhancement so much smaller?
(4) Why does the EPR linewidth grow for decreasing T for x = a and
x = 1, but not for intermediate alloying concentrations?
The answers to these questions should deepen our insights into the
whole class of two-stack organic metals,
4.3. Phonon Anomalies
In this section we summarize the information obtained from diffuse

x-ray diffraction and inelastic neutron scattering in the metallic state of
TTF-TCNQ. We also discuss the x-ray work in TSeF-TCNQ and in the
x = 0,03 alloy, but because these results are preliminary and fragmentary,
The Organic Metals (TSeF).(TTF)I-x- TCNQ 177
and because there are no neutron results for these materials, the discussion
will be deliberately tentative and will draw attention to what we may hope
to learn from more detailed work with x of- O. We start with a brief summary
of the situation in TTF-TCNQ, referring the reader to the article by Comes
and Shirane, Chapter 2 in this volume, for a more extensive presentation.
The belief that the essentially one-dimensional nature of the elec-
tronic motion in TTF-TCNQ would make that system particularly suscep-
tible to a Peierls instability and to its higher-temperature counterpart, a
giant Kohn anomaly, both with wave number along the stacking axis
qb = 2k" led to an intensive search for these phenomena using the diffuse
scattering of x rays. The analogy with the quasi-one-dimensional metal
K2Pt(CN)4Bro.3· H 20 (KCpi 94 - 96 ) spurred the search for the Kohn
anomaly. Furthermore, the metal-semiconductor transition was believed
to be a three-dimensional transition leading to a Peierls-distorted state, a
transition and state which could not be seen in KCP because of the
disordering effects of the random positioning of halogen ions.
After some initial failures,(97) diffuse x-ray scattering along sheets in k
space was seen in the metallic phase by Denoyer et at. (98) using film
techniques and by Kagoshima et at.(99) using counter techniques.
Subsequently, more detailed x-ray investigations have been reported by
both groupsYOO-102) Closely related inelastic neutron investigations by
groups at Oak Ridge l103 ) and Brookhaven(104,105) have also contributed to
our understanding, although the fact that the single crystals are so small
makes these experiments very difficult. The principal results and their
interpretation are the following.
(1) Diffuse scattering at temperatures well above Tel was found along
pairs of sheets in k space displaced by a fixed amount Qb == O,295b*
(b* = 2n/ b) from the a*c* planes of the reciprocal lattice, i.e" along planes
swept out by the vectors G + q, where G is any vector of the reciprocal
lattice and q is any phonon wave vector for which qb = ±Qb. In the con-
ventional interpretation, in which interactions among electrons are
ignored, this scattering is believed to be inelastic x-ray scattering in which
there is emission or absorption of a quantum in a lattice vibrational mode
characterized by the wave vector q. In this view Qb = 2kf, because modes in
one or more vibrational branches with such q's may be softened, i.e., they
may have their frequencies reduced by the virtual scattering of an electron
across the Fermi sea, an excitation having a total wave number ±2kf and
energy =0. This gives rise to stronger x-ray scattering than for modes with
qb ¥ ±2kf because the scattering intensity associated with a vibrational
quantum of frequency w goes as 1/ w 2. The softening for this narrow range
of qb is called the Kohn anomaly. Because of the one-dimensional charac-
ter of the electronic system, the anomaly is expected to be far stronger than
in normal three-dimensional metals and, in contrast to three-dimensional
metals, it occurs for q's lying in planar sheets. At lower temperatures, one
expects scattering also to occur from quasistatic fluctuations having the
same value for qb. These fluctuations, which can be thought of as a collec-
tive mode in the gas of soft phonons, are precursors of a three-dimensional
phase transition to a Peierls-distorted state, and so we call them Pieierls
fluctuations. The fact that 0.295 is not a simple rational fraction supports
the interpretation that some kind of Kohn anomaly is being seen, since in a
Kohn anomaly, Qb is related to kf, the one special wave vector in the
system that is expected to be incommensurate with the underlying
reciprocal lattice. If the electrons are non interacting, then the value
0.295b* for 2kr implies a charge transfer of 0.59 electrons/molecule.
Other possible sources of this anomaly, depending on the electrons being
strongly repelling, have also been proposed, as we shall discuss later in this
subsection.
(2) These diffuse sheets were not observed above 1500 K and their
intensity was found to increase with decreasing temperature, correspond-
ing to the expectation that the softening of modes with qb = ±2kr would
increase progressively as the Fermi distribution in an approximately
noninteracting electron gas sharpens.
(3) The sheets were seen to sharpen, i.e., to become more concen-
trated near qb '" ±Qb, with decreasing temperature, a reflection of increas-
ing correlation length for correlations between displacements of those
scatterers that are separated from one another by a vector with a
component along b. In fact, in the simplest of models, where the molecules
are displaced rigidly along b and the stacks are uncorrelated, the intensity
at a point K = G + q due to the fluctuations of one stack should be pro-
portional to
K~ L e,qb(m-n)b (d(mb )d*(nb)
(m-n)
where d (mb) is the displacement of the mth molecule along the stack.
Thus, if
then,
1/~b
I(qb)oc (1/ ~b)2 + (qb - 2kd
and the half-width at half-intensity is the reciprocal of ~b, the correlation

length. In TTF-TCNQ, Khanna et alYOl) have studied ~b(T) and find a
correlation length of -1.5 molecular spacings at 150o K, growing to -7
spacings at 900 K and -26 spacings at 60o K. It should be remarked that
these results do not confirm the 1/ T dependence of ~b assumed by Ryba-
czewski et al. (69) to explain Xspm (T) in terms of growing Peierls fluctuations.
The Organic Metals (TSeF)x(1TF)l-x-TCNQ 179
Furthermore, if ~b is only =7 b at 90o K, then sliding conductivity, if it

exists, must enter at a lower temperature, which would make the single
analytic form for the dc conductivity from room temperature down to
~6SOK hard to understand.
(4) Similarly, intensity variations were also observed along the sheets,
reflecting correlations between displacements of scatterers with a separa-
tion vector having a component perpendicular to b. Slow variations in
intensity in these directions would correspond to correlations between
displacements of nearby atoms; more rapid variations, to correlations of
displacements of more remote atoms and/or of molecules of different
stacks. With decreasing temperature, these more rapid variations become
more pronounced, especially within 5-10oK of the transition, when the
correlation length perpendicular to the stacks grows rapidly.
(5) General trends in intensity within and between sheets were also
observed which could be interpreted as giving information about the
direction of polarization of the anomalous modes, assuming that these
modes correspond to rigid molecular displacements, i.e., are neither
librons nor intramolecular vibrations. For rigid molecular translations in a
direction u, the intensity corresponding to a scattering vector K = G + q
goes like (u . K)2 so that, from the trends with varying K, deductions have
been made about the polarization u of the most anomalous modes. Initially
it was believed that these modes were longitudinal,(100) then transverse
with displacements along c* ,002) and, most recently, bothyol)
(6) Inelastic neutron scattering experiments by Shirane et aIY04,105)
have shown a pronounced anomaly in an acoustic phonon branch at qb =
Qb having its polarization principally along c*. This anomaly is first seen at
-150 o K and grows with decreasing temperature. In view of the tilts of the
molecules in both kinds of stacks, displacements along c*, as well as along
b, modulate the intrastack transfer integral. This modulation has often
been assumed to be the dominant electron-phonon coupling mechanism.
The first neutron measurements, by Mook and Watson/ 103 ) appeared to
show a pronounced anomaly at room temperature, but after some contro-
versy<104,105) Mook et aly o6 ) have concluded, on the basis of the latest
evidence, that no distinct anomaly has as yet been observed.
(7) The least anticipated and hence most surprising result of the x-ray
investigations has been the discovery of an anomaly on sheets for which
qb = Qt, ==2Qb, the so-called 4k f anomaly, discovered by Pouget et alYoO)
and studied further by Kagoshima et al. (102) and Khanna et aIYO!) At 60 0 K
this anomaly is comparable in intensity to the 2kf anomaly, but it is visible
up to room temperature, is polarized longitudinally (i.e., along b), and
shows a shift of Qt, from O.59b* to O.55b* occurring in the 500 K range
around 200°K. It is also reported(lO!) to have a correlation length that
decreases only very slowly up to 2000 K (where it is still ~ 13 molecular
spacings). The 4kr anomaly has not been seen in a search of the longi-
tudinal acoustic branch by inelastic neutron scattering.
A number of interpretations of the 4kr anomaly have been proposed.
The possibility that it is from a higher harmonic of the 2kr fluctuations, or
from multiple scattering off 2kr fluctuations (i.e., with the emission of two
softened 2kr phonons), seems to be ruled out by the fact that the 4kr
anomaly is much stronger than the 2kr anomaly above 1500 K and its
intensity has a very different temperature dependence. The possibility that
the scattering is with the emission of two softened 2kr librons has been
proposed by Weger and Friedel,(107) who argue that one-libron emission is
not seen because it is weak due to symmetry considerations. Sham(108) has
suggested that 4k r phonons are softened by a mechanism that depends on
the virtual, not real, emission of softened 2kr phonons. Emery<109) has
suggested that 4k r phonons are softened by virtual scattering of two elec-
trons across the Fermi sea.
A variety of other origins of the 4kr anomaly arise from arguments of
Torrance, made(llO) before the 4kr anomaly was seen and elaborated on
subsequently.!7') Torrance argued that if the electrons were strongly repel-
ling, then the same number of electrons as would show an anomaly at Ob if
noninteracting would have a tendency to form an approximately periodic
superlattice, much like a Wigner lattice, with a reciprocal lattice vector
having qb = 20b. Certain limiting cases of this situation have been
analyzed: that of infinite on-site but zero off-site repulsion(111) (in which
the infinite repulsion coupled with minimization of the kinetic energy
accounts for the tendency to form a charge-density wave) and that of
negligible bandwidth but finite on-site and off-site repulsions(112,113) (in
which the repulsions alone account for the charge-density wave). In the
first case, the tendency helps to soften certain phonon modes with qb =
20b ; in the second, there are quasistatic fluctuations, which we might call
Wigner fluctuations, that would account for the anomaly at 20b .
Torrance has also emphasized that, if this regime actually obtains,
there can be anomalies in both x-ray and neutron scattering at Ob, as well.
These can occur because there would be weak antiferromagnetic inter-
actions between the spins of successive electrons which could provide
sources of such anomalies in two ways: (a) an antiferromagnetic magnon
branch having zero-energy magnons at qb = () and qb = ± Ob, the latter being
observable in principle hy spin-flip neutron scattering; and (b) phonon
softening at Oh, due to the emission of two magnons (one at qb = 0, one at
qb = ± Ob) of opposite spin and roughly zero energy, and quasistatic fluctua-
tions related to this softening (i.e., precursors of a spin-Peierls
transition(114)) hoth of which would be observable with x rays and neutrons.
The TTF-TCNQ experiments leave some important questions
unanswered and raise some new ones. First, to what extent is the 2kr
anomaly in the x-ray scattering due to the phonon softening that has been
The Organic Metals (TSeF).(ITF)I_x-TCNQ 181
seen already by inelastic neutron scattering, and to what extent is it due to

other softened modes and/or to quasistatic fluctuations, either Peierls
fluctuations or spin-Peierls fluctuations? More neutron experiments for
T ~ 53°K are needed.
Second, what are the roles played by the two kinds of stacks in the
x-ray 2kr anomaly? Because the intensities of the scattering from the
individual stacks are added when the stacks are uncorrelated, and because
intensities of the two kinds of stacks should have different dependences on
the transverse component Kas* + Kcc* of the scattering vector, it is possi-
ble in principle to separate out the contributions of the TTF and TCNQ
anomalies, but this too is yet to be done. It is expected that it is the TCNQ
stacks that should have the deeper Kohn anomaly because, as we have
remarked in Section 4.2.1 and as we shall see in more detail in Sections
6.1-6.3, it is believed that the TCNQ stacks drive the metal-semiconductor
transition.
Third, how does the degree of softening depend on the stack involved?
The variation of the softening with the transverse component of q is in
principle directly measurable with neutron scattering and can be deter-
mined separately for the two kinds of stacks by looking at zone-boundary
phonons (qa = a* /2) which are on only one kind of stack or the other. All
published neutron studies have been of (0, 1,0) phonons.
Fourth, what role, if any, is played by the softening of librons(107)? An
investigation of Iibron branches near 2kr by inelastic neutrons is needed.
Finally, the discovery of the 4kr anomaly has of course raised many
questions: (a) Does this anomaly involve only one kind of stack (which
one?) or both? For example,(1OI) could the 4kr anomaly be on the TTF
stacks while the 2k, anomaly (or part of it) comes from the TCNQ stacks?
(b) Why do the intensity and correlation length fall off much less rapidly
with increasing temperature than for the 2kr anomaly? (c) Does the 4kr
anomaly arise from mode softening or from quasistatic fluctuations (a more
important question for the 4kr anomaly than for the 2kr anomaly, in view
of the slower temperature dependence and the suggestions based on strong
electron repulsions, which suggest quasistatic fluctuations)?
Experiments on the alloys and on TSeF-TCNQ can throw light on
some of these questions. In the weakly doped alloys, the fulvalene doping
should have a much greater effect in suppressing both kinds of fluctuations
on the fulvalene stacks, if they occur there, than on the TCNQ stacks. A
comparative study of the effect of such doping on the intensity and thick-
ness of both kinds of anomalous sheets, particularly for temperatures at
which the correlation lengths are comparable to the average separation of
dopant (TSeF) molecules, should therefore give information on the relative
roles played by the two kinds of stacks in each kind of anomaly. Very
preliminary results by Kagoshima et alY 15) for x = 0.03 alloys show no
appreciable effect on the values of Qb or Q~, but other effects and other
concentrations have still to be studied. The effect on the 4kr anomaly in the
semiconducting phase of such doping is already significant, as we shall
discuss in Section 6.3.
The results for TSeF-TCNQ by Weyl et a1Y1 6 ) are much more
detailed and striking but still preliminary. Experimentally, the scattering
factor of the Se atoms, being much bigger than that of the S atoms, makes
the TSeF molecule, and especially its Se atoms, an important source of the
diffuse scattering. Several contrasts with TTF-TCNQ are to be noted: (1)
Only one anomaly is observed with either Qb or Q~ = 0.31Sb*. (2) This
anomaly is visible up to 2400 K (cL 1500 K for the 2kr anomaly and at least
room temperature for the 4kr anomaly in TTF-TCNQ). (3) The correlation
length of this anomaly at 90o K, which is far above the metal-semiconduc-
tor transition temperature, is SOb, roughly seven times that for either
anomaly in TTF-TCNQ at that temperature,:!: despite the fact that Tel in
TSeF-TCNQ is barely half that of TTF-TCNQ. (4) A slow variation of
intensity within the sheets is seen, characteristic of the scattering factor of
the four Se atoms in the TSeF molecule, showing that the fluctuations or
softened phonons giving rise to this anomaly are partly if not totally on the
TSeF stacks and involve rigid displacements of the TSeF molecules.
Because the TCNQ molecules are much the weaker scatterers in TSeF-
TCNQ, no conclusions have been drawn about their role in the observed
anomaly. This situation is to be compared with TTF-TCNQ where the
strong 2kr anomaly is on the TCNQ stacks, although the 4kr anomaly is
partly or totally on the TTF stacks.
From this evidence, it is not clear if the observed anomaly in TSeF-
TCNQ is a 2k, or 4k, anomaly.:!: Furthermore, it may be difficult to deter-
mine which it is from differences in charge transfer. If it is a 4kr anomaly
and if Qi, = 1 - 0.315, then the charge transfer per molecule would be
0.685. This is so similar to the charge transfer of 2Qb = 0.63, which would
be inferred if it were a 2kr anomaly with Qb = 0.315, that other measures
of the charge transfer (e.g., certain intramolecular bond lengths) would be
unable to resolve the ambiguity. Nevertheless, it is worth noting that in
either case, the charge transfer would be greater than in TTF-TCNQ (0.59
electrons/molecule), a result that is contrary to any expectations based
solely on the iomzation potential of TSeF being larger than that of TTF. A
third possibility, that it is a 4k f anomaly with Q~ = 0.315, would imply a
charge transfer of 0.315. This would be consistent with the larger ion-
ization potenttal of TSeF and would also be sufficiently different from the
other two possibilities to be distinguishable using bond-length measure-
ments, although such a charge transfer would be anomalously small in the
family of TTF-TCNQ derivatives.
:j: Sec Note 1 Added tn Proof on page 225

The Organic Metals (TSeF).(ITF)l_x-TCNQ 183
At this time it seems m0re likely that the anomaly in TSeF-TCNQ is at

2kf . The apparently continuous behavior of the metal-semiconductor
transition with x would suggest that the source of the anomaly is similar to
that in TTF-TCNQ, i.e., it is a 2kf anomaly. Also, the large increase in
correlation length at -lOooK is consistent with the intensity at - 2000K
having increased from that of the 2kf anomaly in TTF-TCNQ rather than
having decreased from that of the 4kf anomaly. However, it will require a
detailed study of this anomaly over the whole range of concentrations,
which is now in progress by Kagoshima and co-workers (for first reports,
see Reference 10) to settle the question of its identity.
Whether it is the 4kf anomaly or the 2kf anomaly that has disappeared
in going to TSeF-TCNQ has important and entirely different implications.
For example, if it is the 4kf anomaly that has disappeared, then the 2kf
anomaly in TSeF-TCNQ would appear to be much stronger than in TTF-
TCNQ. This would be consistent with the idea that the strength of the
electron-phonon interaction on the fulvalene stacks increases in going to
TSeF-TCNQ, although the size of the increase would be surprising. The
disappearance of the 4k f anomaly would also imply that whatever
mechanism is responsible for that anomaly in TTF-TCNQ could not also
be responsible for properties that are essentially unchanged in going to
TSeF-TCNQ, e.g., the TCNQ spin susceptibility. Thus, if it is argued that
the 4kf anomaly in TTF-TCNQ is due to Wigner fluctuations which can
also account for the large spin susceptibilities by localizing all the elec-
trons,(75) then these fluctuations should be appreciable on both the TTF
and TCNQ stacks, since XF(T) and XQ(T) are so similar, and the disap-
pearance of this anomaly should be accompanied by a big drop in the spin
susceptibilities of both stacks, which is not observed. If, instead, it is the 2kf
anomaly that has disappeared, then the metal-semiconductor transitions in
TTF-TCNQ and TSeF-TCNQ would be fundamentally different, the tran-
sition in TTF-TCNQ involving a 2kf distortion, the transition in TSeF-
TCNQ, a 4kf distortion_ Also, one would have to explain why changing the
fulvalene stacks has such a big effect on the 2kf fluctuations, which are
believed to be mainly on the TCNQ stacks.
5. Metal-Semiconductor Phase Transition
A discussion of the metal-semiconductor transition in TTF-TCNQ,

TSeF-TCNQ, or the alloy system must deal with two sets of questions. First
of all, what is the transition temperature, how does it vary with changing
alloy concentration, and how is this change related to the systematic
changes taking place within the alloy system? Second, what is the nature of
the phase transition itself? If it is second order, what can be learned about
the critical behavior, and what does this tell us about the correlation
lengths and the dimensionality of the transition?
Although these two sets of questions are not completely unrelated, the
universality property of critical behavior(117) allows the effects of different
transition temperatures to be easily parametrized, and makes these two
series of questions somewhat orthogonal. We shall discuss the variation in
transition temperature across the alloy series first. Then we shall review
what has been learned about the nature of the phase transitions in TTF-
TCNQ and TSeF-TCNQ from studies of the specific heat, thermal
conductivity, resistivity, and magnetic susceptibility near Tel. Although the
general emphasis will be on information that can be gained from a
comparison of the transitions in TTF-TCNQ and TSeF-TCNQ and from a
study of the alloy series, the experimental information is not always
complete. Where the information is incomplete, we discuss results on the
metal-semiconductor transition in TTF-TCNQ if these results add
significantly to our knowledge about the physical properties of these
anisotropic systems.
The study of the phase transition region has proved to be fruitful in
general because of the detailed understanding of phase transitions that has
been acquired over the last several years. We are particularly interested in
what these studies of the transition region can reveal about (a) the relative
strength of three-dimensional interactions between the stacks when they
are compared to the one-dimensional interactions along the stacks, (b) the
role of disorder in determining the observed properties of these materials,
and (c) the properties of these materials near Tel.
5.1. Variation of Transition Temperature Tel with Alloying
The value of Tel has been determined by several methods in the pure
materials, TTF-TCNQ and TSeF-TCNQ, but primarily by dc-conductivity
measurements in the alloy series. In TTF-TCNQ and TSeF-TCNQ,
measurements of the specific heat(1l8.119) indicate that the phase transition
coincides with the peak in the logarithmic derivative of the resistivity,
demonstrating the fundamental connection between the basic order
parameter of the system and a changing gap in the electronic density of
states. In discussing the variation of Tel for the alloys, then, we use the dc
conductivity reported by Etemad et aIYS) We also rely heavily upon their
conclusions about the driving forces behind this change in Tel.
The phase diagram illustrated in Figure 4 gives a pictorial represen-
tation of the dependence of Tel on alloy concentration. Qualitatively, the
phase boundary between the metallic and semiconducting states moves
toward lower temperatures with increasing concentration of TSeF in the
fulvalene stack. This decrease is quite gradual until the concentration
The Organic Metals (TSeF)x(ITF)l-x-TCNQ 185
x = 0.90-0.95 when the transition temperature begins to drop pre-

cipitously. At the TTF-TCNQ end, 8Te d8x = -90oK, whereas at the
TSeF-TCNQ end, 8Te d8x = -400oK. In addition to the pictorial
representation in Figure 4, the transition temperature, width of the tran-
sition, and ratio of these quantities are given in Table 3 in order to allow a
more quantitative analysis of this transition.
Although the shape of this phase boundary is complex, Etemad et
al.(IS) noticed an important correlation between the dependences of Tel
and of the lattice parameter b on concentration, which we discussed in
Section 2. They found the linear relation
Tel (0)- Tel (x) = A + B[b(O)- b(x)] (8)
for the whole range O:s: x :s: 0.68. For x> 0.95, they found another linear
relation with a much greater slope. This functional relationship between
Te 1 and b is illustrated in Figure 17.
In addition to the variation of Tel with x, there is also a variation in the
width of the transition. This width, which we define as the full-width at half
the maximum height of the logarithmic derivative of the resistivity, is
greater at the TSeF end of the alloy phase diagram than at the TTF end.
Moreover, this width increases at a faster rate with departures of x from
x = 1 than from x = O. A plot of the width of the transition normalized to
the transition temperature Te I vs. (2x - 1)2 dramatizes the changes taking
place at the ends of the alloy series and is shown in Figure 18.
Analysis of the phase diagram in Figure 4 and of the information
presented in Figures 17 and 18 indicates that there are at least two distinct
regions of this phase diagram and that it is unlikely that the same driving
force is responsible for the changes in Tel at the two ends. One obvious
difference is that small doping at the TTF end lowers Te I while small doping
at the TSeF end dramatically raises it. While we might expect the lowering
of Tel at the TTF end to arise from the disorder introduced by doping, it is
hard to understand a rise in Tc I such as occurs at the TSeF end in terms of
disorder. Noting the similarity in behavior of Tel vs. x and -b vs. x near
x = 0, Etemad et al.,(l'i) have proposed a model for the concentration
dependence of Te I for O:s: x :S 0.9 that is based solely on changes in average
quantities that are related to the changing lattice parameters, e.g., on
changes in electronic transfer integrals and charge transfer. For x> 0.9,
Etemad et al., referring to mechanisms proposed by Schultz(120) and by
Tomkiewicz et al.,(83,87) claim that it is the disorder introduced by the
addition of small amounts of TTF that is playing the dominant role in
changing Tel, although not in the usual way.
Let us consider first the range O:s: x :S 0.9 in more detail. It would
appear that changes in all the transfer integrals. Coulomb interactions, etc.,
......
g;
Table 3. TransitIOn Temperatures, Breadth of Transition, Other DC-Conductivity Results for (TSeF)JITF)I .,-TCNO and for
DSeDTF- TCNO"
b din R!
Tel Tc2 WeI (-We!) xlO 2 T. <TRT Number of
Compound (OK) (OK) (OK) CK) (11 1 cm- l ) samples
Tcl d(1/T)I T "
TTF-TCNQ 52.8 38 0.8-1.5 1.5-2.8 1100-1600 58-60 500 16

x = 0.001 52.8 37.5 U 2.8 700-900 60-62 2
x =0,01 5L5 34 3 5.S SOO-IOOO 62-63 280 4
x = 0.02 51 33 5-6 10-12 900-1100 60-65 4
x = 0.03 50.5 7-8 14-16 800-1000 62-63 340 4
x = 0.18 47 20 42.5 500 65-68 270 3
x = 0.42 42 26-28 62-67 350 62-67 260 2
x = 0.68 41 ±1 20 48-50 300 60-65 350 3 ~
x = 0.95 39 12-14 31-36 350 50-55 410 3
x = 0.98 36 8-10 22-28 400 45-50 640 3 ~
TSeF-TCNQ -29 1.5-3.5 5-12 600-900 37-41 800 17 ~
;:-
DSeDTF-TCNQ -45 large 60-64 550 8 l::
~
s::.
;:
a From Reference 15. b These results taken m warming.
s::..
~
~
Q
s::.
to
;:
'"
The Organic Metals (TSeF)x(TTF)l~x-TCNQ 187
/-U
(;() -15 Ii
i
-20 '-
-25 iL -__ ~ ____- L_ _ _ _~_ _ _ _~_ _ _ _- L_ _ _ _~
o 002 0.04 0.06
Figure 17. Shift in metal-semiconductor transItIOn temperature Tel vs. shift in room-
temperature lattice parameter b as x is changed, in (TSeF)x(TTF)I~x-TCNQ. (After
Reference 15.)
05r---------~-~-~-~-=---
02 X<05
o-U
01
~u
005
TTF-TCNQ~
002L-------~------~--------L--------L------~
00 02 04 06 0.8 1.0
(2X-1 )2
Figure 18. Fractional width WcdTcl of the metal-semiconductor transition vs. (2x-l)2 for
(TSeF)x(TTFh x-TCNQ. (After Reference 15.)
with x might have a complex and unpredictable effect on Tel. There is,
however, considerable evidence from studies of the transport(16.121) and
magnetic properties(74.121) below Tel, some of which we have already
anticipated, that it is the TCNQ stacks that drive the phase transition
(Sections 6.1 and 6.2) in TTF-TCNQ and, at least, in the alloys with a small
concentration of TSeF-TCNQ. This evidence caused Etemad et al. to focus
attention on the changes in to (x ) caused by the changes in b (x), which they
assume are proportional for the small changes involved. Also, assuming
weakly interacting electrons, the charge transfer t' per molecule is related
to the b component of the Peierls vector by v = Obb/7T so that charge
transfers of v(x = 0) = 0.590 and v(x = 1) = 0.635 are obtained from the
observed values of Ob. For simplicity, Etemad et al. interpolate between
these limits by assuming v(x)cx: b(x). They found that the best fit is
obtained using A = 0 and E = - 260 K/ A in Equation (8), which is in
0
surprisingly good agreement with what is obtained using these depen-

dences of to and v on b, on the mean-field Peierls transition temperature
T ml and on simple assumptions about the relation of Tc 1 to T mI. Although
the agreement might be fortuitous, it gives credibility to the importance of
the TCNQ stacks in driving the transition and to the unimportance of the
effect of disorder on Tel for 0 < x :s 0.9.
Now let us consider the precipitous drop in Tel with a small increase in
x at the TSeF-TCNQ end of the phase diagram. Basically there have been
two mechanisms proposed to explain this phenomenon. Both of these rely
on increases in transfer integrals at the TSeF end of the phase diagram, the
first on the increase in tOF and the second on the increase in tF. The first
mechanism, proposed by Tomkiewicz et al.,(87) is based on the idea that an
increasing tOF should increase any tendency to open a gap at the Fermi
energy where the fulvalene and TCNQ bands cross. The competition
between this "hybridization gap" and the Peierls instability should lower
the transition temperature. It has also been argued that impurities such as
occur if x is decreased slightly from x = 1 should sharply reduce the
tendency to form this gap, thereby reducing this competition and raising
Tel.
A second mechanism to explain the x dependence of Tel near x = 1
was proposed by SchultzY20) This idea derives from the increased elec-
tron-phonon coupling in the TSeF stack when compared to the TTF stack.
When coupled with the decreased coupling in the TCNQ stacks, this makes
the susceptibility to Peierls fluctuations more nearly the same on the two
kinds of stacks. In this situation the effective screening by TSeF fluctua-
tions of interactions among TCNQ fluctuations is important and the tran-
sition temperature is thereby lowered. However, a small concentration of
TTF impurities, by pinning the phases of the TSeF fluctuations, can reduce
this screening and thereby raise Te I sharply.
The Organic Metals (TSeF)x (TTF)I- x - TCNQ 189
Both of these mechanisms may playa role in determining the value of

Tc 1(x). Further investigations, in particular investigations with neutron or
x-ray scattering of the changing order parameters with small changes in x
near x = 1, will be necessary in order to determine a better model for this
transition.
5.2. Thermodynamics and Critical Behavior of the Metal-Semiconductor

Phase Transition
5.2.1. Specific Heat
Although there are many indications at this time that the quasi-one-
dimensional crystals of TTF-TCNQ and TSeF-TCNQ undergo a Peierls
transition when the conductivity changes from the metallic state to the
semiconducting state, the most direct thermodynamic information about
the nature of this phase transition comes from measurements of the specific
heat. The first specific-heat measurements showing an anomaly associated
with the transition in TTF-TCNQ were reported by Craven et a1Y18) The
data, shown in Figure 19, exhibit a rather sharp anomaly at Tel, although
the magnitude of this anomaly is small (~2.5%) compared to the large
lattice specific-heat background. There are three important features of
these data. (1) The region in which there is significant excess specific heat
associated with fluctuations in the order parameter is only ten to fifteen
030
cr: 025
'-
tl.
u
0.20
I-
«
w
I o 15
u
LL 0.10
u
w
tl. 0.05
(f)
00
45 50 55 60 65
TEMPE R ATURE (OK)
Figure 19. SpecIfic heat vs. T for TTF-TcNQ. (After Reference 118.)
degrees wide, I.e., it is limited to a range of reduced temperatures t ==

(T - T el )/ Tel of about 0.1. (2) The change in enthalpy at the phase tran-
sition is consistent with what one expects from a Peierls transition in a
system with nearly one-dimensional tight-binding bands. It was considered
unlikely that there is any large, first-order lattice distortion at Tel, and this
has since been confirmed by neutron and x-ray scattering. (3) Most of the
entropy is lost in the transition (for TTF-TCNQ, 95% of the entropy
change expected from a Peierls transition is accounted for by T =
45°K)Y18)
These features indicate that although the phase transition may be
driven by the quasi-one-dimensional nature of the crystalline lattice and
the resultant highly anisotropic nature of the band structure, the phase
transition at Te I is itself a three-dimensional transition, i.e., the correlation
lengths both perpendicular and parallel to the conducting stack must be
diverging. The fact that most of the entropy expected from the opening of a
gap at the Fermi surface can be accounted for near Tel is consistent with
the significant change in resistivity observed at this transition and indicates
that the lower transitions must involve only minor modifications in the
order parameter that characterizes the gap. Although anomalies in the
specific heat associated with these lower transitions were not seen in this
early experiment, a recent paper by Djurek et al.(119) has claimed to see
four anomalies in the specific heat, all associated with phase transitions.
There is certainly some change in the thermodynamic variables at Tel, Te2 ,
T e3 , and T e4 , although the thermal relaxation technique used in these
experiments does not rule out contributions to the signal from anomalous
behavior in the thermal conductivity.
Specific heat measurements have also been made on TSeF-TCNQ by
Craven.(122) These measurements indicate a change in the specific heat
near Tel similar to that seen in TTF-TCNQ. The anomaly is again about
two percent of the lattice background and is slightly less pronounced than
the anomaly at Tel in TTF-TCNQ. The transition in TSeF-TCNQ is again
three dimensional for T near Te I.
It is unfortunate, but the small size of the anomaly in both TTF-TCNQ
and TSeF-TCNQ prevents any quantitative determination of the critical
exponent a of the specific heat [Cp(t) - t- a as t ~ 0+]. If a value for a were
available, it would allow a quantitative comparison between the specific
heat, critical resistivity, and susceptibility discussed later in this section.
5.2.2. Thermal Conductivity
By simultaneously measuring the thermal diffusivity along the b axis

of a single crystal of TTF-TCNQ and the specific heat of the same sample,
Salamon et al.(I21) were able to measure the thermal conductivity K o(this
material in the vicinity of the metal-semiconductor phase transition.

Although this measurement has been performed only on single crystals of
TTF-TCNQ, it is very interesting because it directly examines the ther-
modynamics of the electron transport in the metallic phase above Tel. This
work compares the size of the step like anomaly in the thermal conductivity
which occurs at Tc 1 with the change in K expected from the Wiedemann-
Franz law K/Tu = 7T 2 k1/3e 2 =- Lo, and calculates what contribution to the
thermal conductivity should be made by collective, conductivity-enhanc-
ing, fluctuations, K' / Tu' = 10- 4 L o. The fact that the measured thermal
conductivity is not severely reduced from what the simple Wiedemann-
Franz law predicts, i.e., K/Tu = L o implies that in samples with a conduc-
tivity ratio u(peak)/ u(3000K) = 10-30, essentially all the current is carried
by normal, heat-carrying electrons rather than "superconducting-like"
fluctuations.
5.2.3. Critical Resistivity
Although a measurement of the dc resistivity of a material in the

vicinity of a phase transition is not, in itself, a direct measurement of a
thermodynamic quantity, the experimental ease and high precision with
which this measurement can be made has stimulated a serious effort to
develop a fundamental understanding of the temperature dependence of
the resistivity near phase transitionsY24) In the past, most of the
measurements and analysis of the critical behavior of the resistivity have
been performed on ferromagnetic(125) and antiferromagnetic(126) metals.
The theories, particularly as applied to the antiferromagnetic case,(127) can
be extended to phase transitions involving charge-density-wave fluctua-
tions. In this case, it turns out that although the resistivity itself is not
strongly divergent at the phase transition, its derivative is divergent and is
related to singularities in the temperature dependence of the dynamical
structure factor.
Detailed measurements of the temperature dependence of the resis-
tivity along the b axis of TTF-TCNQ were made by Horn and Rimai. (128)
These measurements were followed by measurements along the a axis in
TTF-TCNQ and along the b axis in TSeF-TCNQ by Horn and Gui-
dottiY29) In Figure 20 we show the behavior of the b-axis resistivity and its
temperature derivative in TSeF-TCNQ for T near Tel. This curve
represents one of the sharpest transitions observed by these workers.
The intent of these measurements was to analyze what effects struc-
tural fluctuations have on the resistivity in the vicinity of the three-dimen-
sional ordering temperature Tel. This analysis was based on a comparison
of the experimental data with the predictions of a model that is based on
noninteracting electrons undergoing scattering from Peierls fluctuations
" 6.0
II
b-axis "
II
00
16 II 4.0
00
TSeF -TeNQ II
,!i
,
,/ 2.0
1.2
~
1.0 '"
:>
0.8 -e
.2
!; I!;; 08 0.6 0:
-/0:
,
I
4--. 0.4
I
/
04 ~
0.2
i
I .-
.-
r"--- \
' ..... ""--
0 2 ' - 4 - - - - - 2 ' - 6 - - - - - 2 ' - 8 - - - - -3.L.O----' 0.1
TEMPERATURE (OK)
Figure 20. In R and ~d In R/ dT vs. Tin TSeF-TCNQ. R is dc resistance along b axis. (After
Reference 129.)
combined with the reduction in the number of carriers arising from the
pseudogap produced by these fluctuations above Tel and from the real gap
produced by a three-dimensional Peierls distortion below Tel. The authors
specifically considered only amplitude, not phase, fluctuations above Tel
and neglected any possible collective contributions to the conductivity,
since they felt that these contributions, if present, would not be dominant
below the peak in the conductivity curves, i.e., in the region of the phase
transition. The authors also hoped to draw conclusions about the nature
and extent of one-dimensional correlations in the metallic region above
Tel, based on their data in the transition region, and to make a comparison
between the critical behavior of TTF-TCNQ and TSeF-TCNQ which
could be related to differences in the interstack interactions discussed in
Section 4.
In order to determine the success of this venture, we have to examine
what the model calculations predict. Following the lead established by the
The Organic Metals (TSeF).(TTF),_x-TCNQ 193
antiferromagnetic theory of Suezaki and Mori,(127) Horn and Guidotti

treated the contribution Pes to the resistivity from critical scattering of
electrons by lattice fluctuations separately from the effects on the number
of carriers due to gap fluctuations. Because of the specific-heat measure-
ments, they assumed that the fluctuations near Tc I are three-dimensional
but that the scattering contribution to the resistivity depends upon D, the
dimensionality of final electronic states (not to be confused with the
dimensionality d of the critical fluctuations, which are three-dimensional).
From this model they obtained
dpcs -(1+,,+'1") ~t -15

-~t D=l (9a)
dT '
dpcs (1+'1") ~t -10 ,
-~t D=2 (9b)
dT
dpcs -(1+'1"-") ~t -0 'i
-~t D=3 (9c)
dT '
where t is again the reduced temperature (T- Tc1 )/Tc1 . The numerical
exponents represent mean-field values for the critical exponents v and 11,
defined, respectively, by qq, T)~ (-" and r(r, Tcl)~ Irld-TJ:
v=!, 17=0 (10)
where f(r) and f(q) are the order-parameter-order-parameter correlation
function and its Fourier transform.
The modification of this critical-scattering resistivity by the decrease in
the number of carriers caused by fluctuations in the gap parameter Il and
the development of a finite average for Ll was also estimated by Horn and
Guidotti(129) under certain assumptions. They assumed that when the
longitudinal correlation length I is much larger than the mean free path A,
the conduction electrons scatter many times in a region of relatively
constant gap parameter so that the density of electronic states and hence
the distribution of thermally excited electrons in a given region is deter-
mined by the instantaneous gap parameter in that region. The local gap
parameter, in turn, was assumed proportional to the local distortion am-
plitude, the effect of fluctuations in the distortion's phase being neglected.
They further assumed that the large anisotropy in the conductivity prevents
the electrons from skirtIng any locally more resistive regions, so that the
resistivities of all regions along a stack can be added in series.
Horn and Guidotti considered the behavior of the resistivity as a
function of temperature in two limiting cases: (1) Tc 1 « T m{, in which range
one expects there to be significant one-dimensional fluctuations. This limit
would be appropriate for an extremely anisotropic system where the one-
dimensional correlations along the conducting direction are very well
established before the three-dimensional transition occurs. (2) Tel:S T mf, in

which limit the one-dimensional fluctuations may occur for T> Tel but the
growth of their correlation length is strongly connected to the onset of a
three-dimensional transition. Since for both TTF-TCNQ and TSeF-TCNQ
the observed resistivity shows a pronounced anomaly in dp/ dT at Tel and
no activated behavior above Tel (barring the mechanisms of ConweU(34»),
Horn and Guidotti concluded from their model calculations that the
second limiting case is the more appropriate for both materials. This is not
to say that there are no one-dimensional fluctuations above Tel, only that
the rapid growth in their correlation length and amplitude occurs only near
Tel when the three-dimensional interactions begin to take over.:j: This is
consistent with both the specific-heat measurements and the neutron(13I)
and x_ray(100-102.116) scattering measurements. As we have seen, the latter
see evidence of one-dimensional fluctuations above Tel, but not strongly
for temperatures much greater than Tel. In TTF-TCNQ for example
T mrl TeI:S 3. (100) Because there is only a weak singularity in the behavior of
the gap, the temperature derivative of the resistivity should be dominated
for T> Tc I by the temperature dependence of the fluctuation-scattering
term according to Equations (9), even for mean-field exponents, so long as
d 2 3. Below Tc I. the resistivity should be dominated by the reduction in
the number of electrons available for transport resulting from the devel-
oping gap.
With the above discussion in mind, it is instructive to look at the
detailed experimental results. For T> Tel, Horn and Guidotti(129) found
that R- I dR/dT - t- I 09±0 15 and _t- 149 ±O.13 for the b axis of TTF-TCNQ
and TSeF-TCNQ, respectively, and that R- I dR/dT_t-1.04±o.17 and
T- o 55±O 2 for the a axis of TTF -TCNQ and TSeF -TCNQ. The agreement
of the b-axis values with the values predicted from Equations (9) is better
for TSeF-TCNQ. In this case we have assumed three-dimensional fluctua-
tions and mean-field exponents, with a one-dimensional (D = 1) density of
final states for the resistive scattering. For TTF-TCNQ, the results would
indicate that D = 2, which is surprising.
Since these results represent a series of measurements on a variety of
samples, one must conclude that the assumptions that were made in the
calculation are not strictly applicable to TTF-TCNQ. The most probable
culprits are the initial assumption that one can isolate the scattering by
critical fluctuations from the effects that these fluctuations have on the
electronic density of states and the assumption that we are in an asymptotic
critical region with well-defined and calculable exponents. If the phase
transition(s) were slightly first order,(132) the critical exponents that are
:j: For a general discussion of phase transitions in quasi-one-dimensional systems, see for
example, W. Dieterich(!30)
The Organic Metals (TSeF).(1TF)l_x-TCNQ 195
determined would have little bearing on a model based on asymptotic

critical behavior. Since the a-axis conduction is expected to be diffusive, its
exponent is expected to be determined by the fluctuations in ~ and their
effect on the number of electrons. For one-dimensional electronic states
and mean-field exponents, the exponent should be -0.5, so that again the
agrement is good for TSeF-TCNQ but bad for TTF-TCNQ. What is even
more surprising is that in TTF-TCNQ tee b- and a-axis conductivities,
which should result from quite different mechanisms, have similar
behavior.
For T < Tel, where the effects of gap fluctuations or an average gap
should dominate, Horn and Guidotti conclude from their model that
R- 1 dR/dT should go like (--tl- 1 for (1~1)~kBT and like (-tra~ for
t ~ 0-. The experimental results of Horn and Guidotti give a value of
{3 = 0.48 in TTF-TCNQ and {3 = 0.26 in TSeF-TCNQ. The value for TTF-
TCNQ is in agreement with what would be expected from mean-field
theory, but the value of {3 in TSeF-TCNQ is too small. This could be
caused by fitting for (3 too close to Tel in TSeF-TCNQ. Again, the results
are mixed and may indicate a breakdown in the initial assumptions.
In summary, the goals that were set for these detailed measurements
of the resistivity were met with only partial success. The qualitative
behavior is exactly what one would expect based on the model, but the
detailed values for the critical exponents do not form complete, self-
consistent sets. The conclusions about the dimensionality and the extent of
one-dimensional correlations are interesting and consistent with other
measurements. It is reasonably certain that three-dimensional interactions
and fluctuations dominate the thermodynamics of the phase transition
itself, although one-dimensional fluctuations are present above Tel.
5.2.4. Magnetic Susceptibility
If one again assumes that the conduction electrons and holes are only
weakly interacting, then at low temperatures the spin susceptibility will be
determined by the reduced density of states within kT of the Fermi energy.
If, as before, the density of states is determined by the presence of a
fluctuating or average gap parameter, then above the transition the spin
susceptibility is expected to decrease with increasing (1~12) and at lower
temperatures X'pln is expected to vary as exp(-(~)kT). Since the low-q
Fourier components of (I.i(q )12) mirror the ordering energy, the tempera-
ture derivative of (1~12> and therefore of Xspm is expected to be proportional
to the specific heat, i.e., ax/aToca(I~(q)12)/aT-t-a. Horn et a1Y33) have
measured the spin susceptibility of TTF-TCNQ in the vicinity of Tel. The
measured susceptibility and its temperature derivative are shown in Figure
21. Their analysis indicates that aX/aT=t- 05 . This is consistent with a
/;
TTF-TeNQ / I
-
,/ '1
/ ,
/
I
;<:
"-
/
E0 / I E "
'"
q-
/
/
'"
q-
Q '0
z I-
a: '0
"-
x'" X
'0
-IL-~ __- L__ ~ __~__L-~_ _~_ _- L__~__~O

46 48 50 52 54 56
TEMPERATURE (OK)
Figure 21. Spin susceptibility and its temperature derivative for TTF-TCNQ in the neigh-
borhood of the metal-semiconductor phase transition. (After Reference 133.)
classical Ornstein-Zernike approximation for the Q,fder-parameter-order-

parameter correlation function, from which one would find a = 2 - (d/2) =
1/2 for d = 3. The conclusion that the effective dimensionality of the
system is three, at least within a 200K width of Tc 1, is consistent with all of the
preceding discussion in this section. High-resolution measurements of Xspin
have not been made in TSeF-TCNQ.
6. Semiconducting Phase, T < Tc 1
The low-temperature phase (or phases), which we have called the

semiconducting phase, has many interesting properties of its own, includ-
ing a multiplicity of phase transitions and elaborate structural properties
which have been treated somewhat phenomenologically. These structural
properties are intimately related to the electrical and magnetic properties
in this phase, which in turn must follow from the same model as would be
used to describe the metallic phase. Thus the connections between the
metallic and semiconducting phases are closer than one might assume. Our
discussion of the semiconducting phase will parallel the discussion of the
metallic phase, treating the transport and magnetic properties first,
although the structural properties are perhaps the more fundamental. With
this in mind, Section 6.3 might be read before Sections 6.1 and 6.2.
6.1. Transport Properties in the Semiconducting Phase

As we have anticipated in our discussion of other temperature ranges,
the study of transport phenomena in the semiconducting region of the
phase diagram provided an early clue to the rich phase structure that is
found in TTF-TCNQ, TSeF-TCNQ, and the alloy series. The majority of
the work has been done with TTF-TCNQ and TSeF-TCNQ, but tracing
the changes in transport in these two materials when either is lightly doped
with the other (i.e., x:S 0 and x:s 1) has led to the discovery of additional
complexity in the phase diagram for x :S 1 and to a better understanding of
which of the stacks drives the structural transitions. In addition to this
study of the phase diagram, considerable work has been done on the
low-temperature properties of the electrical transport. This has focused on
attempts to describe accurately the effective activation energy of the
conductivity and on the anomalous nonohmic conductivity at very low
temperatures in these materials.
6.1.1. Phase Transitions in the Semiconducting Phase for x :S 0

The early measurements of the dc conductivity of TTF-TCNQ were
fraught with controversy and with the difficulties of making measurements
on highly anisotropic materials. At one point, the occurrence of an
anomaly in the temperature dependence of the b-axis conductivity and of
large changes in the conductivity anisotropy (J'b/ (J'a at ~ 38°K were inter-
preted as an indication that poorly aligned contacts were present.(6) After
this, many workers assumed that any anomalous behavior of the conduc-
tivity was due to poor alignment. Etemad(16) clarified this situation by
emphasizing that anomalous behavior could only arise in the anisotropy if
there were an anomaly in the conductivity itself, not vice versa. He went on
to examine other experiments in which anomalous behavior had been seen
at ~38°K, e.g., thermal conductivity,(123) pressure dependence of the
conductivity, (61,62) thermoelectric power, (46) thermoreflectance spectra, (134)
and nuclear-spin-Iattice relaxation.(69) This total analysis firmly established
that the anomaly in (J'b at ~ 38°K was in fact the result of a phase transition
intrinsic to TTF-TCNQ. Later, he and other workers investigated the
effects that doping TTF-TCNQ with TSeF-TCNQ had on this anomaly. As
is seen in Figure 4, the temperature at which an anomaly is seen decreases
with increasing concentration of dopant. With 3% doping, the evidence for
a phase transition in the dc conductivity has been completely washed out,
although x-ray and magnetic susceptibility evidence presented in Sections
6.3 and 6.2 indicate that the transition is still present at this concentration
and that it takes place at T = 29°K. The b-axis resistance of a
(TSeF)o 03(TTF)o 9T TC'NQ crystal is compared with that of a pure TTF-
TCNQ crystal in Figure 22, where the resistances are normalized to agree
7.-'--.--~-r-'--~~--r-~
- I
-2
-3~ __ ~ ____- L_ _ _ _J -_ _ _ _~--J Figure 22. Effect of TSeF-doping in
o 10 20 30 40 ITF-TCNQ on In R vs. T. R is dc resis-
tance along b axis. (After Reference 16.)
at room temperature. This comparison shows that between 38 and 54°K

the conductivity in the pure material is higher than the conductivity in the
doped material. This "excess" conductivity of the pure system is inter-
preted as being contributed by the fulvalene stack, since it is the fulvalene
stack that is disordered by dopingY 6 ) Because the "excess" conductivity is
believed to be on the fulvalene stacks, its drop at 38°K in the pure system is
taken as evidence that these stacks are involved in the 38°K transition, if
not evidence that they are driving the transition.
These inferences, when based on only the dc conductivity, were bold,
but they have been borne out by additional transport experiments, which
we now discuss, and by other experiments, which we discuss in Sections 6.2
and 6.3.
Measurements of the thermoelectric power S in the pure and doped
material gave two remarkable results as shown in Figure 9. (1) In the pure
material, S is positive and large in roughly the range 25-50 o K, in sharp
contrast to its behavior at all other temperatures measured. (2) In the
3%-doped material, S is made small and negative in this temperature
range, but it is only slightly affected elsewhere. The interpretation of these
results requires caution because, in contrast to the metallic state, the
thermoelectric power of a semiconducting state can be very sensitive to
doping, the position of the Fermi level, etc. If we take the simplest view
that SQ and SF are negative and positive, respectively, even below the
metal-semiconductor transition and even in the doped systems, then the
large positive S in TTF-TCNQ reflects a domination by the carriers on the
fulvalene stacks. This domination can result from a much higher conduc-
tivity and/or a much larger thermoelectric power on these stacks. Since
there is no reason to believe that SF is especially large, the fulvalene
domination was attributed to a higher (IF in this temperature range, and
therefore was considered as evidence for a drop in the TCNQ conductivity
in the metal-semiconductor transition. The elimination of the large, nega-

tive S with 3% fulvalene doping was then attributed to a marked reduction
in the fulvalene conductivity, in agreement with the conclusions drawn
from dc-conductivity measurements, although it could also reflect an
impurity-induced change in the density of states near E f in the fulvalene
band.
An additional set of experiments involving transport measurements as
well as an analysis of the magnetic properties was reported by Tomkiewicz
et atY21) Here, methylTCNQ was introduced as a dopant into the TCNQ
stacks of TTF-TCNQ rather than TSeF into the TTF stacks. The doped
material showed drastic changes in the width of the transition at Te 1 and
some change in the transition at Te2 . It was inferred from the conductivity
results of these experiments and the magnetic results discussed in Section
6.2 that the transition at Tc 1 is indeed driven by the TCNQ stacks and that
the transition at Te2 may involve both kinds of stacks.
Following this investigation of the effects of doping on Te2 , detailed
investigation of the TTF-TCNQ by Ishiguro et at. ,(135) and Carruthers and
Bloch(136) discovered thermal hysteresis loops in the conductivity at Te2' a
clear indication that the phase change at Te2 is first-order. This conclusion
is confirmed by the structural investigations. (See Section 6.3.)
After the discovery of this second phase transition in TTF-TCNQ,
extensive investigation of the structural properties led to the prediction by
Bak and Emery(117) of a third transition which is seen in structural experi-
ments at - 49°K. Careful analysis of a great deal of experimental conduc-
tivity data on TTF-TCNQ by Craven et at.(138) confirmed the existence of
this transition in the pure material and very slightly doped material. It
could not be seen for x> 0.01. These workers also pointed out that the
shape of the conductivity curve would be inconsistent with this tempera-
ture being the temperature at which the fulvalene stacks first start to show
long-range order.
6.1.2. Phase Transitions in the Semiconducting Region, x:S 1

A study of the TSeF -TCN Q end of the alloy series has also yielded a
much richer picture of the phase diagram. Craven et at., (17) through a
careful analysis of the derivative of the resistivity near the transition, have
shown that the simple phase transition that has been observed in TSeF-
TCNQ at - 29°K splits into two transitions when TSeF-TCNQ is doped
with small amounts of TTF-TCNQ. As summarized in Figure 4, the higher
transition temperature (== Tc I) increases rapidly with small amounts of
TTF doping, while the lower transition temperature (== Td decreases only
slightly with similar amounts of dopant. The resistive anomaly associated
with the lower transition is. however. drastically broadened by the addition
of only two percent TTF-TCNQ.
I I
800
600
t
.,- f-X · '.O
t:
..,
::::. 400
a::
oS
..,
200
Figure 23. Effect of TIF-doping in

TSeF-TCNQ on dIn R/d(1/T) vs. T for
several TIF concentrations (== 1 - x). R is
20 30 40 50
T (OK) dc resistance along b axis. (After
Reference 17.)
Detailed investigation(17) of d In R/ d (1 / T) vs. T, as shown in Figure

23, indicates that the largest change in resistance, as measured by the
magnitude of this derivative, takes place at Te2 for low doping, but at Te 1
for larger doping. This transfer of importance or effectiveness of the phase
transition occurring on doping is very different from what is seen at the
TTF-TCNQ end. For x 2: 0, the transition at Tel is always the dominant
phase transition, those at Te2 and Tc3 being minor modifications on the new
phase (and at quite different temperatures). In TSeF-TCNQ this dis-
tribution of importance probably arises from the stronger coupling
between the TSeF stack and the TCNQ stack. There is some indication
from analysis of the g shift in material with x = 0.93, reported in this same
paperY7) that below 1',1 there is a larger order parameter on the TCNQ
stack than on the fulvalene stack. This is similar to the situation below
54°K in TTF-TCNQ, but the effects on the conductivity are certainly more
complex for the selenium case. To date, this additional phase structure at
the TSeF-TCNQ end of the phase diagram has only been seen in the
transport measurements.
The effects of fulvalene doping on the thermoelectric power are much
less dramatic than near the TTF-TCNQ end. One sees from Figure 9 that
in pure TSeF-TCNQ S starts to go rapidly negative at ~ 30 o K, while in the
doped systems it exhibits a similar behavior, although the change is less
rapid, commences at a somewhat higher temperature, and leads to a more
negative S. In pure TSeF-TCNQ, it thus appears that the TCNQ stacks
regain dominance of the thermoelectric power without the fulvalene stacks
The Organic Metals (TSeF)x(7TF)l_'x-TCNQ 201
ever dominating, although in view of the uncertainties about thermoelec-

tric power in a semiconducting stack, it could also be a change in the
character of SF itself. A failure of the fulvalene stacks to dominate could be
a consequence of the proximity of Te 1 and Te2 , and the insensitivity of S to
fulvalene doping could be an indication that the electronic motions are less
one-dimensional at the TSeF-TCNQ end, although these are only sugges-
tions.
Investigation of the lattice structure for x:s 1 near 29°K would
certainly enhance our understanding of these transitions and of the
detailed role that each kind of stack is playing in the conduction process.
6.1.3. Measurements of Activation, T < Tel
Measurement of the electrical conductivity below the metal-semicon-

ducting transition in TTF-TCNQ and TSeF-TCNQ has long held the
promise of establishing what effects the transition has on the band struc-
ture. A definitive value for the conductivity gap is difficult to determine.
We discuss the situation in TTF-TCNQ and divide our considerations into
three regions: Tee < T <: T, I, 11 °K:s T < Te2 , and T < 11 oK. To begin with,
we assume a simple model, that is, one in which a gap parameter d can be
defined by the activated temperature dependence of the resistivity, p =
p exp[d(T)/KBT]. In the first region, Tc2 < T< Te!' there are several
measurements of ~(T) that indicate a value A = 250°K.(l6.128.138) This value
is interestingly high when compared with values of the gap parameter in
this region derived from the magnetic susceptibility measurements to be
discussed in Section n.2. and it may indicate that contributions from a
tetnperature-dependent mobility and from fluctuation effects near Tel are
playing the dominant role.
For 11 OK < T < T 2 there have been several values reported for the
gap parameter ~. Part of the trouble stems from the variation in crystal
quality between laboratories, but the largest problem arises from different
methods that are used to analyze the data. Etemad first reported a value of
A = 225°Kyn) This number is derived by treating the energy gap within the
mean-field approximation which gives a BCS-like behavior, forcing d(T)
to be essentially constant for T:s 30 o K. Etemad displays data only to 25°K,
however, for his fit. If we remove the requirement of aT-independent gap
as T 4 OaK and, in"tead. let the gap approach zero above Te 1, then a gap
parameter equal to 2000K ± 25°K is not inconsistent with his data. More
extensive investigation~ of the low-temperature conductivity in TTF-
TCNQ have recently been reported by Cohen and Heeger.(139) Their data
indicate that a constant value for the gap parameter of 1800K can be
established in the region 11 OK < T < 25°K. These gap parameters were
both derived by assumll1g that there is no temperature-dependent mobility,
that is, that the prefactor Po is constant. If we remove this restriction and
assume that p(T)= AT-a exp(6./kBT), then because the temperature is
not small when compared with the gap, a becomes an important
parameter. Using this form for peT), Bloch et al. (140) have found that
6. = 1000K for al\ data between 25°K and 11 oK, although there is still
variation in a (a = 2-4.5) from group to group. The value 6. = 1000 K is also
what is found from susceptibility measurements (see Section 6.2).
None of these measurements preclude the possibility that 6. = 1000 K is
caused by an impurity band within the gap. Recently Eldridge(141) has
made far-infrared photoconductivity measurements at low temperatures in
TTF-TCNQ. He finds that there are two excitation energies for conduction
in the b direction, Eg 1 = lOooK and E g2 = 450oK. The lower energy looks
like a contribution from an impurity band, while the larger signal comes at
E g2 , presumably from the intrinsic gap.
Below T = 1 10 K all workers have found a much smaller gap
parameter. This phenomenon has been ascribed to the presence of soliton-
like excitations in the energy spectrum by Cohen and Heeger,(139) but other
workers(l40) remain skeptical, claiming such nonohmic behavior can come
from hot-electron effects in dirty semiconductors. It does seem, however,
that a common mechanism is responsible for these very-low-temperature
effects, since in addition to the dc conductivity and far-infrared photocon-
ductivity results, measurements of the microwave dielectric constant and
conductivity by Gunning et a1Y42) indicate similar behavior in TTF-TCNQ,
TSeF-TCNQ, DSeTF-TCNQ, and (TSeF)o o3(TTF)o 97-TCNQ.
The measurements of the gap parameter in TSeF-TCNQ under the
simplified assumptIOns of p = Po exp(6./ kBT) has led to remarkable
agreement in the literature. Analysis of the low-temperature data by
Etemad,(It,) Engler et al.,(141) and Horn and Guidotti,029) and the low-
temperature, low-pressure data of Cooper et ai.(62) all indicate that 6. =
115 ± 100K. This IS much smaller than the results for TTF-TCNQ. The
same arguments about impurity bands that we mentioned above can affect
these measurements too. however, An interesting experiment involving
impurity effects was reported by Engler et al.(143) When TSeF-TCNQ was
doped with a methylated TCNQ derivative, the measured low-temperature
gap parameter went from 115 to 40oK. This results in an increase in the
conductivity at T = 4°K by four orders of magnitude, with a doping level in
the host material of less than 10%, and underlines the problems that
impurity bands can cause when transport measurements are used to
measure the energy gap.
6.2. Magnetic Properties

The magnetic properties of the (TSeF)ATTF)I-x-TCNQ system in the
semiconducting phase have been very important for understanding the
The Organic Metals (TSeF)xC1TF)l-x- TCNQ 203
roles played by the two kinds of stacks. They have not posed as many
obviously unsolved problems as those in the metallic phase, although a
completely coherent picture is still lacking. We first review the magnetic
properties of TTF-TCNQ and then discuss successively the effect of 3%
doping by TSeF, the few experiments on the alloys in this phase, and the
contrasting magnetic properties of TSeF-TCNQ.
6.2.1. 1TF-TCNQ
The static susceptibility of TTF-TCNQ was first measured by Tom-
kiewicz et aIYO?) and by Scott et al.,(66) but it was the high-precision
measurements of Herman et at. (144) that first observed and focused on the
38°K transition in the susceptibility. These experiments suffered, however,
from an inability to separate the contributions from the two kinds of stacks.
The distinct roles of the TTF and TCNQ stacks were made unambiguously
evident in the detailed EPR studies of the g shift by Tomkiewicz et al.(74,77)
It was seen that the drop in Xspm,Q that begins at -65-70oK can be fitted
approximately by an exponential at lower temperatures with an activation
energy ~mag.Q = 420°K. By contrast, Xspm,F shows no anomaly at 53°K,
possibly begins to fall at some lower temperature (- 45°K), is falling
rapidly by 38°K, and can be fitted by an exponential with ~mag,F =
125°K(77 or = 140o K(l44) for most of the range below 38°K. From the drop
in Xspm.Q but not in Xspm,F above Tel, from the exponential drop in Xspm,Q
while Xspm,F is hardly decreasing just below Tel, and from the fact that
~mag.Q »~mag.F at low temperatures, it was concluded that the metal-
semiconductor transition is driven by the TCNQ stacks. Furthermore, the
absence of any anomaly in Xspm,F at Tel indicates that the transition is on
the TCNQ stacks alone, I.e., that the two subsystems are, on the average,
decoupled, near the transition, consistent with the approximate symmetry
of the superlattice just below T 1, as we discuss in Section 6.3.
The slow drop \0 X ,pm.F below - 45°K, if present, could be attributed
to an average distortion on the TTF stacks that is driven by the distorting
field on these stacks produced by the three-dimensional TCNQ distortion,
as discu<;sed in Section 6.3. This field begins to grow from zero at Tc3 =
49°K. The more rapid drop in Xspm,F below - Te2 can be attributed, first, to
the larger distorting field from the TCNQ distortion, whose symmetry and,
possibly, magnitude change discontinuously at Te2 and, second, to the
continuing increa~c expected in the susceptibility Xdlst,F of the TTF stacks
to a 2k r distortion. The I1C Knight-shift experiments(69) confirm the rapid
disappearance of Xspm.Q below Tc 1. They also show a break in Xspm,F(T) at
Te3 , but the method does not have the sensitivity of the EPR method in this
range and has been called into question on other grounds as well.(8o,133)
We notice that the activation energy for magnetic excitations on the
TTF stacks is approximately equal to that for dc conduction, - lOooK, as
reported in Section 6.1. This is consistent with the idea that the magnetic
excitations are also the conducting excitations, supporting the model of
weakly interacting electrons, at least on the TTF stacks. In models with
strongly repelling electrons, the magnetic excitations (like antiferromag-
netic magnons) and the current-carrying charge excitations are entirely
dissimilar in character and are expected to have quite different activation
energies. In fact, it was for this reason that the much higher values of Ll cond
inferred by Etemad(16) and by Horn and Rimai(l28) presented a major
challenge to the weakly-interacting-electron models, (77) thereby giving
some support to the picture of strongly repelling electrons.
6.2.2. (TSeF)o 1)1(TTF)o 9T TCNQ
The effect of 3% TSeF doping in TTF-TCNQ on the magnetic

properties was studied in detail by Tomkiewicz et aZY21) using EPR. As
shown in Figure 24, such doping has only a slight effect on Xspm.Q, and
therefore on the behavior near Te 1. By contrast, it has a dramatic effect on
Xspm,F, lowering the activation energy below Te2 and apparently broaden-
ing the Te2 transition completely. Actually, the transition is still seen at
~ 29°K in recent, higher-precision EPR(78) and static susceptibility(89)
measurements. The effect on Xspm,Q near Te2 is unobservable because
Xspm.Q is already very small. This is consistent with the general effect
expected of impurities in a system (in this case the TTF subsystem) with a
Peierls distortion that is induced by a periodic field (in this case arising from
the Peierls distortion of the TCNQ subsystem): a reduction of Xd,st,F, a
consequent decrease in the average distortion amplitude on the fulvalene
,- r T - '-1 ~
4
u;
1- / X spm • F /"
Z / /'
/ /
::::l 3
rn I /
a:: I /
<l / I
c: 2 / I
Q. / I
><'"
Figure 24. Effects of fulvalene doping on
I spin susceptibilities of separate stacks,
/ Spin susceptibility vs, T of fulvalene (F)
/
and TCNQ (Q) stacks in TTF-TCNQ
°2LO~-3-"--0·~~ -'--------'-50--'-----6"--0-----' ( - - ) and in (TSeF)o o3(TTF)o 9T
TEMPERATURE (OK) TCNQ (----). (After Reference 121.)
The Organic Metals (TSeF)x(TTF)l_x- TCNQ 205
stacks and an increase in the distortional fluctuations, and finally a narrow-

ing of the fulvalene Peierls gap due to both these effects. A quantitative
analysis has not been made.
6.2.3. The Alloy Series

A particular use of the low-temperature EPR studies has been to
determine the average fulvalene g value, gF, as a function of x. At
sufficiently low temperatures, the magnetic excitations on the TCNQ stacks
are frozen out. When this occurs, the g value is that of the fulvalene stacks
and further decreases in temperature should result in a negligible change in
g value. This temperature-independent gF was reported for various alloys
in the range O:s X :s 0.93 by Tomkiewicz et al.(87) They found that gF does
not interpolate linearly from x = 0 to x = 1, but rather grows more slowly
for smaller x and more rapidly for larger x. The behavior for small x has
been attributed to an effective reduction in the amplitude of in-band wave
functions (including those near the Fermi energy) on the TSeF sites.(87) For
large x it is argued that the situation is complicated by the interstack
banding. It is nevertheless the quantity gF that is relevant to EPR studies
and which was therefore used in normalizing the Iinewidth in Section 4.2.2.
6.2.4. TSeF-TCNQ
In the semiconducting phase, the EPR line is too broad to be seen at
the TSeF-TCNQ end of the alloy series except at such low temperatures
that the observed spectrum is not a single Lorentzian line. This is
unfortunate because a decomposition of the susceptibility in the region of
the transition would be very helpful in distinguishing the nearly compar-
able, possibly competing, roles played by the two kinds of stacks, especially
for 0.98 < x < 1, where, as we have seen, there appear to be two tran-
sitions. (17)
The static susceptibility measurements(90-92) in TSeF-TCNQ have
included the semiconducting state, however, and show magnetic and
conducting activation energies that are approximately equal (= 125°K),
which is similar to the behavior we have reported for TTF-TCNQ. This
similarity is one example of a behavior in TSeF-TCNQ which, rather than
posing an additional problem by its contrast with TTF-TCNQ, has under
reexamination revealed the expected similarity to TTF-TCNQ. It should
be noted, however, that the idea of distinct magnetic and current-carrying
excitations in TTF-TCNQ but not in TSeF-TCNQ was consistent with the
idea of much stronger electronic repulsions in TIF-TCNQ, an idea that
sprang from two other properties of TSeF-TCNQ, its smaller Xspm,F, as
discussed by Tomkiewicz et at., (86) and the absence of a 4kr anomaly, both
of which must still be dealt with.
A systematic study of the roles played by the two kinds of stacks in the
semiconducting phase, through as much of the alloy series as the observ-
ability of the EPR line will allow, may give some decisive information
about the nature of the magnetic excitations in TTF-TCNQ. Such a study
has been undertaken by Tomkiewicz(78) but has not yet been reported.
6.3. Superlattices and Phonon Anomalies
6.3.1. General Considerations
Below the metal-semiconductor phase tranSItion, the one-dimen-

sional periodic distortions on each stack of wave vector ± Ob, whether they
be Peierls distortions, Wigner lattices, spin-Peierls distortions, etc., are
locked together in a three-dimensionally ordered state. What distinguishes
this state from the thermodynamic state above the transition is not the
correlation between distortions at nearby points nor the rms amplitude of
the distortion, both of which vary continuously through the transition and
continue to grow until T = OaK. Rather, it is the range of the correlations,
which becomes infinite at the transition, and the average amplitude of the
distortion, which begins to grow from zero at the transition. Below the
transition, this average distortion, being periodic, is characterized by a
wave vector Q. which we shall call the superlattice vector, and a set of
average displacement vectors {D,} for the atoms within the unit cell. The
"mode" that develops the average displacement is not in general a normal
mode of the harmonic lattice, i.e., a single vibrational mode. Rather, it can
be viewed as the "condensation" of a collective mode in the gas of
anomalously soft phonons.
Consider the effect of such a superlattice distortion on the x-ray
diffraction pattern If Rn + r, is the equilibrium position of the ith atom in
the nth unit cell, this atom suffers an average displacement
0, exp[ iQ . (Rn + r,)] + 0, * exp[ - iQ . (Rn + r,)] and the intensity of scatter-
ing with scattering vector K due to this distortion is, to the lowest order in
the 0"
Is(K)ocI 8[K -(G-Q)]II (K' 0,) exp[i(K+Q)· r,][.(Kf

G ,
+I8[K-(G+Q)1II(K' D,*)exP[iK-Q)'r,l[.(Kf (11)

G ,
The G's are vectors of the reciprocal lattice and, if we assume the molecu-
lar charge distribution to be a superposition of atomic charge distributions,
the [.(K) are the atomic form factors. Thus the superlattice distortion
produces a set of coherent spots at K = G + Q, satellites of the coherent
The Organic Metals (TSeF)x(ITF)l-x-TCNQ 207
Bragg spots at K = G, with intensities
18(K)<XL o(K - G)IL exp(iK· ri)/i(K)12 (12)

G I
This is to be contrasted with the sheets of anomalous intensity in k space

occurring in the metallic phase that result from long correlation lengths in
only one spatial direction.
Knowledge of the intensities of a sufficient number of satellite spots
should, in principle, allow a complete determination of the displacements
Di. However the intensities are so low that only a few satellite spots have
been observed and no such determination has been possible.
Various simplifying assumptions for the Di have been considered that
could be tested with far fewer spot intensities. For example, if the ath
molecule in the unit cell (a = 1, ... ,4) is assumed to be displaced rigidly by
Da, then
where Fa(K) = Li(aJi(a)exp(iK· ri(a))andra are theform factor and theposi-

tion within the unit cell of the ath molecule, and li(a) and ri(a) are the form
factor and position of the ith atom within the ath molecule. This is to be
compared with the corresponding expression for the Bragg intensities:
(14 )
We see that because of the rapid variation of Fa (K) no simple relation

should exist between the satellite and Bragg intensities of nearby points
unless IK . Da 12 is independent of a, which is likely to be realized, if at all,
only by Da == D.
The one simplifying situation that might actually occur is if all nonzero
Da are parallel, in which case the satellite but not the Bragg intensities
would all vanish for all K normal to that direction. The apparent vanishing
of a set of spots must be treated with caution, however, because most
satellite intensities are very small to begin with. The converse, viz., the
non vanishing of a spot for some K implying that at least some molecular
displacement has a component along K, gives more reliable information.
A similar analysis holds for elastic neutron scattering, the principal
difference being that while the x rays see the displacement of the electron
cloud, which is mostly the molecular "core-electrons," the neutrons see the
displacement of the nuclei. Since the biggest x-ray scatterers are sulfur and,
especially, selenium. while the biggest neutron scatterer is nitrogen, one
might expect the x rays to be more sensitive to distortions on the fulvalene
Table 4. Comparison of Scattering Length a for Neutrons and Atomic Scattering

Factor f(K) for X Rays at K = 1 A- 1
H (' N S Se
a -0.372 0.603 0.940 0.285 0.78

f n.007 111 1.27 3.56 8.70
stacks, and neutrons to probe distortions on the TCNQ stacks. For TSeF-
TCNQ this may be the case; for TTF-TCNQ the detailed interference
embodied in the FaCK) and in La exp(iG· r)F,,(K) are also important. In
Table 4 we have listed the scattering lengths for neutron scattering, which
play the role of K -independent atomic form factors, and the x-ray form
factors, which are of course K -dependent, for IKI = 1 A-I (a typical value).
In addition to the coherent satellite spots, there should also be inco-
herent x-ray scattering around both the Bragg and satellite spots resulting
from the emission (or absorption) of energy into a vibrational mode and/or
from elastic scattering off quasistatic fluctuations. The incoherent scatter-
ing associated with the various satellite spots will be unobservable, because
the coherent superlattice scattering itself is so weak. The incoherent scat-
tering associated with the Bragg points will consist of the diffuse scattering
from normal low-frequency phonons and in addition the scattering from
the quasi-one-dimensional fluctuations on the 2kf sheets. Below Tel the
latter scattering should continue to be prominent, but it is expected to
weaken rapidly with decreasing T as the stiffness of the three-dimensional
superlattice increases with increasing average amplitUde of distortion.
In view of this discussion, the relevant quantities for the semiconduc-
ting state are the positions of the satellite spots, from which the Q's are
determined, the general patterns and possibly detailed relationships among
the satellite intensities and nearby Bragg intensities, from which informa-
tion about the H, can be obtained, and the residual quasi-one-dimensional
anomalies from which inferences can be drawn about what has and what
has not ·'condensed." A comparison of x-ray and neutron results can give
information about the respective roles played by the two kinds of stacks,
although selective doping may be a more sensitive probe of these roles.
6.3.2. TTF- TCNQ
We first summarize our present knowledge about x-ray

d I'ff ractlOn
. (101 .102) an d e Iastlc
. neutron scattenng
. (131 .145)·In t h e temperature
ranges 49-S4°K, 3~-49°K, and ::s38°K. In the range 49-S4°K, satellite
spots begin to develop at K=G±Q, for several G's, where Q=
The Organic Metals (TSeF)xCTTF)I-x - TCNQ 209
Qall* + Qbb* + Qcc* = 0.5a* + 0.295b* + Oc*, but no 4kf satellites are visi-
ble. The 2kf distortion appears to be polarized only along c*, (101) and the
one-dimensional 2kr and 4k f anomalies are still seen.
The value Q a = a'l' /2 and its relation to the apparent decoupling of
the TTF stacks from the TCNQ stacks and from the transition can be
understood in terms of one key observation: If the periodic distortion and
its associated charge-density wave on successive (along a) TCNQ stacks are
out of phase (i.e., Qa = a* /2), then any periodic distortion polarized along
b or c* on the intervening TTF stack will experience no net field from the
neighboring TCNQ distortions, at least to first order in the distortion ampli-
tudes on each stack,i117146l as shown in Figure 25. A complete decoupling
to this order between any such distortion on the TTF stacks and a three-
dimensional superlattice distortion with Qa = a* /2 extending throughout
the whole TCNO subsystem will occur unless(147.148) (1) the TTF distortion
interacts with distortions on non-nearest-neighboring stacks (along a*) and
(2) the molecules tilt out of the ac planes and (3) the crystals fail to be
orthorhombic (i.e., a . c ~ 0). Since all of these conditions are met only
weakly, it is not surprising that the decoupling is nearly complete, i.e., that
predictions based on complete decoupling, such as the lack of a phase
transition on the TTF stacks at 54°K, are reasonably well confirmed.
The fact that the energetically most favorable distortion has Qa =
a* /2 and Qc = 0 just below Tc I is also understandable if the average
distortion is only on the TeNO stacks, because the Coulomb interaction
between CDWs associated with the distortions on different TCNQ stacks,
which falls off exponentially with distances greater than the wavelength of
the distortions, favors successive TCNO-CDWs along both a and c to be
out of phase.o 49 ) In the c direction, since successive TCNQ stacks are
inequivalent, this arrangement is described by Qc = 0 (although DOl =
-002)'
Figure 25. Oecoupling of a periodic lattice dis-

tortion on a single TIF stack (F) from a three-
dimensional superlattice distortion with Oa = a* /2.
Phases of TCNQ distortions are shown by two-
dimensional (i.e., complex) vectors. U l , U 2 • . . . • U6
are fields on central TIF distortion from distortions
of equal amplitude on six neighboring TCNQ stacks
(0 1 , O2 , . . . , 0 6 ), It is seen that the net field would
vanish if (a) the crystal were orthorhombic (i.e.,
(3 = 90°), because then U l =- U 4 , U 2 = - Us and
U 3 =-U6 : or if (b) only nearest neighbors were
interacting, because then [12 = -- U 5 : or if (c) TTF
molecules were to lie in ac planes. because then
U l = -U6 , U 2 = -[i,. and [1, = -[h
Of course, direct Coulomb interactions are not the only forces

between periodic distortions on different stacks. There are, for example,
short-range elastic forces between the distortions, as well. There are also
indirect electrostatic forces arising from the polarization induced on nearby
stacks. In the case of two successive TCNQ stacks along a, Weger and
Friedel(107) have concluded that the polarization of the intervening TTF
stack gives rise to the dominant effective interaction between the TCNQ
CDWs.o°7) We shall, for simplicity, call the net nonelastic force between
two periodic distortions the Coulomb force, and assume that it prefers the
distortions to be out of phase.
The reason why the polarization of the distortion is along c* in any
temperature range is certainly unclear, in view of the 34° tilt angle of the
TCNQ molecules. If the distortion is a Peierls distortion, the direction of
polarization is a result of competing influences: the stiffness of elastic and
Coulomb forces, which might most strongly oppose longitudinal distortions
and the "Peierls force" arising, it is assumed, from the modulation of the
intrastack transfer integral to by the distortion, which would favor some
polarization in the bc* plane that is not along either b or c*. The result of
this competition should not lead to a distortion purely along c*.
In the above, we have implicitly assumed that the relevant order
parameters are the complex amplitudes of acoustic-phonon-like distortions
on each stack polarized along band c*. Of the many other possibilities, two
have been suggested that warrant special comment: acoustic-phonon-like
modes polarized along a, considered by Craven et al. (138) and librons, first
suggested by Morawitz(l ';0) and developed into an elaborate theory by
Weger and Friedel.(107) Distortions of these two kinds are coupled in first
order to a three-dimensional TCNQ distortion with Qa = a* /2, but neither
is coupled in first order to tight-binding electrons. Thus, TTF distortions of
this kind are expected to scatter single electrons only by ± 4kf , not by
± 2k" and then only if the distortions are relatively large. We shall not
consider such distortions further, referring the reader to the papers of
Weger and Friedel for a detailed treatment of these possibilities.
In the range 38-49°K, at least four phenomena are observed: (1) Qa
begins to move from a* /2 (and from the equivalent point -a* /2) accord-
ing to the relation(117 14';) (a*/2)-iQa(T)i OC (Tc3 -T)1/2 although Qc
remains zero within experimental error; (2) at least one satellite begins to
develop at Tc3 (not at T, I)' and this satellite has a polarization component
not along c*(1Ol): (3) a 4k r satellite with Q'(T)= 2Q(T) begins to grow,
although its polarization is not yet established(IOI,I02 l ; and (4) the
one-dimensional 2kf and 4kf anomalies, though weaker, are still
present.(101 102) The fact that a temperature-dependent competition
between interactions within and between subsystems could make the satel-
lites move was noted by seveal authors.(137,146,149) The prediction of how
The Organic Metals (TSeF)x(TTF)l_x-TCNQ 211
this movement would begin and proceed, together with an analytic expres-
sion for Oa vs. T was given by Bak and EmeryY37) They noted that
although a departure of lOa I from a *12 can cost distortion energy for the
TCNQ subsystem, it produces an effective field ABOa proportional
to BQa'="IOal-O.5a* on the TTF stacks, for small BQa. which induces
distortions on the TfF stacks, contributing a negative energy
ex:: -(ABQa)2 Xdlst.F, where X dl>t,F is the susceptibility of the TTF subsystem
to forming a periodic distortion (of wave vector Q) in this effective field.
Assuming that Xdlst F increases with decreasing T, there comes a tempera-
ture Tc3 at which it pays to have BOa ¥- O. Introducing fourth-order terms in
the TTF order parameter with a temperature-independent coefficient to
stabilize the growth of BOa, Bak and Emery predicted a second-order
transition at Tc3 and BOa ex:: (xdlst,F - X~lst,F)I/2 for T < T c3 , where X~lst,F is
the value of Xdlst,F at T". Assuming Xdlst F - X~lst,F to be linear in Tc3 - T,
they obtained BOa ex:: (Td - T)l/2, Such a behavior was then seen to fit
previous neutron data, (1 <; 1) has since been verified with x rays(lOl,102) and
further neutron experimentsY4<;) and considered further theoretically by
Bak.(152) The slowly opening gap expected in Xspm,F has already been
commented on, although no quantitative fit has been achieved.
It should be remarked(l <;1) that other two-stack models, such as the
model considered by Weger and Friedel, give a similar T dependence for
BOa; the good fit of the Bak-Emery prediction is therefore not evidence
for the correctness of their specific choice of order parameters.
What is remarkable about the Bak-Emery prediction is that it fits so
well. If Tc3 were anywhere near the three-dimensional transition tempera-
ture for the TTF subsystem (calculated assuming complete decoupling with
the TCNQ distortion), one would expect Xdlst,F to begin to diverge, so that
a linear temperature dependence would fail and BOa 1:- (Tc3 - T)I/2, The
good fit is therefore suggestive of a TTF tendency toward instability that is
much weaker than that of the TCNQs. Also, as BOa becomes appreciable
we expect the effective field on the TTF stacks to be not ABQa but
(Aa *17T) sine 7TBOal a *). Sy and Mavroyannis have shown(l <;4) that this cor-
rected curve lies within experimental error of the Bak-Emery curve.
Because a 2kf satellite, with some longitudinal polarization, and some
4kf satellites make their appearance at Tc3 , the temperature at which the
fulvalene stacks first show an average distortion, it is tempting to speculate
that both kinds of satellites involve particularly the fulvalene stacks, This
speculation and the great sensitivity of Tc2 to TSeF doping, have motivated
an x-ray diffraction "tudy of TSeF-doped compounds which we report
below,
The basic idea of the Bak-Emery theory, that order parameters asso-
ciated with the TTF and TCNQ subsystems are decoupled as long as
Q a = a* /2, was investigated further for the actual crystal symmetry by
Abrahams, S6lyom, and Woynarovich(147); there, none of the conditions

listed above as being sufficient to decouple the TCNQ and TfF distortions is
fulfilled. Abrahams et at. studied the implications of this symmetry within
the framework of a Landau theory in some detail and generality. (148) In
addition to the absence of complete decouplingwhen Qa = a* /2, they found
that if successive (along c) TCNQ stacks have distortions that are out of
phase just below Te], as is generally believed, then the very small average
distortions on successive TfF stacks should be in phaseJ In the ordered
state below Tc1 as described by Bak and Emery, by contrast, successive TfF
stacks are also out of phase. Thus not only should there be some weak
indication of the metal-semiconductor transition at Tel in the properties of
the TfF stacks, but also the transition at To> should be weakly first order or
nearly so. The possibility that it is first order was seen by Craven et at. in the
dc conductivity, (11H) as we have already mentioned, while the weak
involvement of the TTF stacks in the metal-semiconductor transition is
reported to have been seen in recent higher-resolution g-value measure-
ments by Tomkiewicz. (7H)
The range T:s 38°K is most notable for the first-order phase transition
at 38°K in which Qa jumps discontinuously from 0.3a* to 0.25a*, Q'
jumps correspondingly (i.e., Q' == 2Q), the intensities of the satellite spots
also jump discontinuously, and hysteresis of - 1 K is observed in this
0
transition. From 38°K down at least to 25°K (the lowest temperature

reported,(102) Q remains at 0.25a* +O.295b* +Oc*. While it is generally
agreed that Qa = 0.25a* is stabilized by an energy associated with com-
mensurability in the a* direction, there have been at least four suggestions
as to the nature of the ordered stateY07,137,1'i5,156) In one,(107,155) the
distortIOn IS a superposition of distortions of equal amplitudes with Q± =
±O,2Sa*+O,295b*+Oc* This gives rise to a wave of the form
cos(Q+' r+cf>t-)+cos(Q . r+cf>-}
= 2 cos[O.295b* . r + (cf>+ + cf> - )/2] cos[0.25a* . r - (cf>+ - cf>-)/2]
which is amplitude-modulated with changing r a , in contrast to the wave
assumed above Tc2 with a single Q, which has the form cos(Q± . r + cf>±) and
is therefore phase-modulated with changing ra. The presence of both Q+
and Q_ is consistent with the observation of satellites corresponding to
both Q's with comparable intensities, although this could also arise from
Q+ and Q_ domains (i.e" twins in the superlattice), It might also explain
the observation(99) of a new 4k r satellite, corresponding not to 2Q+ or 2Q_
but to Q+ +Q '
An alternative low-temperature state suggested by Horovitz and
Mukamel(i "6) is fundamentally different from the state above Te2 in having
successive stack., along c now in phase at a cost of interaction energy within
·r See Note 2 Added III Proof on page 22~

TCNQ planes that is more than compensated by the gain in interaction

energy between neighboring TTF and TCNQ planes. This model has the
advantage that the discontinuous change of structure at 38°K need not
result from the presumably small commensurability energy associated with
Qa = a* / 4, but can come from the crossing of free energies of quite
different states, the free energy of the lower-temperature phase falling as
the TTF amplitudes rise. The commensurability energy is still needed to
stabilize Qa at a * /4 in the low-temperature phase. The different dis-
continuities in intensity associated with different satellite spots are adduced
in support of this model.
6.3.3. (TSeF)o o3(TTF)o 9T TCNQ
Let us now turn to very preliminary results on (TSeF)o 03(TTF)o 97-

TCNQ. This work has been motivated by three considerations: (1) to see if
the 4kf satellites arise mainly within one subsystem (if so, which one?); (2)
to understand better the role of the fulvalene stacks in the phase between
49 and 38°K and, in particular, their role in the 38°K transition; and (3) to
begin a study of the whole alloy series that will serve to bridge between
TTF-TCNQ and TSeF-TCNQ with their quite different superlattice
behaviors.
As we have seen in discussing transport and magnetic properties, the
introduction of TSeF molecules into the TTF stacks should have at least
three effects on the development of a superlattice distortion: (1) it should
decrease the tendency of each TTF stack towards a distortional instability,
(i.e., of X d"t,F ~ 00) by smearing the electronic states (if Peierls)(157) or the
magnon states (if spin-Peierls) in k space; (2) it should limit the correlation
length that can develop along an isolated stack because of both this smaller
distortional tendency and the tendency of impurities to pin the phase of a
distortion to a preferred value (say 0 or 7T) at each impurity (thereby
preventing long-range correlation of the phase f <;8); (3) it should inhibit
the growth of the correlation length in the transverse direction because of
the shortened intrastack correlation length and the direct effect of phase
pinning at random points.(I,g lliO) Although the origin of the 4kf
phenomena is uncertain, the doping should have similar effects on the 4kr
satellites. In any case the direct effect of such doping should be much
weaker on phenomena occurring essentially in the TCNQ subsystem alone,
stronger on phenomena involving the TTF subsystem or both.
The effects of fulvalene doping(11 <;) on the diffuse x-ray scattering that
are known at this writing are unfortunately very preliminary, but already
interesting. We summarize and comment on them.
(1) The buildup of the 2k r satellite appears, by extrapolation, to begin
at ~ 50 o K, which is consistent with the lowering of Tc I from 53 to 50 K 0
seen in the dc conductIvity.

(2) The doping has no noticeable effect on the Qb value of this

satellite, nor on Qc, which remains = 0 for all T within experimental error,
as in the pure system.
(3) No sharp transition is seen, and therefore no Tc3 is accurately
determined, where the movement of the 2kf satellites described by Qa(T)
clearly begins, although an estimate of Tc3 = 46°K is not unreasonable. The
movement of the satellite, in fact, appears to be slowed down as a function
of decreasing temperature, as seen in Figure 26. Furthermore, it no longer
shows the simple (Tc3 - T)l/2 dependence [although a (Tc3 - T)1/2 fit, now
poor, would give roughly the same coefficient as in pure TIF-TCNQ]. The
slower movement of Qa can be qualitatively understood, of course, in
terms of an inhibition in the growth of Xdist.F with decreasing T, to be
expected from doping the fulvalene subsystem.
(4) The intrusion of a first-order transition, seen in TTF-TCNQ at
38°K, is also inhibited, so that the transition is now at - 28°K and the
variation of Qa now appears to be continuous through the transition
(Figure 26). This fact alone would argue against the model of Horovitz and
Mukamel, since their model envisages a transition between two states
having a different symmetry along c in addition to having a different Qa, a
transition that must therefore be first-order. For this reason it is interesting
to note the intensity of one of the (021) satellites vs. T as shown in Figure
27. The transition is seen still to be somewhat abrupt, like a first-order
transition with some broadening. This can be taken as support of the
Horovitz-Mukamel model, since the discontinuities in the Bak-Emery and
Bjelis-Barisic models, are related solely to the discontinuity in Qa. Just
TEMPERATURE (OK)
Figure 26. Transverse component Q a of superlattice vector vs. T for TTF_TCNQ(145) (---),
TSeF-TCNQ(l16) (~-), and (TSeFlo.03(TTF)O.9T TCNQ(115) (data points and error bars).
""'- ........... ,
(f)
I-
Z
...... ,
:::) \TTF-TCNQ
al
Il:
<I: \\,
r II
I-
Ui
Z
w
till
II
I- II
,,
Z
-t ... ,
Figure 27. Intensity of superlattlce

satellite at (Qa(T), 2.295b*, CO) VS.
'"..." ,:.
T in TTF_TCNQ(1021 and In 10 20 30 40 50
(TSeF)" 03 (TTF)" 9r TeN Q ( I 1 'J TEMPERATURE ("K)
why the ability of the Horovitz-Mukamel state to compete is so reduced by

doping as to defer the transition to - 28°K, and why the transition appears
to occur just when Qa = 0.25 is still unclear.
(5) If Is.T and I B . T are the total intensities in the satellite and Bragg
spots (the total rather than peak intensity, in order to include the effects of
finite correlations arising from disorder), then the ratio Is.TIIB •T for this 2kf
satellite, measured at 20 o K, is decreased by a factor = 0.4 on doping,
suggesting a decrease in the distortion amplitude on doping by a factor
-0.6, which is comparable to the reduction in the magnetic gap on 3%
doping.
(6) The size and shape of the 2kf satellite spots at 23°K, when account
has been taken both of instrumental resolution and the finite size of the
Bragg spot (due to doping), yield correlation lengths of ga = 9 stack spac-
ings, gb = 12 molecular spacings, and gc = 9 stack spacings. The nearly
isotropic character of these results gives an indication of the three-dimen-
sionality of the ordering interactions in the ordered phase; otherwise, one
might expect the correlation length gb to be longer (mean separation of
impurities along b is 33b) and ga and gc to be shorter.
(7) There is currently a difference of opinion about the existence of
4kf satellites, Kagoshima et al. (115) seeing none and the Orsay group(161)
seeing at least one, although the difference could be related to differences
in dopant concentration. Kagoshima et al. also see a pronounced one-
dimensional 4k r anomaly at temperatures as low as 20oK, which suggests
that, in their crystal, at least, the 4k r satellite does not form and that
whatever happens at 4k r strongly involves the fulvalene stacks. The
presence of a 2kr satellite is an indication that any anomalous behavior at
4k f is not just a harmonic of behavior at 2kf , although the precise relation
of the longitudinal 2k r and 4k f behavior is still to be investigated.
6.3.4. TSeF- TCNQ
The x-ray investigation of TSeF-TCNQ by Weyl et a1Y16) which must

also be viewed as only a beginning, has nevertheless revealed some
important differences from TTF-TCNQ when viewed in the context of
what else we have learned about this material. First, as we have seen in
Section 6.1, there is only one transition (or possibly two very closely spaced
transitions) in which both subsystems develop three-dimensionally ordered
distortions, the situation analyzed by Horovitz and Mukamel. One would
therefore expect a state with Qa of- a* /2. In TTF-TCNQ, in fact, we saw
that the larger the distortion amplitude on the TTF stacks the greater the
deviation of Qa from a */2, reflecting the tendency of distortions in neigh-
boring stacks along a to be out of phase. In TSeF-TCNQ, where the
fulvalene distortions are the largest, the satellites nevertheless appear to
have Qa = a * /2. Schultz has proposed a possible explanation for this(120)
which argues that when both subsystems undergo superlattice distortions
essentially together, one must consider not only terms that are fourth-
order in the order parameter of one subsystem or the other but also
fourth-order terms that are second-order in the parameters of each
subsystem. A phase-dependent term of this structure can lock pairs of
adjacent TCNQ and TSeF distortions together in such a way that Qa =
a* /2 can be optimal, despite the tendency of the bilinear interaction
between subsystems to drive Qa away from a */2.
The value Qa = a* /2 is additionally paradoxical because, as we have
seen, it is for Qa = a* /2 that the two susbsystems are approximately
decoupled to second order in their distortions; yet in TSeF -TCNQ the two
phase transitions occur at or near the same temperature. At present there
exists no theory that predicts only one phase transition and Qa == a* /2. At
least two suggestions of why the transitions may be drawn toward one
another, if not locked together, have been given: One is based on the idea,
discussed in Section 6.2, that in a weakly interacting-electron model,
greater interstack banding is the reason for the low Tel in TSeF-TCNQ. It
is argued(87) that If one subsystem (TCNQ) once begins to suffer a Peierls
distortion, the band gap 2.:l a , which is seen by both subsystems, begins to
shrink and is no longer so effective in lowering the transition temperature
of the other (TSeF) subsystem. The other suggestion(120) is based on the
idea that once one subsystem distorts, fourth-order terms make the
quasistatic fluctuations in the other subsystem sense this distortion, despite
Qa = a* /2. These fluctuations, though incommensurate with the crystal
lattice, are commensurate with the distortion superlattice of the first
subsystem, so their tendency to undergo a three-dimensional transition is
increased. Neither of these mechanisms will lock the two transitions
together, however.
The second principal result of the x-ray investigation is the absence of
The Organic Metals (TSeF).(TTF)H-TCNQ 217
one set of satellites, whether it corresponds to the 2kf or 4kf satellites in

TIF-TCNQ being unclear.t This is consistent with the absence of either the
2kf or 4kf anomaly in the metallic state, and simply serves to deepen the
contrast between TSeF-TCNQ and TIF-TCNQ.
Because of the splitting of the transition into two and the sharp rise of
Tel with doping by TTF, an x-ray investigation of concentrations 0.95 <
x :51.00, and especially x = 0.99 will clearly be of considerable interest.
Also, because of the quite different low-temperature phases of TTF-
TCNQ and TSeF-TCNQ, a study of alloys of intermediate composition will
also be of interest to see how the temperature range for which Oa == a */2
broadens with increasing x, to see if there is a critical concentration Xc
above which the commensurability energy is insufficient to make Oa ==
a */4 for T < T e2 , etc.
7. Summary
In this section we summarize a few of the highlights of the ongoing

investigation of the isostructural system (TSeF)x(TTF)l-x-TCNQ. Broadly
speaking, this system has provided information from three kinds of
comparative studies: (1) direct comparisons between the pure materials
TTF-TCNQ and TSeF-TCNQ, (2) the study and comparison of the effects
of small percentages of doping of each material with the other, and (3) the
systematic study of various physical properties as one goes across the whole
alloy series.
Direct comparisons of TTF-TCNQ and TSeF-TCNQ have provided
the most information, despite the fact that the studies of TSeF-TCNQ are
still much less extensive. Certain broad similarities emerge: Both crystals
have very similar crystal structures and are metallic above a certain phase
transition temperature Te 1 and semiconducting below it. In the metallic
state, their dc conductivities are both very anisotropic and, along the highly
conducting b axis, have similar, if anomalous, temperature dependences.
Their total and separate-stack spin susceptibilities all have similar (and also
anomalous) temperature dependences. At room temperature, the two
compounds have remarkably small EPR Iinewidths, although the smallness
in TTF-TCNQ is the more remarkable. These EPR linewidths both
increase with decreasing temperature. In the metallic state, both materials
show one-dimensional anomalies in the diffuse x-ray scattering pattern
having comparable wave numbers Ob, the so-called 2kf anomaly, and in
the semiconducting state both show a three-dimensional superlattice hav-
ing the dimension Oh along b*. Ob is incommensurate with b*.
t See Note 1 Added In Proof on page 225.

There are certain systematic differences between TTF-TCNQ and

TSeF-TCNQ that are probably to be expected. The TSeF-TCNQ
bandwidth appears to be larger, and the bands are less one-dimensional,
results that are consistent with systematic differences in various lattice
parameters and electron orbital radii. The greater fulvalene bandwidth in
TSeF-TCNQ could account for the more important role played by the
fulvalene stacks relative to the TCNQ stacks in the dc conductivity (as
determined from the thermoelectric power). It is also consistent with a
greater plasma frequency and would imply somewhat stronger coupling of
the electrons with acoustic phonons. The increased banding along the a
axis can explain the much larger EPR linewidth.
There are also some very dramatic differences between TTF-TCNQ
and TSeF-TCNQ. In the metallic state, TSeF-TCNQ apparently:j: has no 4kf
anomaly or low-temperature satellites, yet its 2kf anomaly is much stronger,
being observable at higher temperatures and showing a much larger cor-
relation length:j: at any temperature. There are only one or possibly two very
closely spaced phase transitions in TSeF-TCNQ compared to the three or
four in TTF-TCNQ. Furthermore, the TSeF-TCNQ transition(s) is at a
surprisingly low transition temperature. In this connection, the semicon-
ducting state of TSeF-TCNQ shows a much simpler superlattice behavior,
but not what would be expected from stronger electron-phonon interactions
on the fulvalene stacks.
The experiments with TSeF-doping have confirmed the driving role of
the TCNQ stacks in the metal-semiconductor transition of TTF-TCNQ
and the importance of the fulvalene stacks in the 38°K transition, although
the extreme sensitivity of the latter transition to TSeF doping is not
understood. These experiments locate the 4kr anomalies at least partially
on the fulvalene stacks. The experiments with TTF-doping produce (or
reveal) two transitions in TSeF-TCNQ, with the upper transition rising
sharply with doping, thereby complicating the behavior to be explained in
TSeF-TCNQ.
The alloy experiments have served to follow various phenomena
"continuously" across the alloy series. For example, the transitions at Tel
in both TTF-TCNQ and TSeF-TCNQ are seen to be related. The x
dependence of the various lattice parameters, EPR linewidth, and room-
temperature thermoelectric power also present a consistent picture based
on changing bandwidth. However, some sharper results also emerge. The
variation of g shift with x is important evidence against an interpretation of
the temperature dependence of Xspm based on Peierls fluctuations. The
variation of the temperature dependence of the EPR linewidth with x may
throw light on the anomalous temperature dependences of the linewidth
observed in TTF-TCNQ and TSeF-TCNQ.
:j: See Note 1 Added In Proof on page 225.

The Organic Metals (TSeF)x(TTF)l_x-TCNQ 219
Thus we see that this isostructural system, while contributing answers

to some questions raises others: Which anomaly and satellite, 2kf or 4k" is
absent in TSeF-TCNQ:j: and why? Why does the observed anomaly in
TSeF-TCNQ have a much longer correlation length:!: than either the 2kf or
4kr anomaly in TTF-TCNQ? Why is the temperature range over which this
anomaly occurs much wider than that of the 2kr anomaly (or much nar-
rower than that of the 4k f anomaly) in TTF-TCNQ? How are these
differences from TTF -TCNQ related to the peculiar phase-transition
behavior and superlattice symmetry in TSeF-TCNQ? To what extent can
the similar behaviors (T dependence of Xspin and O"b) be explained in terms
of mechanisms that might also be related to sharply different behaviors (2k r
anomalies, 4kf anomalies)? A fuller study of TSeF-TCNQ, especially with
diffuse x rays and neutron scattering, may provide some of the answers.
In conclusion, the systematic study of the isostructural system
(TSeF)ATTF)! x- TCNQ can serve as a guide to similar studies in other
quasi-one-dimensional metals. The variety of experiments, the richness of
the results, the unusual sensitivity to what was originally conceived as a
small modification-all of these features can be expected to be reflected in
some way in similar studies of other systems.
ACKNOWLEDGMENTS
We are indebted to fellow workers in the field of one-dimensional

organic metals at many institutions for all they have contributed to our
understanding in this field. We are especially grateful to a number of past
and present colleagues whose valuable insights and critical comments have
been so helpful in writing this chapter. It is a pleasure to thank A. N. Bloch,
E. M. Engler, S. Etemad, S. Kagoshima, S. LaPlaca, P. F. Maldague, B. R.
'Patton, P. E. Seiden, B. D. Silverman, Y. Tomkiewicz, J. B. Torrance, and
M. Weger.
1. L. R. Melby, R. J. Harder, W. R. Hertler, W. Mahler, R. E. Bensen, and W. E.

Mochel, 1. Am. Chern. Soc. 84, 3374-3387 (1962).
2. F. Wudl, G. M. Smith, and E . .1. Hiifnagel, J. Chern. Soc. Chern. Commun. (1970),
1425-1436; D. L. Cotfen, J. Q. Chambers, D. R. Williams, P. E. Garrett, and N. D.
Canfield, J. Am. Chern. Soc. 93, 2258-2268 (1971); D. L. Cotfen, Tetrahedron Lett.
(1970),2633-2635; S. Hiinig, G. Kiisslich, H. Quast, and D. Scheutzow, Justus Liebigs
Ann. Chern. (1973), 310-323.
3. M. G. Miles, J. D Wilson, and M. H. Cohen, U.S. Patent No. 3,779,814, December 18
(1913)
:f: See Note 1 Added in Proof on page 225.

4. J. Ferraris, D. O. Cowan, V. Walatka, Jr., and 1. H. Perlstein, J. Am. Chem. Soc. 95,
948-949 (1973).
5. L. B. Coleman, M. J. Cohen. D. J. Sandman, F. G. Yamagishi, A. F. Garito, and A. 1.
Heeger, Solid State Commun. 12, 1125-1132 (1973).
6. Marshall 1. Cohen. L. B. Coleman, A. F. Garito, and A. 1. Heeger, Phys. Rev. B 10,
1298-1307 (1974).
7. E. M. Engler and V. V. Patel, J. Am. Chem. Soc. 96, 7376-7378 (1974).
iL S. Etemad, T. Penney. E. M. Engler. B. A. Scott, and P. E. Seiden, Phys. Rev. Lett. 34,
741-744 (1975).
9. 1. R. Andersen, R. A. Craven,l. E. Weiderbonner, andE. M. Engler,J. Chem. Soc. Chem.
Commun. (1977). 526-527.
10. E. M. Engler, V. V. Patel. J. R. Anderson, Y. Tomkiewicz, R. A. Craven, B. A. Scott,
and S. Etemad. in Conference on Synthesis and Properties of Low-Dimensional Materi-
als, 1977, Ann. N Y. Acad. Sci., New York (1978).
11. E. M. Engler and V. V. Patel. 1. Chem. Soc. Chem. Commun. (1975), 671-672.
12. E. M. Engler. B. A. Scott. S. Etemad, T. Penney, and V. V. Patel. J. Am. Chem. Soc. 99,
5909-5916 (1977).
13. The recent full structural determination for TSeF-TCNQ by P. Corfield and S. La Placa
(private communication) indicates a molecular arrangement essentially identical to that
for TTF-TCNQ
14. T. J. Kistenmacher. T. E. Phillips, and D. O. Cowan, Acta Crystallogr. 830, 763-768
(1974).
15. S. Etemad. E. M. Engler. T. D. Schultz. T. Penney, and B. A. Scott, Phys. Rev. B 17,
513-528 (1978)
16. S. Etemad. Phys. Rev. B 13,2254-2261 (1976).
17. R. A. Craven. Y. Tomkiewicz. E. M. Engler. and A. R. Taranko, Solid State Commun.
23.429-433 (1977).
18. F. Herman. D. R. Salahub. and R. P. Messmer. Phys. Rev. B 16, 2453-2465 (1977);
and F. Herman. Physica Scripta 16.303-306 (1977).
19. S. K. Khanna, E. Ehrenfreund, A. F. Garito, and A. 1. Heeger, Phys. Rev. B 10,
2205-2220 (1974)
20. G. A. Thomas, D. E. Schafer, F. Wudl, P. M. Horn, D. Rimai, 1. W. Cook, D. A.
Glocker, M. 1. Skove, C. W. Chu, R. P. Groff, 1. L. Gillson, R. C. Wheland, L. R.
Melby, M. B. Salamon, R. A. Craven. G. DePasquali, A. N. Bloch, D. O. Cowan, V. V.
Walatka, R. E. Pyle, R. Gemmer, T. O. Poehler, G. R. Iohnson, M. G. Miles, 1. D.
Wilson, J. P. Ferraris, T. F. Finnegan, R.I. Warmack, V. F. Raaen, and D. Jerome,
Phys. Rev. B 13. 5105-5110 (1976).
21. Marshall .I. Cohen, L B. Coleman, A. F. Garito, and A. 1. Heeger, Phys. Rev. B 13,
5111-5116 (1976).
22. R. P. Groff. A. Suna, and R. E. Merrifield, Phys. Rev. Lett. 33, 418-421 (1974).
23. W. N. Hardy. A. J. Berlinsky, and Larry Weiler, Phys. Rev. B 14, 3356-3370 (1976).
24. J. P. Ferraris and T. F. Finnegan, Solid State Commun.18, 1169-1172 (1976).
25. 1. Bardeen. Solid State Commun. 13, 357-359 (1973).
26. D. Allender, J. W Bray, and 1. Bardeen, Phys. Rev. B 9,119-129 (1974).
27. B. R. Patton and L 1. Sham, Phys. Rev. Lett. 33, 638-641 (1974).
28. L 1. Sham and B. R. Patton, Proc. German Physical Society Summer School on One-
Dimensional Conductors. Saarbrucken, July 1974, H. G. Schuster (ed), Springer- Verlag,
Berlin (I97~)
29. P. A. Lee, T. M. Rice, and P. W. Anderson, Solid State Commun. 14, 703-709
(1974).
30. P. E. Seiden and D. Cabib. Phys. Rev. B 13, 1846-1849 (1976).
31. P. F. Maldague (private communication).
The Organic Metals (TSeF)x(TTF)1 x- TCNQ 221
32 E I Blount and C M Varma, Bull Am Phys Soc 20,440 (1975)

33 H Gutfreund and M Weger, Phys Rev B 16, 1753-1755 (1977)
34 E M Conwell, Phys Rev Lett 39,777-780 (1977)
35 A J Epstein, E M Conwell D J Sandman, and J S MIller, Solid State Commun 23,
355-358 (1977)
36 A J Epstein and E M Conwell, Sohd State Commun 24,627-631 (1977)
37 A J Epstein, E M Conwell and J S MIller, In Conference on SynthesIs and Propertles
of Low-DImensIOnal Materials, 1977, Ann NY Acad SCI, New York (1978)
38 J C PhIllips, Phys Status Sohdl (b) 78,171-379 (1976)
39 E I Rashba, A A Gogolin dnd V I Mel'nIkov, In Organic Conductors and SemI-
conductors, Lecture 'Votes In PhYSICS, No 6~, L Pal, G Gruner, A Janossy, and J
S6lyom, (eds), Akademlal Klado Budapest, and Spnnger-Verlag, Berlin (1977)
40 A Madhukar and M H Cohen Phvs Rev Lett 38, 85-88 (1977)
41 J R Cooper (unpuhlI\hed), also J R Cooper and D Jerome (unpubhshed)
42 R H Fnend, M MJlpk D Jerome D L Decker, and D Debray (unpubhshed)
43 D Jerome, ] Phys (Pans) Lett 38, L489-L494 (1977)
44 RobertC Wheland ] 4m Chem Soc 98, 3926-3930 (1976), R C WhelandandJ L
Gillson, ] Am Chem Soc 98 1916-1925 (1976)
45 A Schultz, G D Stucky R Craven M J Schaffman, and M B Salamon, ] Am
Chem Soc 98, ~191-~197 (1976)
46 P M ChaIkin, J F KWdk T E Jones A F Ganto, and A J Heeger, Phys Rev Lett
31,601-604 (1971)
47 J F Kwak P M ChaIkin, A A Russel, A F Ganto, and A J Heeger, Sohd State
Commun 16, 729-7'l2 (197~)
48 P M ChaIkin, R 1 Greene S Etemad, and E M Engler, Phys Rev B 13, 1627-
1632 (1976)
49 P M Chaikin, J F Kwak, R L Green, S Etemad, and E M Engler, SolId State
Commun 19,1201-1204 (1976)
50 H Fntzsche, Solid State Commun 9, 1813-1815 (1971)
51 P E Selden (pnvate communicatIOn)
52 J R Cooper, M MJI]ak G Delplanque, D Jerome, M Weger, J M Fabre, and L
Glral, J Phys (Pans) 38 1097 1101 (1977)
53 N P OngandA M PortIS Phys Rev BI5,1782-1789(1977)
54 P M Grant, R C (,reene G C Wnghton, and G Castro, Phys Rev Lett 31,
1311-1314 (1973)
55 A A Bnght A F Ganto and A J Heeger, Solid State Commun 13,943-947 (1973),
A A Bnght, A F Ganto, and A J Heeger, Phys Rev B 10, 1328-1392 (1974)
56 B Welber, P E ~elden and P M Grant, Phys Rev B (to be pubhshed)
57 W Bludeau, P M Grant, and P E Selden, Bull Am Phys Soc 23, 382 (1978)
58 D B Tanner ( S Jdcobsen A F Ganto and A J Heeger, Phys Rev Lett 32,
1301-1304 (1974)
59 C S Jacobsen, D B Tanner A F Ganto, and A J Heeger, Phys Rev Lett 33,
1559-1561 (1974)
60 B Welber E M Engler, P E Selden, and P M Grant, Bull Am Phys Soc 35,311
(1976), and In Ph}slcS of SemIconductors, Proc 13th Int Conf SemIconductors, Rome,
1976, F G Fuml (ed ) North-Holland Publishing Co, Amsterdam (1977)
61 C W Chu, J M F Harper T H Geballe, and R L Greene, Phys Rev Lett 31,
1491-1494 (1971)
62 J R Cooper, D Jerome S Etemad and E M Engler, Solid State Commun 22,
257-263 (1977)
63 D Debray, R Millet D Jerome S Banslc, J M Fabre, and L Glral, ] Phys (Pam)
Lett 38 L227-1 211 (1977)
64. J. R. Cooper, D. Jerome, M. Weger, and S. Etemad, J. Phys. (Paris) Lett. 36, L219-
L222 (1975); see also D. Jerome and M. Weger, in Chemistry and Physics of One-
Dimensional Metals, NATO Advanced Study Institutes Series (Physics), Vol. 25, H. J.
Keller (ed.), Plenum Press, New York (1977).
65. Y. Tomkiewicz, B. A. Scott, L. J. Tao, and R. S. Title, Phys. Rev. Lett. 32, 1363-1366
(1974).
66. J. C. Scott, A. F. Garito, and A. J. Heeger, Phys. Rev. B 10, 3131-3139 (1974).
67. J. E. Gulley and J. F. Weiher, Phys. Rev. Lett. 34,1061-1064 (1975).
68. P. A. Lee, T. M. Rice, and P. W. Anderson, Phys. Rev. Lett. 31, 462-465 (1973).
69. E. F. Rybaczewski, L. S. Smith, A. F. Garito, A. J. Heeger, and B. F. Silbernagel, Phys.
Rev. B 14, 2746-2756 (1976).
70. Y. Tomkiewicz. A. R. Taranko, and E. M. Engler, Phys. Rev. Lett. 37, 1705-1708
(1976).
71. D. Jerome, G. Soda, J. R. Cooper, J. M. Fabre, and L. Giral, Solid State Commun. 22,
319-325 (1977).
72. U. Bernstein, P. M. Chaikin, and P. Pincus, Phys. Rev. Lett. 34, 271-274 (1975).
73. J. B. Torrance, Y. Tomkiewicz, and B. D. Silverman, Phys. Rev. B 15, 4738-4749
(1977).
74. Y. Tomkiewicz, A. R. Taranko, and J. B. Torrance, Phys, Rev. Lett. 36, 751-754
( 1976).
75. J. B, Torrance, in Chemistry and Physics of One-Dimensional Metals, NATO Advanced
Study Institutes Series (Physics), Vol. 25, H. J. Keller (ed), Plenum Press, New York
(1977); also, J. B. Torrance, Phys. Rev. B 17, 3099-3104 (1978).
76. M. W. Walsh, Jr., L. W. Rupp, Jr., D. E. Schafer, and G. A. Thomas, Bull. Am. Phys.
Soc. 19, 296 (1974).
77. Y. Tomkiewicz, A. R. Taranko, and J. B. Torrance, Phys. Rev. B 15, 1017-1023
(1977).
78. Y. Tomkiewicz (private communication).
79. E. F. Rybaczewski, A. F. Garito, A. J. Heeger, and E. Ehrenfreund, Phys, Rev. Lett. 34,
524-528 (1975)
80. Y. Tomkiewicz, T. D. Schultz, and B. D. Silverman (unpublished).
~l. G. Soda, D. Jerome, M. Weger, J. M. Fabre, and L. Giral, Solid State Commun, 18,
1417-1421 (1976).
82. G. Soda, D. Jerome, M. Weger, 1. Alizon, J. Gallice, H. Robert, J. M. Fabre, and L.
Giral,1. Phys. (Paris) 38, 931-948 (1977).
83. Y. Tomkiewicz, E. M. Engler, and T. D. Schulz, Phys. Rev. Lett. 35, 456-459 (1975).
84. Y. Yafet, in Solid State Physics, H. Ehrenreich, F. Seitz, and D. Turnbull (eds.), Vol. 14,
pp. 74-78, Academic Press, New York (1963).
85. T. D. Schultz, Y. Tomkiewicz, and A. N. Bloch, contribution to International Con-
ference on Magnetism, Amsterdam, 1976 (unpublished).
86. Y. Tomkiewicz. B. Welber, P. E. Seiden, and R. Schumaker, Solid State Commun. 23,
471-475 (1977)
87. Y. Tomkiewicz and T. D. Schultz (unpublished).
88. Y. Tomkiewicz. T. D. Schultz, E. M. Engler, A. R. Taranko, and A. N. Bloch, Bull. Am.
Phys. Soc. 21. 287-288 (1976); contribution to International Conference on
Magnetism, Amsterdam, 1976 (unpublished).
89. K. Singer, T. Wei. and A. F. Garito, Bull. Am. Phys. Soc. 22, 395 (1977).
90. S. Etemad, T. Penney, and E. M. Engler, Bull. Am. Phys. Soc. 20, 496 (1975).
91. J. C. Scott, S Etemad, and E. M. Engler, Phys. Rev. B 17, 2269-2275 (1978).
92. L. I. Buravov, R. N. Lyubovskaya, R. B. Lyuboskaya, and M. L. Khidekel, Zh. Eksp.
Teor. Fiz. 70,1982-1986 (1976) [SOl'. Phys. JETP 43,1033-1035 (1976)].
93. Y. Tomkiewicz (unpublished).
94 R Comes, M Lambert, and H R Zeller, Phys Status Soltdl (b) 58, 587-592 (1973)
95 B Renker, L PmtschovlUS, W Glaser, H Rletschl, R Comes, L Liebert, and W
Drexel, Phys Rev Lett 32, 836-839 (1974)
96 J W Lynn, M itzuml, G Shlrane, S A Werner, and R B Saillant, Phys Rev B 12,
1154-1166 (l97~)
97 R H Blessmg and P Coppens Solid State Commun 15,215-221 (1974), A Schultz,
G D Stucky, R H Blessmg and P Coppens, J Am Chem Soc 98, 3194-3201
(1976)
98 F Denoyer, R Comes A F Ganto and A J Heeger, Phys Rev Lett 35, 445-449
(197';)
99 S Kagoshlma, H AnLal, K KaJlmura, and T Ishlguro, J Phys Soc Japan 39,
1143-1144 (1975)
100 J P Pouget, S K Khanna, F Denoyer, R Comes, A F Ganto, and A J Heeger,
Phys Rev Lett 37 417-440 (1976)
101 S K Khanna, J P Pouget, R Comes, A F Ganto, and A J Heeger, Phys Rev B 16,
1468-1479 (1977)
102 S Kagoshlma, T Ishlguro, and H Anzal, J Phys Soc Japan 41, 2061-2071 (1976)
103 H A Mook and C R Watson, Phys Rev Lett 36,801-803 (1976)
104 G Shlrane, S M Shapiro, R Comes, A F Ganto, and A J Heeger, Phys Rev B 14,
2325-2314 (1976)
105 S M Shapiro, G Shlrane A F Ganto, and A J Heeger, Phys Rev B 15, 2413-2415
(1977)
106 H A Mook G Shlrane and S M Shapiro Phys Rev B 16, 5233-5237 (1977)
107 M Weger and J FnedcL J Phys (Pans) 38, 241-258 (1977), see also the addendum to
this paper, J Phys (Pam) 38 881-882 (1977)
108 L J Sham, Soltd State Commun 20, 623-625 (1976)
109 V J Emery,Phys Rev Lett 37 107-110(1976)
110 J B Torrance (pm ate communIcatIOn)
111 J BernaSCOnI, M J Rice W R Schneider, and S Strassler, Phys Rev B 12, 1090-
1092 (1975)
112 J Kondo and K Yamdjl, J Phys Soc Japan 43,424-436 (1977)
113 J Hubhard Phys Rei B 17 494-505 (1977)
114 E Pytte, Phys Ret B 10 4617-4642 (1974)
115 S Kagoshlma T Ishlguro E M Engler, T D Schultz, and Y TomkieWICZ, Bull Am
Phys Soc 22, 286 (1977) S Kagoshlma H Anzal, T Ishlguro, E M Engler, T D
Schultz, and Y TomkIeWICZ Lattice DynamiCs Proceedmgs of the InternatIOnal Con-
ference on LaUlce DynamlC\, Paw" 1977, M Balkanskl (ed ), pp 591-593, Flammanon
SCiences, Pans (1978)
116 C Weyl, E M Engler K Bechgaard G Jehanno, and S Etemad, Soltd State Commun
19,925-930 (1976)
117 See, for example H E Stanley, IntroductIOn to Phase TranstllOns and Critical
Phenomena, Oxford U nIverstty Press, New York (1971)
118 R A Craven M B Salamon G DePasquah, R M Herman, G Stucky, and A
Schult7, Phy~ Rev Lett 32 769-772 (1974)
119 D DJurek, K Franulovlc M Prester, S Tomlc, L Glral, and J M Fabre, Phys Rev
Lett 38,71';-718 (1977)
120 T D Schultz Solid State Commun 22, 289-292 (1977)
121 Y TomkieWICZ, R A Craven T D Schultz, E M Engler, and A R Taranko, Phys
Rev B 15, 1641-16')1 (1977)
122 R A Craven (unpubhshed)
123 M B Salamon J \\ Bray G DePasquah, R A Craven, G Stucky, and A Schultz,
Phys Rev B 11 619-622 (197')
124. For a review. see R. D. Parks, AlP Conf. Proc. 5, 630-638 (1972).
125. See, for example, F. C. Zumsteg, and R. D. Parks, Phys. Rev. Lett. 24, 520-524 (1970);
J. W. Shacklette. Phys. Rev. B 9, 3789-3792 (1974).
126. R. A. Craven and R. D. Parks, Phys. Rev. Lett. 31, 383-386 (1973).
127. Y. Suezaki and H. Mari, Prog. Theor. Phys. 41, 1177-1189 (1969); and Phys. Lett. 28A,
70-71 (1968).
128. P. M. Horn and D. Rimai, Phys. Rev. Lett. 36, 809-813 (1976).
129. P. M. Horn and D. Guidotti, Phys. Rev. B 16,491-501 (1977); D. Guidotti, P. M.
Horn, and E. M. Engler, in Organic Conductors and Semiconductors, Lecture Notes in
Physics, No. 65, L. Pal. G. Gruner, A. Janossy, and J. S61yom (eds.). Akademiai Kiad6,
Budapest. and Springer-Verlag. Berlin (1977).
130. W. Dieterich, Adv. Phys. 25, 615-655 (1976).
131. R. Comes, G. Shirane. S. M. Shapiro, A. F. Garito, and A. J. Heeger, Phys. Rev. B 14,
2376-2383 (1976).
132. T. F. Carruthers (private communication).
133. P. M. Horn. R. M. Herman, and M. B. Salamon, Phys. Rev. B 16, 5012-5015 (1977).
134. P. I. Perov and J. E. Fischer, Phys. Rev. Lett. 33, 521-523 (1974).
135. T. Ishiguro, S. Kagoshima, and H. Anzai, J. Phys. Soc. Japan 41, 351-352 (1976).
136. T. F. Carruthers and A. N. Bloch (private communications); also T. F. Carruthers, A. N.
Bloch, T. O. Poehler, M. E. Hawley, and D. O. Cowan, Bull. Am. Phys. Soc. 23, 381
(1978).
137. P. Bak and V. J. Emery, Phys. Rev. Lett. 36, 978-982 (1976).
138. R. A. Craven, S. Etemad, T. Penney, P. M. Horn, and D. Guidotti, IBM Research
Report No. RC6098 (1976).
139. Marshall J. Cohen and A. J. Heeger, Phys. Rev. B 16,688-696 (1977).
140. A. N. Bloch, T. F. Carruthers, T. O. Poehler, and D. O. Cowan, in Chemistry and
Physics of One-Dimensional Metals, NATO Advanced Study Institutes Series (Physics),
Vol. 25, H. J. Keller (ed.). Plenum Press, New York (1977), and private communication.
141. J. E. Eldridge, Solid State Commun. 21, 737-740 (1977).
142. W. J. Gunning. S. K. Khanna. A. F. Garito, and A. J. Heeger, Solid State Commun. 21,
765-770 (1977)
143. E. M. Engler, R. A. Craven, Y. Tomkiewicz, B. A. Scott, K. Bechgaard, and J. R.
Andersen. I Chern. Soc. Chern. Commun. (1976) 337-338.
144. R. M. Herman, M. B. Salamon, G. DePasquali. and G. Stucky, Solid State Commun. 19,
137-139 (1976)
145. W. D. Ellenson. R. Comes, S. M. Shapiro, G. Shirane, A. F. Garito, and A. J. Heeger,
Solid State Commun. 20, 53-55 (1976); W. D. Ellenson, S. M. Shapiro, G. Shirane, and
A. F. Garito, Phys. Rev. B 16, 3244-3251 (1977).
146. T. D. Schultz and S. Etemad, Phys. Rev. B 13,4928-4933 (1976).
147. E. Abrahams. 1. S6lyom, and F. Woynarovich, in Organic Conductors and Semiconduc-
tors, Lecture Notes in Physics, No. 65, L. Pal, G. Gruner, A. Janossy, and 1. S61yom
(eds.), Akademiai Kiad6. Budapest, and Springer-Verlag, Berlin (1977).
148. E. Abrahams. 1. S6lyom, and F. Woynarovich, Phys. Rev. B 16, 5238-5249 (1977).
149. K. Saub. S. Barisic. and 1. Friedel, Phys. Lett. 56, 302-304 (1976).
150. H. Marawitz. Phys. Rev. Lett. 34, 1096-1099 (1975); in Organic Conductors and
Semiconductors. Lecture Notes in Physics, No. 65, L. Pal, G. Gruner A. Janossy, and 1.
S61yom (eds.), Akademiai Kiad6, Budapest, and Springer-Verlag, Berlin (1977); and in
Conference on Svnthesis and Properties of Low-Dimensional Materials, 1977, Ann. N.Y.
Acad. Sci" New York (1978).
151. R. Comes. S M. Shapiro. G. Shirane. A. F. Garito, and A. 1. Heeger, Phys. Rev. Lett.
35,1518-1521 (1975).
152. P. Bak, Phy, Rei Lett. 37.1071-1074 (1976).
The Organic Metals (TSeF).(TTF)1 x - TeNQ 225
153 MWeger (pnvate commumcatlon)

154 HK Sy and C Mavroyanms Solid State Commun 23,79-82 (1977)
155 ABJehs and S Bansle Phys Rev Lett 37, 1517-1519 (1976)
156 HorovItz and D Mukamel SolId State Commun 23,285-288 (1977)
B
157 HG Schuster, Soltd State Commun 14 127-129 (1974)
158 J Sham and B R Patton Phys Rev Lett 36 733-736 (1976)
L
1~9 L J Sham and B R Patton Phvs Ret B 13 31~1-31~3 (1976)
160 Y lmry ,md S Md PhYI Rev Lett 35 1199-1401 (1976)
161 J P Pouget (pm dlL commUntLdtlOn)
Note 1 Added In Proof Kdgoshlmd, Ishlguro Schultz, dnd TomkiewIcz (unpublished), m

recent more detailed x-ray expenments, have found the correlation length at 900 K m TSeF-
TCNQ to be compar dble to those m TTF-TCNQ They also observe that With regard to
polanzatlOn the anomaly and satelhtc III TSeF-TCNQ are similar to the transverse 2kf anomaly
and satellite not the longitudlllal 4kf anomaly and satelhte, III TTF-TCNQ
Note 2 Added In Proof Dr Emery has kllldly pOlllted out to us that m the Bak-Emery theory
the decouphng at qa = a* /2 was only between distortIOns of the two subsystems both havlllg
U J = -U2 and that no decoupling WdS dssumed between such d distortIOn III one subsystem and a
distortIOn III the other subsystem havmg UJ = U2 Thus thelT andlysls agrees With the subsequent
work of Abrahams el at (147'
5
Perturbation Approach to Lattice
I nstabili ties in Quasi-One-Dimensional
Conductors
L.l. Sham
1. Introduction
In this chapter, the viewpoint is taken that the famous Frohlich-Peierls(1,2)

lattice phase transition in a quasi-one-dimensional conductor is an effect of
the electron-phonon interaction in a highly anisotropic electronic system.
The theory adopted here to account for the electronic and lattice prop-
erties is essentially that of a conventional field-theoretic treatment of the
electron-phonon system. (3) The model to be borne in mind is a system of
conduction electrons confined to move along parallel chains without
chance of hopping from chain to chain, interacting between the chains via
Coulomb interaction, and interacting with the phonons and with impuri-
ties.
The mean-field theory of the giant Kohn effect(4) and the Fr6hlich-
Peierls transition forms the backbone of our theory and will be elaborated
in Section 2. The fluctuations due to the one-dimensional nature of the
electron dynamics are accounted for in Section 3, in an approximate theory
following the treatments for fluctuations in superconductors of restricted
dimensions.(5,6) From the conventional electron-phonon system viewpoint,
it is simply an approximate self-consistent scheme, taking into account the
electronic contribution to lattice dynamics and the phonon modification of
electron dynamics, In Section 4, the three-dimensional nature of the
phonons and interchain electron coupling are included to give a more
L. ], Sham • Department of Physics, University of California at San Diego, La Jolla,

California 92093,
227
228 L. J. Sham
realistic description of the fluctuation effects. The impurities, which pro-

vide the electrons with a finite lifetime, have drastic effects on the Froh-
lich-Peierls lattice transformation. An account of the giant Friedel oscil-
lations(7) due to impurities and their consequences is given in Section 5.
The intrachain electron-electron interaction is not explicitly consid-
ered here and is simply (perhaps too naively) regarded as renormalizing the
quasiparticle properites of the conduction electrons. This may be a ques-
tionable assumption in view of the results of strong electron-electron
interaction models, (8) where the quasiparticle properties in a one-dimen-
sional system are lost.
However, the theory expounded here, which is based on the works of
Patton and this author,(9-11) has the advantage of starting with a fairly
realistic model, using approximations that are easy to visualize in terms of
simple physical pictures, arriving at conclusions qualitatively in agreement
with more general theory (such as the lack of a phase transition at finite
temperature for a one-dimensional system), and providing explanations for
some experimental observations.
2. The One-Dimensional Electron-Phonon System
We consider first a one-dimensional model in which the electrons are

free to move along the single chain and contribute a term to the Hamil-
tonian:
(1)
where Ck, C~ are the annihilation and creation operators for the conduc-
tion electron with wave vector k and energy Ck. The spin states will be
understood. The bare phonons contribute a term to the Hamiltonian:
(2)
q
where bq , b: are the annihilation and creation operators for the phonon of
wave vector q and frequency Oq (putting Ii = 1). The electron-phonon
interaction term is of the form
Her = L -1/2 I g(q)C~+qCkUq (3)

q.k
where L is the length of the system and
Uq = (1 /20 q )1/2(b ~ + b- q ) (4)
is the Fourier transform of the lattice displacement.
Perturbation Approach to Lattice Instabilities 229
The only difference from the usual electron-phonon problem is the

one-dimensionality. The method that we shall use is the same perturbative
field-theoretic method.(3) Consider first the effect of the electrons on the
phonon dynamics in the simplest approximation. The phonon Green's
function is changed from the unperturbed form
(5)
with WI = 211TT, T being temperature in energy units and I being an integer,
to the form
(6)
where n(q, iWI) is the phonon self-energy.(l2)
In the lowest-order approximation, the phonon self-energy is given by
the excitation of an unperturbed electron-hole pair, as depicted by Figure
la, i.e.,
n(q, iWI) = - 2 T L Ig(q )1 2 Go(k, iEn )Go(k + q, iEn + iWI) (7)

nk
where
(8)
is the unperturbed electron Green's function, and
En = (2n + l)1TT (9)
with n an integer. For the phonon wave vector q near the Fermi diameter
2k p , the electron-hole pairs give rise to the famous Kohn effect,(4) which
lowers the phonon frequency considerably. The approximate evaluation of
(7), neglecting the dependence of the electron-phonon matrix element g
on q in the neighborhood of 2kF and the curvature of the electron energy
in the neighborhood of the Fermi level, yields a real part on analytic
~
(0 I (bl
-c:>- - -c:>- - -c:>- -~

(c)
~ phonon
- - el ectron
Figure 1. Electron and phonon self-energy diagrams. ---- cou lomb in teract ion
230 L. f. Sham
continuation from iw[ to real w, changing the phonon frequency to be given

by
(10)
where A is the dimensionless electron-phonon coupling constant,
A = N(0)lg(2kF t /OZk/ (11)
N(O) being the single spin electron density of states at the Fermi level.
T p =(2YCF/1T)e- 1/ A (12)
Tp is obtained by assuming the electron bandwidth to be twice the Fermi
energy CF; y = 1. 78 is an Euler constant;
(13)
~(3) being the Riemann zeta function and V F the Fermi velocity.
From Equation (10), it is seen that the phonon at wave vector 2kF is
softest at a given temperature T> T p • As T decreases to the temperature
Tp , the frequency WZk F decreases to zero. This signifies a lattice instability
which results in a second-order phase transition to a distorted lattice at
temperatures below T p . Such a distortion of a one-dimensional system was
independently proposed by Peierls(Z) and Frohlich.(l) Frohlich(1) and
Kuper(13) worked out in remarkably complete details the thermodynamics
and transport properties of the distorted phase, which have been repeated
recently during the rise of interest in one-dimensional conductors. The
thermodynamic properties in the distorted phase are very similar to those
of the BCS superconductorY4) The simple approximation of the excitation
of an unperturbed electron-hole pair which leads to the softening of the
2kF phonon and to the lattice instability is called the mean-field approxi-
mation.
The phonon self-energy given by Equation (7) also has an imaginary
part on analytic continuation,
(14 )
which provides phonon damping. Thus, in the mean-field approximation,
there is a temperature range just above Tp where the 2kF soft phonon
becomes overdamped into a diffusionlike mode.
For T -s T p , the lattice distortion has wave vector 2kF with a finite
amplitude (UZk F ). This can also be thought of as phonon condensation into
a coherent state such that (b ZkF ) is not zero. In the mean-field approxima-
tion, in the electron-phonon interaction term Equation (3), all phonons
besides the finite 2kF displacement are neglected, leaving
Hep = L (Il*C~-2kFCk + IlC~k+2kFC-d (15)
k
where
(16)
The problem is reduced to a simple one of electrons in a one-dimensional
periodic potential. It can be solved by a transformation diagonalizing the
quadratic form in the electron operators, as was done by Frohlich(1) and
Kuper,(13) which is analogous to the Bogoliubov-Valatin(15) transformation
for the reduced BeS Hamiltonian. To maintain some uniformity in our
methodology, we shall use the Green's function method, which is analo-
gous to Gorkov's method(3) for the superconductor.
The electron Green's function, defined by
(17)
for imaginary time (3 ) T, has two types of terms: (i) the diagonal term k' = k
and (ii) the off-diagonal term k' - k = ±2kF due to the periodic distortion.
For k close to +k F, the equations of motion are
(iEn - cdG(k, k; iEn) = 1 + il*G(k - 2kF,k; iEn) (18a)

(iEn +ck)G(k -2kF,k; iEn) = ilG(k, k; iEn) (18b)
where we have used the approximation

(19)
measuring energies from the Fermi level. Hence,

G(k. k. iEn) = (iEn + ck)/[(iEn)2 - d - il 2] (20a)
G(k - 2kF.k; iEn) = -il' /[(iEnf - ck - il 2] (20b)
The electronic energy is given by
Ek = ±(d + il 2 )1/2 (21)
The lattice distortion has caused an energy gap 2il in the electron energy
spectrum. The size of the gap remains to be determined by a self-consistent
calculation of the lattice distortion.
From the equation of motion of the lattice displacement, the lattice
distortion (u2k F l is proportional to the electron charge density
\t C~-2kFCk)
By Equations (16) and (20b). then
il= -lg(2k F)/f1 2kF I2T L {il/[(iEn)2-d-il 2]} (22)

n.k
232 L. 1. Sham
Thus, the gap equation is given by(13)

_Al£F tanh[(e 2+il2)1/2/2T]
1- 2 -£F de (e 2 +il 2 )1!2
(23)
This is the same form as the BeS gap equation and il as a function of
temperature has been tabulated. (16) As T approaches the transition
temperature Tp , the gap il tends to zero and Equation (23) yields the
transition temperature, Equation (12). At zero temperature, the gap is
il(0)=2eFe- 1/A =1rTp /y (24)
The lattice distortion and the electron energy gap increases from zero to
the zero-temperature value as the temperature decreases from Tp to zero
in the same functional dependence as the superconductor energy gap
versus the reduced temperature, TITp .
3. Fluctuations in the One-Dimensional System
It is well known that one-dimensional systems do not have phase

transitions at finite temperatures. (17) The structural phase transition
obtained in the mean-field approximation in the preceding section is
smeared out by large fluctuations in one dimension.
Let us first examine the fluctuation effects in the high-temperature
limit. As the one-dimensional system is cooled down from high tempera-
tures, the phonons around 2kF are softened according to Equation (10).
There are finite fluctuation regions where distortions of the lattice have
taken place. In these regions, the properties of the distorted lattice prevail.
For example, the electron energy develops a gap.
The mean lattice displacement of wave vector 2kF over the whole
system is zero but the fluctuation is given by
(u 2) = T L D(q, iW/)
I.q
(25)
using Equation (10), where
e=ln(TITp ) (26)
At temperatures above Tp , this fluctuation is small, say, compared with the
zero-temperature value from Equation (24). A better criterion due to
Ginzburg(1S) will be given below. The fluctuation effects at high tempera-
tures may therefore be included in powers of the soft phonons, i.e., in
powers of (eel e )1/2, where ee is some critical value, below which the
expansion is not valid.
To the lowest order of the fluctuation, then, the electron self-

energy<9,1O) is given by Figure 1b, i.e., for k - kF'
I (k, iEn)= T I Ig(qtGo(k -q, iEn +iwl)D(q, iWI)

q,1
=c. ,~/Go(k - 2kF , iEn) (27)

where ~2 denotes the energy-gap fluctuation,
(28a)
q,l
(28b)
proportional to the lattice distortion.
As the temperature approaches Tp from above, the lattice distortion
fluctuation gets larger and the electron energy gap fluctuation gets larger.
This in turn will affect the Kohn effect, which involves an electron-hole
pair, making it weaker, thus smearing the phase transition. Including finite
powers of the fluctuation in Equation (25) will become an invalid pro-
cedure. We shall now describe two ways of treating fluctuations in the
temperature region around T p .
One way is to start with the Ginzburg-Landau(19) free-energy
functional,
F=const+N(O) J ( 1 ) a~)
dx al~1 +2bl~1 +c ax
2 2 2) (29)
valid near the mean-field transition temperature, taking ~ as the order

parameter. The coefficients a, b, care obtained(20) by comparing with the
expansion of the free energy in the mean-field approximation, which can
be obtained from Equation (20):
a = In(TplT) = -E (30a)
b = 7((3)/87T"2. T~ (30b)
C =~~ (30c)
The partition function and hence the thermodynamic properties of this free
energy in Equation (29) can be solved exactly.(21) This solution has been
applied(22) to the quasi-one-dimensional conductors. However, the free
energy in Equation (29) has the shortcoming that it only applies to a small
temperature range near Tp . The assumed(22) temperature dependence of
coefficients b and a causes some pathological consequences in applying this
procedure to a wider range of temperature.
In order for the high-temperature expansion of the fluctuation effect
in powers of (Eel E )112 to be valid, the Ginzburg criterion(18) says that Ee is at
234 L. 1. Sham
the critical temperature where the free-energy term of second order in

Equation (29) becomes of the same order of magnitude as the fourth-order
term, i.e.,
(ec 1!l12) = !b(I!l12)2 (31)
By Equation (28), this yields
ec -
_[7 (3)J
64
1/3 _
- 0.51 (32)
A second way to treat the fluctuations throughout the temperature

range including Tp is an approximate self-consistent scheme. Assume that
the Kohn effect and the phonon softening are still due to the excitation of
an electron-hole pair as given by Figure la, however, with the electrons
and hole under the influence of large lattice fluctuations as given by Figure
lb. Thus, the phonon Green's function has its self-energy given by Equa-
tion (7) with the unperturbed electron Green's function Go replaced by the
fully dressed one G, which in turn has its self-energy given by Equation
(27). Such an approximation has been called the Hartree-Fock approxi-
mation by Tucker and Halperin,(23) who applied this approximation to
solve the Ginzburg-Landau free energy for fluctuations in "one-dimen-
sional" superconductors (whiskers).
A slight variation ot this method is to use, instead of two fully dressed
Green's functions G· G in the phonon self-energy (7), only one fully
dressed electron, i.e., G· Go. This approximation is equivalent to the
so-called Hartree approximation in superconducting fluctuations intro-
duced by Marcelja.(24) A comparison of the two variants for superconduc-
tors was given by Hassing and Wilkins. (6) For the electron-phonon system
under study here, this Hartree approximation used by Sham and Patton(ll)
has the advantage of (i) taking into account, in a hand-waving manner,
some vertex corrections to the bubble in Figure 1a and (ii) having the limit
of the mean-field theory at low temperatures, and it will be used in the
following. From the Dyson equation
G=GO+GOLG (33)
and the self-energy expression (27), the dressed electron Green's function
is [ef. Equation (20a)]
(34)
Under the influence of the lattice fluctuation, the electron develops an

energy gap near the Fermi level. Substituting this expression into one of
the G's in (7) yields the renormalized phonon frequency:
w ~ = A n~kA 7J + e(q - 2kF )2] (35)
This differs from the first approximation (10) by the additional term
'Y/-E=P(6./T)=27rTt!/
n
I ~()
{En(E~+6.2)1/2[(E~+6.2)1!2+En]}-1 (36)
and by the changed expression for the coherence length ~,
(37)
which differs from Equation (13) by the factor
The electron energy gap fluctuation is, from Equation (28),

6. 2 = (7rVFT/2g)· 'Y/ -1/2 (39)
The set of three equations (36), (37), and (39) can be solved for 'Y/, 6., ~
at a given temperature T. Note that E, given by Equation (26), is just a
measure of T. In particular, from Equation (35), 'Y/ is a measure of the
square of the phonon frequency at 2k F . The condition of lattice stability
then means 'Y/ is greater than zero. It is easy to solve these three equations
numericalIy(11) to show that 'Y/ decreases steadily on cooling but never
vanishes except at T = O.
The low-temperature limit can be solved analytically. The argument of
the P function given by Equation (36), 6./T, tends to infinity if we assume 6.
tends to a constant. Then, from Equation (36),
P(27r0) = (' dE[ E-I tanh 7rE - (E 2 + 02rl/2 tanh 7r(E2 + 0 2)1/2]
Jo
~ In 2yo (40)
for large 8, where y is the Euler constant. Similarly, from Equation (38), as
8 --,> 00,
(41)
and by Equation (37),
e~!(VF/6.)2 (42)
The low-temperature limits of Equations (39) and (36) are
'Y/ 1/2 = 7rT/21/26. (43)
and
(44)
236 L. 1. Sham
The last two equations have the solution
(45)
and
(46)
Thus, the one-dimensional system becomes unstable only at zero tempera-

ture in accordance with the general theory. The energy gap is the mean-
field value, Equation (24). The energy-gap limit would not assume this
value exactly had we used the Hartree-Fock instead of the Hartree
approximation.
In Equation (36), the P function represents the fluctuation effect in the
one-dimensional system, which is neglected in the mean-field approxima-
tion. In that case, 1/ vanishes, and hence the onset of lattice instability
occurs at E = 0, i.e., T = Tp . The fluctuation effects in this theory come
from the modification of the Kohn effect due to the development of an
electron energy gap, which in turn is due to the large lattice distortion
fluctuation. In order to see exactly how this chain of events removes the
lattice instability at finite temperatures, let us approximate P and Q by
their small argument values, i.e., high-temperature limits,
P(b./T) = 7(3)(b./41TT)2 (47)

Q(b./T) = 1 (48)
Thus the 2kF phonon frequency squared is proportional to
(49)
where Ec is given by Equation (32). The second term on the right

represents the contribution to the 2kF phonon frequency from the energy
gap of the electron, which increases with decreasing phonon frequency
(i.e., larger lattice distortion fluctuation). This prevents 1/ from going to
zero and thus stabilizes the lattice.
4. Effects of Interchain Coupling
Consider now a three-dimensional model of a highly anisotropic

conductor. The molecules form chains in the z direction and the chains are
arranged on a square lattice in the x-y plane. The electrons are free to
move along parallel chains without possibility of hopping from chain to
chain. The phonons are, however, now three-dimensional. In the neigh-
borhood of the Kohn singularity qo = (0, 0, 2k F ), the phonon spectrum has
dispersion in the direction perpendicular to the chains as well, viz., instead

of Equation (35)
(50)
where g is the longitudinal coherence length as in Section 3 and g.L is the
transverse coherence length, measuring the transverse phonon dispersion.
We shall now investigate the interchain coupling of the phonons in the
framework of the self -consistent Hartree approximation developed in the
preceding two sections. The three-dimensional nature of the phonon dis-
persion drastically diminishes the fluctuation effects on the electron. When
Equation (50) is used in Equation (2Sa) to calculate the electron energy
gap, it is given by
(51)
where a is the lattice constant in the x-y plane. The soft-phonon fluctua-
tions, as indicated by a small value of 1], no longer give rise to a huge
electron energy gap as in the one-dimensional case, as indicated by Equa-
tion (39). The transverse dispersion, as measured by fda, replaces the
large fluctuation effects.
When Equation (51) for the gap is used in the fluctuation term
P(djT), of Equation (36), the 2kF phonon frequency may now vanish at a
finite temperature Te , lower than T p , the mean-field approximation.
Alternatively, it is possible to start with the one-dimensional results of
Scalapino et al.(21) and treat the interchain coupling in the mean-field
. . (2'i)
approxImation. -
To see that in our Hartree approximation a phase transition is restored
by the interchain coupling, approximate P(~I T) again by the high-
temperature form, Equation (47). Thus,
(52)
If the fluctuation term is small, 1] vanishes at
(53)
If the fluctuation term is not small, the transition temperature is given by
the transcendental equation
(54)
There are two sources of the phonon dispersion perpendicular to the
chains. One is the coupling of lattice vibration between different chains.
This gives a dispersion to the unrenormalized phonon frequency !1q. The
other source is the interchain Coulomb interaction between the electrons.
Let us study this latter contribution to the transverse coherence length
further.
238 L. f. Sham
Evaluate the usual electron-electron interaction term with the elec-

tron wave function taken to be a plane wave in the chain direction (along
the z axis) times a Bloch wave composed of nonoverlapping Wannier
functions 4>(r.L) localized about the chains. Then, the interchain electron
interaction contributes a term to the Hamiltonian of the separable form(26)
Hee = (20r! L V(q)C~C~+qCk'+qCk' (55)
k,k',q
where Ck is the electron annihilation operator associated with the Bloch
wave of wave vector k described above, 0 is the volume of the crystal, and
V(q) is the Fourier transform of the Coulomb energy between the inter-
chain charge-density waves:
L e x, f d 3 rf d 3 r' e - lq z(z-z')I4>(r.L +xl)12[e 2Ilr-r'I]I4>(r~t

2
V(q) = a 1q1
Lz XI
(56)
where Lz is the length of each chain and a 2 is the unit cell area of the
two-dimensional lattice in the x-y plane with lattice vectors Xl. The sum-
mation over Xl excludes Xl = 0, i.e., the intrachain Coulomb interaction.
In the Hartree approximation (or random-phase approximation), the
phonon self-energy includes a series of terms, a typical one of which is
shown in Figure lc. The phonon self-energy is now screened by the
interchain Coulomb interaction:
II(q, iWI)= _ Ig(qtO(q, iWI) (57)
1- V(q)O(q, iWI)
where O(q, iwd is the bubble in Figure 1 a:
O(q, iWI)= 2T L G(k, iEn)Go(k+q, iEn +iWI) (58)

nk
Of course, in the three-dimensional system, the phonons have more than

one branch. For simplicity of exposition, we consider the electron contri-
bution to one branch without the mixing of different branches.
We postulate that, for the system under consideration, the lattice
instability arises out of the electron-phonon interaction and not of the
direct electron-electron interaction. The latter is, therefore, assumed to be
weaker. In quantitative term, if we expand about qo,
(59)
neglecting the variation in the neighborhood of qz = 2kF along the z
direction, then we assume lJ..o much less than the electron-phonon coupling
constant A. Then, the phonon spectrum is given by Equation (50) with
1] = [lJ..o/(A 2-1J..6)] + [d/(A + lJ..o)] + P(~/T) (60)
and
(61)
In the mean-held approxiMlltion, i.I:!., P(il/ T);;;;; 0, 1] vanishes at the

temperature
(62)
We note that by comparison with Equation (12), the Coulomb repulsion
tends to suppress the lattice instability.
The model that we have just considered may be applied to the quasi-
one-dimensional conductor, potassium cyano-platinide (KCP) doped with
bromine or chlorine, K2Pt(CN)4Bro 3o·3.2H20. The platinum atoms form
chains closely spaced along the c axis and the chains form a square lattice
in the x-y plane. The d electrons of the platinum, partially oxidized by
bromine, form the conduction band, in which the electrons run along the
chains freely but not across the chains, thus accounting for the highly
anisotropic conductivity tensor.(26a) At room temperature, there is a Kohn
anomaly in the plane qz = 2kF in the reciprocal lattice space.(35) Neutron
scatterings also show at the same wave vectors the existence of elastic
peaks (central peaks). The reason for these central peaks will be considered
in the next section. The insensitivity of the intensity of the central peak to
the qx, qy components at high temperatures indicate the lack of correlation
between chains or the existence of one-dimension-like behavior.
On cooling to about 100oK, the intensity of the elastic peak at wave
vector (0, 0, 2k F ) actually decreases, but that at the Brillouin zone edge
(1T/a, 1T/a, 2k F ) increases rapidly, indicating the incipient formation of a
lattice distortion with that wave vector. This is explained by Barisic(35) in
terms of the charge-density waves along the chains being out of phase with
their nearest neighbors owing to interchain Coulomb interaction. This
follows naturally from our theory given above. The interchain Coulomb
interaction, V(q), given by Equation (56), is positive (repulsive) for q =
(0,0, 2kF) but most negative (attractive) at q = (1T/ a, 1T/ a, 2kF)' This favors
the instability of the phonon mode at (1T/a, 1T/a, 2k F ) and causes a true
phase transition(11) at about 100°K. The actual observation to about 6°K
shows that the phase transition is smeared out. The explanation for this will
be considered next.
5. Effects of Impurities
Impurities have very interesting effects on the Frohlich-Peierls lattice

instability. The electrons are scattered by the impurity potentials, which
conserve their energies but not momenta. Hence, the electrons acquire a
240 L. f. Sham
finite lifetime. Because the electrons are confined to move along the chain,
for the electrons near the Fermi level there are only two sorts of impurity
scatterings: (i) in the forward direction, which does not change the electron
momentum and will, therefore, be neglected here, and (ii) in the backward
direction with a change of momentum 2k F •
The electron self-energy correction due to impurity scattering is
included here only in the Born approximation for simplicity. Restricting
the impurity scatterings only to the backward direction (change of
momentum 2k F ) means the self-energy term is given by Figure 2a.
Contributions such as from Figure 2b vanish. The electron Green's
function including only the impurity scattering is given by, instead of
Equation (8),
(63)
where 7" is the lifetime given by
7" = VF/CU 2 (64)
c being the impurity concentration and U the backward impurity scatter-
ing amplitude.
In the mean-field approximation for the one-dimensional electron-
phonon system (Section 2), the lattice phase transition may also be
regarded as condensation of electron-hole pairs with momenta kF' -kF'
respectively. The impurity scattering, which changes the electron momen-
tum, therefore, tends to destroy the electron-hole pairs and depress the
,,'"\ ,, ,,,
J\,
I
\
, , I
v:
10) Ib)
Ic) (d)
leI If I
>-- - ~ electron - im pur il y scotle rin g Figure 2. Impurity effects on the electron
and phonon self-energies.
transition temperature(27 9) T p • This is analogous to the impurity effect on

the transition to an excitonic insulator(28) and to the magnetic impurities,
which tend to break up Cooper pairs and depress the superconducting
.. (29)
transItIon temperature.
In the mean-field approximation, the phonon self-energy is given by
Equation (7), except that now the electron is dressed by impurity scatter-
ing, as given by Equation (63). Because an electron on scattering always
changes its momentum by 2k F , vertex correction such as shown in Figure
2c is impossible. Hence, the phonon Green's function is given by(1O)
1/ D(2kF + P, iWl) = wi - Hl~kAE + !/I(!)-!!/I(!+a[l- iT(IWll + VFP»))

--!!/I(!+a[l- iT(lwd - VFP)]I} (65)
where !/I is the digamma function,(30) a = 1/47TTT is the electron-hole

pair-breaking parameter due to impurity scatterings, and € is given by
Equation (26) with the mean-field transition temperature Tp for the pure
system.
From Equation (65), it follows that the 2kF phonon now vanishes at
the temperature TI , given by
In( Tp/ T I ) = !/I(! + 1/ 47TTTJ ) - !/I(!) 66)

TJ is lower than Tp. For example, for weak scattering,(9)
TI = Tp exp(-7T/8TTp) (67)
Thermodynamic properties below the transition temperature TJ can
be calculated by analogy with the case of superconductors containing
magnetic impurities. (31)
Besides giving the electron a finite lifetime, which tends to suppress
the lattice instability, the impurities also induce a static lattice distortion. In
the presence of the electron-impurity scattering term, the time-ordered
lattice displacement correlation function consists of two parts:
1 liT
du'W'(Uq(T)U_q(O» = T- 1 d(q)(\o+ D(q, iWl) (68)
D(q, iwd is the dynamic part of the phonon Green's function considered in
previous sections; d(q) is the static (or zero-frequency) part, given by
Figure 2d:
d(q) = (cU 2 / l)[D(q, O)Il(q, 0)]2 (69)
The neutron scattering cross section is proportional to the lattice
displacement correlation,(32) which is the imaginary part of Equation (68).
For a fixed neutron momentum transfer q, the scattering function as a
function of frequency W measures the phonon spectral density, i.e., the
242 L. 1. Sham
imaginary part of D(q, iWl) plus an elastic peak at zero frequency with
strength d(q). The peak at zero frequency is called the central peak and is
widely observed in systems with displacive phase transitions, such as
ferroelectrics, (33) Nb 3Sn, (34) and a quasi-one-dimensional conductor,
potassium cyano-platinide(35) (KCP), K zPt(CN)4' Bro 30' 3H zO. The width
of the peak is often the instrumental resolution, which means that the
central peak has an elastic component.
The impurity-induced central peak(34.11.33) was first suggested by
Huang. (36) It is easy to understand in the quasi-one-dimensional conduc-
tor.(11) Because of the giant Kohn anomaly,(4) the phonons are particularly
soft at wave vector 2kF along the chain direction. The strong density
response at 2kF (the bubble in Figure 1a), which gives rise to the Kohn
effect, also causes the impurity to induce a strong Friedel oscillation of the
electron charge density with wave vector 2kF in the chain direction (the
bubbles in Figure 2d). This electron charge density wave induces a static
lattice distortion of the same wave vector, at which the lattice is particularly
easy to deform. When the induced lattice distortion is averaged over a
random distribution of impurities, the net distortion is zero but the mean-
square displacement is finite and given by Equation (69).
As a function of the frequency, the central peak due to impurities is a 8
function at zero frequency. As a function of the wave vector, the strength
of the central peak in the neighborhood of the soft mode (0, 0, 2kF ) is,
from Equation (69), proportional to
d(qo + P)oc 1/[ 1'/ + ~2 P; + ~~ (P; + P; )]2 (70)
which is a "squared Lorentzian."
In the self-consistent theory of Section 3, the electron self-energy
includes the fluctuation effect of the soft phonon. Now it should include the
effects of impurity-induced displacement. Besides the terms represented by
Figure 2a and Figure 1b, the electron self-energy now has additional terms
given by Figures 2e and 2f. In Figure 2, it is understood that the interchain
electron interaction screening of the bubble as treated in Section 4 is
included. The self-energy keeps the form of Equation (27), with Go given
by Equation (63) and with the energy gap fluctuation consisting of three
terms:
(71)
Ll6 is the term of Figure 1b, given by Equation (39) in the one-
dimensional system or Equation (51) in the three-dimensional system. Lli
comes from the term represented by Figure 2e:
Lli = 2cU 2 L I1(q, O)D(q, 0) (72a)
q
= 1/4a~~~ (72b)
evaluated in the three-dimensional case. ~~ is represented by Figure 2f:

~~ = cU 2 L [TI(q, O)D(q, 0)]2 (73a)
q
(73b)
evaluated in the three-dimensional case.
In the three-dimensional system, the fluctuations are dampened by the
interchain coupling. ~6 and ~i are independent of 1/ and will simply
depress the mean-field transition temperature without suppressing the
phase transition altogether when substituted into the stability condition,
Equation (36). The fluctuation energy gap contribution, ~2' from the
impurity-induced lattice distortion d(q) is proportional to d(q) and there-
fore inversely proportional to 1/ 1/2 even in the three-dimensional system.
This impurity-induced fluctuation effect is as strong as the fluctuation
effect, Equation (39), in the one-dimensional case and therefore suppresses
the lattice phase transformation.
This result, arrived at by our approximate theory, is quite gen-
eral. (37-39) If a term linear in the order parameter representing the impurity
potential term is added to the Ginzburg-Landau free energy, Equation
(29), it can be shown(38) that, for complex order parameters, long-range
order is destroyed by the impurity term in three-dimensional systems.
A theory including all the components discussed above-the fluctua-
tions of electrons in the chains, the interchain coupling, and the impurity
effects-was put together by Sham and Patton(ll) to apply to KCP. The
bromine ions play the role of the impurities. It explains qualitatively the
observed facts in KCp(35): (1) the existence of the central peak, (2) its
temperature dependence, particularly the lack of a critical behavior signi-
fying the absence of a three-dimensional phase transition, and (3) the
intensity of the central peak as a function of momentum component
perpendicular to the platinum chains, showing a lack of long-range order in
the transverse direction at the lowest temperatures.
ACKNOWLEDGMENTS
I wish to thank Professor J. Bardeen for helpful comments on the

manuscript and Professor Bruce Patton for numerous discussions over the
years. This research was supported in part by the U.S. National Science
Foundation.

1. H. Frohlich. Proc. R. Soc. London Ser. A 223, 296 (1954).
2. R. E. Peierls, Quantum Theory of Solids, p. 108, Oxford University Press, Oxford
(1953).
244 L. f. Sham
3. A. A. Abrikosov, L. P. Gorkov, and 1. E. Dzyaloshinski, Methods of Quantum Field

Theory in Statistical Physics, Prentice-Hall, Englewood Cliffs, New Jersey (1963).
4. W. Kohn, Phys. Rev. Lett. 2, 393 (1959); A. M. Afanasev and Yu. Kagan, Zh. Eksp.
Teor. Fiz. 43, 1456 (1962) [Sov. Phys. JETP 16, 1030 (1963)].
5. B. R. Patton, thesis, Cornell University (1971).
6. R. F. Hassing and J. W. Wilkins, Phys. Rev. B 7, 1890 (1973).
7. J. Friedel, Phil. Mag. 43,153 (1952).
8. V. J. Emery, in Chemistry and Physics of One-Dimensional Metals, p. 1, H. J. Keller
(ed.), Plenum Press, New York (1976).
9. B. R. Patton and L. J. Sham, Phys. Rev. Lett. 31, 631 (1973).
10. B. R. Patton and L. J. Sham, Phys. Rev. Lett. 33,638 (1974).
II. L. J. Sham and B. R. Patton, Phys. Rev. Lett. 36, 733 (1976).
12. See, for example, L. J. Sham, in Modern Solid State Physics, Vol. 2, Phonons and Their
Interactions, p. 143, R. H. Enns and R. R. Haering (eds.), Gordon and Breach, New
York (1969).
13. C. G. Kuper, Proc. R. Soc. London Ser. A 227,214 (1955).
14. J. Bardeen. L. N. Cooper, and J. R. Schrieffer, Phys. Rev. 108, 1175 (1957).
15. N. N. Bogoliubov, Nuovo Cimento 7, 794 (1958); J. G. Valatin, Nuovo Cimento 7, 843
(1958).
16. B. Miihlschlegel, Z. Phys. 155, 313 (1959).
17. L. D. Landau and E. M. Lifshitz, Statistical Physics, p. 482, Pergamon, London
(1958).
18. V. L. Ginzburg, Sov. Phys. Solid State 2, 1824 (1960).
19. See the exposition by N. R. Werthamer, in Superconductivity, Chap. 6, R. D. Parks (ed.),
Marcel Dek'ker, New York (1969).
20. D. Allender, J. Bray, and J. Bardeen, Phys. Rev. B 9,119 (1974).
21. D. J. Scalapino, M. Sears, and R. A. Ferrell, Phys. Rev. B 6, 3409 (1972).
22. P. A. Lee, T M. Rice, and P. W. Anderson, Phys. Rev. Lett. 31, 462 (1973).
23. J. R. Tucker and B. 1. Halperin, Phys. Rev. B 3, 3768 (1971).
24. S. Marcelja, Phys. Lett. 28a, 180 (1968).
25. W. Dieterich, Z. Phys. 270, 239 (1974); D. J. Scalapino, Y. Imry, and P. Pincus, Phys.
Rev. B 11,2042 (1975).
26. See, for example, L..T Sham, Phys. Rev. B 6, 3584 (1972).
26a. K. Krogmann. Angew. Chem. Int. Ed. Engl. 8, 35 (1969); H. R. Zeller, in Fest-
korperprobleme, Vol. 13, p. 31, H. J. Queisser (ed.), Pergamon, New York (1973).
27. H. G. Schuster, Solid State Commun. 14, 127 (1974).
28. J. Zittartz, Phys Rev. 164, 575 (1967).
29. A. A. Abrikosov and L. P. Gorkov, Zh. Eksp. Teor. Fiz. 39,1781 (1960) [Sov. Phys.
JETP 12,1243 (1961)].
30. M. Abramowitz and 1. A. Stegun, Handbook of Mathematical Functions, p. 258, Dover,
New York (1965).
31. See the review by K. Maki, in Superconductivity, Chap. 18, R. D. Parks (ed.), Marcel
Dekker, New York (1969).
32. See, for example. R. A. Cowley, in Modern Solid State Physics, Vol. 2, Phonons and
Their Interactions, p. 43. R. H. Enns and R. R. Haering (eds.), Gordon and Breach, New
York (1969)
33. See Proc. Inl. Con/. Lattice Dynamics, Paris 1977, M. Balkanski (ed.), Flammarion, Paris
(1978).
34. J. D. Axe and G Shirane, Phys. Rev. B 8,1965 (1973).
35. B. Renker. L. Pintschovins, W. Glaser, H. Rietschel, R. Comes, L. Liebert, and W.
Drexel, PhY5. Rev. Lett. 32, 836 (1974); J. W. Lynn, M. Iizumi, G. Shirane, S. A.
Werner, and R B. Saillant, Phys. Rev. B 12, 1154 (1975).
36. K. Huang, Proc. R. Soc. London Ser. A 190, 102 (1947).

37. Y. Imry and S. K. Ma, Phys. Rev. Lett. 35, 1399 (1975).
38. L. J. Sham and B. R. Patton, Phys. Rev. B 13, 3151 (1976).
39. K. B. Efetov and A. I. Larkin. Zh. Eks. Tear. Fiz. (to be published).
6
Theory of the One-Dimensional
Electron Gas
V.1. Emery
1. Basic Physics
1.1. Introduction
The properties of orgamc metals and other highly conducting solids are far
too complex to be understood by means of purely microscopic calculations.
They involve an intricate interplay of electronic, molecular, and lattice
motion, modified by disorder and conditioned by the symmetry of the
underlying structure, all of which must be included in some measure if all
of the experiments are to be understood. At the same time the chainlike
structure of the lattice often confines the motion of the electrons to one
direction in space, and this results in an exaggerated role of the interactions
as well as subtle cancellation between various many-body processes which
are simply not revealed by the approximations usually used to obtain at
least a qualitative description of more isotropic systems. Intuition
developed in this way cannot be trusted.
Faced with this situation, it is fortunate that the restriction to one-
dimensional motion brings along a compensating possibility of finding
exact solutions of certain many-body problems. They are not necessarily
the precise models we want to study for real physical systems, but they do
frequently contain so much of the essence that the solutions are able to
offer some guidance about what is going on as well as to show where more
approximate theories are to be trusted. Often no more is needed, for truly
microscopic calculations rarely confront experiment: it is much more usual
to get a feeling for what qualitative features can appear, to make limited
V J Emery • Department of PhYSICS. Brookhaven NatIOnal Laboratory, Upton, New

York 11971
247
248 V.1. Emery
predictions about the dependence of physical properties on temperature,

frequency, etc., then to introduce phenomenological parameters which are
fitted to experiment. Landau's Fermi liquid theory(1) is a prime example of
this procedure, and it has been employed to great advantage in under-
standing the properties of liquid 3 He, in which strong-coupling effects are
so very important.
It is useful to pursue a similar course in the study of organic metals,
and this chapter is concerned with the first step-a discussion of exact
solutions or controlled approximations of the properties of an electron gas
in one dimension. This is a limited goal. For the most part, the effects of
phonons and molecular vibrations will be ignored and there will be little
consideration of coupling between chains or attempt to use the results to
parametrize the experiments. The latter in particular involves a certain
amount of taste in adapting the results to the physical systems under
consideration and it is best considered case by case. Some of these ques-
tions are addressed in other chapters of this volume.
Three main approaches have been widely used in this field: Bethe's
ansatz, (2) the renormalization group method, and boson representations.
The first has produced many important results, some of which will be used
later, but it does not tell us as much as we should like to know about the
correlation functions which are so important for revealing the nature of the
collective motion or the possible phase transitions which the system may
undergo. On the other hand the other two methods do this very nicely even
though they omit almost all effects of the lattice. Despite the importance of
Bethe's ansatz, this chapter will concentrate largely upon the other two
methods, for they embody the most recent technical developments and also
have a beanng on problems of high-energy physics.
The boson representations are actually much simpler to use than many
of the more common methods of many-body theory and, for this reason, a
good part of Section 2 is devoted to a rather detailed account of the
technique so as to give the interested reader the necessary background to
follow the subsequent discussion and to carry out similar calculations for
himself. Before plunging into the details, however, it is perhaps useful to
give a more qualitative survey of some of the physical questions to be
addressed.:!:
1.2. Phase Transitions and Long-Range Order

Much of the physics of almost one-dimensional metals is concerned
with understanding the conditions for various kinds of phase transitions
that lead to collective states governing the low-temperature behavior. This
:j: The diSCUSSIOn given
In the rest of thiS sectIOn has been taken from the author's lectures(3) at
the NATO school on the Chemistry and PhYSICS of One-Dimensional Metals held In
Bolzano, Italv August. 1976
Theory of the One-DimensIOnal Electron Gas 249
is of central importance in the search for high-temperature superconduc-

tivity, which, for example, could be prevented by the intervention of a
metal-insulator transition. The most frequently discussed transitions are to
states showing singlet superconductivity, triplet superconductivity, static
charge-density waves (COW), or static spin-density waves (SOW). Super-
conductivity requires the formation of bound pairs of electrons which are
(roughly speaking) bosons and form a superfluid. The pairs may form with
singlet spin and even angular momentum as in a metal(4) or with a triplet spin
and odd angular momentum(5) as in liquid 3 He (the latter is a superfluid but
not a superconductor because 3 He atoms are not charged). In a static COW
state, the electron density IS not uniform but has a periodic modulation,
which, in the simplest situation, has the form
n(r)= no+nl cos(q· r+</» (1)
at posItion r. Such states have been studied extensively in layered

compounds. In an SOW state it is the spin density that varies from point to
point to give a loose kind of antiferromagnetism. All of these transitions
have been observed in more isotropic materials, but there may be others,
as yet undiscovered, that are peculiar to almost one-dimensional metals.
The low-temperature phases are characterized by the existence of a
new kind of order which extends throughout the system. They are
described by one or more order parameters-in Equation (1) by nl and </>,
the amplitude and phase of the COW. The magnitude of nl is determined
by minimizing the free energy, but unless there is a specific coupling to
something external (such as the lattice), </> is free to vary and the state is
degenerate. Above the transition temperature, particularly in a nearly
one-dimensional metal, the order can extend over quite long but finite
distances. Within such regions the system has the appearance of the
ordered phase. As the temperature is lowered, the range of the order
increases and it becomes infinite at the transition temperature Te.
However, in a purely one-dimensional system with short-range forces,
thermal or quantum fluctuations prevent the formation of long-range order
and there is no phase transition at all. To see how this comes about,
imagine that the system breaks into two segments with the phase of the
order parameter taking one value to the left of a point P and another to the
right. There are two contributions to the change 8F in the free energy-the
surface free energy (J" due to the mismatch at P and an entropy In M which
arises because P can be at any of the M sites:
8F = (J" - T In M (2)
Since the contact between segments occurs at a point, (J" is finite for large M
and the gain from the entropy far outweighs the cost in surface free energy,
so 8F < o. Then it is favorable to break the system into further segments
250 V. 1. Emery
and to continue to to do this until there is no remnant of the original

long-range order. This argument does not work in two (or more) dimen-
sions because, if the region of fluctuation contains a macroscopic number
AM of particles, the surface energy (T is proportional to M l /2 [or to
M l - lld in d dimensions] and is bigger than In M. The existence of a
fluctuation then depends upon the sign of the surface term. The discussion
is also modified if there are long-range forces which give interactions
between the segments. But none of this concerns us here; the point is that
fluctuations must be taken seriously and it is not useful to do a rough
calculation of the properties of a one-dimensional system. This is true even
at T = 0, where the entropy contribution to Equation (2) vanishes,
because, in general, the transition is removed by quantum fluctuations
which, so far, have been ignored.
Even though there is no long-range order, it is possible for very-Iong-
range (but finite) correlations to build up, and they trigger a phase tran-
sition in real materials by enhancing the coupling between the chains. The
important question then becomes the determination of what kind of cor-
relations grow the most rapidly as the temperature is lowered, for they, in
conjunction with the details of the interchain coupling, control the nature
of the phase transitions that finally take place.
There is a further effect severely limiting the applicability of simple
theories. At low temperatures, the various kinds of phase transitions are
consequences of interactions between particles primarily on the Fermi
surface, but in one dimension, the Fermi "surface" consists of the two
points k = ±kF' where kF is the Fermi wave vector. Then the available
phase space is so small that the various kinds of singular process cannot
help but interfere with each other, so it is not useful to do a theory of one
kind of transition and to include fluctuations in a self-consistent way-the
situation virtually demands an exact solution, or at least controlled
approximations which can perhaps be obtained in certain limiting cases.
These are the questions with which we shall be concerned.
1.3. Mathematical Model
The discussion will be based upon the Hamiltonian

H= Ho+Ht +H2 (3)
where
(4)
(5)
and
H2 = V L Pmu{Jm+l,o-' (6)
m,CT,O"
where I/I~,o- creates a Fermion of spin (T (of value i or t) at a lattice site n

and Pmo- = '" :o-I/Imo- is the number operator. This Hamiltonian describes
what is often called an extended Hubbard model, with Ho producing
hopping from site to site and Hl and H2 giving an on-site or an intersite
interaction, The couplings U and V come partly from Coulomb inter-
actions and partly from exchange of phonons or molecular vibrations, but,
for the most part, retardation will be ignored and the interactions will be of
short range in order to focus on the Fermi-surface effects, which are so
important in one dimension, Even with these restrictions there is a rich
variety of physical phenomena, and the theory forms a good basis for
extension to more complicated situations, Later, in discussing the
continuum limit, it will prove useful to generalize H by allowing a different
coupling constant for different scattering processes, but Equations (3)-(6)
will always be kept in mind as the more physical model that has to be
solved,
There are certain kinematic properties of the non interacting system
Hl = H2 = 0 which are of importance in interpreting the physical effects.
Imagine that there are M molecular sites in a length L == Ms, beyond which
the system is extended periodically, It is assumed in Equations (3)-(6) that
each site has just one spatial state for an electron to occupy. This is an
oversimplification, but it seems to be a good approximation for many
physical systems,(til The free Hamiltonian Ho may be diagonaJized by
Fourier transformation
1/1:0- = M~l/2 L e ,krnsa Lo- (7)

k
where the wave vectors k have values 27r1l/ L with II an integer. The energy
spectrum is (-2£ cos ks) and the ground state is a Fermi sea with two
electrons in each state of wave vector k less than the Fermi wave vector
kF == 27T'No/ L. The value of kF is related to the number Ne of electrons by
(8)
252 v. J. Emery
which may be expressed in terms of the reciprocal lattice vector G == 27T/ s

as
(9)
This relationship may be used to calculate Ne once kF and G are obtained

from x-ray and neutron scattering experiments. An important special case
is a half-filled band Ne = M for which G = 4kF .
1.4. Strong Coupling
A particularly simple picture of the possible states of the system may

be obtained in the large-lUI limit, in which Ho and H2 are treated as
perturbations. It is not necessarily assumed that large IUI corresponds to
any particular material, and for the present purpose it is merely a limit in
which the properties of one-dimensional conductors can easily be visu-
alized. The discussion will largely be qualitative at this stage and the
mathematical basis will be given in Section 3. The properties depend upon
the sign of V
1.4.1. U<O
When Ho = 0 = H 2 , the electrons occupy molecular sites in pairs to

take advantage of the on-site attraction. This is illustrated in Figure la,
where the crosses represent molecules and the arrows refer to up or down
spin electrons. The ground state is very degenerate because the energy
does not depend upon which sites are occupied.
CD W states occur in an extreme form when there is an intersite
repulsion VII but still no hopping. To minimize the energy, the pairs are
equally spaced, as shown in Figure Ib for a half-filled band. The charge
(0 ) )( )( )( )( )( )( )( )(
II II II II
(b) )1
II '" '"
II '" '"
II '" '"
II
)E
Figure 1 Some configurations of

electrons for the strong couphng
(c)
'" '" hmlt with U < 0 The arrows show
)( )( )( )( )( )1
II II the electron spm

density varies periodically from I to 0 in a distance LINe so, using Equa-

tion (8), the wave vector is just 2kF . This is a possible ground state because
the system is classical when Ho = o. More realistically, when Ho i' 0, the
picture is not so static and there is a much smaller modulation of the charge
density. Also, if Nel M is irrational, the electrons cannot be distributed so
neatly amongst the sites. However, it will be seen that the wave vector 2kF
always characterizes the cow.
Singlet superconductivity can arise when hopping (Ho) is included. The
electron pairs are bosons, bound in a singlet state, and it is possible that
they become superfluid (and hence superconducting since they are
charged) at low enough temperatures. Actually this does not happen in a
purely one-dimensional system since quantum mechanical fluctuations
prevent superconductivity even at zero temperature. When hopping
between chains is allowed, a transition does take place at a temperature Tc
determined partly by the value of the perpendicular hopping amplitude. In
general, Tc is not the same as the temperature lull kB' which characterizes
pair formation, a distinction which persists for weak coupling. In this
respect nearly one-dimensional conductors differ from isotropic three-
dimensional metals, where, in weak coupling, the pairs form and Bose-
condense at the same temperature.(4)
Triplet superconductivity will not occur because the electrons are
bound into singlet pairs before long-range triplet correlations can build up.
The excited states are of two kinds. Charge-density wave excitations
require the movement of pairs from site to site and they can be phonons or
plasmons according to whether the neutralizing background (ions or holes)
moves in phase or out of phase with the electrons. However, in a spin-wave
excitation, a spin is turned over and a pair must be broken since two
electrons of the same spin cannot occupy the same site. This costs an
energy lui, which appears as a gap in the spin-wave spectrum. A static
version of such an excitation is shown in Figure lc. In the same way, the
Pauli susceptibility X is proportional to e -iUi/ k B T at low temperatures since
only thermally broken pairs can respond to a weak magnetic field. The
spin-wave gap and exponential susceptibility are general features of one-
dimensional systems with attractive interactions and they occur for weak
coupling also. Later it will be shown that, in the field theory models, they
are related to the excitation of soliton-antisoliton pairs.
1.4.2. U>O
In this case the electrons avoid the strong repulsive on-site interaction
and the ground state has no doubly occupied sites. (The discussion will
assume N e :::; M but it is easily adapted to Ne > M.) When HI = 0 = H 2 ,
254 V. 1. Emery
(0) ~ ~ )E ~ )E )E )E )E
! I ! !
(b) ~
"
~ ~ ~ ~ ~ ~
Figure 2. some static configurations

(e) ~ )E )E )E )E )E
for the strong coupling limit with U>
)' ~
! !I ! I O. The arrows show the electron spin.
there is degeneracy because neither the spin directions nor the locations of
the unoccupied sites have any effect on the energy.
CDW states occur again in an extreme form when there is an intersite
repulsion V mn but not hopping. In contrast to U < 0, however, single
electrons rather than pairs are equally spaced so the period of the CDW is
halved and its wave vector is 4k F . This is shown in Figure 2a for a
quarter-filled band, which is very close to the situation in TTF-TCNQ. (No
particular significance is attached to the spin orientations at this stage.)
Once again, hopping makes the CDW weaker and less static. It also mixes
in doubly occupied sites and restores the Fermi sea, which may lead to an
additional 2kF periodicity by the mechanism described for U < O.
The SDW instability is most clearly visualized for a half-filled band,
which has exactly one electron per site so that only the spin degrees of
freedom have to be considered. Virtual hopping produces an effective
antiferromagnetic exchange interaction and the ground state has a modu-
lation of the spin density, which is illustrated in Figure 2b. (This picture is
correct if only the z components of the spins are coupled.) The period is
2L/ Ne and the wave vector is 2k F . For weaker coupling and a different
number of electrons, the state is more dynamic, but the wave vector is still
2k F .
Superconductivity will occur when Ne ¥- M and V is attractive. The
electrons form Cooper pairs(4) with either singlet(4) or triplet(S) spin. It
requires a coupling to the lattice or to collective modes of the electrons
themse!ves(7) to overcome the intersite Coulomb interactions and make
Vmn attractive. As for U < 0, hopping between the chains is required for
the phase transitions to take place.
The excited states are rather different in character from those
described for U < O. When Ne ¥- M, there is no difficulty in rearranging
spins and charges to produce an excited state but, for a half-filled band, the
CDWexcitations have an energy gap, for they require the double occu-
pancy of a site as shown in Figure 2c. The spin waves are known exactly for
this case and have no gap.
It is very useful to keep this overall picture in mind when following the
mathematical description of one-dimensional conductors since it gives a
concrete feeling for what is going on, even though, for small IVI, the states
are not so sharply delineated.
2. Spin less Fermions

2.1. Definition of the Continuum Limit
This section will be concerned with the theory of spinless fermions in

one dimension for which it is possible to find an exact solution in the
continuum limit s ~ O. It is interesting to consider this limit because dis-
tances are given by ps with p ~ 00 as s ~ 0 and thus the correlation
functions reflect the long-range behavior (large p) of the lattice system, and
this is just what is required for a discussion of long-range order.
The starting point is the lattice model introduced in Section 1.3 and,
initially, the continuum limit will be described for the free fermion Hamil-
tonian
(10)
which is the same as Ho in Equation (4) except that the spin label has been
omitted. The fermion creation and annihilation operators obey the anti-
commutation relations
(11)
and in the Heisenberg picture (with Ii = 1), the equations of motion are
(12)
which can readily be solved by Fourier transformation, using Equation (7),

to obtain
(13)
256 V.I. Emery
Thus the time evolution is determined by the cosine spectrum of the

discrete lattice.
In taking the continuum limit, it is necessary to take account of the fact
that, in Equation (12), the time derivative of fields on odd (even) sites is
related to the values of the fields themselves on even (odd) sites. This
property may be preserved by introducing two functions I/IE(X, t) and
I/Io(x, t) such that, as s ~ 0,
1/12n(t)~ i2n(2s)1/2I/1E(X, t)
(14)
1/12m+l(t)~ i2m+I(2s)I121/10(Y, t)
where nand m ~ 00 as s ~ 0 in such a way that 2ns ~ x and (2m + l)s ~ y.

The factors s 1/2 are introduced to turn Equations (11) into continuum
anticommutation relations with 8mn ~ 8(x - x'), and the phase factors
change (I/In+1 +I/In I) into in+I(l/In+I-I/In_l) which will become a derivative
in the continuum limit. Then if 2£s tends to a finite limit VF, Equation (12)
becomes
(15)
(16)
which is the two-component Dirac equation for massless particles. The

velocity VF will play the role of the Fermi velocity for the one-dimensional
electron gas, but it is analogous to the velocity of light in the Dirac
equation and, in this sense, the whole theory will turn out to be Lorentz
invariant. Equations (15) and (16) may be separated by defining
- 1/2
1/11 = (I/IE + 1/10)/2 (17)
Ji2=(I/IE-I/IO)/2 1/2 (18)
so that
aJiI aJiI
-=VF- (19)
at ax
aJi2 aJi2
-=-VF- (20)
at ax
The new fields Jil and Ji2 define fermions moving to the right (Jil) or left
(Ji2) with velocity VF. The energy spectra are ±v~ instead of 2£ cos ks as
found in the lattice model, and there is no lower cutoff so the energy of a
particle becomes -00 when k ~ +00. This implies that there is no ground
state and, as usual in Dirac theory, it requires the vacuum to have all
Theory of the One-DImensIOnal Electron Gas 257
negative energy states filled. This method of going between lattice and
continuum forms of the Dirac equation was introduced by Kogut and
Susskind.(8) It should be mentioned that it is not obvious that taking the
limit s ~ 0 in the equations of motion will give the same results as taking
that limit in correlation functions and, indeed, there are cases where it is
not true. For the present purpose Equations (19) and (20) define the free
fermion part of the model and then the question of interchange of limits will
not arise.
For a many-particle problem it is necessary to fill positive energy states
up to k = k p for right-going particles and k = -kp for left-going particles.
Then it is convenient to introduce new fields
(21)
(22)
which satisfy Equations (19) and (20) and the anticommutation relations,
so that the phase factors explicitly eliminated from 1/11 and 1/12 will no longer
appear in the manipulations of operators. This simplifies the discussion.
The presence of the time-dependent factors in Equations (21) and (22)
implies that vpkp(N I + N 2) has been subtracted from Ho (Nl and N2 are
number operators for right- and left-going particles) and its continuum
form is
(23)
In the absence of spin, the interaction HI does not occur, and intro-
ducing the definitions of HI and H2 into Equation (6) gives the continuum
limit in the form
(24)
apart from a multiple of the number operator. In Equation (24), p, =

1/1: (x )1/1, (x) and the couplings VI = V and Vz = 2 V have been labeled so
that their different roles in the calculation may be distinguished. The
continuum limit for spinless fermions given by the Hamiltonian Ho + H2 is
essentially the model proposed by Luttinger(9) for a one-dimensional
system, and it was shown by Mattis and Lieb(IO) that the partition function
can be evaluated exactly. Their method is the first step in introducing
boson representations of Fermion operators.
258 V. I. Emery
2.2. Boson Representations and the Free Energy
In general, it is expected that density operators commute with each

other because they are functions of the coordinates of the particles and not
their momenta. However, in a second-quantized theory, the commutation
relations may depend upon the form of the vacuum and they can be
different if the negative energy states are occupied or unoccupied. To see
how this comes about, consider first the right-going branch for which the
Fourier-transformed annihilation operators are defined by the continuum
version
alk=L ~I/2 f dxe Ikx I./I[X
() (25)
with the aid of Equation (14). They satisfy the commutation relations
(26)
The canonical transformation to operators with a vacuum containing filled
negative energy states is given by
iilk=aik. k<O
(27)
k>O
Suppose that the momenta k lie in a range -K ~ k ~ K where, eventually,
K will become infinite. Then the Fourier transform dxpI(x)exp(ipx) J
becomes
K~p t
Pip = L
k~~K
al.k+palk
K~p
(28)
Pl,~p = L
k~~K
a ikal.k+p
It is then straightforward to use Equation (26) to evaluate the commutator

of Pip and PI.~p· for p ~ pi> 0 and obtain
K~p+p' ~K+p'
[pp, p~p,] = L
k~K~p
a r.k+p~p,a Ik - L
k~~K
a i.k+p~p'al.k (29)
The right-hand side of this expression has annihilation operators standing

to the right of creation operators so it gives zero when acting upon the
vacuum state or, when K -7 00, upon any state with a finite number of
particles at large momenta. In this sense, the right-hand side of Equation
(29) becomes zero when K -700. On the other hand, it may be expressed in
terms of iik. using Equations (26) and (27) to obtain
K~p+p ~K+p' ~K+p'
[Pip, PI.~p'] =
k~K
L p
iii.k+p~p'iilk +
k~~K
L iir.kiil.k+p~p,-8pp' L
k~~K
(1) (30)
Theory of the One-Dimensional Electron Gas 259
Once again, as K ~ 00, the operator expressions on the right-hand side of

this equation give zero in the same sense as before, but the final c-number
summation tends to a finite limit -8 pp pL/21T since each k is of the form
21Tl!/ L, with v an integer. Then as K ~ 00 the commutation relation
becomes
(31)
A similar calculation shows that the same result is true for pi 2: P > °and
also that [Pip, PiP'] = 0, [P2p, P2p'] = 0, whereas
pL
[P2p, P2,-p'] = 21T 8pp' (32)
°
for p > 0, pi > 0. The change in sign occurs because, for the left-going
branch, it is the states with p > that are filled in the new vacuum. Thus it
can be seen that the change in vacuum has a drastic effect on the com-
mutation relations and that, apart from the numerical factor pL/21T, the
density operators are bosons. Indeed Pip and P2p may be combined into
one canonical boson operator bp by defining
p>o
(33)
( 21T )1/2 p<o

= -i IplL P2p,
and the combined commutation relations of the PIP become
(34)
The phase factors ±i in Equation (33) have been included in anticipation of

the requirements of Section 4. Physically b;
creates a density wave of
momentum p in the original Fermi gas and it is a boson for all p, in this
model.
The interaction part of H may be expressed immediately in terms of
the bp by substituting Equation (33) into the Fourier transform of Equation
(24) to obtain, apart from constants,
In this equation, the operators b :vb±p come from PipPi.-p whereas b ~ ~p

and its Hermitian conjugate arise from products such as P2PPI.-p involving
one left-going and one right-going density operator.
260 V. I. Emery
The representation of the kinetic energy requires more discussion. In

terms of the variables ak, Equation (23) may be written
(36)
and, using Equations (26), (28), and (36), it is straightforward to verify that
[Ho, Pip] = VFPPlp
(37)
[Ho, P2p] = -VFPP2p
It is important to notice that these equations, as well as Equation (34), are

true only if k ranges from -00 to +00 for, if there were a cutoff Ko in H o, all
momenta k > Ko would disappear from the left-hand side of Equations
(37), and the equations could not be correct since it is necessary to have
K ~ 00 in Equation (28). It may be that the results obtained by the boson-
commutation relations are a good approximation for a system with a large
but finite cutoff, but this should be verified in each particular case.
Generally speaking, if a cutoff K is included, any physical quantity that is
independent of K should be obtained correctly by the boson approach but,
overall, the limit K ~ 00 must be taken.
The significance of Equations (37) is that, if 10) is the ground state for
non interacting fermions, then PIPIO) is an eigenstate with excitation energy
±VFP, as can be seen by applying both sides of Equation (37) to 10). This is
a consequence of the linear dispersion relation used in Equation (36) for,
otherwise, it would be true only if P ~ O. Similarly it can be seen that other
eigenstates may be obtained by acting on 10) with products of the P,p' Once
a set of eigenstates and eigenvalues are known, it is always possible to
rewrite the Hamiltonian in terms of them. This may be accomplished here
by using Equation (33) to rewrite Equations (37) as
(38)
and then, from Equation (34) it is clear that the same equation is satisfied
by the boson representation of Ho:
HOB == VF L Iplb;b p +!J.E (39)

p
where !J.E is a c number to be determined in a moment. Thus states with

any number of charge-density waves are identical to states with the same
number of bosons created by the b;
and their excitation energies are given
exactly by HOB. It would be an enormous simplification if HOB could be
used in place of Ho for calculating the dynamical and thermodynamic
properties of the system, for both HOB and H2 are quadratic forms and
hence their sum can be diagonalized by canonical transformation. So far,
however, it is not clear that the boson states are complete, because excited
states of the ideal Fermi gas are composed of all allowed combinations and
numbers of particle-hole pairs obtained by applying operators such as
a;+kak to 10), and it has not been shown that they are all included in the set
of boson states. This question will be addressed in the next section by
showing that "': (x) can be expressed in terms of the b; and bp , in such a
way as to obtain the correct correlation functions.
Some feeling for the validity of using the boson representation may be
obtained by evaluating the constant ilE appearing in Equation (39). It
should merely be the difference of the ground-state energies and should
have no thermodynamic significance. This may be checked by equating th~
free energies Fo and FOB of the ideal Fermi gas and its boson represen-
tation. The boson free energy is just the usual expression
FOB = ilE + kBT 2: In{l + e ~13v~pl) (40)

p
where T is the temperature, kB is Boltzmann's constant and f3 = (kBT)~l.

Changing sums to integrals via ~p ~ (L/21T) Jdp, the integral may be
evaluated to give
1TL 2
FOB = ilE --(kBT) (41)
6VF
The fermion form of free energy may be evaluated in a similar way, using
the operators ilk defined in Equation (27), and remembering that FOB refers
to a fixed number of particles, since the density operators do not change
the number of fermions. The result is
(42)
Equating FOB and Fo gives ilE = (LvFk} )/21T, which is just the ground-
state energy of the free fermions, and is independent of temperature, as
required.
Assuming that HB = HOB + H2 does represent the interacting system,
the thermodynamic properties may be obtained by diagonalization, using a
canonical transformation. From Equations (35) and (39), HB may be
written
HB = 2: HB(p)+const (43)
p>O
where
(44)
262 V.I. Emery
and this quadratic form may be diagonalized by the transformation
e,Sb; e~,S = b; cosh <pp + b~p sinh <pp

(45)
e,sb~p e~,S = b; sinh <pp + b~p cosh <pp
with <pp = <p~p. Equations (45) are identical to the transformation used by
Bogoliubov(11) in the theory of a Bose gas and, by using the commutation
relations (34), it may be verified that
S=i I (b;b~p-bpb~p)<pp (46)

p>O
Using Equations (44) and (45), it can be seen that the off-diagonal terms of
the transformed Hamiltonian (those proportional to b;b~p and b~pbp)
vanish provided
=g (47)
with V, = s V./ 1T'VF. Then the full transformed Hamiltonian becomes
(48)
with
(49)
and
C = I [ep - (1 + Vl)PVF]+~E (50)

p>o
In this equation, the sum over p must be cut off for p larger than O(s ~1) to
obtain a finite result. This reflects the lattice origin of the theory and C is
O(S~l).
Thus the maIO effect of the interactions is to modify the energy of the
density waves, and the free energy may be obtained from Equation (40)
with vFlpl replaced by ep.
As frequently happens in these circumstances, there is a much more
significant change in the correlation functions but, in order to understand
how this comes about, it is necessary to introduce the boson represen-
tations of the !/I, (x ).
2.3. Boson Representations of Fermion Fields
The method of evaluating the Luttinger model partition function

described in Section 2.2 is essentially the one used by Mattis and Lieb,(lO)
but the justification is somewhat different and requires the introduction of
boson representations of the fermion fields 1/1, (x). The latter also are
important in themselves, since they allow us to evaluate correlation
functions in a rather direct way and to find special solutions of the model
with spin which will be introduced later. Boson representations of 1/1, (x)
were introduced first by Schotte and Schotte(12) in order to calculate x-ray
threshold transition rates. Later Schotte(13) and also Blume et al. (14) applied
them to Kondo's model of a magnetic impurity in a metal and derived an
expression for the partition function. These problems required a represen-
tation of 1/1, (x) at a single point, say x = 0, but it is a simple matter to allow
for arbitrary values of x, either directly or by a gauge transformation, as
required for the electron gas in one dimension. Following this route, the
correlation functions of Luttinger's model were evaluated by Luther and
Peschel,(15) and a solution of the model with spin was obtained for a
particular value of a coupling constant by Luther and Emery,(16) who also
pointed out the close analogy to the Kondo problem. In describing these
developments, the boson representations of 1/1, (x) will be used in a manner
that requires a more extensive justification than the original one given by
Schotte and SchotteY2 l For this reason, a rather detailed discussion will be
given and, in particular. the roles of cutoffs and limiting procedures will be
emphasized in order to avoid some of the confusion that has arisen in the
literature.
The form of 1/1. (x) is determined almost completely by requiring it to
have the correct commutation relations with the density operators. From
Equations (25), (26), and (28), it follows that
(51)
and, using Equation (33), this equation may be rewritten in terms of boson
operators b;:
[I/I,(x), b;] = i I~~ e,pxI/I,(x) (52)
where p > 0 for 1/1, (x) and p < 0 for 1/12 (x ). These equations are satisfied by
the boson form
(53)
where
(54)
264 V.I. Emery
and <P 2 (x) is given by a similar equation with the sum taken over k < O. The
factor exp(-alkl/2) has been included(15) in Equation (55) as a cutoff for
divergent integrals which will occur in intermediate steps of the cal-
culations. Physically, a is something like the lattice spacing and there will
be numerical factors 0'1/2 replacing SI/2 which occurs in Equation (14).
Results that are derived from Equations (53) and (54) are only correct if
a ~ 0, and this limit will always be taken. The coefficients A, in Equation
(53) are not determined by Equation (52) since it is homogeneous; but they
may be obtained from the fermion commutation relations amongst the
I/I,(x). From Equations (53) and (54),
I/Ii(X)I/II(X ' )=IA I 2 e- Ft e F ex p [27T L e-"\e'k(X-X')_l)]

1 (55)
L k>O k
where
2 1/2
F -- (--.!!.)
L
'L\' -ak/2
_e_
kl/2
b (
k e
,kx _
e
'kx')
(56)
k>O
On the right-hand side of Equation (55), the operators have been normally
ordered, i.e., annihilation operators collected into e F have been moved
to the right of creation operators which appear in e -Ft. This has been
accomplished by using the formulas derived in Appendix A. Changing
sums to integrals, the last factor on the right of Equation (55) gives
0'/[0' - i(x - x')] and, evaluating 1/11 (x')I/Ii (x) in the same way, it can be seen
that
As a ~ 0, the term in square brackets becomes 2m5(x - x'), so it is possible

to set x = x' in F and Ft and they vanish. Thus the right-hand side of
Equation (57) is a c number, and it is equal to 8(x - x') provided
27Ta1A112 = 1 (58)
In a similar way it can be shown that [1/11 (x), 1/11 (x ' )]+ = 0 and that 1/12 (x ) and
1/1; (x') satisfy fermion anticommutation relations. Finally it is necessary to
arrange for 1/11 (x) and I/Ir
(x) to anticommute with 1/12(X') and I/I~ (x'). This
may be accomplished by choosing
e'A N2 J
Al = (27Ta)I/2
,A 2 N,
(59)
A __ e_-;--=
2 - (27Ta )1/2
with
(60)
and (A 1 - A2 ) an odd multiple of 1T, as may be verified by direct substitution.

Thus, despite the fact that the t/Ji (x) are made up of boson operators, they
satisfy fermion anticommutation relations.
As an illustration of the use of this representation of the t/Ji (x), it is
instructive to evaluate a single-particle correlation function for the nonin-
teracting system:
(61)
where (... ) denotes thermal average for free fermions or bosons as
°
required, and t/Ji(X, t) is the time evolved operator eiHtt/Ji(x ) e -IHt. Since C O
I•
is a function of (x - x') and (t - t'), it is possible to set x' = 0, t' = and then,
from the fermion form of t/Ji(X, t),
C 1.0 (x,t)=L I L e 1k (X--v Ft)(1-fd+L- 1 L eik(X-vFt)fk (62)

k>O k<O
where A is the fermion occupation number, [exp({3vF!k!)+ 1]-1. The right-

hand side of Equation (62) may be evaluated by changing sums to integrals
to obtain
(63)
where
(64)
On the other hand, in the boson picture, Equations (53), (54), (59), and
(61) give
c..O(x, t) = - 12 ("",(x.t) e -i<l>,(O») (65)

1TCl
This is the thermal average of exponentials of linear combinations of

creation and annihilation operators and this is the general form that cor-
relation functions will take. It is the principal advantage of the boson
representation that such expressions may be evaluated quite simply as
266 V.I. Emery
described in Appendix A, and CI,o(x, t) is given by

1 1 2
CI,O(X, t) = 2mx exp -2{([O k
+(2nk + l)[cos k(x -UFt)-In) (67)
where nk is the boson occupation number. As L ~ 00, the sums in this

equation become integrals which may be evaluated in closed form to give
Equation (63) with K (g, a) replaced by
KCg a)=_i_lfC1-ig/{3VF+a/{3VFt (68)

, g + ia r2(1 + a/ (3VF)
The factor i comes from the commutator in Equation (66) and the rest
t
from the thermal average. Letting a ~ 0 and using Iruy = 1T/(Y sinh 1TY)
gives K (g, a) = K (g, 0) and the fermion result, Equation (64), is recovered
exactly. When (1/11 (x, t)I/II(O, 0)) is evaluated in the same way, the signs of
sin kg and cos kg in Equation (67) are changed and the integral diverges at
the lower limit to give zero for the correlation function, as it should.
This method can easily be extended to evaluate expectation values of
products of any number of 1/11 (x, t) and 1/1; (x', ('). Using Equations (53) and
(54) and the results of Appendix A, the various factors may be combined
into one exponential of the form exp[iLnC±)<PI(xn, tn )] and, as in Equation
(66), the correlation function becomes the exponential of a sum of pair
functions. This is precisely the product of pair functions which is obtained
directly from the initial fermion variables, and the correct overall sign is
guaranteed because the anticommutation relations are satisfied.
These results are sufficient to demonstrate that the boson represen-
tation is correct for the interacting system, at least in perturbation theory,
since, order by order, the correlation function or the partition function are
integrals of a product of free-particle correlation functions. This is where
the true value of the boson representation is realized, for it enables us to
obtain results for the interacting system with much greater ease than by
using the fermion variables directly.
Finally, another view of Equations (53) and (54) may be obtained by
differentiating <P, (x) and using Equations (33) to find
d::X) = (-1y+ 1 21TPI(X) (69)
so that <PI is the integral of a physical quantity, the density. In other words,
<1>1 is a non local quantity and it is essentially this feature that allows 1/11 (x) to
satisfy anticommutation relations. Also, from Equations (55), (56), (58),

and (69),
t 1 1
1/11 (X)I/II(X')= --2'
'TTl X - X
'+'la +Pt(X) (70)
for x = x'. The first term is undefined as x ~ x' and a ~ 0, but it has to be
there in order to give the 8 function in the anticommutation relation (57),
and it is the usual infinite term obtained from the product of two quantum
fields at the same point. There is no inconsistency, since the singular term
contributes to the p = 0 component of the Fourier transform of Pi(X) and
this does not occur in Equation (33). For p -c;C 0, the correct relation
between Pip and bp is obtained. Indeed Equations (53) and (54) could have
been used as a starting point for the discussion and the physical inter-
pretation of the bp deduced from that.
2.4. Correlation Functions of the Interacting System
In this section, the boson representations will be used to evaluate the

one-particle Green's function and various two-particle correlation
functions from which the physical properties of Luttinger's model may be
obtained. The interesting physical questions concern the existence of a
Fermi surface and the tendency towards superconducting or charge-density
wave order as the temperature is decreased, but the mathematical results
will also be of use in considering fermions with spin. The correlation
functions have also been evaluated by directly using fermion variables,(17)
but it will be seen that the boson representations, which were used for this
purpose first by Luther and Peschel,(lS) give a very simple method of
calculation.
The first quantity to be evaluated will be the retarded one-particle
Green's function
G~ = -iO(t - t')([ l/Ii(X, t), 1/1; (x', t')]+> (71)
where OCt) is the unit step function. Translational invariance implies that
Gf is a function of (x - x') and (t - t') so it is possible to set x' = 0, t' = 0
without loss of generality. It can be seen that Gf(x, t) contains the cor-
relation function C.(x, t) defined in Equation (61) and, for a noninteracting
system, using Equations (63) and (64) together with Equation (71),
G~(x, t)= -2iO(t) Re C.,o(x, t)
= {-iO(t)8(X - VFt), i= 1
(72)
-iO(t)8(x + VFt), i= 2
268 V. 1. Emery
In the presence of interactions it is necessary to diagonalize the Hamil-

tonian with the aid of the transformation (45) and, to obtain Gf(x, t), it is
merely necessary to find the effect on <l>i(X). From Equations (45) and (54),
~(x) = elsl(x) e -IS = kt (I:~) 1/2 1

cosh lPk [b e-ikx + bk e lkX ] e -"lkI/2
+ '\ (21T)1/2. h'" [bt e-ikx+b eikx] "lkl/2

k~O IklL sm 'f'k k k e
(73)
Similarly ;(x) is obtained by interchanging sinh lPk and cosh lPk. For
practical calculations, it is simplest to assume that Vp and lPP are constants
V and lP and then from Equations (47) and (49), Ep = v~lpl and sinh2lP ==
'Y, where
I
VF= VF [(1 + V-)2
1 -
V- 2]1/2
2 (74)
and
(75)
However, to obtain finite values of integrals, it is necessary to introduce a

cutoff. This does not spoil the boson representation and it is the one
allowed way of introducing a cutoff in using this method. Conceptually, the
simplest assumption is to take V p = V for IP1':5 Po and V = 0 for Ipi> Po,
but this can introduce awkward nonanalyticities, and the alternative form
sinh2lPp = 'Y e -rp , used by Luther and Peschel,(15) is much more convenient
and will be adopted here. Wherever possible, r will be set equal to zero and
then it need not be introduced into Equation (74), serving mainly to define
otherwise singular quantities.
Using Equation (73) and the corresponding one for <l>2(X), it can be
seen that, wlren lPk is a constant, the canonical transformation actually
mixes up the correlation functions C 1 •0 (x, t) and C 2 •0 (x, t). From this fact,
the Green's function for the interacting system can be written down
immediately in terms of that of the non interacting system, as in Equations
(63), (66)-(68). From Equation (73) the contribution from k >0 to Equa-
tion (66) will involve cosh2 lP P = 1 + sinh 2 lP P = 1 + 'Y e- rp • The first term (for
'Y = 0) will lead to C1.o(x, t) and the second term will give Ci,o (x, t) with a
replaced by r, since e- rp takes over the role of cutoff and a may be set
equal to zero. Similarly the k < 0 contribution from Equation (73) gives
CIo (x, t). In al\ of these functions, VF is replaced by v~. The net result is
that
G R( ) i8(t) { I 2 *
1 x,t =---Re K(x-vpt,O)[r K(x-vFt,r)K (x+v~t,r)r}
I
(76)
1T
where K (~, a) is given in Equation (64). To obtain Gf (x, t), it is necessary

to interchange x + v ~t and x - v ~t. It can be seen that, for large x, Gf (x, t)
falls off exponentially with a correlation length proportional to T- 1 ,
whereas, at T = 0, it becomes a power law:
G 1R (x, t) ~ {I
i8(t) 1 m , [ , r2
, ] Y} (77)
T~O x-vFt+iO (x-vFt+ir)(x+vFt-ir)
This result is exact for Luttinger's model but it is also correct asymptotic-
ally for a lattice model and is independent of the details of the cutoff. It is
characteristic of the boson method that, in contrast to the usual methods of
many-body theory, the correlation functions are obtained most naturally in
real space and time. The Fourier transformation of Gf(x, t) for T = 0 is
described by Luther and Peschel,(15) and, in particular, they find that
1 R r(wr)Y-l
--ImG 1 (q=O,w+iO)=, - , f(l-y)sin'TTy/2
'TT VF 2VF
+ -2r, ( -wr,)2
Y-l [[(1 - y) sin 'TTy]2 (78)
YVF 2VF
for w« v~/r. This is to be compared to the usual form Z8(w)+b(w)

obtained in three dimensions, where the 8 function comes from the
quasiparticle pole. It is clearly absent from Equation (78) and, as first
pointed out by Gutfreund and Schick,(18) there are no fermion quasiparti-
cles in Luttinger's model. In this sense, the fermions are "confined," and
this is true for all w when r ~ O. This is consistent with the conclusion that
the Hamiltonian may be expressed completely in terms of density waves. If
the right-hand side of Equation (77) is expanded in powers of y, every
order will contain powers of In w, which diverges as w ~ 0, and they must
be summed to obtain an accurate result. This feature is common to all
one-dimensional systems and is not a consequence of the continuum limit.
It is clear that selective summations, such as the random-phase approxi-
mation, do not give correct results in these circumstances.
The other interesting correlation functions are the 2kF susceptibility
(79)
describing the response of the system to a perturbation which transfers an

electron from the neighborhood of one fermi point to the other (i.e.,
right-going fermions to left-going fermions and vice versa) and the pairing
correlation function
(80)
270 V. 1. Emery
Physically, divergences in the Fourier transforms X R(w, 0) or pR (w, 0)

are a sign that a many-chain system may undergo a transition to a CDW
state or a superconductive state as described in Section 1.
The procedure for evaluating x R (x, t) and pR (x, t) is exactly the same
for the single-particle correlation functions. Introducing the boson
representation for tfJ,(x) gives an expression similar to that in Equation (66)
except that a product of four exponentials has to be evaluated and the
mixing of <1>1 (x) and <1>2 (x ) leads to factors cosh cPP ± sinh cPP= exp(±cPp)
which replace the separate cosh cPk and sinh cPk factors in Equation (73).
Then it is convenient to introduce a slightly different cutoff exp(±2cPp)~
1 + [exp( ± 2cPp ) - 1] exp( - rp), which again enables integrals to be evaluated
simply. As a result
X R(x,t ) =-2
0 (t) [2 ( *
2 1m r K x-vFt,r)K (x+v#,r)]",
I I
(81)
1T r
and
pR(X, t) = O(? Im[,zK(x -v~t, r)K*(x + v~t, r)]I//L (82)

21T r
where
-2</>
P, = e
=(~)1/2 (83)
l+g
and g is given by Equation (47). Using Equation (83) for p" it can be
seen that X R (x, t) and pR (x, t) differ only by the sign of g. The Fourier
transforms of x R and pR have been evaluated by Luther and Peschel.(15)
For T = 0, they find
0, Iwl<v~lql
1m xR(q, w) = (1T2Iv~)r2(1- y) sin 2 1T(1- y) (84)
x r(r2/4v~)(w2-c2q2W-l, Iwl>v~lql
Here, because of the phase factors which have been removed in Equations
(21) and (22), the wave vector q is actually measured relative to 2kF • Again
there is a power law singularity at the edge of the continuum, which can
only be found by summing all orders of perturbation theory. There is a
divergence for p, < 1 or g > 0 (corresponding to repulsive interactions)
and this will lead to a CDW instability for coupled chains at sufficiently low
temperatures. On the other hand, for attractive interactions, p, > 1 and
X(q, w) vanishes as w ~ 0, q ~ O. The pairing correlation function pR (q, w)
is obtained by changing p, to p, -I in Equation (83) and this is divergent
when g < O. Here the phase factors of Equation (20) cancel and q refers to
the total momentum of a pair. Physically, a divergence in P(q, w) implies a
tendency towards superconductive pairing when there is an appropriate
coupling between chains.
These are the qualitative features of the correlation function of
Luttinger's model. A more extensive discussion may be found in the
literatureys.17) However, to get closer to the real physical situation, it is
necessary to take account of the electron spin, and this produces a qualita-
tive change in the physics and in the mathematics that are required to
describe it. The problem becomes much more difficult because the Hamil-
tonian, now given by Equations (12)-(15), can no longer be written as a
quadratic form in boson variables, and the remaining sections are con-
cerned with the various limiting situations in which some progress can be
made. They have the common feature that the charge and spin degrees of
freedom separate, but in a way that usually does not allow an explicit
solution to be found as simply as for Luttinger's model.
3. Large "On-Site" Interaction
The simplest situation for fermions with spin occurs when the "on-
site" interaction U, defined in Equation (5), is the largest energy in the
problem, and it will be considered in some detail, for not only is it the basis
of the simple physical picture described in Section 1, but it may be relevant
for real materials in which there are strong Coulomb forcesY9) To be
complete, it is necessary to consider U < 0 also, and this has some
pedagogical value for, even if large negative U is unlikely to be found in
nature, a feeling for what happens at smaller negative values is useful,
because they may well arise when electrons de localize on large molecules
(thereby reducing the Coulomb force) and attractive couplings are induced
by molecular polarizability or other collective effects. (7)
The method of calculation will be to regard Ho and H2 of Equations
(4) and (6) as perturbations, and to work to the lowest nontrivial order. As
described in Section 1. the unperturbed states depend upon the sign of U.
3.1. Attractive Interaction

Here the unperturbed ground state has N el2 of the sites occupied by
two opposite-spin electrons. These are tightly bound pairs which replace
the Cooper pairs(4) occurring for weak coupling, and they are responsible
for superconductivity. The excited states have one or more pairs broken
and are separated in energy by multiples of lUI. When NeI2<M, some
272 v. J. Emery
sites are unoccupied and there is a substantial degeneracy because the

energy does not depend upon how they are distributed, but the inclusion of
Ho and H2 splits the levels into nonoverlapping bands of charge-density
waves in which pairs of electrons move from site to site. Interband tran-
sitions with one or more pairs broken correspond to spin-density wave
excitations. The discussion will be restricted to the lowest band, which is
sufficient for a calculation of the properties of the system at low tempera-
tures.
In first-order degenerate perturbation theory, H2 has to be diagonal-
ized in the space of paired states and, within that space, it is an effective
Hamiltonian from which the thermodynamic properties may be obtained.
But Ho breaks pairs and so has no first-order matrix elements. It must
therefore be calculated to second order, allowing virtual transitions into
the next band of states which have one pair broken. Suppose H2 = 0 for the
moment, and let the various degenerate ground states of HI be denoted by
la) with energy £ I = NU/2. The Schrodinger equation is
(85)
and dividing both sides by £ - HI and rearranging gives
1'1') = I la) (aIHol'l') +_P-Ho!'I') (86)

" £-£1 £-H 1
where P = 1 -let )(a I projects out of the unperturbed ground states. By
substitution, it can be seen that Equation (85) is equivalent to
(87)
where
(88)
and
a" = (a IHol'l')/(£ - £1) (89)
To first order, Equation (88) is
1'1',,)= la)+[P/(£-H1 )]Hola)
= la)(l/U)Hola) (90)
The last line follows because Ho breaks exactly one pair to give an excita-
tion energy -U and P is irrelevant because Hola) has no component in the
ground states. Then, substituting Equations (87) and (90) into Equation
(89) gives
(91)
which is a Schrodinger equation in the 10') subspace with effective Hamil-

tonian H~/ U. To ensure that Ho acts between the ground states, it is
necessary that, if the first application of Ho transfers an electron with spin (T
from site n to site m, then the second application of Ho either returns the
electron to its original site or transfers the electron with spin (-(T) from n
to m. Thus the effective Hamiltonian is
(92)
Since it is sufficient to work to first order in H 2 , the total effective Hamil-

tonian is
H'=H~ +H2 (93)
which may be used for temperatures T« IVI where the probability of
having a broken pair is very small.
A feeling for the physics represented by this Hamiltonian may be
obtained by introducing boson variables that are quite different from those
used in Section 2. Define
Pn = ~(Pn i + Pn t - 1) (94)
bn = !/Ini!/lnt (95)
(Tn = !(Pni - Pnt) (96)
where Pna = !/I~a!/lna as in Equation (6). Then
(97)
+ V L (2Pm + 1)(2Pm+l + l)+const
m
The operators (Tn commute with the bn and Pn and give zero when
applied to an empty site or a doubly occupied site, so they may be ignored
for the lowest band. The variables b m and Pn satisfy the commutation
relation
mf:-n
b~ =0 (98)
[Pm, b~l = b~8mn
which imply that they may be interpreted as boson operators and their
number operators, provided there is a hard-core interaction preventing
two or more particles from occupying a single site. The question of the
existence of superconductivity in the original system is now rephrased as
274 V. f. Emery
the existence of a condensation of the bosons and, as mentioned in Section

1, this is quite independent of the formation of pairs, a distinction that
persists for weak coupling, in contrast to the behavior of more isotropic
materials, where the BCS picture(4) takes over as the coupling is reduced.
In fact, when the correlation functions are evaluated, it will become clear
that quantum fluctuations prevent Bose condensation even in the ground
state of a one-dimensional system and the existence of an energy gap is a
rather local phenomenon related to bound pairs rather than to long-range
order.
3.2. RepuLsive Interaction
Here the hopping term Ho breaks the degeneracy in first order, and
although it is quite easy to deal with the translational motion, it is much
more difficult to incorporate the spin degrees of freedom. An exception is
the half-filled band, Ne = M for which electron hopping leads to doubly
occupied sites. The effective Hamiltonian for this case can be obtained
from Equation (97) by using the canonical transformation
(99)
which does not affect H o, but changes the sign of U in Equation (5), while
adding a multiple of the number operator. In the unperturbed state, sites
that contained a down-spin electron are now empty, whereas the other
sites are doubly occupied. Thus the path traced out for U < 0 may be
followed once again and only the physical significance of the variables is
changed-now the boson operators of Equations (94)-(96) represent spin-
density waves, and the energy gap immobilizes the charge-density degrees
of freedom at low temperatures. This interchange of the roles of charge-
and spin-density waves is not peculiar to strong coupling and appears in
the continuum limit to be considered in Section 4.
Since 'iuPmcr is unity for every site m, it follows that H2 is a constant in
lowest order and the effective Hamiltonian is given by Equation (97) with
V=O.
3.3. Correlation Functions
The bosons of Equation (94)-(98) are physically appealing, but the

hard-core interaction makes it difficult to evaluate the correlation
functions. A more useful representation may be obtained by noticing that
Equation (98) are the commutation relations of spin-half operators
s;", s;;" S:" with the identification

bm =S-;"
b~ =S-:;' (100)
Pm =S:"
where S!. = s;" + is:''. In this notation, the Hamiltonian may be rewritten
H' = - L (JxS;"S;"+! +IyS;;'S;;'+! +IzS:"S:"+d (101)

m
where
Ix =Iy
=4e 2 /IVI
(102)
Iz=-4e 2 /IVI-4V
with V = 0 for V> O. It is simplest to carry out calculations in the grand
canonical ensemble for the original problem, and since the number opera-
tor is essentially ~mS:", the chemical potential becomes an effective
magnetic field. The discussion here will be restricted to zero field, cor-
responding to a half-filled band.
As described in Section 2, the trend to various kinds of long-range
order is expressed in the correlation functions generated by t/I ~ut/lnu' or
t/lmut/lmu" When V < 0, the wave functions for the lowest band are linear
combinations of the 1'1',,) of Equation (90) and they, in turn, are linear
combinations of states with all sites occupied by pairs of electrons of
opposite spin or by pairs broken without spin-flip. Only t/I ~tt/lni' t/I ~~t/ln~ or
t/I~tt/l~L have matrix elements within this space, and all other combinations
connect to states separated from the ground state by an energy gap which
prevents divergences at low frequency. It is sufficient to use the bare
operators to generate correlation functions, since the admixture of broken
pairs in 1'1',,) is of order Ie/VI, so 1'1',,)=10') and the role of Ho and H2 is
merely to determine the coefficients an that appear in Equation (87). From
this and Equations (94), (95), and (100), it follows that only the operators
(103)
and
(104)
need be considered. In the pseudospin picture, for V < 0, a charge-density
wave state corresponds to antiferromagnetic order in the z direction and
superconductivity to ferromagnetic order in the x-y plane. The analogous
picture for V> 0 may be obtained by substituting Equations (99) into
276 V.I. Emery
Equations (103) and (104), leading to operators that generate longitudinal

and transverse spin-density waves, respectively.
For IJz / Jx 1:$ 1, the asymptotic forms of the correlation functions have
been obtained by Luther and Peschel,(20) who combined results known for
the lattice model with scaling laws derived for the continuum limit to find
that, if x = ms and T = 0,
(105)
(106)
Here, r is the cutoff introduced in Section 2 and

() = ~ - 1T -1 arcsin(Jz/ Jx ) (107)
For large x, the right-hand sides of Equations (105) and (106) are
proportional to x 1/8 and x 8 , respectively, and, since 0:$ ():$ 1 for the
range of Jz / Jx under consideration, it follows that there is no long-range
order in the ground state-a consequence of quantum fluctuations rather
than the classical thermal fluctuations described in Section 1. If there is an
appropriate coupling between the one-dimensional segments of the
system, the existence of a phase transition requires much less of the
correlation functions. Their Fourier transforms are proportional to wI'-
[with (f-l. = () -1 - 2, q = 2kF ) for Equation (l05) and (f-l. = () - 2, q = 0) for
Equation (l (6)]. and merely have to be sufficiently large at low
frequency,(21) which is true if f-l. <0. The resulting physical picture is
precisely the one set forth in Section 1.
A more detailed account of the strong coupling limit may be found in
the literature(21 23) In particular, a discussion of the region Jz < -Jx = -Jy
[or, by Equation (102) V> 0, U < 0] has been given by Fowler,(23) who
finds that the ground state has long-range charge-density wave order and
that there is a gap in the charge-density spectrum as well as the one in the
spin-wave spectrum discussed above. This situation may also be realized
for weaker coupling, as will be seen in Section 4.
4. Continuum Limit-Energy Gaps
In this section, the continuum limit will be introduced at the outset

with a view to obtaining the long-range behavior of the correlation
functions in much the same way as for spinless fermions. It turns out,
however, that the introduction of spin produces a number of significant
modifications. As might be expected from the discussion of the strong
Theory of the One-Dimenswnal Electron Gas 277
coupling limit, the spectra may now have energy gaps, and this is related to
the fact that the Hamiltonian is no longer quadratic in boson variables as it
was in Section 2. Moreover, the continuum limit does not exist for all
values of the coupling constant, since ultraviolet divergences may drive the
free energy to minus infinity. Fortunately the use of boson representations
is nicely supplemented by the renormalization group method (Section 5), in
that each works where the other fails, but it will be necessary to wait until
the end of Section 5 for a full discussion of the possible phases of the
system.
4.1. Separation of Charge and Spin Degrees of Freedom
If the original Hamiltonian of Equations (3}-(6) is rewritten in terms

of the right-going and left-going operators 1/110- and 1/120-. the boson vari-
ables that may then be introduced acquire a spin label, and a significant
simplification is achieved by introducing charge- and spin-density operators
C pt=(b pi
t +b pt )/2 1 / 2
j
(108)
in terms of which the problem will prove to separate into two independent
parts. First of all the quadratic, Luttinger model contribution HLM must
have the general form
SUII
+ L [
t
vsp(spsp + s-ps-p)+-
t t t
(sps-p +s~psp)
]
(109)
p>O 21T
;c
in which there are no products such as s p , involving one charge and one
spin variable, because the Hamiltonian should be invariant under reversal
of the spins which, by Equation (108) changes sp to - sp and leaves cp
unchanged. As a consequence, the charge- and spin-density waves are
Table 1 Relatwns between Couplmg Constants
QuantIty 'I:-ology" pIcture Extended Hubbard model
Vc VF + g4/ 7T VF +SU/27T + 2sV/ 7r

WI (2g2~gd/s U+6V
W.l 1:3/S U-2V
V, IF vF-sU/27r
UII gds U-2V
U.l 1:1/S U-2V
278 V.I. Emery
uncoupled in H LM . Table 1 shows the relationship between the notation of

Equation (109) (which reflects the differing roles of various terms in
determining the solution) and that of the original Hamiltonian. Also
included is the so-called "g-ology" notation, which has frequently been
used in this context.(25)
The other parts of the Hamiltonian which come from taking the
continuum limit are backward scattering
(110)
in which electrons of opposite spin cross the Fermi surface in opposite

directions, and umklapp scattering
Hu =sW"- f dxe4Ikp<!{!;i!{!i~!{!I~!{!lf+H.c. (111)
in which electrons of opposite spin cross the Fermi surface in the same
direction. In Equation (111), 4kF == 4kF - G and the momentum transfer is
absorbed completely by the lattice for electrons at the Fermi surface when
there is a half-filled band G = 4k F . In the absence of spin, Hu vanishes
[because !{!i(x) = 0] and, by permuting operators, the integrand in H BS may
be rewritten as PI(X)P2(X) and incorporated into the Luttinger model
Hamiltonian. [In fact this is why V 2 = 2 V rather than just V in Equation
(24).] Apart from the factor s, the notation for the coupling constants is
that of Luther and Emery(16) and of Emery et a1Y4) The relation to other
descriptions is given in Table 1.
In order to obtain a boson representation of H BS and H u , it is
necessary to add a spin label to Equations (53) and (54); then, using
Equation (108) it is found that
!{! ; tl/!;tl/!l ~1/!2i = (2 7T0' r 2 exp[ - 2 I/2 i (<PIs - <P 2s )] (112)
and
(113)
where <P,S and <P'C are obtained by substituting sp and C p , respectively, for bp
in Equation (54). From Equations (112) and (113) it is clear that backward
scattering affects only the spin-density waves and that umklapp scattering
affects only the charge-density waves, so the Hamiltonian completely
separates for the two kinds of collective mode.
4.2. Reduction to Spin less Fermions
The possibility of obtaining a solution now relies upon the way in

which the right-hand sides of Equations (112) and (113) behave under the
Theory of the One-DImensIOnal Electron Gas 279
canonical transformation (45) and (46)

els(<PlA -<P 2A ) e- ISo = e-"'(<p lA -<P 2A ) (114)
where A represents either the s or c subscripts. Then by choosing l/J =
(In 2)/2, it is possible to transform the right-hand sides of Equation (112)
to exp[-i(<P 1s -<P2s)1 and Equation (113) to exp[i(<plc -<P2c)], and from
Equation (53) they are boson representations of products of two spin less
fermion operators:
15. 1,+ t -IS (2 )-1 t
(115)
e 'I' lirjJ2~rjJUrjJ2i e = 1TlX rjJlsrjJ2s
and
(116)
Notice that the phase factors involving fermion number operators, which
must be introduced to ensure the full anticommutation relations as in
Equation (59), do not appear in Equations (115) and (116), as shown in
Appendix B.
If the canonical transformation is now applied to the remainder of the
Hamiltonian and the result is rewritten in terms of the spinless fermion
variables rjJlc and rjJ20 it is found that
(117)
where
Hs=i(10+30)J,L
II
dX(."t arjJls_. " t arjJ2S)
'1'1 s a 'I' 2s a
Vs 8 0 x x
(118)
and
with
U-11--'
_ SUII - sU~
u~=--
1TVs 1TVs
(120)
Tlfs"'ll - sW~
YYII=-' W~=--
1TVe 1TVs
280 V. 1. Emery
As shown in Section 2, the bare Fermi velocity is given by 2ES (where

E is the hopping amplitude) and, using the relations given in Table 1, the
lattice spacing cancels out of the variables in Equation (120). Then for
01. = 0 = W1., since Hs and He are quadratic forms in boson variables, the
energy levels are linear in the Fermi velocities and hence are proportional
to s. In order to define the continuum limit, this factor of s must be
absorbed either IOto the time and temperature (which multiply H in the
partition function and correlation functions) or into all of the coupling
constants. These alternatives are equivalent since one of the coupling
constants may be chosen as a unit of energy. On the other hand it will be
seen that the continuum limit in general requires that 01./2a and W1./2a
vary as powers of s, since the energy levels are not linear in these quan-
tities. Using Equations (120) and bearing in mind the factors of a (which
are proportional to s), this implies that V1. ~ 0 and W1. ~ 0 as s ~ 0, which
does not correspond to the initial extended Hubbard model or g-ology
picture, as can be seen from Table 1. It would make no difference if the
results were valid for small but finite s, which would allow V1. to be set
equal to VII, for the analytical form of the energy scale will turn out to
correspond to IV, / viil« I and IW 1./ wIII« 1. It is essential to bear this in
mind when making comparisons between the results of continuum theories
and those of other approaches which retain a discrete lattice.
The problem has been reduced to two sets of independent fermions,
representing the collective modes of the system. This transition back and
forth from fermions to bosons is a characteristic feature of one-dimen-
sional system~, and since the fermion operators are nonlinear functions of
the original boson collective mode variables, even a fairly simple solution
in the new representation constitutes a qUite elaborate solution of the
original problem This feature is made possible by the kinematic con-
straints of one-dimensional motion, but, at the same time, it is demanded
by the strong correlations which occur for the same reason.
4.3. Solution of the Spin less Fermion Problems
The arguments will be given in detail for the spin-wave Hamiltonian

and the results for the charge degrees of freedom may then be obtained by
substitution.
Two approaches have been used to obtain exact solutions. The first,
introduced by Luther and EmeryY6) is to choose a particular value of 011:
(121)
in order to remove the second term from Hs and make it a free fermion
Hamiltonian, which is solvable by Fourier transformation, using Equations
(25) and (36) to obtain
Then the canonical transformation
(123)
leads to
(124)
with
(125)
when (h satisfies
tan 2(h = 50~/8ka (126)
The transformation (27) should be made to ensure that, at T > 0, all
negative energy states are full and all positive energy states are empty.
Using Equations (124) and (125), it can be seen that, now, there is a gap
~ == Vs 10 ~1/2a in the spin-wave spectrum, and this will have important
consequences for the correlation functions. A finite gap requires 0 ~ pro-
portional to s in order to cancel the factor a, which pushes us to the
anisotropic limit V ~/ Vn = 0 as s ~ O. This implies that the energy gap of the
lattice model must vanish, which is quite usual when taking the continuum
limit for, otherwise, the correlation functions will decay exponentially
along the lattice and shrink to nothing in the continuum.
The charge-density Hamiltonian He may be diagonalized in exactly
the same way<24-26) except that 011 is replaced by (- WIi ) in the solubility
condition (121),0 1 is replaced by WJ. in Equations (125) and (126), and
the factors exp(±4ikFX) in Equation (119) prevent the gap from appearing
in the low-lying levels unless there is a half-filled band (kF = 0).
These results agree qualitatively with the calculation of Section 2.1.
States with one or more spinless fermions correspond to excitations into the
separated bands which occur for strong coupling, and, using Table 1, it can
be seen that the energy gap appears in the spin excitations when (U-
2 V) < 0 (attractive coupling if 21 VI < 1Vi) or in the charge-density modes
when (U + 6 V) > 0 (repulsive coupling). With the right combination of
on-site and near-neighbor coupling, these conditions can be satisfied
simultaneously to produce a gap in all collective modes. A comparison of
correlation functions will be made later.
282 V. I. Emery
Another method of obtaining the energy spectrum for all

< 0 or WI >
o has been given by Luther,(27) who used the fact that Hs and He are the
continuum limits of lattice-fermion Hamiltonians, which are related to a
spin-chain model, whose energy spectrum has been found by Johnson et
al.(28) The spin Hamiltonian is given by Equation (101) and, in the usual
way, the Jordon-Wigner transformation
(127)
may be used to rewrite it in terms of fermion variables !/1m. Then, taking the
continuum limit as described in Section 2 [except that !ii2 in Equation (18)
is replaced by - i!iiijleads to Equation (118) with
Vs - s ( Ix +Iy +2Iz)
-(lO+3UII)=- -
8 2 7T
7TVs -
~(6+5UII)=4sIz (128)
The "free" solutions of the spin chain, analogous to a particle-hole

excitation of H" have minimum energ/ 28 )
(129)
with
(130)
and
(131)
where (J is defined in Equation (107). The equation for Il is quoted for
small I, since the requirement Il- s may be satisfied by choosing /- s".I7T.
The details of the calculation may be found in Luther's paper.(27) To be
consistent with the canonical transformations carried out in obtaining Hs
and He. it is actually necessary to use the continuum form(2D) of (J:
2 (Ix+Iy-2Iz/7T)1/2
(132)
(J = Ix +J +6Jz/7T
y
since exponents depend upon the details of the model. This illustrates once
again the difference between taking the continuum limit in the Hamil-
tonian and in the equations of motion. According to Equations (121) and
(128), the solvable value of 011 corresponds to the x - y model, Jz = 0, and, at
this point, Equations (128)-(132) show that 11'1/.L = 2 and A = vs 01-1 a, or
twice the value obtained from Equation (125), as it should be. For other
values of 011, Equations (128) and (132) give (J = (J., where
/2
(J = ( 2+0)1
_ _II
(133)
s 2- 011
and, by Equations (129)-(131), A~ 0~2(H1), which gives the entire
dependence of the gap upon O~. This result has previously been derived(24)
from an exact homogeneity relation of the partition function, which will be
proved in Section 4.5. When 1011 1« 1, the scale becomes A~ O~IIOII, and
this result as well as the remaining dependence upon 011 may also be
derived from the renormalization group method as described in Section 5.
A qualitatively different effect appears when (J:::::; 1/2. The spin waves
form bound states of energy
(134)
with n = 0, 1, 2, ... , no:::::; (J-I - 1, implying that the spinless fermions form
bound states in this region, which corresponds to 011:::::; -6/5. Since the
continuum theory is Lorentz invariant, the momentum dependence of the
excitations is given by (E~+e)1/2 as in Equation (125) and this enables us
to obtain the free energy.
4.4. Correlation Functions
The principal difficulty in evaluating the correlation functions of the

model is that the various transformations used to diagonalize the Hamil-
tonian do not lead to expressions that may readily be evaluated. This is in
fact quite usual; however, in this case, it is possible to exploit the existence
of energy gaps to evaluate correlation functions for T» A and for T« A,
and this is sufficient for many purposes.
When T» A, the backward and umklapp scattering contributions to
Hs and He may be ignored and the Hamiltonian reduces to the Luttinger
model form HLM of Equation (109), whose correlation functions may be
evaluated by the method of Section 2, with the difference that the fermions
now have a spin. The correlation functions are of the form
x~ = - i(J(t) < [01(x, t), 0 . . ]) (135)

284 V. f. Emery
where the charge-density wave response is generated by the operator

Ocow(x) == [I/Ii r (x )1/11 r(x) + I/IL (x )1/1 1l (x )]/21 12 (136)
the transverse spin-density response by
OSDW(X) == 1/I;(T(x )1/11 .-(T(x) (137)
singlet pairing by
(138)
and triplet pairing by
(139)
Using the representation (53) and (54) (with an added spin label) and
transforming to charge- and spin-density waves with the aid of Equation
(108), the correlation functions separate and may be evaluated by follow-
ing the route that led to Equations (81) and (82) to find
O(t)
XA =-2-Im
{2
[r K(x-vet,r)K
I *(x+vet,r)]
I O±l/2
c
2rr r
x [rK(x - v~t, r)K*(x + v~t, r)]O:1/2} (140)
where v~ and Oe are obtained by setting VI = 0, V 2 = WII/2, and VF = Ve in
Equations (47), (74), and (83) and similarly for v~, Os:
Oe
(2- WII)1/2
=
- (141)
2+WII
_ /
2+°)1
0= (_ 11 2
(142)
s 2-V[[
with VII and W: defined in Equations (120). As in Equation (84), the

I
Fourier transform of XA is a power law at T = 0, and the overall exponents

corresponding to the operators in Equations (l36}-(139) are f..tA' where
f..tCDW = Oe + Os - 2
f..tSDW=Oe+0~1_2
T»A (143)
f..tsp=0~1+0,-2
f..tTP=O;1 +0~1-2
which specifies the appropriate combinations of O~I and 0:
1 for Equation
(140). Once again the phase factors removed in Equation (21) imply that
the wave vector q is measured relative to 2kF for the CDW and SDW
correlation functions. The factor ~ in the exponents (140) and the plus-or-
minus signs in Equation (140) are consequences of the factor rl/2 and the
minus sign in the transformation (108). Thus, at high temperatures, the

correlation functions have power law singularities, similar to those found
for spinless fermions.
The low-temperature properties are significantly affected by the exis-
tence of an energy gap. As in the strong coupling limit, it is plausible that a
spin-wave gap prevents a singularity in the response to a perturbation
which flips a spin, whereas for a perturbation that treats the spins sym-
metrically there is a power law singularity determined solely by the charge-
density degrees of freedomY6.29) To see that this is so, Equation (135) is
written in terms of the exact eigenstates 1m) and eigenvalues Em of the full
Hamiltonian:
XAR -iO(l)
=-- L e- /3£~ +If(£~- £)"(m IOA(x)n)(n
A I I"OA Im)+c.c. (144)
Z m.n
where Z is the partition function. The states separate into a charge-density

part together with a spin-density part whose contribution to 1m) in Equa-
tion (144) is forced by the factor exp( - (3Em) to be the ground state at low
temperatures, when there is an energy gap. The crucial question now is the
existence of a finite ground-state expectation value for the spin-wave part
of OA, because transitions to excited states will have a minimum energy d
and give an oscillating factor exp(idt) which will prevent the divergence of
the Fourier transform of X~ at low frequencies. This can be examined(29)
most easily for the solvable case 011 = - 6/5. In the boson representation,
the spin-wave parts of OSDW and OTP are proportional to exp[r I/2 i(ls +
<P2s)], whereas DclJW and Dsp contain exp[2 1/2i(<P 1s -<P2s)]' The trans-
formation exp(iS) used in Equation (114) removes the factor rl/2 in the
first of these factors and gives the boson representation of t/!1st/!2s. which has
a zero ground-state expectation value because it does not conserve particle
number. The second factor (related to OCDW and Osp) becomes
exp[i(ls -2s)/2], which is (crudely speaking) proportional to (t/!ist/!ls)l/2
and has a finite ground-state expectation value. In that case the spin
degrees of freedom are frozen and do not contribute to the power law
singularity, but they do not inhibit the charge-density waves. A similar
argument gives the consequences of a charge-density wave gap for W II > 0,
and a half-filled band, which removes the singularity in X:p and X~p,
The overall conclusions for 0: 1< 0 are then as follows.
4.4.1. Band Not Half-Filled
Here the umklapp scattering is ineffective because kF:;i. 0 and the

contribution from the charge-density waves is the same as at high
286 V.I. Emery
temperatures. Thus, X:DW and X:p are nonsingular whereas

f..LCDW=Oc }
T«a (145)
f..Lsp=O~1
4.4.2. Half-Filled Band, WI;> 0

H ere, XSDW,
R R
Xsp, an d XTP
R h
ave no d·Ivergence b ut XCDW
R .
IS a constant
at large x and t since both ground-state matrix elements are finite. This
implies that the ground state has long-range CDW order even for the
one-dimensional system, a consequence of the inhibition of quantum
fluctuations by the energy gaps.
Although this discussion was restricted to the solvable values of 011
and WII , it is plausible that the energy gaps established for other values of
011 ( < 0) and WII (> 0) have the same consequences, particularly as they are
in qualitative agreement with the strong coupling results.
The picture IS not yet complete because the methods used so far tell us
nothing about solutions for 011 > 0 or WII < 0 (unless umklapp scattering is
ineffective). The solvable values of the coupling constants have the
opposite signs and the spin-chain equivalence may not be used because, as
will be seen in Section 4.5.1, the continuum limit does not exist, at least as
it has been defined so far. The renormalization group method described in
Section 5 will enable us to fill in the missing information, and a further
discussion of the phases and correlation functions will be deferred until
then.
4.5. Relationship to Other Problems
At this point the development of the theory will be interrupted in

order to discuss the relation between the spinless fermion Hamiltonians
and a number of other problems-the sine-Gordon equation, classical
Coulomb gas, x-y model, and interface roughening.
4.5.1. Sine-Gordon Equation
Suppose the HLM in Equation (109) is diagonalized by canonical

transformation, as in Section 2. Then the boson form of the backward and
umklapp scattering terms (110) and (111) may be rewritten (for kF = 0)
H Bs = sO.L
-2-
27T a
f dx COS[(87TOs) 1/2
<l>s(x)] (146)
(147)
where Os> Oe are defined in Equations (141) and (142) and, using Equations
(112) and (113), <Ps = (<p 1s -<P2s)/(47T)1/ 2, <Pe = (<p 1e -<P2c)/(47T)112 are
canonical boson fields. [The factors (47T )112 come from comparison of
Equation (54) with the standard definition of a Bose field.] In this form, Hs
and He are sums of a free boson term and the nonlinear cosine contribu-
tions of Equations (146) or (147). They constitute sine-Gordon Hamil-
tonians, and the relationship with results that were obtained independently
and almost simultaneously in elementary particle physics becomes evident.
Coleman(30) started from the sine-Gordon equation and used perturbation
theory in 01- or W..L to derive the relationship to spinless fermions. This
stimulated Mandelstam(11) to rediscover the boson representation of
fermion operators as a simpler method of carrying out the calculation! (See
also the work of Mattis.(32))
The connection between these two lines of development was noticed
by Peschel(33) and by Heidenreich et al.,(34) and it had a number of
immediate implications. The sine-Gordon equation had been studied
extensively in both classical and quantum mechanics, and it was known to
have particlelike solutions called solitons, which were just the spinless
fermions of Section 4 and had a mass equal to the energy gap. The
existence of soliton bound states and the expression (134) for their spec-
trum had been derived by a semiclassical approximation.(35) It was clear
that much of the behavior of one-dimensional systems could be regarded as
soliton processes, and reviews of this point of view have been given by
Emer/ 3) and by Luther.(36)
One further result, which will be useful later and is most easily
proved(30) by applying the variational principle to the sine-Gordon equa-
tion, is that the ground state is unbounded below for 011 > 0 or Wil < 0 (Le.,
Os> 1, Oe> 1) when the continuum limit is taken in the manner described in
Section 4.3. This implies that the connection to the spin-chain problem
breaks down in this region, since it relies on the continuum limit, and
indeed, at present, we do not know the precise range of values of Os or Oe
for which the whole procedure is well defined. The same difficulty does not
occur if there is a finite bandwidth (as already known for the analogous
problem of magnetic impurities in metals(37)) but then it is necessary to
ascertain which of the results proved in the continuum limit are still
correct, or to use a quite different approach. This question will be taken up
in Section 5.
4.5.2. Classical Coulomb Gas

The equivalence to the classical Coulomb gas(24.38) may be obtained by
expanding the partition functions Zs for the spin-density waves in powers
of U1-' Every order is an integral of the expectation value of a power of
288 V. f. Emery
cos[(87TOs)I/2<Psl and therefore is a correlation function of Luttinger's
i 2m i
model. Then, using the method of Section 2, it is not difficult to obtain
ex: 1 [[; J2m L V
;(32m
Z,=ZO L - ( ,)2 - 4~2 ndx, ndv~tJ
m ~() m. 7Ta 0' 0 J
x exp( ¥' (- I t-qcpq)

p.lj
(148)
where v~ is the transformed Fermi velocity, Zo is the partition functiOl, for

U~ = 0 and Cpq == C(x p - Xq, tp - tq) with
e C(x.r)
= IK- (x-lVst,a
. I . I )1 8
)K- ( -X-lvst,a' (149)
and K(~, a) given by Equation (68). Equation (148) may now be given a
quite different interpretation by regarding v ~t as a distance and v'J3 as the
total length in that direction. Then Z, is the grand partition function of a
classical system with fugacity proportional to U~. The thermodynamic
limit requires (3 -> OJ when, using Equations (64) and (149),
C(x, t)= Os In{[x2+(v~t)2l/a2} (150)
becomes the Coulomb potential in two dimensions. Thus the ground-state

properties of the quantum system are related to the thermodynamic
behavior of the classical problem, and variation of the coupling constants
corresponds to a change in fugacity and temperature. Since the thermal
average of an odd number of fermion fields vanishes, only even powers of
U ~ occur and the Coulomb gas is neutral. There is a similar series for Zc.
This equivalence gives an appealing picture of the energy gap as a
screening length/ 1S ) but it adds little to our mathematical understanding of
the initial quantum system since much of what is known about the classical
Coulomb gas In two dimensions comes from renormalization group equa-
tions(39) that are one order lower than those to be obtained in Section 5. At
the same time, the problem is sufficiently delicate for intuition developed
from experience with the three-dimensional gas to be untrustworthy.
Nevertheless it is possible to give a simple derivation of the scaling rela-
tions of the continuum limit, and, in fact, they were first obtained in this
way.(24.'l7) They are a consequence of the fact that Zs is a homogeneous
function of degree zero in (3, L, and a since, if these quantities were divided
by a factor A, It would be possible to recover the original form of Zs by
t:
changing integratIon variables to x', = Ax, and = At,. Further, for small a,
all factors a and [;. appear in the combination [;~a2(8,-1). It follows that
the continuum limit may be taken by absorbing s into tl while requiring
[;J ~a2(1-8< and that. since the energy gap ~ is an intensive quantity and
Zs is a function of (3~, homogeneity implies ~ ~ a -lor, more completely,
~ ~ a -1 [ ; :/2(1 8,) These results are in complete agreement with Equations
(128}--(131) and the definition of the continuum limit given in Section 4.4.
Although the argument presented here relied on taking s ~ 0, it is also
possible to relate the fermion system with finite bandwidth to a classical
gas, but one with complicated many-body interactions, except at large
distances where they separate into two-body Coulomb forces. To the
extent that the long-range parts of the interactions determine the asymp-
totic behavior of correlation functions, it might be expected that general
properties do not depend sensitively upon the magnitude of the bandwidth,
but this does not necessarily mean that the continuum limit is harmless, and
it is necessary to be aware that taking the bandwidth to infinity can have
significant effects. This will become still more evident in Section 5.
4.5.3. Interface Roughening and the Classical x-y Model
Finally it is of interest to mention that the Coulomb gas is equivalent

to the discrete Gaussian model of interface roughening(40) for arbitrary
dimension, and dual to Villain's form of the classical x-y model in two
dimensions.(41-41) The rough phase of the former(40-44) and the low-
temperature phase of the latter(39.41) model constitute a line of critical
points and have analogs in the one-dimensional quantum system which
will be explored in Section 5.
It is important to keep in mind these relationships between completely
different physical systems, since progress in anyone field is likely to have a
substantial impact on all of the others.
5. Renormalization Group Method
The methods described in Section 4 have enabled us to obtain

solutions for VII < 0 and for arbitrary W II unless there is a half-filled band,
when they are restricted' to Wi! > O. In order to deal with other couplings
and to get a better appreciation of the significance of the continuum limit, it
is necessary to turn to some other approach, and in this section it will be
shown that the renormalization group technique, to a great extent, is able
to fill in the missing information, although by itself it is not able to provide
a complete solution. The idea is to keep a finite bandwidth or lattice
spacing and to vary it while making a compensating change of the coupling
constants so as to leave the partition function unaffected. In this way, one
physical problem may be scaled step-by-step onto another, which may be
easier to solve. The most complete discussions of the scaling equations and
their consequences have been given by Menyh<ird and S6lyom (45) and by
S6Iyom,(46) but their approach will not be followed here because it is
possible to derive more general equations by a simpler method and also to
290 V. 1. Emery
find more accurate solutions, making use of results that have been obtained
in the preceding sections.
5.1. Scaling Equations
The consequences of varying the lattice spacing in Equation (148) with

C(x, t) given by Equation (150), may be obtained at once from the scaling
law 0 1- ~ a 2(1 -0,) derived in Section 4.5. On defining I == In a and
differentiating, it is found that
dO~
di= -
2(1- ()s)U1- (151)
where the notation 01- has been used for the scaled value of 01-.
This is a complete renormalization group equation for Equation
(148) in which a has been set equal to zero wherever possible, but it
certainly is not the whole story for the original Fermi system with finite
bandwidth. The right-hand side of Equation (151) is linear in 01- because a
variation of a has produced a mere multiplicative change in the correlation
functions which may be absorbed order by order, and the coefficient is
correct because all such changes must be compensated by choice of 01- if a
well-defined continuum limit is to be obtained. When the omitted
dependence on lattice spacing is restored, a variation in s in one order must
be compensated by adjusting 01- in a different order, and this adds higher
powers of 01- to the right of Equation (151) as well as generating a similar
equation for 011.
For present purposes, it is sufficient to know the scaling equations to
third order in VI! and V1-, and they may be obtained from Equation (151) if
it is supplemented by simple symmetry arguments. First of all, the sign of
01- in Hs [Equation (118)] may be reversed by a gauge transformation
which has no physical effect and should therefore leave the scaling
equations unchanged. Then, to third order, using Equation (133) to expand
(}s in powers of Oil'
--=
dl
-
d U 1-
- 2
UII -
-U- 1- (UII+-+CU
-
4
2)
1-
(152)
Now, if Equation (109) were rewritten in terms of the original fermion

variables, the VII term would differ from H BS in Equation (110) only in that
all fields would have the same spin and, wherever both interactions play the
same role, they will give the same numerical coefficients, so C = 1/4 in
Equation (152). The equation for 011 may be written down immediately
from the requirements that (i) it is even in 01-, (ii) there is no
renormalization when V_ = 0, and (iii) it becomes identical to the equation
for V1- when VII = V-,-. The latter follows because spin rotational invariance
implies that, physically, VII = V1- (see Table 1) and this symmetry is not
spoiled by renormalization. Thus, to third order, the scaling equations are
_ -2 -2
dU ee _ (- UII + U 1-)
Tt= - U ee U II + 4 (153)
dVI = -V 2
dl 1-
(1+ VII)
2
(154)
In the usual way, the equivalent equations for the charge-density waves
may be obtained by substituting (- WII, W1-) for (VII, V1-), although there is
renormalization only for a half-filled band. The advantage of the notation
is apparent for, not only does it allow an enormous simplification of the
derivation, but also the separation of the scaling equations into
independent sets comes about quite naturally, whereas, in other
. . (45-47)..
defimtIons, thIS IS apt not to be the case.
5.2. Trajectories and Energy Scales
The scaling trajectories may be obtained by eliminating I from

Equations (153) and (154) and solving to obtain
_2 _2
U I - U 1-
- =a (155)
2+UI
where a is a constant. This equation describes families of hyperbolas
displaced along the VII axis, as illustrated in Figure 3. The low-order scaling
equations may be used to solve the problem in region I, in which VII ~ 1V 1-1
for, according to Equation (153), dIV1-ldl-sO, so IV1-1 will continue to
decrease until it meets the "fixed line"
V~ =0 (156)
for which the right-hand sides of Equations (153) and (154) vanish and
II II
Figure 3. Scaling traJectones in the

(VII' V.L) plane. In region I (VII "" 1V.L I) I
the renormalization stops at the fixed
line V.L = 0 whereas in regions II and
III it goes out to strong coupling. ------'-------'''--------'----- 0 II
292 v. J. Emery
renormalization stops. The asymptotic value of VII is VR where,

substituting the initial and final coupling constants into Equation (155)
(157)
In this case, an initially weak coupling problem scales to still weaker

coupling, so the trajectories stay within the region of validity of Equations
(153) and (154), and the asymptotic Hamiltonian is a Luttinger model
(011 = OR, O~ = 0), which is solvable. The correlation functions and their
physical consequences will be given in Section 5.3.
The fixed line (156) is a property of the exact scaling equations, since
there is no renormalization when V1- = 0, and it corresponds to the critical
·
Ime common to t he x-y mo d e1(39,41) an d t he mter . f ace roug henmg
.
models,(40,44) for which the correlation functions are power laws and there
is no long-range order.
On the other hand, in regions II and III of Figure 3, VII and V1- scale
towards strong coupling and the behavior is governed by the "fixed point"
(158)
but this is outside the range of validity of the scaling equations. There is a
fixed point at infinitely strong coupling (corresponding to infinite
temperature in the Coulomb gas) but the approximation has placed it at the
finite value (158) giving the impression of a critical fixed point. The actual
"phase transition," or change in ground state of the quantum system, takes
place when the coupling constant values cross the line 011 = IOJ This is
strongly suggested by the general features of Figure 3, with one state
corresponding to a flow to strong coupling and the other to a flow towards
V1- = 0, and it is completely consistent with the results of Sections 3 and 4,
where it was found that, for particular locations in regions II and III of
Figure 3, there is an energy gap that is essentiaIly the finite coherence
length of the disordered phase. Specifically, this was shown for the
continuum limit when IV1-I« - VII and for strong coupling when VII = V1- <
0. Also, at the end of Section 3, it was pointed out that, in the strong
°
coupling limit with V < 0, a gap appears in the charge-density wave
spectrum when V passes through and, using Table 1, this is the same as
crossing the isotropic line from region I to region II in the ( - WI W1-) plane.
Although it is necessary to consider all values of WII and W1-' the
situation is somewhat simpler for the spin-density waves because,
physicaIly, VII = V., so that scaling proceeds along the isotropic lines
VII = ± 1V1-1 and the transition takes place at VII = 0. Also, since the
°
continuum limit requires that V ~ ~ and W~ ~ 0, it will give a transition at
UI = 0 or WI = 0, which is consistent with the explicit solution found in

Section 4.
There remains the problem of finding a solution for an arbitrary point
in regions II and III. One possibility is to scale onto the strong coupling
limit and then to use the method of Section 3, but this must be done
numerically. Another that has been consl·d ere d(16).IS to seaIe onto t he
solvable value VII = --6/5, but this gives no more than a qualitative
indication of the existence of a gap for several reasons-Equations (153)
and (154) are not accurate at such a large value of IUIII, the scaling
procedure generates many-body forces that are not likely to be negligible,
and, finally, UJ. becomes large, which is incompatible with the continuum
limit. Nevertheless, all of the evidence points to the existence of energy
gaps in regions II and III or effective Luttinger models in region I and, as
argued in Section 4.4, this is sufficient to determine the correlation
functions at temperatures that are much higher than or much lower than
the energy gap.
If this information were supplemented by a determination of the
crossover temperature, the picture of the physical behavior of the one-
dimensional electron gas would be reasonably complete. At weak coupling,
the third-order scaling equations are sufficient for this purpose. Using
Equation (155) to eliminate U. from Equation (154) and integrating gives
f Oil
_ dU{(I+U/2)[U 2 -a(2+U)Jr 1 =ln(a/&)
UII
(159)
where & is the scaled value of a. If the indefinite integral on the left of this
equation is denoted by cfJ (U), it can be seen that UII is a function of
a-I exp[-cp(UII»), which gives the dependence of physical quantities upon
the original parameters of the problem once they have been evaluated in
terms of UII • In other words, it defines the energy scale. The integral can be
evaluated in closed form for arbitrary a but there are two situations
of particular interest. For the spin-density waves, the physical value is
a = 0 [or UI = OJ. in Equation (155)], which leads to cfJ(U) =
_U- 1 _ (lnIUI)/2 + O(U) and an energy scale
Es = A ivI 11/2 e O'I ' , (160)

a
This result is identical(24) to that obtained from exact solutions of the
Hubbard model,(48) and a slightly more detailed analysis enables the
coefficient A to be obtained. (49) The other interesting region is IVJ.I« IVIII«
1, which may be compared to the solutions in the continuum limit. In this
case,
(161)
294 V.1. Emery
The exponent of 01- is the same as the first two terms in the expansion of
(1- Osr 1 /2 and, to this order, it agrees with the form of the continuum
limit given by Equations (129) and (131) or by scaling the Coulomb gas
free energy. However, the remaining dependence on 101 1in Equation (161)
is not the same as in Equation (129) for small I0 1 1, and it is clear that taking
the continuum limit as prescribed in Section 4 has nontrivial consequences.
The higher orders in IV1-1 which were introduced in going from Equation
(151) to Equations (153) and (154) affect the energy scale even when 101-1
is small. In considering the significance of this result, it is necessary to bear
in mind that it is not known if the continuum limit exists for all negative
values of IVIII and, whenever it does not exist, the spin-chain solution is not
directly applicable.
5.3. Low- Temperature Properties
We are now in a position to evaluate high- and low-temperature

°
correlation functions for Oil> and Wil < 0, filling in the gaps that were left
at the end of the discussion of the continuum limit. The physical properties
may be inferred from the correlation functions, which have the power-law
behavior of Luttinger's model, modified by the existence of an energy gap
and renormalized exponents. A different approach was used by S6Iyom,(46)
who wrote down scaling equations for the correlation functions and evalu-
ated them entirely within the renormalization group method, without
recourse to the exact solution of Luttinger's model. This has the disad-
vantage that one is forced to rely on the fixed point (158), which is a fiction
and does not reveal the energy gap. Furthermore, the exponents cannot be
found to a higher accuracy than the scaling equations, and in this respect
are clearly less satisfactory. However, the scaling equations for the cor-
relation functions do expose logarithmic corrections, (46) which usually may
be ignored (and will be ignored here) but become important when the
exponents vanish.
The correlation functions may be specified by giving their exponents as
in Equations (141 )-(143) and (145). They are listed in Table 2 for high and
low temperatures for a band that is not half-filled, and in Table 3 for low tem-
peratures and a half-filled band. The functions OcR and OsR are obtained by
replacing WII and 011 by WR and OR in Equations (141) and (142), respec-
°
tively. For physical systems, 011 = 01- and OR = so OsR = O;~ = 1. Where
no exponent is quoted, it is implied that the power law singularity has been
removed by an energy gap. In this connection, notice that Table 3 has no
CDW entry for WI >-IW1-1 and 0 1 2:101-1, whereas the argument of
Section 4 leads to an exponent of OsR - 2. The reasons for this conclusion
are given in Appendix C.
Table 2. Correlation Function Exponents: Band Not Half-Filled
Low temperature
T«t:.
Correlation High temperature
function T»t:. UII<IU.c1 UII2:I U.c1
CDW Oc+ O,-2 Oc -2 Oc+O,R -2

SDW Oc+0~1-2 Oc +(J~~ -2
SP 0~1 +0,-2 0~1 -2 0~1 +O,R-2
TP 0~1+0~1_2 0;1 +(J~J-2
When q = 2kF for the density waves or q = 0 for pairing states, the
correlation functions vary as w'-', where f.L is tabulated. Negative values of
f.L imply that long-range correlations will develop at low temperatures, as
shown in Figure 4 for a band that is not half-filled and Figure 5 for a
half-filled band. Both figures assume the physical condition 011 = O.c.
Figure 4 uses WII and 011 = 01. as variables but, since 1W1.I is significant for a
half-filled band, it is more appropriate to display WII and IW1.I sgn(OIl) in
Figure 5. The "phase diagrams" in terms of the variables of the original
models may be obtained with the aid of Table 1.
It is interesting to compare Figure 5 to the conclusions of the strong U
limit which, using Table 1, corresponds to the neighborhood of the iso-
tropic lines in the first and third quadrants. Then U > 0 gives SOW cor-
relations and U < 0 gives COW correlations if V> 0, or a combinations of
COW and SP correlations when V < O. This is in agreement with the
discussion at the end of Section 3.
An issue of considerable practical interest is the possibility of making a
high-temperature superconductor out of organic metals. Since coupling
between chains commonly favors a charge-density wave state which creates
a single-particle gap and destroys superconductivity, it is desirable to have
a system that inclines to pairing but not COW correlations. Figure 4 and 5
both show that O!l> 0 and Wil < 0 is the most favorable case, and according
Table 3. Correlation Function Exponents for a Half-Filled Band at Low

Temperatures
ltIl:s-IW.c1 WlI>-IW.c1
Correlation
function UII<IU.Li UI 2:I U.L1 UII<IU.c1 UI 2:I U.c1
CDW OcR -2 OcR +O,R-2 -2

SDW OcR +(J~~ - 2 O~~ -2
SP OcJ -2 o;J +(J,R - 2
TP 0~J+0~J-2
296 V. 1. Emery
SP COW
TP sow
-6/5 6/5
- - - - - - - . - - - - - + - - - - - - . - - - - WI!
SP cow cow cow

SP SP
Figure 4. "Phase diagram" in the (WII , VII) plane for a not half-filled band, showing regions in
which the various correlation functions diverge as w -+ 0 and T -+ O.
S DW
SP
TP
WI!
/ COW
SP I SP
/
/
/
/
/
/
/
/ C OW
/
/
Figure 5. "Phase diagram" in the ( WI , IWJ.I sgn (VII)) plane for a half-filled band showing
regions in which the various correlation functions diverge as w -+ 0 and T -+ O. The dashed line
is WR = -6/5.
to Table 1, this requires U > 0, V < - U/6. that is, a short-range repulsion
and a longer-range attraction. This interaction, which also is characteristic
of superfluid 3 He, (5) may be achieved if the Coulomb force is offset by
exchange of phonons or electronic collective modes at long distances, but it
does require a fairly strong attraction if U is not reduced by molecular
polarization.
5.4. Four-Particle Functions
An important feature of the boson representations is that it is not too

difficult to evaluate many-particle correlation functions, and the arguments
given in Section 2 show that the method is quite generally valid for this
purpose. Indeed it has already been used in Section 4.5 to derive the
Coulomb gas partition function. Up to this point, the discussion of physical
properties has concentrated upon two-particle correlations, but x-ray scat-
tering experiments(50'1) appear to have shown signs of divergences in
certain four-particle functions,(52) and therefore it is worthwhile showing
how this comes about.
Experimentally, the existence of a charge-density or spin-density wave
state is most clearly seen in x-ray or neutron scattering experiments which
usually see directly the lattice effects produced by the electron-phonon
coupling
Hep = f dx 4> (x )[g2 e21kpXOcow + g4 e41kpX04] + H.c. (162)
where OCDW is given in Equation (136),

04(X) = t/J;j (x )t/J;t (x )t/Ju(x )t/Jl j(x) (163)
and 4> (x ) is the phonon field. The factors exp(2ikFx) and exp(4ikFx) have
been exposed by the transformations (21) and (22). Physically, Hep
originates in the modulation of the electron hopping Ho and Coulomb
interaction H2 which occurs when a phonon produces a local variation of
the lattice spacing. Only large momentum transfers have been retained, in
order to focus on the processes in which one or two particles are transfer-
red across the Fermi surface, giving rise to the singular effects.
The electrons and phonons form a coupled system and, in reality,
neither one can be treated in isolation from the other, but space does not
permit an adequate discussion of this problem, and, for present purposes,
Equation (162) is merely introduced to indicate that the lattice couples to
two-particle and four-particle operators in such a way that a tendency to
form long-range electronic correlations will produce a measurable effect on
the lattice. It is clear from Equation (162) that OCDW and 0 4 have their
singular behavior at q = 2kF and q = 4k F, respectively, reflecting the
298 V. f. Emery
change in momentum when two or four electrons are scattered across the
Fermi surface. The momentum balance has to be taken up by the phonons
and they will respond at the corresponding wave vectors. The operator 6 4
has already appeared in the umklapp scattering term given in Equation
(111), and the boson representation (113) involves only the charge-density
wave operators. It is then straightforward to show, as before, that the
corresponding correlation function behaves as
R 48-2
X4 ~w c (164)
for q = 4k F . From the definition (141) of ()o it can be seen that xf diverges
when w ~ 0, provided WI > 6/5, which requires a fairly strong repulsive
interaction. The low-temperature state which occurs when there are cou-
pled chains is a correlated state of charge-density waves mixed with a
lattice distortion, and it is much more complicated than states that
normally are envisaged for more isotropic systems.
For a half-filled band, Equation (164) gives only the high-temperature
behavior, and the normalization group argument indicates that, at low
temperatures, if WI,::; -I W1-1, WI must be replaced by WR , otherwise there
is a gap that freezes out the charge-density waves and gives an exponent of
(-2).
It is also interesting to compare the exponents of X~DW and xf in
order to see which one dominates at any particular temperature. For
repulsive interactions and the physical situation 011 = 01-, Table 2 shows
that X~DW has an exponent which crosses over from ()e + ()s - 2 to Be - 1 as
the temperature decreases, and the CDW response becomes more diver-
gent. At low temperatures, X~DW is more divergent than X: when WII <
8/5, and there can be a crossover from a dominant 4kF response at high
temperatures to a dominant 2kF response at low temperatures, provided
011 and hence (), is large enough. It is possible that this mechanism(52)
accounts for the observed x-ray scattering(50,51) at 4kF observed in TTF-
TCNQ and, if so, it indicates that the electron-electron coupling is repul-
sive and fairly strong.
The transformation (114) changes the four-particle part of Rep into an
interaction between a phonon and two spinless fermions and therefore it
should be possible to construct essentially the same theory to describe the
interplay between the lattice and either kind of correlated state. (52)
There are divergences in other four-particle correlation functions,(53)
but, so far, no experimental effects have been found that might be
attributable to them.
The dynamical role of phonons is not the only significant topic that has
been omitted from the discussion. The production of phase transitions by
interchain coupling, properties of the ordered phases, effects of impurities,
intramolecular vihrations, and transport properties are important subjects
that have been omitted, although some of them are discussed in other
chapters of this volume. I am aware that I have not made reference to
many important contributions to the subject, but my objective was not to
write a comprehensive review but rather to give an integrated account of
one area in which there has been some progress in the past few years, and I
apologize to those authors whose work was not mentioned.
ACKNOWLEDGMENTS
I am grateful to Dr. J. N. Fields for a careful reading of the manuscript.

This research was supported by the Division of Basic Energy Sciences,
Department of Energy. under contract No. EY -76-C-02-0016.
Appendix A: Some Results That Are Useful for Working with

Boson Representations
There are two results that are extremely useful in evaluating cor-
relation functions and commutators when the boson representations of
fermion fields are used:
(1) If the commutator [A, B] commutes with A and B, then
(A.I)
(2) If! is any linear form in boson operators, its thermal average for a
free boson distribution is given by
(exp(/) = exp(~(l») (A.2)
The first of these results may be proved by introducing a time-ordering
label and writing
e A+B = e AT exp [11 dt e ~AtB e At] (A.3)

()
Then, since [A, B] commutes with A, it follows that e ~AtB eAt =

B - t[A, B] and, substituting this into Equation (A.3), using [B, [A, Bll =
0, gives Equation (A. 1).
Equation (A.2) may be proved by diagonalizing the free-boson
Hamiltonian and deriving the result for each mode separately. The general
form is
Ab++A*b)=
( ~{3wbtb e Ab++A*b)
T re
(
e - Tr(e {3wb+b)
300 V. 1. Emery
To obtain the last line, the trace has been evaluated in eigenstates of b t b,
Equation (A. 1) has been used to separate exp(Ab t + A*b) into a product
of exponentials and the boson commutation relations have helped to
rearrange «bt)'{bn as (n!/(n-r)!) where n==btb. If the summation is
rearranged so that
OCn 00 00
L L a nr ~ L L a nr
n~Or~O r~On~r+l
then the summation over n can be carried out to give
= exp[!IA 1\2n + 1)] (A.5)
where n == (el3 w _1)-1 is the Bose occupation factor. Now since

«Ab t +A *b)2) = IA 12«b tb +bb t))
= IAI2«2b tb + 1)) (A.6)
Equation (A.5) is just
(eAbt+A *b) = exp[!«Ab t + A*b )2)] (A.7)
Then taking the product over all modes gives Equation (A.3).
Appendix B: Anticommutation of Different Fermion Fields
This appendix will consider the method of arranging for different

fermion variables to anticommute and its consequences for the trans-
formations used in Section 4. According to the discussion of Section 2, the
boson representation ensures that !/lUI and !/I1<7' satisfy fermion anticom-
mutation relations for i = j and (J' = (J", but, for i r' j or (J' r' (J", it is necessary
to include additional phase factors involving number operators as in Equa-
tions (59). To specify the convention for operators with spin, the operators
will be labeled !/Im(x) with m = 1, 2, 3, 4 denoting !/lIt, !/I2j, !/IH, !/I2~ in that
order. Then the phase factor for !/1m is taken to be exp[{i1T/2)(L~~I£mnNn)],
where Emn is an antisymmetric matrix with Emn = 1 for n > m. Then, com-
muting !/Ip!/lq for p r' q gives a factor exp(i1TEpq) = 1 as required. Inserting this
convention into the backward scattering operator (112), the phase factor is
exp[{i1T/2)(Nu + N2t - NJi - N2j)] whereas, for the umklapp scattering
operator (111), it is exp[(i1T/2)(Nlt + N2t + NI~ + N2~)]' Since both of these
exponents commute with H and all answers depend only on the modulus of
the coupling constant, the phase factors have been omitted from Equations
(112) and (113) and, by a similar argument, from Equations (115) and
(116).
Appendix C: Charge-Density Wave Gap and CDW Correlations
The arguments of Sections 4 and 5, if followed literally, would lead

to a COW exponent of OsR-2 for WII > -IW1-1 and On 2: 101-1 when there is a
half-filled band, and this would allow the formation of long-range COW
correlations despite the presence of a gap in the charge-density spectrum,
in qualitative disagreement with the conclusions for the strong coupling
limit.
This question may be examined in another way for the Hubbard
model, which has U, = U ~ = ~I = W 1- = sU (see Table 1) and falls in the
region of consideration for U > O. Since OR = 0 in this case, the straight-
forward argument gives a COW exponent of -1. On the other hand, for a
half-filled band, the transformation
(C.l)
has the effect of reversing the sign of U in Equations (4) and (5) without
changing the number of particles. The same transformation effects the
following interchanges of correlation functions:
SP~transverse SOW
(C.2)
COW~longitudinal SOW
Table 3, as written, satisfies this symmetry since OcR = OsR = 1 for the
Hubbard model, but it would not have done so if the original naive COW
exponent for U > 0 had been used.
Physically it is reasonable that, if an SOW gap prevents an SOW
power law for U < 0, then a COW gap should have the same effect on the
COW correlations when U > 0, and this conclusion should not be special to
the Hubbard model. In the strong coupling limit, spin symmetry is satisfied
by having long-range correlations in the up-spins and d.Jwn-spins
separately, but they are exactly out of phase so that one compensates the
effect that the other produces on the charge density. For weaker coupling,
it is easier in general for COW and SOW order to coexist, but there is little
doubt that the argument for the effect of a gap is less fragile for the SOW
302 V.I. Emery
correlations (with U < 0) than for the COW correlations (with U > 0). The
former rests on the argument [following Equation (144)] that, in the
solvable case, 011 = -6/5, the spin-wave contribution to OSDW is trans-
formed to l/llsl/l2s, which is straightforward to deal with, but the latter
involves a transformation to (l/licl/lIS!2, which is difficult to define and
certainly does not allow a convincing discussion of the (possibly delicate)
role of phase factors
Thus the absence of a power law for the COW correlation function in
the case of a half-filled band and U > 0 appears to be the correct
conclusion
References
A A Abnkosov, L P Gorkov, and I E Dzyaloshmskl, Methods of Quantum Field

Theory In StatistICal PhYSICS, Chap 1, Prentice-Hall, Englewood Chffs, New Jersey
(1963 )
2 E H Lleb and F Y Wu, Phys Rev Lett 20, 1445-1448 (1968)
3 V J Emery. m Chemistry and PhYSICS atOne-DimensIOnal Metals, pp 1-23, H J Keller
(ed ), Plenum Press, New York (1977)
4 J Bardeen. L N Cooper, and J R Schneffer, Phys Rev 108,1175-1204 (1957)
5 J C Wheatley, Rev Mod Phys 47,331-470 (1975)
6 I P Batra, B I Bennett, and F Herman, Phys Rev B 11, 4927-4934 (1975)
7 W A LIttle Phys Rev 134A, 1416-1424 (1964)
8 J Kogut and L Susskmd, Phys Rev D 11, 395-408 (1975)
9 J M Luttmger, J Math Phys 4,1154-1162 (1963)
10 D C MattIS and F- H Lleb, J Math Ph ys 6, 304-312 (1965)
11 N Bogohubov, J Phys USSR 11,23-32 (1947)
12 K D Schotte and U Schotte, Phys Rev 182,479-482 (1969)
13 K D Schotte Z Phys 235,155-165 (1970)
14 M Blume V .I Fmery, and A Luther, Phys Rev Lett 25,450-453 (1970)
15 A Luther dnd I Peschel, Phys Rev B 9, 2911-2919 (1974)
16 A Luther and V .I Emery, Phys Rev Lett 33,589-592 (1974)
17 I E Dzyaloshmskl and A I Larkm, Zh Eksp Tear F,z 65,411-426 (1973)[Sov Phys
JETP 38,202-208 (197,)]
18. H Gutfreund dnd M SChiCk, Phys Rev 168,418-425 (1968)
19 J Torrance m Chemistry and PhYSICS ot One-DimensIOnal Metals, pp 137-166, H J
Keller (ed ), Plenum Press, New Yark (1977)
20 A Luther and I Peschel, Phys Rev B 12, 3908-3917 (1975)
21 V J Emery Phys Rev B 14, 2989-2994 (1976)
22 K V Efetov and A I Larkm, Zh Eksp Teor F,z 69,764-776 (1975) [Sov Phys JETP
42,390-396 (1976)]
23 M Fowler, Phys Rev B 17 2989-2993 (1978)
24 V J Emery A luther, and I Peschel, Phys Rev B 13,1272-1276 (1975)
25 H Gutfreund dnd R A Klemm, Phys Rev B 14, 1073-1085 (1976)
26 G E Gurgemshv!ll, A A Nersesyan, G A Kharadze, and L A Cobanyan, Physlca
848,243-248 (1976)
27. A. Luther, Phys. Rev. B 14, 2153-2159 (1976).

28. J. D. Johnson, S. Krinsky, and B. M. McCoy, Phys. Rev. A 8, 2526-2547 (1973).
29. P. A. Lee, Phys. Rev. Lett. 34, 1247-1250 (1975).
30. S. Coleman, Phys. Rev. D 11, 2088-2097 (1975).
31. S. Mandelstam, Phys. Rev. D 11, 3026-3030 (1975).
32. D. C. Mattis, 1. Math. Phys.15, 609-612 (1973).
33. 1. Peschel (private communication).
34. R. Heidenreich, B. Schroer, R. Seiler, and D. Uhlenbrock, Phys. Lett. 54A, 119-122
(1975).
35. R. F. Dashen, B. Hasslacher. and A. Neveu, Phys. Rev. D 11, 3424-3450 (1975).
36. A. Luther, Phys. Rev. B 15,403-411 (1977).
37. V. J. Emery and A. Luther, Phys. Rev. Lett. 26, 1547-1549 (1971); Phys. Rev. B 9,
215-226 (1974).
38. S. T. Chui and P. A. Lee. Phys. Rev. Lett. 35, 315-318 (1975).
39. J. M. Kosterlitz, l. Phys. C 7,1046-1060 (1974).
40. ~. T. Chui and J. D. Weeks, Phys. Rev. B 14, 4978-4982 (1976).
41. J. V. Jose, L. P. Kadanotf, S. Kirkpatrick, and D. R. Nelson, Phys. Rev, B 16, 1217-1241
(1977).
42. R. Savit, Phys. Rev. Lett. 39, 55-58 (1977).
43. H. J. F. Knops, Phys. Rev. Lett. 39, 766-769 (1977).
44. V. J. Emery and R. H. Swendsen, Phys. Rev. Lett. 39, 1414-1417 (1977).
45. N. Menyhard and J. S6lyom, 1. Low Temp. Phys. 12, 529-545 (1973).
46. J. S6lyom, l. Low Temp. Phys. 12, 547-558 (1973).
47. M. Kimura, Pro gr. Theor Phys. 53, 955-969 (1975).
48. A. A. Ovchinnikov, Zh Eksp Tear. Fiz. 57, 2137-2143 (1969) [Sov. Phys. lETP 30,
1160-1163 (1970)].
49. A. 1. Larkin and J. Sak, Phys. Rev. Lett. 39, 1025-1027 (1977).
50. J. P. Pouget, S. K. Kahanna, F. Denoyer, R. Comes, A. F. Garito, and A. J. Heeger, Phys.
Rev. Lett. 37, 437-440 (1976).
51. R. Comes and G. Shirane, Chapter 2, this volume.
52. V. J. Emery, Phys. Rev Lett. 37,107-110 (1976).
53. R. A Klemm (to he puhli'hed)
7
The Prospects of Excitonic
Superconductivity
H. Gutfreund and W. A. Little
1. Introduction
The present era of intensive research in the field of one-dimensional

conductors that started with the work of the Penn groupO) on TTF-TCNQ
was preceded by the discussion of the possibility of superconductivity in
one-dimensional organic materials which was suggested by one of US(2) in
1964. This suggestion was based on a new mechanism of superconduc-
tivity, the so-called exciton mechanism. The term "exciton" applies here
broadly to any electronic excitation. In this new mechanism the effective
attraction between electrons at the Fermi surface is induced by the
exchange of excitons, rather than by phonons as is believed to be the case
in all presently known superconductors. The specific one-dimensional
structure proposed for the realization of the exciton mechanism consisted
of a spine of conducting electrons with organic dye molecules chemically
bound to this spine at regular intervals. Under favorable circumstances one
may expect that the excitons propagating along the array of the dye
molecules will induce an effective attraction between the electrons on the
spine. The exciton mechanism is in principle not restricted to one-dimen-
sional structures. Ginzburg(3) has discussed the possibility of excitonic
superconductivity in two-dimensional systems consisting of thin metallic
films sandwiched between, or coated on one side, by layers of a highly
polarizable material. Allender et al.(4) discussed along the lines of Ginz-
burg's ideas a specific system consisting of a thin metallic layer coated by a
---------~ -----
H. Gutfreund • Racah Institute of Physics, The Hebrew University, Jerusalem, Israel.
W. A. Little • Department of Phy~ics, Stanford University, Stanford, California 94305.
305
306 H. Gutfreund and W. A. Little
semiconductor with a high dielectric constant. Their work was followed by

an unsuccessful attempt to find superconductivity with an enhanced tran-
sition temperature in such a system.(5l The common feature of the pro-
posed models, based on one- and two-dimensional structures, is the dis-
tinction between the electrons that are expected to form Cooper pairs and
those that participate in the virtually excited excitons. These two types of
electrons are confined to two spatially separated regions in close contact
with each other.
One of the main attractions of the proposed exciton mechanism is the
apparent possibility of higher transition temperatures. This follows from
the BCS formula for T,
(1)
where A characterizes the strength of the attractive interaction necessary

for the creation of Cooper pairs and f..t * measures the repulsive Coulomb
interaction. In the case of the phonon mechanism hWD is approximately the
Debye temperature, which is of the order of several hundred degrees. For
the exciton mechanism, on the other hand, hw is expected to be a typical
electronic excitation temperature of the order of 104 _10 5 oK, thus leading
to much higher values of Te. This argument is oversimplified, and the
conclusion is in general not correct but depends upon the details of the
system. While it appears that under special conditions the exciton
mechanism may lead to higher values of Tn the high temperature of an
electronic excitation is by itself not sufficient for that. This will be discussed
later in greater detail.
The proposed exciton mechanism raises many questions that have to
be examined with great care, for one is attempting to apply the BCS theory
in a predictive manner outside of the area where it has achieved its most
spectacular successes. First, one has to face the general problems that arise
when one replaces phonons by excitons, then one encounters a host of
other problems that are associated with the restricted geometries of the
proposed models. In addition one should bear in mind that there is still no
experimental evidence for the existence of the excitonic mechanism of
superconductivity and thus no empirical input to guide the theory. On the
positive side, we feel that the progress in the understanding of super-
conductivity and of the above problems together with the accumulated
experience gained from the theoretical and experimental research on
quasi-one-dimensional metals makes the time right now for a critical
reevaluation of the whole idea of excitonic superconductivity and an
examination of the arguments and counter arguments used in the contro-
versy that this idea has provoked. This is the purpose of the present review
article.
The Prospects of Excitonic Superconductivity 307
The subject has a rich history, and many of its facets have been
discussed in review papers over the past decade. A summary of the early
work and an introduction to the basic problems is contained in the Pro-
ceedings of the International Conference on Organic Superconductors (6)
and in excellent reviews by Keldysh(7) and Ginzburg.(s.9) Contributions to
the understanding of several of the key theoretical problems are contained
in later reviews by Ginzburg, (10) Ginzburg and Kirzhnits, (11) and Bulaevsky
et at. (12) This early work of Ginzburg and his co-workers made a partic-
ularly illuminating contribution to the understanding of superconductivity
as a whole by establishing the connection between the most complete
theoretical treatments and the simple form of the BCS equation for Tc
[Equation (1)]. With suitable redefinition of the parameters of that equa-
tion the expression represents correctly the essence of the complete theory.
More recently there has been a review by Andre et al. (13) This contains
a summary of recent theoretical developments using renormalization group
methods and exact field-theoretical results, which have contributed
significantly to our understanding of idealized one-dimensional systems.
This review also contains a discussion of many of the materials of a
filamentary or one-dimensional nature whose properties are relevant to the
model discussed here and an extensive list of references. A comprehensive
review of synthetic endeavors directed toward the preparation of a one-
dimensional organic superconductor is given by Yagubskii and
Khidekel.(14) The broader question of conductivity in complexes of TCNQ
and the platinum square-planar compounds is contained in a review by
Shchegolev,(15) while Zeller(16) had given an overview of the present
understanding of the mixed-valence square planar system. A comprehen-
sive source of information on other mixed-valence inorganic complexes is
contained in a recent review by Miller and Epstein.(17) A feel for
developments in one-dimensional metals and some background on exci-
tonic superconductivity can be got from the proceedings of the two NATO
ASI meeting edited by Keller.(1R 19)
In this review we have attempted to present the material at two levels.
First, much of the initial section is presented at an introductory level for the
newcomer to the field who is not familiar with all the techniques of
many-body-theory. Second, we discuss the most recent advances at a more
sophisticated level where we assume the reader is familiar with that theory.
However, in discussing the latter we have not given the usual formal
derivations of the results discussed but instead have presented the essence
of the underlying physics. By so doing we believe it will be useful to both
groups of readers.
We begin with a nontechnical discussion of the nature of super-
conductivity emphasizing the basic ideas of the BCS theory and pointing
out the limitations on Tc imposed by the ordinary phonon mechanism. We
then introduce the exciton mechanism, which we hope is not subject to

these limitations, but at the same time we point out the problems posed by
this mechanism. The next two sections are devoted to the discussion of
these problems. First we discuss some general questions which are due to
the difference between phonons and excitons. These include the role of
exchange, vertex corrections, and the difference in local corrections or
umklapp processes. Following that, we consider the problems introduced
by the one-dimensional structure of the proposed system. The emphasis in
this review is on a one-dimensional model, and this emphasis is based on
our belief that the one-dimensional structure has the best prospects for a
successful realization of the exciton mechanism. The reasons for this
opinion will be presented in the course of this article. The answers to many
of the questions raised in these two sections depend on the specific features
of the particular model considered. In Section 5 we discuss in detail one
such model. It consists of a linear chain of partially oxidized platinum
atoms closely similar to the structure of the much-studied KCP system(15)
[K2Pt(CN)4BrO " 3H 2 0], but with cyanoligands replaced by polarizable
cyanine-dye-like ligands. For this model we perform a detailed calculation
of the various interaction parameters which determine the superconducting
phase transition. The model proposed raises important questions of
synthetic and structural chemistry. Among these are: Will such a structure
be stable? Will the monomers stack? Can the Pt spine be oxidized without
destroying the ligands? These and other questions of chemical nature will
be discussed in a later section.
It should be stressed that there must surely be other systems based or
other metals such as Ni, Jr, Os, etc. with similar or with other types of
polarizable ligands that might exist as crystalline stacks or as true polymers
and that could also satisfy our theoretical criteria for superconductivity. It
is not claimed that the model discussed in Section 5 is unique but rather
that it is a prototype of the type of system in which excitonic superconduc-
tivity might occur. The common features that all of these systems would
need to have are the following: (a) a dense packing of the polarizable
substituents, (b) a polarizability of the ligands comparable to that of a
cyanine dye, (c) a strongly concentrated conduction-electron density in the
center of the plane of the ligand system, and (d) the possibility for these
electrons to move more or less freely along the axis of the stack perpendi-
cular to the plane of the ligands. The value of the specific model discussed
in Section 5, then, is that it provides a guide as to how dense the substi-
tuents must be packed, how polarizable the ligand must be, how concen-
trated the electron density, etc. If comparable values for these factors can
be achieved in other systems then they too would be expected to satisfy the
criteria for superconductivity. Within these constraints, demanding though
they may be, the synthetic chemist is free to innovate.
The conclusion of the discussion and calculations described in this

article is that the exciton mechanism does not violate any physical principle
and its existence or nonexistence will depend on the specific properties of
the system under consideration. We shall specify the requirements for the
realization of this mechanism and show that they are exacting and not
easily met. This may explain why it is so hard to "construct" an excitonic
superconductor and why a lot more effort in that direction is needed before
one can hope to test a real system that meets the conditions of the
proposed model.
2. The Nature of Superconductivity
2.1. Background
The superconducting state occurs in a very large number of metals and

alloys. As one cools such a metal through a temperature that is charac-
teristic for the material, one finds that the electrical conductivity changes
abruptly to an infinitely large value. The temperature at which this occurs is
known as the critical temperature Te , and typically this lies at a tempera-
ture within a few degrees of absolute zero. The distribution of the tran-
sition temperatures of several hundred of the known superconducting
alloys is illustrated in Figure 1.(20) At present the highest known super-
conducting transition temperature occurs in the alloy Nb 3 Ge at 23.2°K.
Most others lie below the boiling point of liquid hydrogen (20.4°K) in the
liquid-helium range (=4.2°K).
en 250
~
~
~200
1£
~
en 150
!
lC 100
~
a:
~ 50
Figure 1. Plot of the number of known
superconducting compounds as a ~
function of their transition temperature
(1968). A small number of compounds o 20
are now known up to 23 2°K
Our understanding of superconductivity is based on the highly suc-

cessful theory developed by Bardeen, Cooper, and Schrieffer(21) in 1957-
the BCS theory. It has successfully explained and predicted a vast number
of effects and phenomena in the field. The original theory has been expan-
ded upon, reformulated, and developed enormously over the years, but the
essence of the theory has remained unchanged and thus one still refers to
the present theory as the "BCS theory." The one area where the theory has
not played a significant role is in the prediction of new superconducting
compounds or compounds of high Te. Here the experimentalists have
continued to play the key role. This has often been cited as a failure of the
theory. However this is not a fair criticism of the theory of superconductivity
itself, for it is our inability to predict the normal-state properties of new
materials that limits the predictability of the superconducting properties.
Where the normal-state properties can be predicted, values of Te for the
superconducting state can be predicted correctly to within a few percent.
Our discussion of excitonic superconductivity will be based on the BCS
theory but certain features of the exciton mechanism introduce new types
of problems, the understanding of which is far less securely based than
those of conventional superconductivity. Thus, not only are we limited by
our lack of knowledge of the properties of the normal state of the materials
one hopes to synthesize, but in addition these theoretical problems make
calculations of T, even less certain. For this reason our calculation of Tc in
Section 5 is not to be taken as of great numerical significance but rather as
an estimate of the order of magnitude of Tc and a guide for the design of
new materials.
To understand the concept of excitonic superconductivity one must
appreciate the essential elements of the BCS theory. It is based on two key
features. First it requires the conduction electrons to form bound pairs, and
second these pairs must form a Bose-Einstein-like condensate. It is illu-
minating to use a simple mechanistic model to explain how the pairing
interaction comes about and how it is related to the existence of the
condensate. On the basis of this one can give a crude plausibility argument
to explain the limitations on To. to gain some insight into the role played by
the Coulomb repulsion and the possibility of other mechanisms, such as the
exciton mechanism, leading to high- Tc materials.
2.2. Phonon Mechanism
Consider a small portion of a metal with the positive ions located at

more or less regular sites within the crystalline lattice and the conduction
electrons moving among them (Figure 2). It should be recalled that the
Coulomb interaction between the electron and the ion, and between the
• ••• •
I.
• • • •
Figure 2. Schematic representation of the electron-
• ..•
. .' . '.
electron attraction resulting from the polarizabiIity
of the lattice. The large circles represent the positive ' "
ions and the small circles the electrons. As # 1
electron moves through the lattice the ions move
1'
. .' . '.
towards its instantaneous position. A second elec-
tron can then move into the distorted part of the '
lattice left behind and benefit from the excess posi-
tive charge density of this region. An effective
2.
•
attraction between the electrons thus results from
the momentary lattice distortion. ' '
electron and another electron, is screened by the collective motion of the

other electrons, so the effective interaction is one of short range, (e 2 Ir) e -.cr.
As one electron moves through the lattice the short-range screened inter-
action will result in the electron giving each ion an impulsive force as it
passes by. The ions will begin to move towards the instantaneous position
of the electron. However, the electron, which moves at the Fermi velocity,
moves much more rapidly than do the ions, so that the ions will reach their
maximally distorted position when the electron is some distance away.
Thus some distance behind this electron there will be a puckered region of
the lattice containing a slight excess density of positive ions. A second
electron will be attracted to this region and thus, indirectly, will be attrac-
ted to the first electron. This indirect attraction is what is known as the
phonon-mediated electron-electron interaction. It derives its name from
the fact that the distorted lattice may be described in terms of the phonon
coordinates.
This model is too simple as it stands, for one can easily show that a
large number of electrons will share the same region of the lattice with the
"first" two electrons described above. If the motion of these other elec-
trons is random and uncorrelated with respect to the first two, then they
will create random fluctuating forces which will interfere with the inter-
action between the first two electrons. On the other hand, if their motion is
coordinated with that of the other two, then the interaction will be con-
structive, the interaction between the first two electrons will remain, and
they in turn will not interfere with the motion of the other coordinated
electrons. This harmonious coordination can be achieved if the pairs form a

Bose-Einstein-like condensate in which the center of mass and internal
states of each pair of the condensate are identical to those of all others.
At low temperatures it is energetically favorable to form this coherent,
coordinated state in the metal. This is the superconducting state. At higher
temperatures thermal agitation causes the disruption of the pairs. The
disrupted pairs now make no further coherent contribution to the inter-
action and the effective binding energy of the other pairs is reduced. The
weakened binding then makes the pairs more vulnerable to thermal dis-
sociation, with the result that above a critical temperature Tc the condensate
disappears and the system reverts to the normal state.
The BCS theory shows us that within the approximations of a simple
model where we neglect the Coulomb interaction the value of this tran-
sition temperature is expressible in the form
(2)
where hWD is a characteristic energy of vibration of the ions of the lattice,

and A = N(O) V where N(O) is the density of states at the Fermi surface and
V is the effective electron-electron attraction. The factor hWD in (2) comes
from the width of the energy region within which the electrons can make a
coherent contribution to the pair binding energy. It is thus a measure of the
number of electrons which make such a contribution.
Also the expression (2) is valid only for weak coupling of the phonon-
electron system. With stronger coupling, W itself is affected by the inter-
action with the electron system and A is also renormalized by this inter-
action. Indeed, as will become more apparent in our later discussion, the
parameters w, N, and V are not independent of one another, but each
influences the other in a complicated manner. However, Equation (2) with
the parameters suitably defined does give one a useful expression for the
critical temperature. It may also be used to illustrate why there is a natural
limit on Te.
2.3. Limitation on Te
An understanding of the limitation on the value of Tc can be obtained

by using our simple mechanistic model and the above expression for Te .
Let us suppose that each ion is attached to its lattice position by a simple
spring of spring constant k. Then the phonon spectrum will be the single
Einstein frequency W = (k/ M)l/2, where M is the mass of the ion.
Figure 3. Variation of Tc as a function of effective spring constant k which describes the

elasticity of the lattice.
Let us suppose that we can vary the strength of the spring constant k
and then consider the value of Tc as a function of 1/ k. The point of this
argument is that one can consider small variations of composition and
structure of a series of alloys as variations in the stiffness of the lattice with
only a modest variation of the mass of the ions. So our ~imple model can
give us insight into the variations that can occur in Tc for such a series of
alloys. If the spring is extremely stiff (k large) then the electron moving
past the ion will be unable to displace it and, thus, there will be no
phonon-mediated attractive interaction. Consequently A in (2) will be zero
and, even though liw would be large [w = (k/ M)l/2], Tc will vanish because
of the dominance of the exponential term. As we make the spring less stiff
by reducing k, a small attraction will begin to appear and Tc will become
finite and will rapidly increase with the decreasing negative exponent. This
tendency is illustrated in Figure 3. If we continue to make k smaller, then
eventually the variation of the exponential term will flatten out and the
reduction in the pre-exponential factor w = (k/ M)1/2 will begin to contri-
bute. Ultimately the decrease of this factor will cause Te to reach a
maximum and then to decrease. Thus we see even within this simple model
one should expect a natural limit to occur in the value of Te.
One other factor needs to be taken into account to understand more
fully the limitations on Te. As we decrease the value of k then eventually
the electron-phonon interaction itself will begin to contribute to the
effective restoring force of the lattice. This interaction has the effect of
renormalizing the phonon frequencies. One finds that in the simplest
model of this effect(22)
(3)
where Wo is the "bare" phonon frequency, defined as the phonon
frequency in the absence of the electron-phonon coupling, and w is the
renormalized frequency resulting from this interaction. The coupling

parameter A is the same as that occurring in (2).
Now as k is reduced the attractive interaction increases and thus A
increases and approaches unity, causing w 2 to fall and eventually to
become negative. This yields an imaginary value for lV, which gives a
nonoscillatory solution for the motion of the ions. This signals the onset of
an instability resulting in a distortion and change of symmetry of the lattice.
(This simple model would require A ::; n
Such lattice instabilities impose
even stricter limits on the maximum attainable value for Te. In a more
sophisticated treatment of the problem where one includes the effects of
the Coulomb field and the effects of renormalization on both the phonons
and the electrons and upon the electron-phonon coupling, a more compli-
15
14
2240
13
cL
~
UJ
I- 12
1150
30 "To Re 75 " lII:
~
u
l-
T 10
T I
I
I I
I I 9
I .l
J,. I
...
8
Tc OF BULK
T DEPOSITION (OC)
Figure 4. Tc vs. film deposition temperature for Moo 3sReo.62' Insert shows part of the phase
diagram . Note the large enhancement of Tc for films prepared at temperatures close to the
structural transformation temperature at 1150°C (Reference 23).
cated criterion for the stability of the lattice would result. This problem has
not been analyzed completely as yet, but there is ample empirical evidence
to show that high transition temperatures are limited by the stability of the
lattice.
On the basis of the above argument one might expect Tc to vary
strongly as a function of composition in the vicinity of a structural phase
boundary. This is well illustrated in Figure 4, where Tc is plotted as a
function of the deposition temperature of an alloy in the vicinity of the
structural transformation temperature.(23) It explains also why certain
alloys such as Nb 3Ge can be prepared with high Tc by suitable treatment
which stabilize the incipient lattice instability while without such treatment
Tc is substantially lower.(24)
The above argument of a limitation on Tc assumes, of course, that
there is some limit on the Debye frequency. This does appear to be the
case, for this frequency depends upon such factors as crystal structure,
bond strengths, and ionic masses which combine to restrict hWD :S 10 3 oK.
On the other hand a different type of interaction such as the exciton
interaction which we will describe shortly would be subject to somewhat
different limitations. It is on this basis that hope exists for attaining
substantially higher transition temperatures. In principle, at least, a possi-
bility exists for devising a system in which the effective interaction can be
maintained at a large value over an energy range substantially greater than
the Debye energy hWD which occurs in (2). These are systems in which an
electronic or excitonic interaction is used to provide the attractive inter-
action instead of the phonons, and in these it has been argued that one
should be able to ohtain substantially higher transition temperatures. Our
simple model may be used again to illustrate the physical basis for these
arguments.
2.4. Isotope Effect
Consider again Figure 2. As the electron moves past an ion of the

lattice it gives one such ion a momentum p. This ion is then set into motion
having a kinetic energy p2/2M. After the electron has passed, the ion
continues to move until it is limited by the elastic restraints of its neighbors.
The original kinetic energy of the ion is then converted into potential
energy. If the maximum displacement is X then ~kX2 = ~p2 I M.
So for a given momentum transfer p and spring constant k, the dis-
placement is inversely proportional to the square root of the ionic mass. As
explained earlier this displacement is responsible for the attractive
component of the electron-electron interaction. The larger this displace-
ment the greater this attraction can become. The attraction depends on
both the coupling strength and the value of the phonon frequencies w. The
net effect of both these factors is to make the term A in (2) independent of
the ionic mass.
McMillan(2'i) has shown that the coupling constant A has the form
N(O)J 2
A = M(w 2 ) (4)
where J is a matrix element for the phonon-electron interaction averaged

over the Fermi surface which is independent of the mass of the ions M, and
(w 2 ) is an average of the square of the phonon frequency. Recalling that
w =M- 1 /2, as in our Einstein model, we see that A is independent of M.
Because of this, the only mass-dependent term in the expression (2) for Tc
is that of the pre-exponential term hw, which then gives a mass dependence
for Te of M- 1 /2. This result is almost exactly true in the absence of the
Coulomb interaction. When the Coulomb interaction is included, devia-
tions from a simple power dependence occur. The reason for this is that the
electrons can interact via the Coulomb interaction with all the other
electrons regardless of their energy, whereas the interaction via the phonon
mechanism is limited to electrons within ±hWD of one another. It is
possible to replace the real Coulomb interaction J.t by an effective or
pseudointeraction J.t * which has the same energy cutoff ±hWD as the
phonons. When this is done one finds that(26)
(5)
and (2) becomes
kBTe = hw exp(~)
A-J.t
(6)
Now, however, A - J.t *, unlike A, does depend upon w through (5) and thus
upon the mass M. The contribution from this term within the exponential
causes deviations from the simple isotope effect.
The reasons for this deviation can also be seen from our model. Let
us assume that one can vary the value of the ionic mass. If one makes it
smaller, then the resultant larger ionic displacements J/M 1 / 2 contained in
(4) will result in a stronger coupling constant for the electron-phonon
interaction. However, the ionic motion also becomes more rapid because
the frequency w in our simple Einstein model depends on the mass through
the expression w = (k/ M)I /2. Because of this, the most heavily distorted
region of the lattice will now lie closer to the "first" electron which is
responsible for this distortion than it would for the case of a heavier ionic
mass. The second electron is attracted to this region, but some of the
enhanced attraction due to the larger lattice distortion will be offset by the
greater direct Coulomb repulsion of the nearby "first" electron. Thus the
contribution of the Coulomb repulsion causes a deviation from the simple
isotope effect, giving a reduced isotopic dependence,(27) Tc =M-l/ 2 (1-I;)
We shall see that an appreciation of the factors that lead to the isotope
effect can give insight into the possibility of obtaining high Tc through the
exciton mechanism.
2.5. Exciton Mechanism
Consider now a system illustrated in Figure 5. Let A be a conductive

polymeric chain which we call the "spine" and attached to this spine is a
series of polarizable molecular side-chain substituents, B. As an electron
moves along the spine it will induce a movement of charge in the side
chains. This induced positive charge will reach its maximum value not at
the time the electron is adjacent to the side chain but a short time later.
A B
• P --{'-_ _ _-'
p' - - - { ' - - - - - -
• · --{---------)
• · --{~---~
• · ---{---------)
• • - - - { ' -_ _ _ _.J
• · ---{'------'
• · --{---------
· . --{'-----~
a b
Figure 5. (a) Schematic diagram of an excitonic superconductor. A is the conductive "spine"
and B the polarizable substituents to provide the excitonic attraction. (b) Suggested realiza-
tion of the model (Reference 2),
The reason for this delay lies in the finite frequency of oscillation of the
charge in the side chain, just as in the analogous case of phonons the finite
frequency of oscillation of the ions give a similar delay. Consequently the
region of maximum induced charge will lag some distance behind this
electron. A second electron will be attracted to this induced charge and,
just as for the phonon case, one might expect pairing to occur between the
electrons. However, in this case the interaction is mediated by the virtual
electronic excitations of the side chains, which may be described in terms of
"exciton" coordinates rather than phonon coordinates. The characteristic
energy, hw, associated with these excitations is much higher and can be of
the order of several electron volts rather than a fraction of an electron volt
characteristic of phonon frequencies. Thus, provided one can find a system
where the exciton-mediated electron-electron interaction is comparable to
that of a phonon system and the electron density of states is also compar-
able, then expression (1) would lead one to expect substantially higher
values of Tc because of the large pre-exponential factor hwv.
Alternately one may arrive at the same conclusion using the
arguments of the isotope effect. In the proposed excitonic system the
interaction is mediated by the movement of an electron in the side chain
rather than by the much heavier ion of a phonon superconductor. Provided
the momentum transfer p and effective "spring constant" k can be kept
comparable to those in the phonon case, then the transition temperature Tc
for such a hypothetical superconductor would be scaled up from that of
conventional superconductors by a factor of order (MlOnl Me )1/2, which is
typically of order 300. So if the above conditions could be met then one
would have hope of finding superconductors of substantially higher tran-
sition temperature than those known today.
There are good physical reasons for believing that in an excitonic
system it should be possible to make the momentum transfer p and the
effective "spring constant" k comparable to the phonon case. To see this
one notes that the momentum transfer p in the exciton case is related to the
distance of the nearest approach of the conduction electrons in the spine to
an electron in the side chain. This can be of the order of an angstrom or
two, which is somewhat greater but of the same order of magnitude as the
distance of nearest approach of a conduction electron and an ion core in a
conventional metal. So one would expect the resultant momentum transfer
for the two case~ to be somewhat similar. Provided the above distances lie
within the screening length of the two systems then one can show that the
matrix element J of Equation (4) will be approximately proportional to the
square of this momentum transfer. Second, an examination of typical
electronic and Debye energies shows that their ratio is of the order of 100.
This shows that the effective spring constants k (=Mw2) for the electronic
and vibrational systems are of the same order of magnitude. One should
note also that this factor Mw 2 is the denominator of (4) and thus the
effective coupling constant for a tightly coupled electron-exciton system
could be comparable to that of a phonon system. On the other hand, where
the polarizable substituents are separated by several angstroms from the
conductive spine we would conclude that A should become very small. This
is borne out in our detailed calculations of Section 5.
This is the dogma of those who believe in the possibility of achieving
high-temperature excitonic superconductivity. However, these arguments
have been known for over a decade, yet no such high-temperature super-
conductor has been found to date. Indeed, no excitonic superconductor of
any kind is known. We believe, and will present arguments to back this up,
that the reason for this is that only in very special structures can the
excitonic interaction reach a magnitude large enough to overcome the
Coulomb interaction and that until now none of these structures have been
prepared. They may be impossible to make but the indications are that
though difficult they should not be impossible to prepare.
The above mechanistic model, while useful as a guide to thinking,
must not be taken too literally. For one must remember that in discussing
the superconducting ground state one is discussing a single quantum-
mechanical state. A single such state is a stationary state and cannot
describe electrons and ions that are moving, for to do so would require a
wave-packet formalism involving several quantum-mechanical states.
Nevertheless the physical effects of motion in the classical model show
themselves in the ground state as an admixture of virtual configurations.
Keeping this caveat in mind, we can use the model in establishing the
physical principles.
We shall endeavor to show that the above crude arguments are essen-
tially correct and are supported by the detailed sophisticated analysis, and,
further that the electronic and structural criteria that must be met to obtain
a sufficiently strong interaction can only be realized in a small set of highly
contrived systems. We believe that only by the deliberate synthesis of these
systems can one hope to find such excitonic superconductors.
3. Problems of Superconductivity Unique to the Exciton Mechanism
In Section 2 we introduced the idea of the exciton mechanism of

superconductivity. We emphasized that this mechanism raises several
questions stemming from the qualitative differences between an electron-
phonon and an electron-exciton system. In the present section we examine
these questions.
3.1. Exchange
The most obvious difference between an electron-phonon and an

electron-exciton interaction is that in the first case electrons interact with
ions, whereas in the second case they interact with other electrons, namely,
with identical particles. This results in exchange terms in the interaction
between electrons in the conducting and in the excitonic regions. These
exchange terms are of opposite sign to the direct Coulomb terms and may
reduce substantially the effective coupling between the electrons and the
excitons. In view of the delicate balance between the Coulomb repulsion
and the exciton attraction this reduction of the latter may be disastrous.
The importance of the exchange terms was first pointed out by one of
US.(28) Since the exchange interaction is very short-range while the direct
interaction does not change greatly over distances of the order of chemical
bond lengths, we find that a spatial separation between the conducting
electrons and the excitonic medium of about a bond length optimizes the
electron-exciton coupling strength for a given system. Such a separation
was required by Little(2) in his proposed one-dimensional structure and at
least in principle would be possible in the two-dimensional structures of
Ginzburg.(3) In a three-dimensional system the elimination of exchange
terms would require a very special structure in which the two kinds of
electrons would have to occupy orthogonal orbitals such that the "excitonic
electrons" would not be hybridized with any of the states within the
conduction band. This is certainly not the case in the model discussed by
Geilikman,(29) who suggested the possibility of pairing in the s band of a
transiton metal induced by the interaction with d electrons. Another model
in which the exchange interaction may be destructive is the model sugges-
ted by Allender et al. (4) There the electrons of a thin metallic film spend
part of their time in a semiconductor layer where they interact with
electron excitations across the semiconductor gap. This interaction deter-
mines the electron-exciton coupling strength. Exchange effects were not
considered in the estimate of the electron-exciton coupling constant, so
that the number A = 0.5 quoted for this model is probably significantly
overestimated.
3.2. Apparent Limitation on A - j.L *

Another argument that has been raised(3o.31) asserts that the electron-
exciton coupling constant Aex is restricted to significantly smaller values
than the electron-phonon coupling constant Aph ' This argument is based on
the assumption that the net electron-electron interaction may be
represented in the form

47Te 2
V(q, w = 0)= 2 ( 0) (7)
q E q, W =
where E (q, w) is the dielectric function of the medium at momentum q and
frequency w. It contains the direct Coulomb interaction as well as the effect
of all the excitations of the system that contribute to the effective inter-
action between two conduction electrons. This effective interaction may be
separated into a Coulomb part and an indirect part due to the exchange of
some virtual excitation. This separation will be demonstrated explicitly
later. The product of the averages over the Fermi surface of these two parts
and the density of states at the Fermi surface N(O) yield the parameters f.L
and A, respectively, namely, N(O) V(q, 0) = f.L - A. In a simple model of an
electron liquid immersed in a uniform background of positive charge,
causality and stability arguments imply(32) that E(q, O)~ 0, namely, As f.L.
This does not exclude superconductivity in such a model because the
relevant parameter is A - f.L * [see Equation (6)], which may still be positive,
but because of the close connection between f.L and f.L * [Equation (5)] the
above inequality restricts severely the value of A - f.L *. The stability
argument is based on the fact that a static modulation of the background
charge by 8p (q) would modify the total energy of the system by
47Te 2 18p (q )1 2
8£=-2- '--'--..:..~ (8)
q E(q, 0)
If E (q, 0) < 0 for some q, the uniform background would be unstable
against such a distortion. This would result in a static charge-density wave
and a gap in the single-electron energy spectrum. One should, however,
keep in mind that this result is derived in a model in which one neglects the
change in the kinetic energy that arises from a change in the zero-point
motion of the positive background. In a real material E (q, 0) may be
negative and thus A may exceed f.L as is the case in most ordinary super-
conductors.(25) For Pb, for example, A = 1.3, f.L * = 0.1, and f.L = 0.6. This
fact is commonly assigned to local corrections or, synonymously, to
umklapp processes. These corrections come from the fact that the local
fields at the ions are considerably different from those seen on the average
by the electrons. It has been argued(31) that the local field corrections
would be very ineffective in producing the same effect for the exciton
mechanism, because the exciton, being an electronic excitation, occupies a
large part of the unit cell and thus there is not much difference between the
average field seen by the electron and that seen by exciton. However, for
the systems we have considered(2,33) this argument is not correct. For in
these systems the unit cell is very large-much larger than an atomic cell;
and while the exciton system fills a large part of it, the region of interaction
between the excitonic system and the electron is concentrated in a very
small portion of it. The local corrections are therefore very large and the
average field seen by the electron and that seen by the exciton are very
different. Indeed, as is obvious from the details of the earliest cal-
culations, (2) the spatial separation between the spine electrons and the
excitonic system is essential in order to get a net attractive interaction.
Should the conduction electrons be distributed uniformly throughout the
unit cell the attractive component can be shown to be negligible.
3.3. Vertex Corrections
The formulation of the theory of superconductivity, both in the weak

coupling regime leading to the BCS equation or in the strong coupling case
resulting in the Eliashberg equations(34l depend on the validity of Migdal's
theorem, which asserts that vertex corrections to the electron-phonon
interaction are small and may be neglected. The lowest-order correction to
the electron-phonon vertex is shown in Figure 6. For an incoming phonon
of a phase velocity wi q much smaller than the Fermi velocity VF, this
correction is of the order of (wDI E F ) = 10- 2 • Most of the phonons involved
in conventional superconductivity have momentum q = PF and, therefore,
a very small phase velocity . This is the basis of Migdal's theorem . If this
argument is carried over blindly to the exciton mechanism we would have
to replace WD by an electronic frequency and this would result in large
vertex corrections and would raise questions as to the applicability of the
usual theory of superconductivity to the exciton mechanism .
To see that this does not in fact have serious consequences, one has to
consider again the specific properties of the model proposed by us. The
basic difference between this model and a conventional superconductor is
that in the latter the electron-ion interaction is strongly screened, giving
rise to a phonon-electron interaction of a very short range. The Fourier
p+q
Figure 6. Lowest-order correction to the electron-phonon

p vertex .
transform of this interaction results in an electron-phonon coupling

constant that depends only weakly upon momentum. On the other hand, in
the proposed exciton system the bulky polarizable ligands make it
impossible for the transition charges in the ligands to come very close to
the charges in the spine, and this, together with reduced screening in such
filamentary compounds, makes the electron-exciton interaction one of
long range. This results in a strongly momentum-dependent electron-
exciton coupling constant 10(q )1 2 , which falls off sharply at values of II b,
where b is of the order of the distance of the nearest terminal group of the
ligand system to the spine. The qualitative conclusion is borne out by
detailed calculations (Section 5, Figure 21).
It thus follows from the preceding paragraph that the distinctive
feature of the model discussed here is that only excitons with small
momenta, and hence phase velocities much greater than VF, are involved in
the conjectured superconducting transition. It was shown by Engelsberg
and Schrieffer(3'i) that for phonons with a high phase velocity the vertex
correction in Figure 6 is not given by UJDI EF but rather is of order
g2 N(O)/ UJD, where g is the electron-phonon coupling constant. This is a
crude estimate of A and can be of the order of unity in strong coupling
superconductors. However, this result was obtained assuming that g is
momentum independent. This is certainly not the case in our model, as
O(q), the electron-exciton interaction, is strongly peaked around q = O.
Assuming that O(q) is a constant for q :::; qc, where qc is an average width of
O(q), and zero outside this range, one finds that the lowest-order vertex
correction is
r = 1012 N (0)(!lE) (9)
lEE
where E is a typical exciton energy and !lE = E (PF + qc) - E (PF ), E (PF )
being the electron energy at the Fermi momentum PF. Explicit calculations
on our model show that 101 2N(O)I E = 0.3 and that the factor !lEI E
reduces this by about an order of magnitude, rendering the vertex cor-
rections small. Thus, the strong momentum dependence of the electron-
exciton interaction, which as mentioned previously is due to the separation
between the electrons in the spine and the excitons, is responsible for
strongly reduced vertex corrections. One should again point out that this is
not the case for the three-dimensional s-d model(l6) or the two-dimen-
sional metal-semiconductor model, (4) mentioned in Section 2.1.
3.4. Equation for Tc
We shall now prepare the ground for a numerical estimate of the

transition temperature in our model. To this end we adopt the method
developed by Kirzhnits, Maximov, and Khomskii(36) (to be referred to as

KMK). This method applies to a weak coupling superconductor and it
results in a simple BCS-like equation for the gap function and for Te. Its
merit is that it brings out explicitly and in a convenient form the relation-
ship between the kernel of this equation and the microscopic properties of
the system such as the electron band energies, the exciton band energies,
and the electron-exciton coupling matrix elements.
Before we continue our discussion we would like to point out that the
equation described in what follows lies within the framework of the mean-
field approximation, and the temperature Tc obtained from it is not the
actual transition temperature for a system of weakly coupled linear chains.
The special problems associated with the dimensionality of the system and
the relevance of the mean-field transition temperature to the supercon-
ducting behavior of such a system will be discussed in Section 4.
Keeping in mind these words of caution we can now summarize briefly
the KMK method in general and then discuss it with regard to the system
under consideration. The starting point is the integral equation for the
anomalous Green's function F, which at T = Tc reads
(10)
where Wn = (2n + 1 )1TTc . In this equation it is already assumed that the

vertex corrections are small, which was shown in the last subsection to be
also true in our model. At Tc the "imaginary time" Green's functions G of
the superconducting state are replaced by their counterpart in the normal
state and approximated by the normal functions Go
1
Go(p, iw n )= iW n -g(P) (11)
where g(p) = e(p)- CF is the single-electron energy measured with respect

to CF. It is this approximation that restricts the present treatment to weak
coupling superconductors.
The essential feature of the KMK method is the use of the Lehmann
representation for the effective electron-electron interaction. The latter
can be written in the form
V(q,W)= Vo(q,w) (12)

e(q, w)
and the finite-temperature analog of the Kramers-Kronig relation for the

reciprocal of the dielectric function leads to the spectral representation

.
V(q,IWn)=VO(q) 1-2 [ foOO'P(q,w ') d
W
2_ ,2
Wn W
W ']
(13)
The spectral density p (q, W ) is related to the dielectric function by
p(q, w)= -~
1T
Im[_l_]
c(q,w)
(14 )
It is also convenient to write the anomalous Green's function in the

spectral representation
. )=
F( p,IWn icc f(p,. x) dx (15)
-cc IWn-X
Substituting Equations (13) and (15) into Equation (10) one can perform
the frequency sum explicitly, and after some manipulations and plausible
approximations, described in Reference 23, one obtains the equation
( =-f
</J p)
d 1 k U(p,k)tanhlg(k)1/2Tc (k)
(21T)3 2g(k) </J (16)
where
OO
</J(P)=2Ig(k)lf F(p,x)dx
o
and
[ rOO p (p - k, w) dw ]
U(p, k) = Vo(P- k) 1- 2 o J W + Ig(p)1 + Ig(k)1
(17)
To obtain an equation similar to the BCS equation, one usually assumes

that the material is either isotropic or that the "dirty" limit approximation
is valid. This allows one to replace the three-dimensional momentum
integral by an integral over the energy variable ~ and over the "angle"
variables in momentum space, leading to an integral equation in a single
energy variable
(18)
where U (g, f) is essentially U (p, k) integrated over the angle variables of

k in momentum space. Comparison with BCS theory shows that </J(g) =
Re d[w = Ig(PF )1]. Without discussing in detail the form of U(g, f) in the
general case, we point out its two most important properties: (a) The
kernel U(~, f) is a smooth function of the variables g, f, unlike the
interaction itself, which has a complicated resonant structure; (b) the
magnitude of U (~, f) decreases when either one of the variables ~, f

departs from the Fermi energy. To make the analogy with the BCS equa-
tion complete, one has to separate the contributions of the phonon and the
Coulomb interactions. The first has an effective cutoff at wv. The latter
extends over a much larger energy range; however, it can be replaced by an
effective interaction [Equation (5)] with the same cutoff, so that one finally
gets a BCS-like equation with the f integration extending from -Wv to
+wv·
One of the significant differences between the phonon and the exciton
mechanism is that in the latter case the direct Coulomb interaction and the
exciton-induced interaction extend over the same energy so that Equation
(18) would have to be integrated over the whole energy band. We note in
passing that if the condition A :-::; f..t would hold for the exciton mechanism,
there would be no hope for a superconducting transition in this case,
because when A -s f..t, superconductivity is possible only as a result of the
large difference in the cutoffs of the two interactions which replaces f..t by
the significantly reduced f..t *.
Since we intend to apply the integral equation to a one-dimensional
system and since in our case there is no advantage to transforming to
energy variables because there is no small-energy cutoff, we prefer to leave
Equation (16) a~ an equation in the momentum variable. Extending the
integration over the entire Brillouin zone, we write this equation in one
dimension:
( __ fIT;a dk U(p,k)tanh[/~(k)/!2Tc] )
1J (19)
p)-
IT/a 2 7T 2e(k)
~
1J(k
3.5. The Kernel U(p, k)
We shall now discuss the kernel U(p, k). The spectral density p(q, w)
in Equation (14) may be separated into the contribution of the exciton-
induced interaction Pex and the contribution of the Coulomb interaction Pc-
We can similarly separate U(p, k) into its two components
__ ) f"" Vo(p - k)Pex(P - k, w) dw (20)
Uex(p, k)- - () w+i,(p)i+i,(k)i
[ f
pc(p-k, w)dw]
e('
Uc(p, k)= Vo(p-k) 1-2 () w+i,(p)i+i,(k)i (21 )
The numerator in the integrand in Equation (20) is related to the imaginary

part of the effective exciton-exchange interaction. We write the latter in
the form
(22)
This corresponds to the exchange of an exciton of momentum q, frequency

w, and band index a (in case there are several exciton bands). QOI(q) is the
coupling strength between the electron and the exciton and DOI(q, w) is the
exciton propagator. We assume that this propagator has the form of a
boson Green's function
2EOI(q)
(23)
where E,,(q) is the exciton energy. This assumption deserves a brief dis-
cussion. The electron-exciton interaction may be represented by a form
similar to the Frohlich electron-phonon interaction
(24)
where at( ad and b L(bd are the creation (destruction) operators of the
electrons and excitons, respectively. The exciton propagator has the form
given in Equation (23) only when the boson commutation relation,
[b q, b:l = 1, is satisfied. For exciton operators this is an approximation and
it remains to see how good this approximation is. The propagating exciton
is a linear combination of electronic excitations on the single molecules,
and we can write
(25)
where b creates an excitation on the ith molecule. A typical property of

l
the dye molecules discussed by us is that the transition to the lowest

electronic excitation exhausts most of the oscillator strength and the higher
states playa negligible role. Hence, the excitons in our model propagate in
an array of coupled two-state units. For such a system the local destruction
and creation operator~ <;atisfy mixed commutation relations(37)
[b bl I = [ b;. b; I = 0
l•
[b b; I =
l• () for i ¥- j (26)
{bl, b;} = I
where { } denotes the anticommutation relation. These are the com-
mutation relations of spin-:t raising and lowering operators and are there-
fore called the Pauli commutation relations. It is now easy to calculate the
commutation relations for the propagating exciton operators in Equation
(25). One gets
[ b'I' btl
q
= [)
qq
- 2,),
N~e
I(q-q')r'b t
I
b
I
(27)
The second term on the right-hand side is a correction to the boson

commutation rule. This correction is small as long as the number of
virtually excited excitons is small, which is the case when the electron-
exciton coupling is not too strong. In the particular calculation we describe
in Section 5 we show that in this case the coupling is weak and in fact we
find it would be difficult to conceive of a system where the coupling is
strong.
It has been argued recently by Chaikin et al.(3K) the the electron-
exciton interaction renormalizes the electron bandwidth resulting in a
drastic band narrowing. An examination of the analysis in Reference 25
reveals that a rather strong electron-exciton interaction is required for that
to happen. Chaikin et al. assumed that the excitons behave like bosons;
however, in the region in which their results are valid the deviations from
the boson commutation relations are becoming important. It was shown by
Bad'C) that when these are taken into account the band narrows at most by
a factor of 2.
After this digression we return to the exciton contribution to the
kernel in the equation for T. On account of Equations (14), (22), and (23),
we obtain
(28)
and the exciton-exchange contribution to the kernel becomes
(29)
3.6. Effects of the Phonons on the Exciton Mechanism
Our discussion thus far has treated the excitonic system without
considering the effects of the phonons. The phonons can contribute in two
ways: first, virtual phonons will give rise to an attractive contribution to the
net interaction ju~t as in conventional superconductivity, and second, real,
thermally excited phonons can also affect the system in a variety of ways. If
the phonon contribution to the coupling parameter A is Aph and that of the
excitons is Aw then any limit on A imposed by the stability of the system
would impo~e a more ~tringent limit on Acx. At an early stage, when it was
thought that A :S ~ wa~ a real limit on A set by the lattice, it was argued(40)
that this would limit Aex:S d - Aph ) and thus would keep Aex small and so
prevent T from attaining a large value. The reason the two contributions
can give a limIt on the stability of either can be understood as follows. The
combined cffect of the phonon and exciton interactions can create a
charge-den~ity wave in the conduction electron system. The charge-density
wave in turn can act back on both the phonon and exciton system and cause
each to distort. In this way the exciton and phonon systems are coupled to
one another via the electron system. This constraint does not appear to be
important for we now know that A can attain much higher values than!, so
Aex is not limited significantly by this factor. Moreover, as we shall see in
the discussion of the lattice stability (Section 4), the stability is not deter-
mined by a single gross parameter like A but by the details of the terms that
contribute to it. How the phonon and exciton terms contribute to the
stability of the entire system needs to be determined in each case.
A second objection was raised relating to the damaging effects of
phonons on the electron-electron interaction for energy changes just
greater than the Debye energy. This followed from the use in the kernel of
the integral equation of the Bardeen-Pines form of the phonon-mediated
electron-electron interaction,
(30)
For energy changes t(p)-t(k) greater than the phonon energy E(p-k)
the interaction changes sign and becomes repulsive. However, Ginzburg(9)
showed that it is not this interaction that should appear in the kernel of the
integral equation but rather an effective interaction of the form of Equa-
tion (29). This is a monotonically decreasing function of It(k)1 and It(p )1,
which is attractive at all energies and does not have the resonant form of
Equation (30). Because of this all virtual phonons make a positive contri-
bution to the net attraction.
It has also been argued(41) that in the presence of phonons the excitons
could not give rise to high values of ~" This argument was based on the
results of solving the BCS integral equation for the gap with a form for the
kernel obtained by approximating the phonon and exciton contributions V
by two wells, with the phonon contribution cut off at Wo and the exciton
contribution extending from n to n'. Solving the gap equation at Tc
I =- r' N(O)Vtanh({3cEk) d 3k
(31 )
Jo 2lckl 2
then yields a value for Tc « hWD and hence in spite of the exciton contribu-
tion does not give a large critical temperature. The physical reason for this
is that the major contribution to the integral comes from the region where
lE(k)1 is small and in the above approximation for V it was assumed that
the exciton system makes no contribution to this region. If the interaction
with the exciton system i~ treated correctly(42) rather than by the above
two-well approximation thcn one finds that the exciton system does
contribute in the region of If'k I = 0 and can lead to high values of Tc-
We see then that the virtual phonons do not affect adversely the
superconductivity, but what of the thermal phonons? It was conjectured(43)
some years ago that at temperatures greater than or comparable to the
Debye temperature the presence of these thermally excited phonons would
have a devastating effect on T( and would probably preclude the possibility
of superconductivity occurring by any means at temperatures appreciably
above the Debye temperature. This problem has had an interesting history,
but we now know from a rigorous treatment by Bergmann and Rainer(44)
that this is not the case. To appreciate this point it is instructive to consider
first why the customary neglect of the thermal occupation of the phonons in
the equation for the gap gives reasonable results. The Eliashberg equa-
tion(25) for the gap function, using McMillan's terminology, is as follows:
1 Jd I
~(w) = - - ~ Re[~(w')l
{WO dWq a 2(w q)F(Wq)
Z(w} W 0
X ([N(w q)+ [(-W')][(W ' + Wq + Wr 1+ (Wi + Wq - Wrl]

-IN(wq )+ [(W')][(-W ' + Wq + W r l +(-w' + Wq - w)]} (32)
where we have simplified the expression by the neglect of the Coulomb

interaction. Here N(w q ) and [(Wi) are the phonon and electron occupation
numbers, respectively. The function Z(w) is the renormalization
parameter for the electron energies, which is obtained from an equation of
somewhat similar form that also contains factors of N(wq ) and [(Wi).
We note that the major contribution to the integral for the gap ~(w)
comes from Wi"" 0, so if we neglect Wi in the curly bracket, the terms in
N(w q ) cancel paIrwi~e, leaving us with the familiar tanh(f3w ' /2) from the
combination of [(-Wi) and [(Wi). The only contribution from the phonon
occupation that remains comes from corrections to this approximation and
the terms in Z(w). These can be shown to be of the same order of
magnitude and small. Appel(4'i) went beyond this approximation to include
the effects of phonon scattering on the electron lifetime. He claimed that
the dynamical character of the phonon interaction should affect the energy
gap and Tc becau~e even though the Frohlich Hamiltonian commutes with
the time reversal operator, this Hamiltonian contains time-dependent
phonon operator~. Therefore the theorem of Anderson(46) and Maki,(47)
which states that T, should be unaffected by a perturbation that conserves
time-reversal symmetry and is static, would not apply and Tc should be
affected. Appel calculated the effects of phonon scattering on the electron
self-energy and showed that it reduced Tc but that this effect would not
preclude the possibility of room-temperature superconductivity on these
grounds. AlIen,(4X) working from McMillan's equations, included the resi-
dual contributIOn from the phonon occupation terms but neglected the
The Prospects of Excltomc Superconductivity 331
above lifetime effects, and he also showed an apparent repulsive or

destructive effect upon the gap due to thermally excited phonons. In this
the temperature dependence of Z(w) was ignored. Bergmann and
Rainer(44) then showed in a rigorous proof using the Matsubara formula-
tion of the Eliashberg equations, which include all effects of the thermal
excitation of the electrons and the phonons, that the thermal phonons are
never repulsive and that the contribution to Tc of all phonon modes is
positive. Karakozov et al.(49) helped to explain this by noting that for
low-frequency phonons the terms in Z(w) in (32) exactly cancel the low-
frequency contributions to the numerator. An additional factor that has to
be considered i~ renormalization of the electron mass with increasing
phonon excitation. The electron self-energy depends explicitly upon the
phonon occupation numbers. One can readily show that this leads to a
renormalization of the electron mass m* relative to its band mass m:
(33)
The electron energies Ek that result from solving for the normal part of
the self-energy X(p) in the Nambu formulation(49) of the Eliashberg equa-
tion contain this explicit phonon occupation-dependent contribution. This
is neglected in the approximation that leads to the gap equation used by
McMillan. Including this phonon contribution changes the effective density
of states N(O) at the Fermi surface. The effect of this is to offset partially
the pair-breaking contribution of the phonons.
Thus we see that in a gross sense treating the phonons and excitons as
independent subsystems, the phonons serve only to enhance Te. At the
microscopic level, where on~ treats the coupled exciton-phonon systems as
we do in Section 4 in the "g-ology" picture, we find that the phonons can
have a subtle role on Tc and in determining the type of pairing in the
system. We discus!> that role briefly later.
4. Effects of Limited Dimensionality
We have emphasIzed in the Introduction our belief that the one-

dimensional model proposed for excitonic superconductivity has the best
prospects of success. We have already pointed out in Section 3 how the
particular one-dimensional structure proposed here helps to reduce vertex
corrections and to minimize the harmful exchange-interaction terms. In
Section 5 we shall show that the one-dimensional structure is almost
essential in order to achieve an effective exciton-induced attraction that is
sufficiently strong to overcome the Coulomb repulsion. These arguments
follow from the examination of the general properties of excitonic super-

conductivity and they clearly favor the one-dimensional models. The price
one has to pay for these apparent advantages is that the one-dimensionality
poses a series of new questions and difficulties that had previously received
scant attention in the context of conventional superconductivity.
Of particular significance is the problem associated with the absence of
phase transitions in a one-dimensional system. It was well known(51-53) that
phase transitions involving a classical order parameter, such as the liquid-
gas transition in a system of particles with finite-range interactions, cannot
occur in one-dimensional systems. Stimulated by the suggestion of the
possibility of superconductivity in polymeric systems it was shown by
Ferrell(54) and more rigorously by Rice(55) and Hohenberg(56) that this
could be extended to include the superconducting phase transition. Real
systems, however, are not isolated one-dimensional systems but are
composed of parallel chains that are in some sense coupled. Because of this
such systems can undergo phase transitions at a finite temperature exhibi-
ting three-dimen~ionallong-range order, although this will usually occur at
a temperature significantly lower than the mean-field transition tempera-
ture. In addition, currents with very long (practically infinite) decay time
may occur in one-dimensional systems even without a real phase transition,
because the fluctuations that destroy off-diagonal-long-range-order (the
signature of the ~uperconducting state) affect differently the conductivity
and the order. Thus, the absence of true phase transitions in strictly
,me-dimensional systems does not exclude the possibility of superconduc-
tivity in real filamentary systems, even at high temperatures, as will be
discussed in detail. The notion that it did, however, cast a pall on the early
development of the ideas of a one-dimensional superconductor, which has
only been removed in recent years. The principal effect of this problem was
to stimulate interest in other types of systems of twO(3) and three(29)
dimensions that might be developed to utilize the same exciton mechanism.
(These early developments are well reviewed by Keldysh.(7)
The limited dimensionality also plays a key role in the question of
stability. The connection between superconductivity and lattice stability
was already mentioned in Section 2.3. This connection is demonstrated
particularly clearly in the high- Tc compounds of the f3-tungsten struc-
ture,(57) which usually undergo a transition from a cubic to a tetragonal
phase at temperature~ somewhat greater than Tc- One possible effect of a
structural instability in the neighborhood of Tc is due to the fact that near
such an instability one of the phonon modes becomes soft. This, in turn,
increases the electron-phonon coupling constant, and Tc may come closer
to its maximum value determined by the dependence of Equation (1) on
the phonon frequencies, as explained in Section 2.3. Such an effect would
account for the relatively high values of T. in the f3-tungsten compounds.
A different point of view, which emphasizes the destructive effect of lattice

instabilities, asserts that the structural phase transition occurs because the
electron-phonon coupling constant in what would be the high- Tc phase is
too large. In the new phase the lattice becomes stable with a lower A.
Therefore, if it were possible in some way to suppress the structural
transition, the superconducting transition temperature would be larger
than the observed one. This point of view on the restrictions on the values
of Tc set by lattice instabilities has been frequently expressed by
Matthias. (5X) The effect of lattice instabilities on superconductivity becomes
even more dramatic in a one-dimensional system. In three dimensions such
an instability merely changes the density of states at the Fermi surface, but
in a linear chain it produces a gap at the Fermi surface driving the system
into the insulating state. Thus, a discussion of superconductivity in a
filamentary compound require~ a careful examination of other competing
instabilities. This problem is discussed in detail in the present section.
Another objection raised against the idea of excitonic superconduc-
tivity which was based on dimensionality is that of screening. The screening
of thc Coulomb interaction in a system of restricted geometry is less
effective than in three dimensions, and it was conjectured(59) that in a
filamentary system it could never be of sufficient strength to allow the net
interaction to be attractive. Although the discussion of this problem
involves several general arguments of principle, the ultimate answer
depends on detailed calculations for a particular model. The problem of
screening will be di~cus~cd in Section 4.0.
4.1. Effects of Fluctuations
The absence of phase transitions in one-dimensional systems results

from fluctuations of the order in the system. In a one-dimensional system a
fluctuation that disrupts the order at one point in the system breaks the
long-range order of the system as a whole. At finite temperature it is
energetically possible for such a disruption to occur, and hence the system
shows no long-range order. In two or three dimensions, however, for the
long-range order of the system to be destroyed the order must be disrupted
across a line or a plane. respectively. Such a fluctuation is very much more
costly in energy and thus less probable than that in the one-dimensional
case.
Ferrell(,)4) extended these classical arguments to the superconducting
case. He showed that in a one-dimensional system the zero-point or
thermal fluctuation~ of the superfluid density caused fluctuations of the
phase of the order parameter, with the result that the so-called off-
diagonal-long-range-order (ODLRO), which is descriptive of the order of
the superfluid state. was destroyed. This ODLRO was believed to be an
essential characteristic of the superconducting state. This argument was

clarified and put on firmer grounds by Rice,(55) and later Hohenberg(56)
showed rigorously that ODLRO could not occur in -any infinite one- or
two-dimensional systems.
These rigorous mathematical results, however, raised a number of
physical paradoxes. Using Ferrell's argument, DeWames et al.(60) showed
that for a finite but long one-dimensional system a superconducting phase
could exist but its transition temperature would be depressed below that of
a short specimen! The paradox that this raised can be seen as follows.
Consider a small bead of superconducting metal. According to experience
and the above arguments, such a finite sample can superconduct. Consider
now a long string of such beads. If this string is made long enough, then
according to the arguments of Ferrell, Rice, and Hohenberg such a chain
would not superconduct. This raised the problem of how the local conduc-
tivity could be changed by the addition of distant beads. This paradox was
re~olved by the arguments of Little, (6\) who considered the effects of
fluctuations upon the apparent resistivity of a one-dimensional "super-
conductor." The arguments relating to the resistivity can be appreciated by
first considering Ferrell's arguments. The parameter that describes the
order in a superconductor is the function t/J, which is complex, t/J = !:J.e'<P.
Fluctuation~ of the local density of the electrons at some region in the
system due to zero-point or thermal motion must be accompanied by a
fluctuating in-flow or out-flow of current to or from this region. From the
quantum-mechaOlcal current operator,
(34)
one can show that these fluctuating currents cause fluctuation of the phase
<p of the order parameter. If one considers two points x and x' in the
superconductor, then as one increases the separation between these points,
the fluctuation of the relative phase of the order parameter between the
two points must increase. Thus, for a sufficiently large separation, the
relative phase of the two points will fluctuate over an angle greatly in
excess of 27T. In this case the value of the average (t/J(x )t/J*(x'» will clearly
vanish. However, the existence of a finite value for this average in the limit
that Ix - x'I ~ CD i~ the requirement for ODLRO, so in such a one-dimen-
sional system ODLRO cannot occur.
Rice())) used the Ginzburg-Landau equations and the above
arguments to calculate quantitatively the range of order for both one- and
two-dimensional systems. Using this formalism Little(6l) then considered
the effects of fluctuations on the lifetime of "persistent" currents in a
"one-dimensional system" bent into the form of a closed ring. He showed
that while such fluctuations destroyed ODLRO, in themselves they did not
REAL
8(r)
Figure 7. Plot of the real and imaginary parts of the

order parameter "'(r) as a function of position r in a
one-dimensional sample carrying a current j(r).
cause the decay of the persistent current. This can be shown on the basis of
a topological argument.
If we plot the complex order parameter as a function of position as
illustrated in Figure 7, then the value of !/I(O) must equal its value at !/I(L),
where L is the perimeter of the ring. This follows because !/I must be single
valued. Thermal or zero-point motion causes !/I to fluctuate in amplitude
and phase at each point. But, provided it remains finite at all points, then
the total phase change round the loop must be an integral multiple of 27T,
i.e., 27Tn, in order to satisfy the boundary conditions. We can classify the
different !/In into different subensembles given by the value of this integer n.
Within a given subensemble fluctuations can occur that cause the represen-
tative path of the order parameter to wind and unwind about the origin line
of the Argand diagram. One can readily show, as argued above, that such
fluctuations destroy ODLRO, but equally well one can show that the net
circulating current is fixed by n (i.e., the total phase change round the
ring), and, provided thi~ does not change, no current decay is possible. The
essential requirement for the current to decay is for a fluctuation to occur
between subensembles such as illustrated in Figure 8. Clearly from the
topology of the diagram the only type of fluctuation that can allow this to
occur is one in which the order parameter drops to zero at some point,
value for n. Becau~e of the finite coherence length '0

passes through the origin, and emerges on the other side with a different
in a superconductor
(i.e., the effective size of the Cooper pair) the order parameter must be
driven to zero over a region whose length is of this order. Consequently,
the cost in free energy 11[ for such a fluctuation to occur is finite for
temperatures below the mean-field Teo and the probability of such a
fluctuation occurring IS proportional to exp( -11[1 kT). This becomes
Figure 8. Illustration of the type of fluctuation in the

complex-order parameter which can cause the decay of a
persistent current in a one-dimensional sample.
exponentially small at low temperatures and consequently the resistivity

below Te should also fall exponentially.
These arguments show that as far as fluctuations are concerned a
one-dimensional superconductor cannot exhibit true long-range order
(ODLRO) in the strict mathematical sense, nor can it show a sharp phase
transition, but nevertheless it can be expected to exhibit a conductivity that
approaches an infinite value at temperatures somewhat below Te.
The rare fluctuations that lead to dissipation also introduce a special
problem in considering the statistical mechanics of a one-dimensional
superconductor. In most problems in solid-state physics the assumption is
made that the physical properties of a material can be described by taking
an average over an ensemble of identical systems. Such an ensemble
average is equivalent to a time average of one system provided the time
average is taken over a sufficiently long time-strictly speaking, over an
infinite time period. However, in this particular problem involving the
decay of persistent currents the time for fluctuations to occur between the
subensemble~ discussed earlier can be shown to be so long(61,62) that a true
time average cannot be obtained even over time periods as long as the age
of the universe. It is therefore misleading to use an ensemble average, for
this would not describe the properties of the system that would be observ-
able over any reasonable time period. In particular, the proof of the
absence of ODLRO in a long but finite one-dimensional system is based on
such an ensemble average. If the time evolution of such a system is studied
one finds that, in fact, the system does indeed exhibit ODLRO for extremely
long periods of time but that over an infinite time period it does not.
Caution must thu~ be used in assessing the physical consequences of the
limited dimen~ionaJity in calculations where conventional methods of sta-
tistical mechanics are used.
The above arguments are all based on the assumption that within the
superconductor a local order parameter !/J(x) can exist. For the case of a
macroscopic bead of superconductor discussed earlier this assumption is
reasonable. On the other hand, for a molecular chain of atomic dimensions
it is not so obvious that it is valid and requires further justification. One
must examine with care the microscopic theory of the condensate. This will
be discussed in Section 4.2. However, as stressed earlier any of these
results or their extensions that do not take proper account of the possible
metastability of any fluctuating condensate can give results that would not
correspond to that which would be observed in the laboratory. For exam-
ple, this metastability imposes upon the microscopic theory the require-
ment that the expectation values for the gap function or pair condensate
used in the gap equation itself be taken within the subensemble and not
over the whole en~emble. The problems that this raises are not trivial, for it
requires the explicit examination of the validity of the ergodic hypothesis
..
Rn : 05 11
A: 2' 10-"crn'
I - 0 .2 ,.A
10~
Figure 9. Comparison of McCumber-Halperin theory with experimental data points of

Newbower, Beasley, and Tinkham for resistance of tin whisker below Tc due to ther-
modynamic fluctuations. The small "foot" is believed due to contact effects (Reference 62).
for each system and temperature, and thus a knowledge of the "master
equation.,,(63) For this reason further work is needed to justify the rele-
vance of recent rigorous results to the physical observables.
An offshoot of these sometimes esoteric arguments has been the
explanation of the perfect conductivity of a superconductor, a subject that
remained obscure long after the development of the microscopic theory.
As pointed out by Little(!>]) and Langer and Ambegaokar,(64) the fluctua-
tions that change the phase of the order parameter by 27T at some point in a
wire are the source of dissipation. Langer and Ambegaokar justified the
consequences of the electrodynamics of the fluctuations and established a
detailed theory of the decay. This work was expanded upon by McCumber
and Halperin(h5) and has been studied extensively by several experimental
groups, notably at Cornell and at Harvard. This experimental work is
reviewed by Tinkham{(2) The essential exponential temperature depen-
dence of the resistivity whose origin was outlined above is beautifully
illustrated by the results of Newbower et at. (Figure 9).(66)
These arguments and their experimental verification give one
confidence that the essence of the problem is understood and that states of
very high conductivity could occur in filamentary materials in spite of their
limited dimensionality.
4.2. Types of Order in a One-Dimensional Electron Gas
Although a one-dimensional electron system cannot undergo a phase

transition at any finite temperature, it may exhibit long-range order at
T = O. An understanding of the conditions for the onset of a particular type
of order at T = 0 and the growth of the corresponding condensate is
important for the discussion of the behavior in real filamentary materials.
We can base our discussion of this problem on a system described by the
electron gas Hamiltonian
(35)
where a1,(J"(ak,,,) creates (destroys) an electron with spin projection ( I in the
state of momentum k and energy Ek. This Hamiltonian contains only the
electron-electron interaction V(q), and it is assumed that the electron-
phonon (or electron-exciton) interaction is effectively included in the
latter. A discussion of one-dimensional systems based on a Hamiltonian in
which the coupling to phonon (or exciton) modes is explicit will be
mentioned briefly in Section 4.4. Another line of approach to one-dimen-
sional systems is based on the Hubbard model(67) and it is particularly
appropriate in the case of Uj4t» 1 (where U is the interaction between
two electrons on the same atomic site and 2t is the bandwidth). We shall
not discuss this line of approach because the systems we are interested in
are more properly represented by a band picture as in Equation (35).
A one-dimensional Fermi system is characterized by a Fermi "sur-
face" consisting of two points at ±kF' This results in a striking difference
between the three- and one-dimensional Fermi system. While in the first
case there exist low-energy electron-hole excitations with momenta
between 0 and 2kh in the second case one finds such excitations only in the
neighborhood of q = 0 and q = 2k F• This means that the important elec-
tron-electron interaction processes are those that involve momentum
transfers in the neighborhood of these two values. There are four relevant
interaction processes, which are represented in Figure 10. The q = 0 pro-
cesses may involve electrons either on one side of the Fermi "surface" (g4)
: :c~::
: :r :
Figure 10. Diagrammatic representation of the four interaction processes in a one-dimen-
sional Fermi system. The forward scattering processes are denoted by g2 and g4, the back-
scattering by gl and g3' The latter exists only for a half-filled band.
a
'Ll m
!(I)
92
CK)
9,
b
b ill
j
92
c
'L5
j(t)
(I) (])
92
~
9,
+
d '(:)
!(I) ..
(])
92
CD 9,
+
Figure 11. Diagrammatic representation of the zeroth and first -order contributions to the
generalized susceptibilities X(2k F , w) corresponding to (a) COW, (b) SOW, (c) SS, and (d) TS.
The +, - signs denote the two sides of the Fermi "surface," and the vertical arrows indicate
the possible relative orientations of the spins.
or on both sides (g2). In both cases the electrons do not change their
direction of motion as a result of the interaction between them. For this
reason, these processes are referred to as forward scattering processes.
When electrons on the two sides of the Fermi "surface" exchange momen-
tum q = 2k r they change their direction of motion. This is therefore a
backscattering proce~s (gd. In the case of one electron per atom (half-filled
band) there is also an umklapp process in which two electrons on the same
side of the Fermi "surface" scatter together to the other side (g3)' The
study of instabilities in a model in which the interaction is reduced to these
coupling constants is generally referred to as one-dimensional" g-ology."
We have mentioned already that the question of order in a single chain can
be discussed in terms of its tendency towards long-range order as T ~ O.
Any type of order would show up in the response of the system to a
corresponding generalized external field. There are four possible types of
ordering when T ~ 0: charge-density wave (CDW), antiferromagnetic or
spin-density wave (SDW), and singlet and triplet pairing (SS and TS).
Other types of order are in principle possible for very large coupling
constants,(hX) but then the band picture assumed in Equation (35) breaks
down. To each of the types of order there corresponds a generalized
susceptibility, (h9.7()) for example.
Xcow(q, w)= -if' -(X'

dIe'"'' L, L, (T[a;""(t)ap+q<T.(t)a;'<T,(O)ap'-q<T'(O)])
PP (TIT
(36)
where T is the time-ordering operator. There is a similar expression for

each of the other types of order. A divergence of one of these functions

signals the onset of the corresponding type of order. In one-dimensional
systems such divergences may occur only at T = 0, which is equivalent to
the statement that a phase transition cannot occur in such a system at any
finite temperature.
As a first step it is instructive to calculate up to first order in pertur-
bation theory the susceptibilities defined in Equation (36). The processes
that contribute in each case are shown in Figure 11. The basic effect of
one-dimensionality is that the integration over the internal momenta of
each electron-electron and electron-hole pair gives a logarithmic contri-
bution of the form p 10g(T/ D) (where p is the density of states and D is an
energy cutoff) to the susceptibility at q = 2k p and w = O. This is in contrast
to the three-dimensional case in which the electron-hole propagator is not
singular. The expressions for the four susceptibilities defined in Equation
(36) at w = () at q = 2k l . for CDW and SDW and at q = 0 for SS and TS
are(69.70)
(37)
(38)
(39)
(40)
r
where v is the Fermi velocity and (1TV I is the density of states at the
Fermi level. In the random-phase approximation one picks from each
order in perturbation theory that term which is merely a repetition of the
first-order processes, so that the expression for the susceptibilities in this
approximation becomes a geometric series which can be summed to give
the general form
x, ()
(41)
x, = I-A, In(T/D)
where the index i denotes one of the possible types of order and the A, are
the corresponding effective constants, which can be read off from Equa-
tions (37)-(40):
Acow = ~(2gl - g2)
AsDw = ~g2
(42)
Ass = +~(gl + gz)
ATs=+~(g2-gd
TS sow
SS cow
Figure 12. RegIOns In the (gj, g2) plane with the

highest mean-field transition temperature to the
indicated type of order.
where g, is g, in units of 1TV. In the random-phase approximation one finds

a phase transition for negative A, given by the BCS-like expression
~ = Del/A, (43)
which is the mean-field transition temperature. If one now asks in what
regions of the (gj, g2)-plane is one of these transition temperatures higher
than all the others one gets the picture in Figure 12. Although the mean-
field treatment is insufficient, the results obtained indicate the regions in
the space of the coupling parameters in which the system tends to develop
large fluctuations corresponding predominantly to one of the possible types
of order, and these regions are remarkably similar to the qualitative picture
obtained from the more rigorous treatment.
The basic difficulty in the theoretical treatment of order in one-
dimensional Fermi systems is due to the divergence in the bare electron-
electron and electron-hole propagators. These cause divergences in the full
electron-electron interaction in the particle-particle channel (Cooper
channel), which indicate the onset of pairing instabilities and in the parti-
cle-hole channel, which indicate the onset of spin- and charge-density
wave instabilities. The <;implest examples of processes in these two chan-
nels are shown in Figure 13. In higher-order processes these two channels
become coupled and one has to treat them simultaneously. This was first
done by Bychkov el al.,1711 who considered a model in which gl = g2 = g.
neglecting g3 and g4. They calculated the effective interaction between two
electrons in the parquet approximation. In this approximation one solves
the coupled equations for the vertex functions (effective interactions) in the
particle-particle and particle-hole channel; however, in each channel one
keeps only the terms that appear in the random-phase approximation.
Therefore, this is still a mean-field approximation and one finds a phase
(a) (bJ
Figure 13. Diagrammatic representation of the simplest interaction processes in the particle-
particle channel (a), and in the particle-hole channel (b).
transition at a finite temperature. Bychkov et al. found that the particle-

particle (r pp) and particle-hole (r ph) vertex functions are both of the form
(44)
r ex:: I + gin (DI T)
which for negative g gives a critical temperature Tc = D exp( -1 Ilg!). They
concluded that at this temperature the system goes over to an ordered state
of condensed quartets of two electrons and two holes. This is a pioneering
contribution to the field of one-dimensional systems and attracted a lot of
attention; however, it suffers from two significant shortcomings. One is the
restriction of the interaction parameters to gl = gz, and the other is the use
of the mean-field approximations.
As one goes beyond the parquet approximation and includes addi-
tional terms which also diverge logarithmically, one finds that all suscep-
tibiities or vertex functions change dramatically from the logarithmic form
in Equations (41) and (44) to a power law behavior TO<. A negative value of
the exponent a indicates the corresponding type of ordering at T = 0. We
shall summarize the basic results of the extensive study of the generalized
susceptibilities carried out in recent years.
Another treatment of the competition between the superconducting
and the Peierls state in the mean-field approximation was done by Levine
et al. (72.73) They concluded that these two states do not coincide except in
very exceptional circumstances.
The first discussion of order in a one-dimensional system that went
beyond the mean-field approximation is due to Menyhard and
S6Iyom.(69.74) They applied the renormalization group method to calculate
the susceptibilities corresponding to COW, SOW, and SS, and found that
for gl > 0, the line gl = 2g2 separates the region of superconducting and
charge-(spin- )density wave behavior (only gl, g2 were taken into account).
This approach wa~ extended by Fukuyama et al., (70) who also calculated the
TS-response function. The renormalization group method fails for gl < 0,
except in the neighborhood of the origin.
For negative values of the backscattering coupling constant gl the

results are based on the work of Luther and Emery,(75) who found a
remarkable solution for a particular value of negative gl' They showed that
the one-dimensional Hamiltonian can be described in terms of two kinds of
degrees of freedom: charge- and spin-density oscillations. The first gives
rise to gapless phonon like excitations, while the latter results in an excita-
tion branch with a gap. Luther and Emery were able to diagonalize this
Hamiltonian for the case g, = -6/ 5 and to calculate explicitly the low-
frequency behavior of the susceptibilities. Their results were extended to
all values of g] < 0 by Lee.( 7n ,
Our present understanding of the types of order in a single metallic
chain is summarized in Figure 14 taken from Reference 77. In the lower
half-plane there is a region in which the CDW and SS susceptibilities both
diverge; however the CDW divergence is stronger on the right-hand side of
the line g] = 2g 2 and the SDW divergence is stronger on the left-hand side.
Note that the SDW and TS types of order can exist only in the upper
half-plane. The picture in the lower half-plane of Figure 14 differs in some
details from other treatments of this region (for example, Reference 76)
because it contains the effect of the g4 process (in a real system g2 = g4)'
The general conclusion one can draw from the picture illustrated by
Figure 14 is that the regions of the various types of order at T = 0 are
strikingly similar to the regions determined by the highest mean-field
transition temperature (Figure 12). The basic feature of the diagram in the
(g" g2) phase is that the sign of 2g 2 - g] distinguishes between super-
conducting (negative sign) and charge- or spin-density wave (positive sign)
behavior. One can understand why this combination of the coupling
constants is in one dimension the effective interaction that determines the
low-temperature (and low-frequency) behavior of the response functions.
TS
SS
Figure 14. Regions in the (g" g2) plane in

which the response function for the indicated
types of order diverge. The dotted and
dashed lines represent the boundaries of the
'- -- -- -- 91 =- 2
region in which the model can be solved.
First, one can show that the g4 process only serves to modify the "sound"
velocity of the charge-density oscillations. The g2 process occurs with a
factor of 2 because of the two possible relative spin orientations of the two
electrons. The back scattering process is different. If the spins of the two
electrons are parallel, then backscattering in one dimension is indis-
tinguishable from a forward scattering process with exchange. Therefore it
appears in the effective interaction with a minus sign. Backscattering with
anti parallel spins, on the other hand, does not contribute to the low-
temperature behavior of the response functions because this process is
described by the spin-density degrees of freedom and these have a gap in
their excitation spectrum.(7'))
4.3. Relevance of "g-ology"
In Section 4.2 we described the possible types of ordering at T=O in a

single metallic chain. This description is based on a very simple model of an
electron gas with two coupling parameters gl and g2' These parameters
contain all the contributions to the electron-electron interaction without
distinction between the direct Coulomb interaction and the interaction
induced by phonons or excitons. In the case of phonons the basic difference
between these two interactions is the large difference in their energy
cutoffs. This problem has been discussed recently,(78.79) in the context of
the one-dimensional electron gas, but it should not concern us here,
because the exciton-induced interaction has the same cutoff as the Cou-
lomb interaction. Another subtle feature of the phonon- (exciton-) induced
electron-electron interaction is the effect of retardation, which is ignored
when the interaction i~ replaced by a static interaction. The simple model
described above also neglects any effects that might arise from the presence
of impurities. It is not clear how these two effects would modify the details
of the picture represented in Figure 14 for single chains, but some work on
retardation effects in coupled chains has been done and will be discussed in
Section 4.4. Nevertheless, we believe that the gross features of this picture
do represent "real physics," in the sense that they indicate the relation
between the properties of the interaction and the type of condensate which
the system tends to develop.
One reason why it is hard to relate the g-ology picture to real materials
i~ that there i~ no way known at present to measure the interaction
constants gl' Nevertheless, without being too specific, we believe that all
the presently di~cussed charge-transfer and metallo-organic compounds
are represented by points in the (gl, g2) plane lying on the right-hand sides
of the line 2g 2 c- g I. Let us now discuss the location of the proposed
excitonic system in this plane. The explicit expressions for these two
parameters in our model are [see Equation (29)]
g, = V(2PF)- E(2pF)+2/g(PF)/ (45)
2/Q(OW
(46)
where the first term is the screened Coulomb interaction. We have already
mentioned that the electron-exciton coupling constant decreases rapidly
with increasing momentum and is negligible at q = 2PF' This results in a
small but positive value of g,. If the balance between the Coulomb and the
exciton-induced interaction at q = 0 is favorable, gz becomes negative.
Calculations on the model discussed in Section 5 show that this is the case
and that the system considered is represented by a point on the left-hand
side and quite remote from the line 2gz = g! (Figure 12). This puts our
system in the region in which both the rigorous treatment and simple
mean-field theory Equation (42) predict triplet superconductivity.
The occurrence of triplet rather than singlet superconductivity in our
model calls for an interesting comparison with He 3 . This is one case where
superftuidity ("superconductivity" in a neutral fermion system) was pre-
dicted successfully by the BCS theory in a regime that is completely
different from its usual area of application. If He 3 were a one-dimensional
system we would place it in the (g" gz) plane in the neighborhood of the
point representing our model of the excitonic superconductor. The inter-
action in He 3 has a short-range repulsion (hard core) and a long-range
attraction. In momentum space this corresponds to negative gz and positive
g" which places it in the region of triplet superconductivity, and indeed,
He 3 is a triplet "superconductor."
There are two effects that might change the above prediction of triplet
and not singlet pairing. First, if there is also a phonon-induced electron-
electron interaction then this will affect mainly the value of g!. If this
interaction is sufficiently strong it will result in a negative value of gj. The
system will then lie in the region of singlet rather than triplet superconduc-
tivity. Second, the effect of impurities may result in the suppression of
triplet pairing so that singlet pairing will become relatively more favorable.
This will be discussed in Section 4.5.
4.4. Interchain Coupling
We have argued that extremely long-lived currents may exist in linear

chain systems even without a phase transition. However, if one wants to
discuss real phase transitions in such systems one has to include interchain
coupling. The nature and strength 01 this coupling determines the actual
transition temperature and may also affect the g-ology diagram in the
plane of the single-chain interaction parameters.
One approach to the problem of interchain coupling is to start from

exact solutions for the single chain and to introduce interchain coupling as
a perturbation. In thi5. approach the longitudinal correlations are treated
exactly and the transverse correlations in mean-field theory.(RO) A typical
expression for any response function in this approximation is
Xld(qll)
X(qll,q~T)=l_V( ) () (47)
q~ Xld q"
where Xld is the exact one-dimensional response function, qll (q~) are the
longitudinal (transverse) momentum components, and V(q) is the inter-
chain interaction.
The problem of weakly coupled linear chains was investigated by
Klemm and Gutfreund(Hl) using the single-chain solutions found by Luther
and Emery.(7'i) Scaling arguments make it possible to generalize these
results to the gl = 0 half-plane. Two types of interactions were considered.
One may be referred to in a broad sense as an interchain Coulomb
interaction in which the interacting electrons are confined to their chains,
and the other is the interchain single-particle tunneling interaction. In the
Coulomb interaction, one may distinguish between two processes: inter-
chain forward scattering and interchain back scattering. Inclusion of the
first of these two proce~~e5. still gives an exactly soluble model; however,
this type of interchain coupling does not give rise to finite transition
temperature~, but merely modifies the exponents of the response functions.
It also changes somewhat the boundaries between the regions correspond-
ing to the different type~ of order. The backscattering Coulomb interchain
coupling and "mgle-particle tunneling arc represented by the terms
Jeh ,= V
nn \s
f
i: 2.: dx"'::(x)"';~'(x)"'7:'(x)"'~s(x) (48)
where "'~(x) is the field operator which destroys an electron with spin
component s on one or the other side of the Fermi "surface" (i = 1 or 2).
The index n denotes the chain and the summation extends over all nearest
neighbors. The Coulomb interaction [Equation (48)] may be decoupled
into a product of two term~ descnbing the order parameter for a charge-
density wave on <,eparate chains, (",i,(X)"'2s(X)/. There is no way to get
from this term, in first order, coupling between the pairing order
parameters, (",i,(x)",;,(x», on two chains. Therefore, this interaction can
only give a finite tran~ition temperature to the charge-density wave state.
The tunnelling interaction [Equation (49)] may be decoupled, in second
order of perturbation theory, to represent an interchain coupling between
order parameters of both kinds. Therefore, motion of electrons from one

chain to another may result in a phase transition at finite temperature
either to a charge density wave or to a superconducting state. In the model
of weakly coupled chains one gets a real phase transition to a particular
type of ordered state only if the single-chain response function cor-
responding to this type of order diverges. As a result of this, one finds that
for g] - 2g 2 < 0, the system will undergo only a CDW phase transition. In
the region where the one-dimensional SS response function is divergent but
the CDW response function is not divergent, the tunneling interaction
ensures that the system will be superconducting. In the intermediate region
where both one-dimensional responses are divergent but the SS function
predominates, there are competing effects, as the Coulomb interactions
favour CDW, but the effect of tunneling is to favor the SS transition. The
transition temperature itself depends on the strength of the interchain
coupling. There is no such detailed treatment of the upper half plane,
gl > O. Mihaly and S6Iyom(H2) have discussed this region for the case of
interchain Coulomb coupling and they find essentially the same results. A
similar treatment which studies the transition between one- and three-
dimensional behavior in a system of parallel chains coupled by the Cou-
lomb interaction was presented by Menyhard.(83)
Another approach to the problem of interchain coupling is to consider
the coupled chain system as a highly anisotropic three-dimensional system.
This anisotropy may be represented by an electron energy dispersion
relationship of the form
F (k) = E (k z )+ 1/EF(COS akx + cos aky) (50)
where a is the distance between neighboring chains and 1/ is a measure of

the interchain coupling, which is given essentially by the ratio of the
interchain and intrachain transfer integrals. The advantage of this
approach is that for sufficiently strong interchain coupling, that is, when the
anisotropy is not too large, the mean-field theory is a good approximation.
One can therefore compute the transition temperature in the mean-field
approximation and then check a posteriori, by calculating the effect of
fluctuations, when this approximation breaks down.
Maniv and Weger(X4) have estimated along these lines the effect of
fluctuations on the superconducting mean-field transition temperature
using the Ornstein-Zernike approximation and find that Tc is reduced
significantly only when 1/ < O. I. This approach was also used to discuss(85)
the effects of fluctuations on the PeierIs transition in an electron-phonon
system described by the Frohlich Hamiltonian
(51 )
where ab b k are the electron and phonon destruction operators, Ek and Wq

are their respective energies, and Gq is the electron-phonon coupling
constant. The previous approach based on exact solutions for the single
chain cannot be applied in this case, because no such solutions are known
for the Frohlich Hamiltonian, which contains retardation effects. We shall
summarize the main results of Reference 85. The mean-field transition
temperature Tp to the Peierls state is almost independent of 11 up to a
certain critical 11c depending on the nature of the electron dispersion E(k z ),
at which it drops rapidly to zero. There exists a characteristic tempera.ure
To = 11 T F /4, which plays an important role in the model. At this tempera-
ture the inverse transverse correlation length crosses the Brillouin-zone
boundary. If the mean-field transition temperature Tp happens to be
smaller than To, the thermodynamic fluctuations will have a small effect,
suppressing the actual transition temperature by at most 20%. If, however,
the mean-field Yr, exceeds To, we expect large fluctuations, and mean-field
theory breaks down. Whether a given system may be described reasonably
well by mean-field theory depends in this model on the values of 11 and the
electron-phonon coupling. This picture was found useful to explain recent
pressure experiments on KCp.(R6) Pressure is expected to increase 11 and
these experiments show that Tp is in fact enhanced by pressure. The Peierls
transition in the Frohlich Hamiltonian has also been treated in a model in
which the interchain coupling is provided by the Coulomb interaction(87)
and by the three-dimensional phonons.(88)
The model of the anisotropic electron dispersion relation was also
used(il'),90) to investigate the competition between the Peierls and the
superconducting transition in the region when mean-field theory applies,
namely, when T/ is sufficiently large but there is still nesting of the Fermi
surface so that an uncoupled channel calculation would yield a Peierls
transition. The relevant effective electron-electron interactions are those
induced by the exchange of phonons with momentum q = 0 and q = 2k F• In
the language of g-ology they are described by negative values of gJ, g2,
defined as
gl = 2G~kN(O)/ W2k F
(52)
g2 = 2G~N(0)/ Wo
The calculation was performed in a double Nambu formalism which treats

the two transitions on the same footing. It turns out that the boundary line
separating the two transitions in the (gJ, g2) plane depends on the phonon
frequency Wo (an Einstein phonon was assumed) (Figure 15). The line
2g2 = gl is obtained as Wo--> 00 corresponding to the static interaction in the
electron gas picture. The deviation from this line for finite Wo represents
the retardation effects of the electron-phonon interaction. It is interesting
+--t- -+ + +--+-~ i
S
g2
P wo/Ec = 0.0.3(0)
wo/Ec= QIQ(b)
b wo/Ec = 0..3 (c)
c
wo/Ec- 00 (d)
Figure 15. Effects of retardation on the phase diagram of the superconducting (S) and Peierls
(P) phases for a coupled linear chain system. Ec is the Coulomb cutoff energy and Wo the
phonon cutoff.
to note that they tend to increase the region of the Peierls ordering.
Another result of this investigation is that high superconducting transition
temperatures are possible only for high values of wo, of the order of an
electronic cutoff energy.
4.5. Localization and Impurities
The discussion in thc preccding subsections assumes that the electrons

occupy band states with definite momentum and propagate coherently
throughout the whole system. However, it is known that any disorder
caused by the presence of impurities and defects, which are unavoidable in
a real material, results in a spatial localization of all the states in the
band. (91) The localization length, namely, the distance over which the
amplitude of the wave function drops to e·· I of its maximum value, depends
on the energy of the state, on the nature of the impurities, and on their
distribution.
We shall sketch briefly a proof(92) of localization based on the transfer
matrix approach introduced by Borland.(93) Let us assume a simple model
of a linear crystal represented by random potentials concentrated around
equally spaced lattice sites, so that between two adjacent lattice sites there
is a free region in which the wave function is given by
(53)
The transfer matrix connects the amplitudes A and B in two consecutive

intervals
(54)
Borland has shown that the transfer matrix Tn in the case of equally spaced
potentials is
T = ((1- iAn) e 'kl
-iA ) (55)
" iAn (1 + iAn)ne -Ikl
where I is the lattice spacing and An is equal to ILn/2k, where ILn is a

parameter that characterizes the potential at the nth lattice site and k =
E1/2. The relation between the wave-function amplitudes in the first and
(n + 1)th interval is
(56)
In a band state the wave-function amplitudes in the intervals between

different sites are related by phase factors. This is the case when the overall
transfer matrix connecting any two such intervals can be diagonalized by
a unimodular transformation to give eigenvalues e±18. For a periodic lattice
when all the An are equal the condition for this is Tr(T)s 2, where Tr(T) is
the trace of the matrix T This is the band condition and determines the
band energie~. When this condition is satisfied by the single-step transfer
matrix, it will also hold for the n-step transfer matrix Tn for any value of n.
In the case of random values of An the band condition will be violated for a
sufficiently long segment of the chain even if every single-step matrix
satisfies Tr( Tn) S 2. The reason this happens is that the square of deviations
from the mean of the parameter An accumulate on the diagonal of the
n-step transfer matrix. It can be shown that the trace of the ensemble-
averaged n-step transfer matrix grows as one moves along a chain of
random potentials. The diagonalization of this matrix then results in the
two eigenvalues
E\~d = exp[±arccosh(T(n)/2)] (57)
so that one gets an exponentially growing and an exponentially decaying
state. There is also an extensive literature on localization in one-dimen-
sional systems ba~ed on a different approach,<94.95) which relates the local-
ization length to the density of states. We have chosen to describe here this
particular proof hecause it will help us to discuss another point that will
come up later.
In discussing the effects of localization on the properties of one-
dimensional systems one should first of all worry about their influence on
the normal hehavior, hecause localization is also harmful to normal
conductivity. It was recently proved rigorously that a one-dimensional

system with localized states has a zero static conductivity independent of
the localization length.(96,97) However, in real systems there are several
effects that allow the flow of electric current. One is interchain coupling,
which must have a delocalizing effect because the possibility of hopping
from one chain to another provides alternative paths of propagation and
enables the electron:-. to avoid defects or impurities. As far as we know
there is no quantitative study of this effect. Another effect that may cause
de localization is the interaction with phonons; this has recently been
studied by Gogolin et al. (98) In addition, electron-electron interactions may
prevent localization by random lattice potentials. In any case there is ample
empirical evidence that the filamentary conductors do conduct and can
have metallic conductivity in spite of a significant number of impurities and
lattice defects.
In addition to the effect of impurities on the conductivity of the normal
state one might expect the impurities to favor or inhibit the tendency for
the formation of different types of condensates. This problem was first
considered by Zavadovski(99) for the charge-density wave and supercon-
ductive phases. Zavadovski considered a model of nonmagnetic, weak
scattering center~ characterized by a potential strength much smaller than
the Fermi energy. The key assumption of this model is that impurity
scattering does not mix left- and right-going states. Within this approxima-
tion he showed that such impurities have no effect on either the Peierls or
superconducting transition temperature. The formal reason for this is that
the vertex function~ whose divergence determines the transition tempera-
ture in the Bychkov et at. model(71) [Equation (44)J are merely multiplied
by pha~e factors whIch do not change the location of the poles. This result
can be understood by reference to the transfer matrix formulation of the
problem of impurities. The basis of Zavadovski's result is that to first order
in An one may neglect the off-diagonal terms of the transfer matrix because
in diagonalizing this matrix they only occur in the square. The diagonal
terms 1 ± iAn can be ab<.,orbed in the exponential and this simply multiplies
the state function by a random phase factor at each step. In one dimension
this cannot cause interference and hence the correlation functions of the
phases are unchanged by such scattering. The error in this approach is the
result of the buildup of the diagonal and off-diagonal terms by multiple
scattering which eventually results in localization and thus a mixing of the
left- and right-going states,
The effect of impurities on the Peierls transition was considered by
Patton and Sham(IO()) using a diagrammatic approach. They showed that
impurities tend to suppre:-.s the Peierls transition and argued further that
this prevented the three-dimensional ordering of the Peierls state between
strands of KCP.
Larkin and Mel'nikov(IOJ) studied the effects of impurities on the

Peierls and superconducting states in a system of coupled chains. They
concluded that impurities tended to suppress the transition to the Peierls
state more strongly than the superconducting state. They also suggested
that in the region where both condensates tend to develop (for weak,
negative coupling constants) deliberate doping could be used to suppress
the formation of a static COW and thus give the superconducting state.
According to the theorem of Anderson(46) nonmagnetic impurities do
not affect superconductivity. For scattering within a chain the only effect
such impurities should have is a change of coherence length and a change
of Tc due to a reduction in the density of states N(O) by such scattering. By
the same reasoning nonmagnetic impurities should have a dramatic effect
upon the transition temperature of a triplet state superconductor.(102) For
this reason triplet state superconductivity may be suppressed by deliberate
doping with nonmagnetic impurities to yield singlet superconductivity.
Finally one should comment on the effects of localization by impurities
of Cooper pairs. No work on this has been reported to date, but it is
plausible to assume that once Cooper pairs are formed in a certain local
region they could be delocalized by Josephson tunneling to neighboring
regions. One could conjecture that if the localization length is not so small
as to prevent the formation of Cooper pairs then complete delocalization
would occur at a sufficiently low temperature.
4.6. Effects of Screening
From the earliest work on polymeric systems it was apparent that the
degree to which the Coulomb field of the electrons would be screened in
such materials would depend upon the direction considered relative to the
conductive spine. Rough estimates were made using the Thomas-Fermi
approximation, which suggested that the screening length along the spine
would be of the order of 3 A and appreciably greater in the transverse
direction. This wa~ criticized by Kuper,(59) who considered a model system
consisting of a single isolated spine with appendant side chains filled with a
uniform electron gas. On the basis of this model and a detailed application
of the Thomas-Fermi method to it he concluded that no significant screen-
ing of the Coulomb field would occur and that oscillations of charge in the
side chains would not provide a net attractive interaction. However, these
conclusions are not generally true and result from the model, which is a
poor description of the molecular system originally suggested. In partic-
ular, the latter conclusion of the absence of a net attraction results from
the treatment of the charge oscillations in the side chains as plasma oscil-
lations in the electron gas and the violation of the condition requiring the
separation of the conduction and excitonic system. Treating the electrons
The Prospects of EXCItOnlC SuperconductIVity 353
(?)))«)]T?\<]?~n)222)))~
...... .................... ..... .................................. ............ .
, ,',
.... ,." .. ..
.. ....
-:.:':':':-:":':'"1---;..---+-1
:::::::::::::::::
.........
...... .. ....
..... .......
....... ....
..............
..............
-:.>:->:.:- -:-:-:.:-:-:-:-:
:::;:;::::::: ::::::::::::::::
I \!
R
......
, ...... .
...
.... .....
.·.....
.. .
. . ., ...
.. ...... .
........
..... . .. ..
........ .
· . .. , . . ,.
. ........... .. .......
..· . ,. .... .,. , ... . .
........
.... ...
.......
..........
.......... .....
.......
........ .. ............
.. .....
............ ..
Figure 16 Two zone geometry
used by DavIs (Reference 104) to
- ~&:~~ >~:~:»~
...............
calculate screenmg m filamentary
compounds
In a polarizable molecule dS a plasma grossly overesttmates the energy of

characteristic absorptIOn of the molecule Secondly, the absence of
substanttal screenIng In the model at long wavelengths results from the
filament beIng treated In IsolatIOn In any real material one would not have
an Isolated filament but rather a compound with many such molecular
filaments parallel to one another Dzyaloshmskll and Kats(I03) showed that
m such d configuratIOn '>creenmg m the long wavelength hmlt approaches
that of a homogeneous metdl This has the Important effect of weakemng
the Coulomb mteractlOn dt small momenta, where, as we shall show, the
exciton mteractlon make" ItS maximum contributIOn
Numerlcdl estimate'> have been made of the screenmg 10 hlamentary
compounds of the Krogmann type by DaVIS (104) He considered a model 10
which the contribution of neighboring filaments were sImulated by replac-
109 them by an outer cyhnder of a low-densIty electron gas separate from
the central filamentary '>pIne as shown In FIgure 16 USIng the Thomas-
FermI approXimatIOn he found nearly ISOtroPIC screen 109 10 the regIOn near
the spIne of the usual form V(r)=(e 2 /r)exp(-Kr) but With d screenIng
length 1/ K substdntlally greater than that for a bulk metal Values for thIs
screen 109 parameter were calculated for a model deSigned to represent the
partially OXIdIzed Pt-cham compounds HIS results for KCP are gIVen In
Figure 17 In DdVI~' model only static screenmg was considered, and
clearly thIS IS msuftlclent for the full calculatIOn of the net effective mter-
actIOn over an energv [dnge of the order of the eXCIton energy =2 eV
1.r-~~----~-----r----,-----.-----r----'
/COULOMB POTENTIAL
I I/R
~(R)
Z=O
FILAMENT AND
CYLINDER SCREENING
.01 Keff=O.I11 A-I
RI<R<R2
FULLY SCREENED
K=0.88A- 1
,
.0001 IL. _ _ _L..-_ _ ~ _ _L -_ _'---_ _L -_ _L - _ - . . . . J
o 2 4 6 • 8 10 12 14
R(A)
Figure 17. Results obtained by Davis (Reference 104) for the screened potential cf>(R) along
a line perpendicular to the filament axis (Z = 0) for the model system illustrated in Figure 16.
Bush(105) has extended this model to calculate the dynamic screening

of the Coulomb field both in the Thomas-Fermi and the random-phase
approximation. He finds that because of the large density of states at low
momenta in a one-dimensional metal, extremely strong screening occurs at
low Fermi energie~ and thus at low electron densities. In fact, over a
limited range the screening in one dimension is inversely proportional to
the clectron density. Also, the frequency dependence of the screening has a
considerable amount of structure in it due to the various plasmon modes of
the interacting filaments. He showed that if the response function of a
given spine is known, one may compute the lattice sums exactly to calculate
the response of the sample as a whole without resorting to the concentric
cylinder approximation of Davis.
Thus, while the essential problem of screening in the filamentary

compounds has been solved, many details remain to be clarified. In partic-
ular one would expect that structural imperfections, a Peierls distortion,
or interrupted strands would severely modify the low-frequency screening
in these compounds. It is not clear, though, whether this would have any
impact on factors responsible for superconductivity because such materials
would still have a large real dielectric constant(lOo) and the long-wavelength
electron-electron interaction would be severely reduced without neces-
sarily being screened in the strict sense. In fact, in the model discussed in
Section 5 one finds that the principal reduction in the Coulomb field does
not come from metallic screening but from the dielectric function of the
organic material within which the conductive chain is located. The complex
behavior of the dynamic screening in filamentary compounds does not
allow one to draw many general conclusions on the strength of the screen-
ing but rather requires one to consider each class of materials individually.
In an excitonic system both the polarizable ligands and the conduction
electrons would contrihute to the screening and each would affect the
other, ~o one really needs to ~olve the problem of screening in a self-
consistent manner and not by treating the screened excitonic interaction
and the screened Coulomb field as simply additive interactions. In cal-
culations of the effective interaction in these materials the greatest
uncertainty lies in our lack of adequate knowledge of the screening.
Measurements that could reveal the strength of the screening by optical,
electron energy los~, NMR, or other experiments would be of great value
in establishing an experimental reference point against which to compare
the theoretical predictions.
The related problem ot screening in layer compounds or two-dimen-
sional structures ha~ also been studied in some detail. Interest in this lies in
understanding of the inversion or accumulation layers(107) in semiconduc-
tors and surface electrons on liquid He or Ne.(108) Again screening is
virtually absent in any single isolated layer because the interaction is
dominated by contributiom from the electromagnetic fields which extend
out into the surrounding nonconductive medium. However, in layered
structures of many layers one again recovers three-dimensional-like
screening with the Yukawa form (e/ r) e Kr for the screened interaction, but
the screening parameter K differs from the hulk Thomas-Fermi value.
Visscher and Falicov ' Iii") have considered the layered structures. A
complete treatment of the full electrodynamic behavior and local screening
of both the isolated ~heet and the periodic array of such sheets has been
given by Fetter. (110 I II) The plasma o~cillations differ from those of an
isotropic bulk material with one branch having a finite energy and all
others tending to zero in the long-wavelength limit. Knowledge of this
screening would be Important in considering any proposed two-dimen-
sional excitonic sy"tem.
5. Real Models
The preceding theoretical arguments lead us to the conclusion that

there are no known matters of principle that could prevent an excitonic
system from superconducting at temperatures at least as high as the Debye
temperature. Whether or not this can be realized in practice then depends
upon the strength of the coupling that can be realized in a given system. To
examine this point one has to be specific about the systems. Thus far only
two systems have been discussed in any detail taking into account recent
development~ in theory. These are the two-dimensional layered system
discussed by Allender et al.(4) and a filamentary model discussed by Davis,
Gutfreund, and Little (DGL).(33) We have mentioned earlier that the
two-dimensional systems look less attractive now because of problems of
exchange and the probable failure of the Migdal approximation in the
systems considered thus far, and perhaps the most serious problem is the
weakness of the interaction. In all systems where reasonably detailed
calculations havc been carried out it appears that the interaction is
insufficient to overcome the Coulomb repulsion.(112.113) The only two-
dimensional systems where a reasonably strong excitonic interaction
appears to be possible and that would be free of exchange problems are
those with a very small Fermi energy,o 13) The reason fOT this is essentially
one of a matter of scale. In order for the excitonic system to have modes of
low enough energy to give a coupling constant of sufficient size the excitons
must have physical dimensions of the order of loA or thereabouts. For
these modes to couple to a significant fraction of the electrons, the Fermi
wavelength must be of somewhat similar magnitude. For one dimension,
only scattering processes involving very long (q = 0) and very short (q =
2k p ) wavelength~ are important in scattering between points on the Fermi
surface, while in two dimensions all momenta up to 2k p are important. So
for the excitons to have a significant effect on superconductivity in the
latter case, the coupling constant for all momenta up to 2k p must be large,
and this is only possible if k p is small.
To get a bctter idea whether high-temperature superconductivity
could occur in ~uch two-dimensional systems one needs to define the
system more precisely than has been done to date and then to examine in
detail the questions discussed above together with questions relating to the
screening of the ('oulomb field, the strength of the coupling constants, and
the stability of the system etc. Based on the calculations done to
date(112.113) and the absence of any indication of an excitonic contribution
to superconductivity in experiments reported thus far/ 5 ) we feel the pros-
pects for finding such superconductivity in the two-dimensional systems
are slim. On the other hand the prospects of satisfying the various
theoretical criteria for superconductivity in filamentary systems appear to
(0) (b)
Figure 18. Proposed model of the structure of an excitonic superconductor. (a) Top view of
square planar phenanthroline-dye ligands complexed to Pt. Double bonds in the chromo-
phore are omitted for simplicity. "Et" stands for ethyl. (b) Side view of chain.
be much better. The reason for this is simply that one can pack more of the
excitonic material more closely round a conducting thread than round a
conducting plane and thus in one dimension a stronger net attraction can
be expected. In addition, one can satisfy the condition for the validity of
the Migdal approximation, and for the stability of the superconducting
phase, furthermore. the effects of fluctuation and localization, while
important, do not appear to be disastrous to the superconducting state in
these systems. In the following section we discuss the essential elements of
the model proposed by DGL stressing the essential physical requirements
of the model, sketching the method by which Tc was calculated, and
discussing the elements of the chemistry that would be required to prepare
material of this general class.
5.1. Model of a Filamentary Excitonic Superconductor
The model that was proposed is illustrated in Figure 18. It consists of a

chain of Pt atoms surrounded by a sheath of highly polarizable dyelike
molecules. Each unit of the conductive chain of Pt atoms is a bis-phenanth-
roline ligand system complexed to the metal atom. To each of the 1-10
phenanthroline Iigand~ are attached two cyanine-dye chromophoric units
at the 4 and 7 po~itlon~.
In the proposed structure the polarization of the chromophore results
in the movement of a positive hole from one nitrogen atom remote from
the Pt site to the nitrogen adjacent to the Pt. Because of the large move-
ment of charge a '-.trong electron-exciton interaction can be expected.
Figure 19. Simplified version of the structure

of Figure 18 for which detailed calculations
were presented by DGL.
The d z 2 orbitals of the Pt atoms of the chain overlap with one another
to give a linear conductive pathway. Because of the repulsion between the
7T electrons of the bulky ligands the Pt atoms along the chain can only come
to within about 3.4 A of one another. It has been assumed that sufficient
overlap occurs at this separation to yield a conductive chain. A justification
for this is given later.
Four counter ions (CI -) for each square-planar ligand system are
required and the Pt chain needs to be partially oxidized to give a partially
filled d z 2 conduction band as in K2Pt(CN)4Bro3·3H20.:j: There is some
question whether this is possible in such a system, but this and other
variations of the model will be discussed later. Instead of working with the
large phenanthroline groups a skeleton structure of the chromophore units
alone was used, as illustrated in Figure 19. Calculations were presented on
this simplifled model. This gave the essence of the results of calculations of
the more complex system done earlier. The transition temperature Tc was
calculated by the method adopted from Kirzhnits et al.(36) This method
applies to a weak coupling superconductor and results in the BCS-like
equation for the gap function 4>(k) of Equation (19). The interactions were
represented by the kernel of Equation (29).
5.1. 1. The Interactions

The calculation of the kernel U(p, k) required a knowledge of the
electron-band energies ~(p), the exciton band energies E,,(q), the coupling
constants Q,,(q), and the Coulomb interaction between two electrons on
the spine. The actual values and variations of the first two quantities where
shown by DGL to have only a minor effect on the results within a wide
range of reasonable parameters. The interaction energies, however, are
crucial and had to be estimated in a reliable way.
:j: The paper by DGL contains an error here in stating that six negatively charged counter ions
would be required instead of four.
The electron-exciton coupling parameter has the form
O"(q)=O,,q, k-q/VIO, k) (58)

which corresponds to the \cattering of an electron from momentum k to
k - q, accompanied by the creation of an exciton of momentum q and band
index l¥. When the electron states on the spine are described in the
tight-binding approximation and the exciton-band states by a linear
combination of terms in which only a single molecule is in an excited state,
one obtains (DGL)
O,,(q) = (~) Ii 2 m~." f 14>(rdI 2 V(r" r2)C: n(q) e ,qRm pv(r2, Rnm) d 3 r1 d 3 r2
(59)
W
where 14>(rl is the density of the electron on the spine, Pv(r2, Rnm) is the
transition density between the ground state and the excited state v of the
dye at site n in the mth unit cell, and c: n are the coefficients of the exciton
band states. The method used for the calculation of p was a simple
extension of the Pariser-Parr-Pople semiempirical technique, (114.115) which
had previously been shown to give accurate values for the molecular
energy states and the oscillator strengths.(llh) The result for the lowest
excitation band of the pyridine cyanine is represented in Figure 20. It
shows a clear oscillating dipole pattern. The typical dependence of 1O,,(q )1 2
upon q for the mode l¥ which couples most strongly to the spine is
illustrated in Figure 21. Its distinctive feature is the sharp falloff in 1O,,(q t
with increasing q. This occurs at values of q = 1/ d, where d is of the order
of the distance of the nearest terminal group of the dye from the spine as
discussed earlier. This single low-lying level provided the excitonic inter-
action in the model. The other higher excitations of the ligands contributed
to Pc of Equation (21), as a correction to the Coulomb interaction.
The Coulomb interaction was found to be the hardest quantity to
estimate in a reliable way. The starting point was the parameters 'Ym which
measure the bare interaction between electrons on two atoms on the spine
separated by a distance 'n
= na, where a is the interatomic spacing. The
Nishimoto-Mataga(117) form was used for these parameters:
(60)
The parameter b, which corresponds to the interaction of two electrons on

the same atom, was derived from experimental values of the ionization
energy and the electron affinity, and for platinum this yielded 'Yo = 6.03 eV
giving b = 2.4 A. The bare interaction is modified by the screening of the
electrons in the same filament and in neighboring filaments. In addition, it
-0.0647
~-0.0274
Figure 20. Linear-combination-of-atomic-orbitals (LCAO)

0.0467 calculated values of the transition density for the principal
low-lying absorption band for the pyridine cyanine.
is reduced by the dielectric constant of the surrounding organic medium.

The first of these contributions to screening was estimated on the basis of
the calculation of Davis(104) on filamentary compounds like KCP. His
results were used to obtain the screened parameters
(61)
The Fourier transform of the partially screened interactions [suitably

corrected at large q for errors introduced by the discrete atomic represen-
tation (see DGL)] are plotted in Figure 22 (upper curve).
The higher-energy excitations were shown to reduce the long-range
Coulomb interaction by about a factor of 2. This is shown in the lower
curve of Figure 22. For the particular model used this reduction comes
from the 'Tr electrons alone. The inclusion of the (T-electron contribution
would reduce this interaction further. For q = 0 the total reduction factor
>
Q)
Figure 21. Calculated electron-exciton interaction IQ(q)12 as a function of momentum

transfer q.
UPPER CURVE· T F SCREENI NG

A:014
a : 3.4 (LATTICE CONSTANT)
b:2.4
6 LOWER CURVE . WITH HIGHER
EXCITATIONS
(4 PYRIDINE CYANINE DYES /CELL)
5
>
'"
I~
I
I
~~
o "TrIa 2 ..../0
q
Figure 22. Screened Coulomb interaction calculated USing Thomas-Fermi screening due to
electrons In the same and nelghbonng filaments (upper curve), and with the addition of
dielectric screening from the neighboring organic environment (lower curve).
should be of the order of the dielectric constant E of the surrounding
medium. This can be estimated from the refractive index n of a similar
unsaturated hydrocarbon through the relation n 2 = E. From this estimate
one would expect the Coulomb interaction at small q to be reduced by
about 2.6, so perhap!> (DGL) have overestimated the Coulomb repulsion
by as much as 3()°/t,.
5. 1.2. Calculation of T,
Using the geometry of the system illustrated in Figures 18 and 19

DGL calculated numerically the transition temperature Te. It proved to be
convenient to use, instead of Equation (19), the zero-temperature equation
for the gap
f
"/(l dk U(p-k)<fJ(k)
<fJ(p)=- ,,/a 41T [e(k)+<fJ 2 (k)]l/2 (62)
and to obtain Te from the gap at T = 0 through the relation

kTc = 3.5<fJ(k F )1 (63)
T=O
For singlet superconductivity the spins of the electrons of the pair are
antiparallel and <fJ(k) = <fJ(-k). Equation (62) can thus be written as
, 1 1"/<1 [U(p, k)+ U(p, -k)]<fJ'(k)
(64)
<fJ (p)= - 41T I) dk {g2(k)+[<fJ'(k)]}I/2
0 2k r
>., 2.0 +I
,
0-
",,'"
",,'"
:; Figure 23. Plot of U (kF' kF - q) vs. q for the
q
model system. The major contributions to
superconductivity come from the interaction
at q = 2kF (gl) and at q = 0 (g2)'
-4.0
On the other hand, for triplet superconductivity the orbital symmetry

requires cjJ(k)= -cjJ(-k) and Equation (62) gives
, ___1 [1T/a k [U(p, k)- U(p, -k)]cjJ'(k)
cjJ(p)- 47TJo d {g2(k)+[cjJt(k)]2}1/2 (65)
It should be noted that because of the denominator, the integral in

both cases is dominated by the contribution of the kernel at k = kF and
thus by the term~ U(PF, k F ) and U(PF' -kF)' In Figure 23 we show a plot
of U(PF' k F - q) vs. q calculated for the model of Figure 19. We see that
U(PF' k F ) is negative, owing to the strong exciton contribution of Equation
(29) resulting from the large value of IO,,(q)1 2 near q = O. On the other
hand, U(Ph -k r ) i~ positive because of the dominance of the Coulomb
repulsion for large momentum (2k F ) transfers. Because of this the
combination I U(PF' kF)- U(PF, -k F)], which occurs in the triplet case,
yields a more attractive contribution than does [U(PF' kp)+ U(PF' -k F )],
which occurs in the singlet case. Consequently we expect a larger zero-
temperature gap and 1~ for triplet-state superconductivity compared to
that for the "inglet case. This is borne out by the numerical result for the
two gaps iIlu~trated ID Figure 24.
GAP
0.020
0.0 10
--~-----+-7~---0 1-
Tr/O
- 0.01 1
k-
Figure 24. Calculated results for the
-0.02 singlet (symmetric) and triplet
(anti symmetric) gap functions for the
model system of Figure 1.
T.S.
(5.5.)
~ 1752K(T.S)
. - ' 1 1278K(S.S)
5.5. C.D.W.
(T.S.)
Figure 25. gl-g2 plane showing region of charge-density wave instabilities (CDW) and singlet
(triplet) superconductivity S.S. (T.S.). The calculated values of Tc for the model are indicated.
This also provide~ justification for using a single-channel calculation

rather than a coupled two-channel approach involving the competition
between both superconductivity and charge-density wave condensates, for
the representative point on the g-ology plane can be seen from Figure
25 to have gl(q = 2k r ) small and positive, and g2(q = 0) large and negative.
This puts it well away from the line g2 = 2g 1 which separates the charge-
density wave region from the superconducting region and thus in a region
where the competition between them should be negligible.
5.1.3. Results
The value of Tc for singlet superconductivity was calculated for

bandwidth between 2 and 4 e V and for Fermi energies corresponding to
1/2 and 5/6 filling of the band. These values of Tc ranged between 100 and
1280o K, the larger values occurring with increasing density of states at the
Fermi surface. The value of Tc for triplet-state superconductivity was
always found to be higher, in agreement with the predictions of g-ology
and our discussion of Section 4.2. DGL also noted that no superconducting
solution could be found for a similar model with only two polarizable ligands
per Pt atom instead of four or when the ligands were moved so that the N
atoms were 3 A rather than 2 A from the Pt.
Again one should note that these calculations are of the mean-field
transition temperature. This i~ not the temperature of any true phase
transition but the temperature at which one can expect the pair condensate
to begin to grow and for marked changes in the conductivity to occur.
Whether or not bulk superconductivity would occur near this temperature
would depend on interchain coupling and the degree of perfection of the

individual strands. None of these factors are considered by DGL.
While this calculation was done for a specific model, one can draw
some general conclusions from it. First it shows how sensitive Tc is to the
distance between the polarizable ligands and the spine. This follows from
the form of Equation (59) for the coupling term Oa(q) and the fact that the
superconductivity is determined by the difference between a term pro-
portional to 100,(q)/" and the Coulomb term [Equations (20) and (21)].
Both are large terms. If the exciton term is larger, one gets high-tempera-
ture superconductivity, but a small change in Oa(q) can reverse the
inequality and give Tc = O.
Second. it 'ihows that a very dense packing of the polarizable ligands
around the spine is essential. Reducing the number of polarizable groups
from 4 to 2 weakens the net O( q) too much to give a finite value for Teo
Third, the value of O(q) would be greatly reduced if the conduction-
electron charge were not concentrated in a small region close to the
polarizable ligands. For this reason it does not appear possible to obtain
excitonic superconductivity in a TCNQ type of salt where the 7T' electrons
are distributed over a relatively large area.
Fourth. we see that a comparatively narrow band with its large density
of states at the Fermi surface is helpful to superconductivity and results in
large values for T.
The specific model discussed above suffers from a number of short-
comings in the chemistry. First, it appears most unlikely that one could
form the bi~-phenanthroline structure with a charge of +4 on the ligand
system and have it stack. On the other hand, the doubly charged structure
would be formed more readily. but it would have a significantly smaller
transition charge and this would lead to weaker coupling. Second, the
model requires the d pz band to be partly filled, so the Pt chain must
Z 2 -
be oxidized. In this process the relatively sensitive double bonds in the

chromophorc units must not be attacked. This is a general problem and
attention must be given to the relative oxidation and reduction potentials
of the spine and ligands. Third, the ligand system may not stack in the form
postulated and then no metal-metal interaction would occur. This too is a
general problem and a better understanding of the factors that are
important for qacking is essential for further progress. Fourth, one is
relying on a significant metal interaction between the metal atoms even at a
metal-metal separation of 3.4 A determined by the van der Waal's
contacts between the ligands. This is substantially larger than the metal-
metal separation in KCP. Earlier theoretical estimates,(118,119) however,
and a recent band calculation(l:'O) along the lines of an earlier calculation
on KCp(1211 confirm that a substantial interaction would remain, so this is
perhaps les~ of a problem than was originally thought.
There are a number of alternative structures that retain the essential

elements of the above model but avoid some of its limitations. One could
replace the charged cyanine chromophores with uncharged merocyanine
chromophores. This should increase the probability of stacking and would
allow one to compensate for the lack of symmetry between the Pt and ethyl
terminal groups. The latter has the effect of reducing the oscillator strength
of the transition.(lW Alternately the use of an anion-cation structure as in
magnus green salt would favor stacking because of ionic forces.
5.2. Discussion
It is evident from the above that the most important factor in deter-
mining Tc is Q" (q). While it is difficult to estimate accurately the Coulomb
term, whatever it is, it can be expected to be more nearly the same for
different systems as it is determined, at least in the long wave limit, by the
gross features of the system. On the other hand it is the microscopic
properties of Q,,(q) that determine its behavior in this same limit. Thus
Q,,(q), particularly in the region q = 0, and £,,(q) in the denominator of
U(p, k) [Equation (29)J will determine whether or not a given system will
superconduct. For this term to be large the conduction electron density
11>(rl)1 2 in the spine must be near the positive (negative) part of the
transition density p" and far from the negative (positive) part. [See Equa-
tion (59).] In this expression V(r], r2) is the partially screened Coulomb
interaction. To maximize Q,,(q), Pv should be concentrated at the extreme
end of the polarizable ligand and oriented perpendicular to the chain axis.
In the above model the near part of p" is concentrated in the ring structures
adjacent to the Pt. The prospects of getting high-temperature supercon-
ductivity would be enormously enhanced if a structure could be devised in
which even closer approach of the transition charge and the spine could be
achieved. One pos~ibility is the use of some heteroatom immediately
adjacent to the metal atom of the spine to concentrate Pv in its neighbor-
hood. An even more promising possibility may exist in the d K compounds
where the spine atom could act both as the conducting chain and as the
terminal group of the polarizable ligand. This may appear to violate much
of what we have said in regard to exchange in Section 3.1. However, this is
not necessarily so, for if the 7T electrons of the ligand system are conjugated
to the d xz and d yz orbitals of, say Pt, then these electrons will be approxi-
mately orthogonal to all the 5d z 2 - 6pz states of the conduction band. The
orthogonality in this case is maintained by the different symmetry of the
dxAdyz ) states and both the d z2 and pz orbitals across the plane of the
ligand. So even though the hybridization of the d z2 - pz states will vary
through the band, orthogonality is maintained and the exchange terms will
vanish. For the dxAdvz) orbitals to participate with the ligands in this way
the Fermi energy would have to intersect both of these levels and the d z 2
level. If this situation could be realized in some dB system, it would open up
a rich field of structures free of many of the obvious chemical and steric
difficulties of the model of DGL.
6. Summary
In all superconductors known at present the phonon interaction is

believed to be responsible for the superconducting state. In principle,
however, other interactions are capable of yielding a similar state, and
some might do so at relatively high temperatures. In this review we have
examined in detail the possibility of the exciton interaction leading to a
relatively high-temperature superconducting phase. We find that it does
appear capable of doing so. However, we find that a simple application of
the theory of superconductivity to this problem cannot be justified.
Instead, one must examine or reexamine each of the basic assumptions
customarily used in this theory. These involve questions of stability, vertex
corrections, and the form of the gap equation, together with an ab initio
determination of the interactions. In addition, certain other factors must be
considered that are unique to the exciton mechanism, such as the effects of
exchange between the exciton and conduction subsystems, and the effects
of phonons on the exciton interaction. We have discussed each of these in
turn.
We have argued that the most promising systems in which one could
hope to find excitonic superconductivity are filamentary or one-dimen-
sional system~. Such systems, because of their limited dimensionality, do
not exhibit any true phase transitions, but instead their behavior is
dominated by fluctuations that persist down to T = O. In spite of this, high
conductivity or other partially ordered phases can occur for certain values
of the electron-electron coupling constants. We have discussed these
factors and the effect of coupling between the filaments upon these phases.
In addition, the effects of impurities and random faults in the filaments
have been considered.
Finally, one ~pecific model of a proposed excitonic superconductor is
examined, which, if it could be synthesized, can be expected to have a high
superconducting transition temperature. We believe that such high tran-
sition temperatures should occur and that all known objections to the use
of this mechanism to obtain superconductors with substantially higher
transition temperatures than those known today have been answered.
Several important theoretical questions remain and the field can be expec-
ted to continue to generate a rich source of novel problems. The problem

of the design of a tractable material and its synthesis which would bring
~uch matmial5 to reality remain~ a challt:mge to both the phY5ici5t and
chemist.
ACKNOWLEDGMENTS
We are indebted to the National Science Foundation (Grant No.

DMR-74-00427-A(3), the National Aeronautical and Space Agency
(Contract No. JPL 953752), and US-Israeli Binational Science Foun-
dation (Grant No. 215) for financial support during the period when much
of the above analysis was done. In addition we owe our thanks to the
NATO Scientific Affairs division (Grant No. 918) for travel support which
made this collaborative work possible. To Professor Heimo Keller, of the
Anorganische Chemisches Institute der UniversiUit, Heidelberg, we wish
to express our gratitude for the warm hospitality of his Institute, where the
review was completed.
References
1. L. B. Coleman, M. J. Cohen, D. 1. Sandman, F. G. Yamagishi, A. F. Garito, and A. 1.

Heeger, Superconducting fluctuations and the Peierls instability in an organic solid,
Solid State Commun. 12,1125-1132 (1973).
2. W. A. Little, Possibility of synthesizing an organic superconductor, Phys. Rev. 134,
A1416-A1424 (1964).
3. V. L. Ginzburg, Concerning surface superconductivity, Zh. Eksp. Teor. Fiz. 47, 2318-
2320 (1964) [Sov. Phys.-JETP 20, 1549-1550 (1965»).
4. D. Allender, J. Bray, and 1. Bardeen, Model for an exciton mechanism of superconduc-
tivity, Phys. Rev. B 7, J020-1029 (1073).
5. M. Strongin, A search for excitonic superconductivity, Solid State Commun. 14, 88
(1974).
6. Proceedings International Conference on Organic Superconductors, W. A. Little (ed.)
1. Polymer. Sci. Pt. C No. 29 (1970).
7. L. V. Keldysh, Superconductivity in nonmetallic systems, Usp. Fiz. Nauk 86, 327-333
(1965) [Sov. Phys. Usp. 8, 496-500 (1965)].
R. V. L. Ginzburg, The problem of high temperature superconductivity, Con temp. Phys. 9,
355-374 (1968).
9. V. L. Ginzburg, The problem of high-temperature superconductivity. II, Usp. Fiz. Nauk
101, 185 (1970) [Sov. Phys. Usp. 13 (3), 335-352 (1970)].
10. V. L. Ginzburg, The problem of high-temperature superconductivity, Ann. Rev. Mater.
Sci. 2, 663-696 (1972).
11. V. L. Ginzburg and D. A. Kirzhnits, On the problem of high temperature superconduc-
tivity, Phys. Rep. Phys. Lett. C (Netherlands) 4,343-356 (1972).
12. L. N. Bulaevsky, V. L. Ginzburg, D. 1. Khomskii, D. A. Kirzhnits, lu. V. Kopaev, E. G.
Maximov, G. F. Zarkov, and G. P. Molulevitch, The problem of high temperature
superconductivity, ParI', I and II, Prcprints N45 and N74, Lebedev Physical Institute,
Moscow (1974).
13. J. 1. Andre, A. Bieber, and F. Gautier, Physical properties of highly anisotropic systems:
radical ion salts and charge transfer complexes, Ann. Phys. (Paris) 1, 145-256 (1976).
14. E. B. Yagubskii and M. L. Khidekel, High temperature exciton superconductivity:
synthetic aspects, Russian Chem. Rev. 41, 1011-1026 (1972).
15. I. F. Shchegolev, Electric and magnetic properties of linear conducting chains, Phys.
Status Solidi (A) 12,9-45 (I 972).
16. H. R. Zeller, Electronic properties of one-dimensional solid state systems, Adv. Solid
State Phys. 13. 31-58 (1973).
17. J. S. Miller and A. J. Epstein, One-dimensional inorganic complexes, in Progress in
Inorganic Chemistry, Vol. 20, pp. 1-151, Stephen 1. Lippard (ed.), John Wiley & Sons,
New York (1976).
18. H. 1. Keller. Low-Dimensional Cooperative Phenomena, NATO-ASI Series B7, Plenum
Press, New York (1975).
19. H. J. Keller, Chemistry and Physics of One-Dimensional Metals, NATO-ASI Series
B25, Plenum Press, New York (1977).
20. B. W. Roberts, Superconductivity, in Handbook of Chemistry and Physics, 48th Ed., pp.
E75-E90, R. C. Weast and S. M. Selby (eds.), The Chemical Rubber Co., Cleveland,
Ohio (1967).
21. J. Bardeen, L. N. Cooper, and J. R. Schrietfer, Theory of superconductivity, Phys. Rev.
108,1175-1204 (1957).
22. S. V. Tiablikov and V. V. Tomachev, The interaction of electrons with lattice vibrations,
Zh. Eksp. Teor. Fiz. 34, 1254-1257 [Sov. Phys. JETP 34,867-869 (1958)].
23. L. R. Testardi Structural instabilities in A-15 compounds, Rev. Mod. Phys. 47, 637-
648 (1975)
24. J. R. Gavaler, Superconductivity in Nb-Ge films above'22 K, Appl. Phys. Lett. 23,
480-482 ( 1973).
25. W. L. McMillan, Transition temperature of strong-coupled superconductors, Phys.
Rev. 167,331 344 (1968).
26. P. Morel and P W. Anderson, Calculation of the superconducting state parameters with
retarded ekctron-phonon interaction, Phys. Rev. 125, 1263-1271 (1962).
27. G. Rickayzen, III Superconductivity, Vol. I, pp. 72-115, R. D. Parks (ed.), Marcel
Dekker, New York (1969).
28. W. A. Little, J Polymer Sci. Pt. C 29,17-26 (1970).
29. B. T. Geilikman, A possible mechanism for superconductivity in alloys, Zh. Eksp. Teor.
Fiz.48, 1194-1197 (1965) ISov. Phys.-JETP 21,796-798 (1965)].
30. M. L. Cohen and P. W. Anderson, in Superconductivity in d- and f-Band Metals, D. H.
Douglass (cd.). AlP. New York (1972).
31. J. C. Phillips. Superconductivlly mechanisms and covalent instabilities, Phys. Rev. Lett.
29,1551-1'\54 (1972)
32. D. Pine, and P NOZieres, The Theory of Quantum Liquids, W. A. Benjamin, New York
(1966).
33. D. Davis, H. Gutfreund, and W. A. Little, Proposed model of a high temperature
excitonic superconductor, Phys. Rev. B 13,4766-4779 (1976).
34. 1. R. Schrietfcr. Theorv of Superconductivity, W. A. Benjamin, New York (1964).
35. S. Engelsberg and 1. R. Schrietfer, Coupled electron-phonon system, Phys. Rev. 131,
993-100H (196,).
36. D. A. Klrzhnit,. E. G. Maximov, and D. I. Khomskii, The description of superconduc-
tivity in terms of dielectric response function, J. Low Temp. Phys. 10,79-93 (1973).
37. D. B. Chesnut. fxciton renormalization in conducting molecular solids, Mol. Cryst. 1,
351-375 (1966)
38. P. M. ChaiklO, A. F. Garito, and A. J. Heeger, Excitonic polarons in molecular crystals,
Phys. Rev. B 5. 4966-4969 (1972).
The Prospects of EXCltOnlC SuperconductIVity 369
39 R A Ban, Excltomc poldron~ m molecular solids, Phys Rev Lett 30,790-794 (1973)
40 Yu M Balkarel and D I Khomskll, Lattice stability m the phononless mechamsm of
superconductiVity, JETP Lett 3, 181-183 (1966)
41 J P Hurault, SuperconductIVIty m small crystallites, J Phys Chern Solids 29, 1765-
1772 (1968)
42 V L Gmzburg, Mamfe,tatlOn of the exciton mechamsm m the case of granulated
superconductors, JE7P Lett 14, 196-199 (1971)
43 P W Anderson, Edltondl comment, PhYSICS 2, 1')1 (1966)
44 G Bergmann and D Ramer, The sensitIVIty of the transitIOn temperature to changes m
a 2 F(w), Z Phys 263 ')9-68 (1973)
45 1 Appel, Role of thermdl phonons m high temperature superconductiVity, Phys Rev
Lett 21, 1164-1167 (1908)
46 P W Anderson, Theorv of dirty \uperconductors, J Phys Chern Solids 11, 26--30
(1959)
47 K Mdkl, Gdpless ,uperconductlvlly, m SuperconductIVity, Vol 2, pp 1035-1105, R D
Parks (ed), Marcel Dekker, New York (\969)
48 P B Allen, Repulsive effect of low frequency phonons on superconductiVity, Solid State
Cornrnun 12,379-381 (1973)
49 A E Kdrakozov E G Mak'imov, dnd S A Mashkov, Effects of the frequency
dependence of the electron-phonon mteractlOn spectral functIOn on the thermodynamiC
properties of superconductor" Sov Phys -JETP41, 971-976 (1976)
50 J R Schneffer D 1 SCdlapmo, and 1 W Wllkms, Effective tunneling denSity of states
10 superconductors PhYI Rev Lett 10,136-139 (1963)
51 H A Kramers dnd G H Wdnmer Statistic, of the two-dimensIOnal ferromagnet Part
I, Phys Rev 60, 2 ')2-270 (1941 )
52 L D Ldnddu and E M Llf~hltz Stallsllcal PhYSICS, Pergamon Press, London (1959)
53 L V dn Hove, 'iur I mtegrale de conhguratlon pour les systemes de particles a une
dimenSion, PhY~lca 16 117-141 (1950)
54 R A Ferrell, Pos'lhllitv of one dimenSional superconductiVity, Phys Rev Lett 13,
330-11') (2964)
55 T M Rice ,Superconductivity 10 one dnd two dimenSions, Phys Rev 140, A889-A 1891
(196<;)
56 P C Hohenberg, EXistence of long rdnge order m one and two dimenSions, Phys Rev
158,183-186(1967)
57 M Weger dnd I B Goldberg, Some lattice and electromc properties of the l3-tungstens,
10 Solid State PhYSICI, Vol 28, pp 1-178, F Seltz el al (eds), AcademiC Press, New
York (1971)
58 B T Matthias, Higher temperatures dnd mstabllitles, 10 SuperconductIVity In d- and
[-Band Metals, pp 167-17') D H Douglass (ed), AlP, New York (1972)
59 C G Kuper, Little'.; proposdl for a superconductmg orgamc polymer, Phys Rev 150,
189-192 (1966)
60 R E DeWdme" G W Lehmdn, and T Wolfram, SuperconductiVity m macroscopIc
one dimensIOnal system" Phys Rev Lett 13,749-750 (1964)
61 W A Little, DecdY of per'l,tent currenh 10 ,mall ~uperconductors, Phys Rev 156,
396-403 (\ 967)
62 M Tmkham, The electromagnetic properties of superconductors, Rev Mod Phys 46,
587-')96 (1974,
63 K Huang, Statistical MechaniCS, p 203, 10hn Wiley dnd Sons, New York (1963)
64 1 S Ldnger and V Ambegdokar, Intnnslc reSistive tranSitIOn 10 narrow superconduc-
tmg chdnnel" Phys Rev 164,498-5 \0 (1967)
65 D E McCumber dnd B I Halperin, Time scale of mtrmslc reSistive fluctuations 10 thm
superconductmg wire, PhYI Rev B 1, 10')4-1070 (\970)
66. R. S. Newbower, M. R. Beasley, and M. Tinkham, Fluctuation effects on the super-

conducting transition of tin whisker crystals, Phys. Rev. B S, 864-868 (1972).
67. V. Emery. Basic aspects in the physics of one dimensional metals, in Chemistry and
Physics of One-Dimensional Metals, H. J. Keller (ed.), NATO-ASI Series B25, Plenum
Press, New York (1977).
68. B. Horovitz, Instabilities of electron systems with nesting fermi surfaces, Solid State
Commun. 18,445-448 (1976).
69. J. S6lyom, Application of the renormalization group technique to the problem of phase
transition In one-dimensional metallic systems. II. response functions and the ground-
state problem. J. Low Temp. Phys. 12,547-558 (1973).
70. H. Fukuyama, T. M. Rice, C. M. Varma, and B. I. Halperin, some properties of the
one-dimensional Fermi model, Phys. Rev. B 10, 3775-3780 (1974).
71. Yu. A. Bychkov, L. P. Gorkov, and I. E. Dzyaloshinskii, Possibility of superconductivity
type phenomena in a one-dimensional system, Sov. Phys. JETP 23, 489-501 (1966).
72. K. Levin, D. L Mills, and S. L. Cunningham, Incompatibility of BCS pairing and the
peierls distortion in one-dimensional systems. I. mean field theory, Phys. Rev. B 10,
3821-31BI (1974).
73. K. Levin, S. L. Cunningham, and D. L. Mills, Incompatibility of BCS pairing and the
Peierls distortion in one-dimensional systems. II. fluctuation effects, Phys. Rev. B 10,
3832-3843 (1974).
74. N. Menyhard and J. S6lyom, Application of the renormalization group technique to the
problem of pha,e transition in one-dimensional metallic system. I. invariant couplings,
vertex and one-particle Green's function, J. Low Temp. Phys. 12, 529-545 (1973).
75. A. Luther and V. J. Emery, Backward scattering in the one-dimensional electron gas,
Phys. Rev. Lett. 33, 589-592 (1974).
76. P. A. Lee, Comments on a solution of a one-dimensional Fermi-gas model, Phys. Rev.
Lett. 34, 1247·-1250 (1975).
77. H. Gutfreund and R. A. Klemm, Order in metallic chains. I. the single chain, Phys. Rev.
B 14, 1073-1085 (1976).
78. S. T. Chui, T. M. Rice, and C. M. Varma, Coulomb effects on the Peierls transition,
Solid State Commun. IS, 155-1559 (1974).
79. G. S. Gre'l. F Ahraham" S. T. Chui, P. A. Lee, and A. Zawadowski, Two-cutoff
scaling for the one-dimensional electron gas, Phys. Rev. B 14,1225-1232 (1976).
80. D. J. Scalapino, Y. Imry; and P. Pincus, Generalized Ginzburg-Landau theory of
pseudo-one-dimensional systems, Phys. Rev. B 11, 2042-2048 (1975).
81. R. A. Klemm and H. Gutfreund, Order in metallic chains. II. coupled chain, Phys. Rev.
B 14,1086-1102 (1976).
82. L. Mihaly and J. S6lyom, Renormalization group treatment of three-dimensional
ordering in a syqem of weakly coupled linear chains, 1. Low Temp. Phys. 24, 579-596
(1976 ).
83. N. Menyhard, The effects of I D correlations on the phase transition in quasi-l D
metallic system" Solid State Commun. 21, 495-498 (1977).
84. T. Maniv and M. Weger, The superconducting transition in coupled linear chain
systems, 1. Phys. Chem. Solids 36,367-376 (1975).
85. B. Horovitz, H. Gutfreund, and M. Weger, Interchain coupling and the Peierls tran-
sition in linear-chain systems, Phys. Rev. B 12,3174-3185 (1975).
86. D. Jerome and M. Weger, Electronic properties of organic conductors: pressure effects,
in Chemistry and Physics of One Dimensional Metals, H. J. Keller (ed.), NATO-ASI
Series B25, Plenum Pres" New York (1977).
87. S. Barisic and K Saub, Selfconsistent calculation of the Peierls instability in quasi-one-
dimensional conductors, 1. Phys. C 6, L367-370 (1973).
88. W. Dietrich. Fluctuations and three-dimensional ordering in weakly coupled linear
conductors. Z. Phvs. 270. 239-243 (1974).
89. B. Horovitz and A. Birnboim, Superconductivity and Peierls instability in coupled linear
chain systems, Solid State Commun. 19,91-95 (1976).
90. B. Horovitz, Solid State Commun. 18,445-448 (1976).
91. N. F. Mott and W. D. Twose, The theory of impurity conduction, Adv. Phys. 10,
107-163 (1961).
92. W. A. Little and H. Gutfreund (to be published).
93. R. E. Borland, Existence of energy gaps in one-dimensional liquids, Proc. Phys. Soc. 78,
926-931 (1961).
94. D. 1. Thouless, A relation between the density of states and the range of localization for
one-dimensional random systems, f. Phys. C 5,77-81 (1972).
95. D. 1. Thouless, Electrons in disordered systems and theory of localization, Phys. Rep.
13,93-142 (1974).
96. Yu. A. Bychkov, Frequency dependence of the conductivity of one-dimensional sys-
tems, Sov. Phys.-fETP 38,209-213 (I 974)[Zh. Eksp. Tear. Fiz. 65,427-438 (1973)].
97. V. L. Berezinsky, Kinetics of a quantum particle in a one-dimensional random potential,
Sov. Phys.-fETP 38,620-627 (1974) [Zh. Eksp. Tear. Fiz. 65, 1251-1266 (1973)].
98. A. A. Gogolin, V. I. Melnikov, and E. I. Rashba, Conductivity in disordered chain
caused by electron-phonon interaction, Sov. Phys.-JETP 42, 168-178 (1975) [Zh.
Eksp. Tear. Fiz. toQ 327·-349 (1975)1.
99. A. Zavadovskii, Effect of impurities on superconductivity-like phenomena in one-
dimensional systems, Sov. Phys. JETP 27, 767-771 (1968).
100. B. R. Patton and L. J. Sham, Conductivity, superconductivity, and the Peierls instability,
Phys. Rev. Lett. 31, 631-634 (1973).
101. A. I. Larkin and V. I. Mel'nikov, Effect of impurities on the phase transitions in the
quasi-one-dimensional conductors, Zh. Eksp. Tear. Fiz. 71, 2199-2203 (1976) [Sov.
Phys.-fETP44, 1159-1161 (1976)].
102. I. F. Foulkes and B. L. Gyorffy, Jrwave pairing in metals, Phys. Rev. B 15, 1395-1398
(1976).
103. I. E. Dzyaloshinskii and E. I. Kats, Superconductivity in quasi-one-dimensional (thread-
like) structures, Sov. Phys.-fETP28, 178-182 (1969).
104. D. Davis, Thomas-Fermi screening in one dimension, Phys. Rev. B 7,129-135 (1973).
105. B. Bush, Theory of the screened coulomb interaction in quasi-one-dimensional metals,
Ph.D. Thesis, Stanford University (1974).
106. A. S. Berenblyum, L. I. Buravov, M. L. Khidekel, I. F. Shchegolev, and E. B. Yakimov,
Zh. Eksp. Teor. Fiz. Pis'ma Red. 13, 619-622 (1972) [Sov. Phys. JETP Lell. 13,
440-442 (1973)].
107. F. Stern, Polarizability of a two-dimensional electron gas, Phys. Rev. Lett. 18,546-548
(1967).
108. T. R. Brown and C. C. Grimes, Observation of cyclotron resonance in surface-bound
electrons on liquid helium, Phys. Rev. Lell. 29,1233-1236 (1972).
109. P. B. Visscher and L. M. Falicov, Dielectric screening in a layered electron gas, Phys.
Rev. B 3, 2541-2547 (1971).
110. A. L. Fetter, Electrodynamics of a layered electron gas. I. single layer, Ann, Phys.
(N. Y.)81, 367-393 (1973).
111. A. L. Fetter, Electrodynamics of a layered electron gas. II. periodic array, Ann, Phys.
(N. Y.) 88,1-25 (1974).
112. L. N. Bulaevskii and Yu. A. Kukharenko, Effectiveness of the exciton mechanism of
superconductivity 10 layered compounds with molecules, Sov. Phys. JETP 33,821-824
(1971).
113. W. A. Little, The effectiveness of the exciton mechanism in superconducting layered
compounds, I Low Temp. Phys. 13,365-369 (1973).
114. L. Salem, The Molecular Orbital Theory of Conjugated Systems, W. A. Benjamin Inc.,
New York (1966).
115. H. Gutfreund and W. A. Little, Correlation effects of 7T electrons. II. Low lying
excitations of polycyclic hydrocarbons, 1. Chern. Phys. 50, 4468--4477 (1969).
116. W. A. Little and H. Gutfreund, Dynamic effective electron-electron interaction in the
vicinity of a polarizable molecule, Phys. Rev. B 4, 817-823 (1971).
117. K. Nishimoto and N. Mataga, Electronic structure and spectra of nitrogen hcterocyoles,
Z. Phys. Chern. 13, 140-157 (1957).
I U'. A. M. Abarbanel, An energy band calculation of linear chain transition metal
complexes, Ann. Phys. (N. Y.) 91, 356-365 (1975).
119. D. M. Whitmore, A one-dimensional band calculation of a linear, square planar pla-
tinum complex, Phys. Rev. Lett. 50A, 55-56 (1974).
120. R. P. Messmer (private communication).
121. R. P. Messmer and D. R. Salahub, Importance of chemical effects in determining the
free-electron-like band structure of K2Pt(CN)4Bro.,·3H20, Phys. Rev. Lett. 35, 533-
536 (1975).
122. K. Mccs. The Theory of the Photographic Process, The MacMillan Company, New York
(1966).
8
Recent Developments and Comments
John Bardeen
1. Introduction
In this concluding chapter we shall attempt to give a brief account of

several topics concerning charge-density waves in quasi-one-dimensional
conductors that have not been discussed in the preceding chapters, make
some comments on some of the controversial points that have been raised,
and indicate directions for future research. In spite of the large amount of
experimental and theoretical work that has been done during the past few
years, there is no general agreement on a number of basic questions. These
include the nature of the phase transitions, the importance of Coulomb
interactions between electrons, explanations for physical properties,
including the temperature dependence of the magnetic susceptibility,
infrared optical properties, electron and nuclear spin resonance, specific
heats, and perhaps, most important, the role of collective excitations in
transport.
This book has been devoted largely to the compounds most exten-
sively studied, KCP, TTF-TCNQ, its alloys with TSeF-TCNQ, and related
materials. It is these materials that have been studied most extensively by
x-ray and neutron scattering, as described by Comes and Shirane in Chapter
2. This work has been essential to our present understanding. Many
different sorts of measurements have been made by various groups on
carefully prepared and highly purified single-crystal materials. As
emphasized in Chapter 4. there are severe demands on a theory that
attempts to fit them all. If one can be found, there is added confidence that
the explanation is correct. If there are two or more possible explanations, it
John Bardeen • Department of PhYSICS. University of Illinois at Urbana-Champaign,

Urbana, Illinois 61 SO 1.
373
374 John Bardeen
may be possible to devise experiments to decide between them. As is

evident from the discussions presented, many open questions remain.
In Chapter 3, Heeger presents strong evidence that the high static
dielectric constant observed below the metal-insulator transition tempera-
ture, Te , in TTF-TCNQ is associated with weakly pinned charge-density
waves. Because of the low pinning frequency, E(W) decreases rapidly with
increasing w. The nonlinear increase in conductivity with increasing electric
fields observed at very low temperatures also suggests collective behavior.
More controversial is his explanation of the conductivity in the metallic
range above Tn which he attributes to fluctuating Frohlich modes, i.e.,
moving CDWs. He makes use of a phenomenological theory of Rice based
on large fluctuating Frohlich modes, as suggested by the model of Lee,
Rice, and Anderson. This model is based on the assumption that one-
dimensional effects predominate, with the mean-field Peierls transition
temperature, Tp , much higher than the three-dimensional phase transition
at To presumably caused by weak interchain coupling.
Counterarguments are given by Schultz and Craven in Chapter 4, who
point out that such properties as electronic thermal conductivi.ty and
thermoelectric power are difficult to reconcile with collective transport.
They also give evidence that the transition at Te and fluctuations in the
neighborhood of T, are more nearly what is expected for a mean-field
transition (as described theoretically by Sham in Chapter 5). Specific heat
and magnetic susceptibility data, they suggest, give additional support for
an independent-particle model.
X-ray and neutron scattering data indicate that there are both one-
dimensional and three-dimensional fluctuations above Tn with only the
former extending to high temperatures. At least two order parameters may
he required to describe the one-dimensional and three-dimensional effects.
An important question is then whether or not the one-dimensional fluctu-
ations are large enough to give an appreciable average gap (of the order of
kB T or larger) above Tc in TTF-TCNQ and related compounds.
Several groups have studied the systematics of changes in chemical
composition. For example, Bloch and co-workers(l) have varied the cation
in TCNQ salts, replacing S by Se in TTF and changing the terminal protons
with other groups. They find compounds ranging from ones that remain
metallic to 7;. = 0 and others with Tc's covering a wide temperature range.
In the metallic range above T" there is a similar temperature dependence
of the resistivity and magnetic susceptibility for nearly all compounds
studied. An interesting compound in which charge transfer is complete or
almost complete is N-methylphenazanium (NMPt(TCNQf. Coulomb
effects appear to he large in this material with antiferromagnetic
ordering. (2)
Recent Developments and Comments 375
Compounds of ITF in which the anion is varied have also been

studied. One of considerable interest is ITF-CUS4C4(CF3)4 and its Au
analog.(3) Charge transfer is complete with antiferromagnetic coupling
between neighboring spins. In the Cu compound, there is a spin-Peierls
transition at 12°K, below which spin zero dimers are formed in the chain
direction, with greatly reduced magnetic susceptibility.
Another way the parameters can be varied is by application of hydro-
static pressure.(4) The resistivity and magnetic susceptibility change rapidly
with modest pressure of the order of 10-20 kg/cm 2 , indicating large
increases in the transfer matrix elements.
Studies of the chemistry and physics of organic conductors have
yielded considerable information about how charge transfer and charge
transport take place in organic systems and about Coulomb and magnetic
interactions between electrons.
Inorganic quasi-one-dimensional conductors that have been studied
recently include TaS" which undergoes a metal-semiconductor tran-
sition, ('i) and NbSe3' which remains metallic as T ~ 0 but has two resistivity
peaks in the p vs. T plot from incommensurate CDWs.(6) These peaks are
of considerable interest because they disappear in very small electric fields
(-I V /cm) and also at microwave frequencies, phenomena very suggestive
of Frohlich modes. Another interesting compound(7) is Hg3- x AsF 6 in which
there are mutually perpendicular Hg chains threading the AsF 6 lattice with
an incommensurate spacing. As discussed in Chapter 2, there is evidence
for one-dimensional phonons on the Hg chains.
Theories used to interpret the data are based on model systems that do
not attempt to reflect all of the complexities present in real materials. Even
though much oversimplified, they are valuable in indicating the sorts of
phenomena that can be expected. Emery in Chapter 6 discussed some of
the vast amount of theoretical work that has been done on ideal one-
dimensional systems, In which there are no phase transitions at finite
temperatures. Fluctuations are large. The behavior as T ~ 0 indicates the
nature of the ground state: charge-density waves, spin-density waves,
singlet or triplet superconductive pairing, Wigner lattice with anti fer-
romagnet ordering, etc. Renormalization group methods he describes are
valuable in determining the range of parameter values for which a given
type of order can occur. Interchain coupling or interaction with three-
dimensional phonons give transitions at finite temperatures.
Of the one-dimensional models, one often used is the Hubbard tight-
binding model in which the important parameters are U, the intrasite
Coulomb energy, and til, the transfer matrix element between neigh-
boring molecules. A generalized Hubbard model with off-site Coulomb
interactions is the one discussed in considerable detail by Emery. In the
376 John Bardeen
extended Luttinger-Tomonaga model, the Hamiltonian includes the kinetic

energy of the electrons and an interaction between them which may be spin
dependent. In the Frohlich Hamiltonian, the interaction between electrons
and phonons is included, but a direct interaction between electrons is
omitted. Impurity scattering, which takes an electron from one side of the
Fermi surface to the other (wave vector 2k F ) is often included and can have
important effects. With a wide variety of models and parameters covering
wide ranges, many different types of order can occur.
Following Landau's approach, one would like to know the nature of
the ground state and of the elementary excitations of the system. In a
normal metaL one can start with a sea of noninteracting electrons, in which
the excited states are described by giving the wave vectors and spins of the
electrons excited above the Fermi level and the holes below. The elemen-
tary excitations of the lattice are phonons, which again can be designated
by appropriate quantum numbers and wave vectors. In a real metal there
are interactions between electrons and between electrons and phonons, but
the low-lying states are in one-to-one correspondence with those of the
noninteracting system and can be designated by the same quantum
numbers. Interactions may be taken into account by a perturbation expan-
sion, provided, of course, that the expansion converges. The quasiparticle
states and phonons defined in this way are in general not eigenstates of the
system because there are residual interactions between the elementary
excitatiol\s. An excitation is reasonably well defined if its lifetime 7 is
sufficiently large so that hi 7 is small compared with its energy, E. Landau
showed that in Fermi systems 7 ~ CXl as T ~ 0, so that the low-lying excita-
tions (E ~ kB T) arc well defined at sufficiently low temperatures.
If impurities are present, the wave vector k is no longer a good
quantum number, but may be used approximately if the scattering time 7i
is sufficiently large. One may consider that the states are broadened in
energy by hi 7, where 7 includes scattering by impurities as well as by other
excitations.
When ordering is present, one must start the perturbation expansion
from an initial state different from that of a non interacting Fermi gas, one
defined by appropriate order parameters. Such order parameters are
generally associated with a broken symmetry in the new ground state, one
that does not have the complete symmetry of the original Hamiltonian. If
the nature of the order parameter is known, one can usually determine its
magnitude empirically.
Much controversy has arisen in various cases over what should be
taken for the appropriate ground state, which in turn determines the
excitation spectrum. Renormalization group methods indicate the possi-
bilities, but one must choose the correct model and parameter values to
describe a particular physical system. For example, for TfF-TCNQ there is
no agreement about the magnitude of the intrasite Coulomb energy V

relative to the tight-binding bandwidth, W = 4(11, of the Hubbard model.
Some (Chapter 3) assume that V is sufficiently small so that the system is
qualitatively the same a~ that for U = O. Others (see the discussion in
Chapter 4) assume that UI W> 1 so that the system is qualitatively similar
to an antiferromagnetic chain. In the limit VI W = <Xl, the spin and orbital
motions are completely decoupled.
In the small- U limit, the ground configuration corresponds to each
electron state for Ik I <c kF doubly occupied by electrons of opposite spin.
For large V, two types of ground configurations are possible which differ in
the excitation spectrum. In one each electron has its own space orbital so
that the Fermi surface i~ at ±2kh giving a space wave function antisym-
metric on interchange of the space coordinates of any two electrons. The
other is the Wigner lattice with the spins in an antiferromagnetic array.
Argumenb can be givcn for both the small- V and large- V limits, as
discussed in Chapter'> 3 and 4.
Organic molecule,> have many different vibrational modes that couple
to the conduction electrons. As described in Chapter 3, Rice and co-
workers have shown that the electron-phonon interaction parameter, A =
LA" is a sum over the couplings to the various modes. The frequencies of
the modes are modified by interaction with the conduction electrons. From
an analysis of the dielectric function E(W) one can determine coupling
constants and uncoupled intramolecular frequencies.
An important rea~on for interest in quasi-one-dimensional conductors
is the possibility of designing a system that exhibits superconductive
behavior, hopefully at high temperature~. Many years ago Little suggested
that one might be able to get an effective attractive interaction between
electrons by an excitonic mechanism in which polarizable molecules sur-
round a conducting spine. Thi~ hopefully would lead to BCS-type pairing.
He and Gutfreund give in Chapter 7 a careful analysis to find under what
conditions superconductive pairing might be observed. They find that the
spine must be surrounded by arrays consisting of at least four polar mole-
cules, a difficult problem in molecular engineering. To get true super-
conductivity, the conducting spins ~hould be locked in phase so that there is
a reasonable coherence distance transverse to the spine. This would imply
Josephson tunneling of electrons between spines, so the spines cannot be
too far apart. It would be ot great scientific intere~t to observe pairing even
without this phase coherence, but one cannot expect to observe extremely
high conductivity without it.
Another pos~ibility i~ to observe Frohlich superconductivity from
moving charge-density waves. This would mean finding a system in which
the COWs are not pinned by the lattice or by impurities. Again, it would be
necessary to have pha~e coherence between chains to observe very high
378 John Bardeen
conductivity. Although COWs are pinned below the three-dimensional

phase transition, one-dimensional fluctuations above may be mobile and
contribute to the conductivity. If the pinning is weak below Teo Frohlich
modes may contribute to the static dielectric constant and account for the
large values of E observed in such materials as KCP and ITF-TCNQ. As
discussed in Chapter 3, there is evidence in TTF-TCNQ for a low-
temperature electronic contribution to the conductivity which increases
nonlinearly with electric fields in the range 10-100 V / cm. These are attri-
buted to phase modes of an otherwise pinned COW.
In the following sections we shall discuss the nature of the phase
transitions (mean tield or one dimension), the role of Coulomb interactions
and the Hubbard model (large or small U?), evidence for contributions to
the dielectric function and frequency-dependent conductivity from Froh-
lich mode~, field-dependent conductivity, and the possible role of
nonlinear excitations (solitons) in the COWs and methods for getting
information about intramolecular vibrations and coupling constants from
optical data. Finally, we shall make some concluding remarks in regard to
the future of the lield.
2. Nature of the Phase Transitions
Phase transitions at finite temperatures may occur from interchain

Coulomb interactions, from a transverse transport matrix element t~,
which gives a three-dimensional Fermi surface, or from three-dimensional
phonons. The molecular displacements and resulting energy gaps are
normally enhanced below Te from the interactions responsible for the
transition. In many cases the increase in d with decreasing temperature just
below Te is given approximately by the BCS mean-field theory.
This will be true, for example, if the transition is caused by a phase
locking of COWs with alternate chains 180 out of phase. This enhances
0
the attractive Coulomb interaction energy, as originally suggested by

Barisic for the (1Tja, 1Tja, 2k F) displacements in KCP. An attractive elec-
tron-electron interaction proportional to d 2 and a positive elastic dis-
placement energy both add terms proportional to d 2 to the quasiparticle
free energy and lead to a gap equation similar to the BCS equation. The
change in sign of the electron-electron interaction corrects for double
counting in the ~um over quasiparticle energies. Proportionality to d 2
implies a BCS-Iike gap equation and a transition temperature, To given by
the BCS relation 2d(O) = 3.SkBTo provided that the average gap above Te
from one-dimensional fluctuations is small. In many cases there are fmther
interactions that add to the gap below Tn so that 3.SkBTc> represents only
the change in gap associated with the transition at Teo Further, there may
be large contributlon~ from one-dimensional fluctuations above Te. The
total gap at T = 0 is usually as large as (8-10)kBTc in quasi-one-dimensional

conductors.(8)
An example where the displacements follow the BCS relation at all
temperatures is the spin-Peierls transition in TIF-CuS4C 4(CF 3 )4' From
x-ray data, Moncton et al.(9) have measured the intensity J of the spots
corresponding to a period 2a in the chain direction. They find that J (T) is
proportional to LlBcs(T)2, showing that mean-field theory applies. Fluc-
tuations of the soft phonon are three dimensional above Te, but the motion
is along the chain direction below Teo It is the three-dimensional nature of
these phonons that i~ responsible for a mean-field transition.
2.1. Transition in Kep
The first studied and one of the best understood of the quasi-one-
dimensional conductors is KCP. As described in detail in Chapter 2, x-ray
and neutron scattering data have provided a great deal of information
about the incommensurate 2k r CDWs on the chains of Pt atoms and their
variation with temperature. A phase transition broadened by impurity
scattering occurs at temperature Tc of about 125°K. There are one-
dimensional fluctuations with random phases on the chains extending up to
room temperature. Below Tn three-dimensional order begins to set in with
diffraction peah at (71'/ a, 71'/ a, 2k F ), as shown in Figures 10 and 11, Chap-
ter 2. Associated with the transition is a resistivity anomaly indicated by a
broad peak in a plot of-d In R/ dT vs. T
The elastic neutron ~cattering intensity is proportional to the square of
the displacement. In a rough approximation, one may define order
parameters Al and A, for the rms gaps associated with the one-dimensional
and three-dimensional displacements. The transition involves not only
phase ordering, but also an enhanced gap. Since the gaps add in random
phase, the total gap Ll i<, given hy
Ll" = <Llf>+<Ll~> (1)
The intensity 1 of the (71'/ a, 71'/ a, 2k r ) peaks, as shown in Figure 10,
Chapter 2, varies linearly with T,. - T in a range below Te. This implies that
Ll3 is proportional to (Tc - T)I I", as expected for a mean-field transition.
The temperature variation of the intensity is not far from LlBcs(Tt
The way the total displacement, including that from one-dimensional
fluctuations, varies with temperature is shown in Figure 12, Chapter 2.
Although the transition itself is mean field, the one-dimensional fluctua-
tions are large and decrease slowly with increasing temperature. In fact, the
increase in amplitude below T, is only about 20%. This shows that even
though the transition at T, is mean field, because of the three-dimensional
nature of the phonons involved, the one-dimensional fluctuations above Tc
need not he small.
380 John Bardeen
2.2. Transitions in TCNQ Compounds
Careful studies have been made of phase transitions in TTF-TCNQ,

TSeF-TCNQ, and alloys of the two, as described in extensive discussions in
Chapters 2 and 4. We shall give here only some additional confirmation
that the 53°K transition in TTF-TCNO and the 29°K transition in TSeF-
TCNQ are essentially mean field and associated with the 2kF anomalies on
the TCNQ chains. In addition to the x-ray and neutron scattering data,(1O)
specific heat, magnetic susceptibility, (II) and resistivity data all support this
picture.
Very complete and detailed resistivity data, taken in directions both
parallel and transverse to the b axis, recently reported by Horn and
Guidotti,(121 are reviewed in Chapter 4. As is typical of many quasi-one-
dimensional conductors, there are sharp peaks in a plot of -d In R/ dT vs.
T [or better, d In R/ d (T I) vs. T I) at the metal-semiconductor phase
transition.
A plot of Horn and Guidotti for the b-axis resistivity of TSeF-
TCNQ is reproduced in Figure 20, Chapter 4. As for the TTF compound,
the transition i~ associated with Peierls transition at the wave vector
(0.5a*, 2kF-, 0). The drop in resistance below the transition is consistent
with assuming that
R = Ro exp[-a(T)/kBT) (2)
where Ro is a slowly varying function and a(T) follows the BCS tempera-
ture variation, which is proportional to (Tc - T)I/2 just below Te.
·1 6 .0
I'
b - cXls .1
1.6 I' 4 .0
TSeF - TeNO 1
C 1.2
-.......... ..... 2.0 Figure 1. Estimate of change In
gap parameter, fJA., from the 29°K
:l
phase transition in TSeF-TCNQ.
1.0 Extrapolations of In R from above
0::/1- o.a .ci;:; and below are made to the tran-
"U '0 0.8 sition temperature, Tc- It is
0.6 0::
-10:: assumed that the difference
-I 0.4 between the two extrapolations
I
gives the change in logarithm of
0.4
_.-' 0.2
the carrier concentration resulting
from the transition, or fJA./ k8 Te.
The experimental data are from
2'-:
4------=2'-6-----:"
:- 2'a-=-------="
3Lo--' 0 .1 Horn and Guidotti, Reference 12,
shown in Figure 20, Chapter 4.
TEMPERATURE (K)
One could estimate the total change in gap associated with the tran-
sition, 8~, by integrating over the peak in the derivative plot. A better way
is to extrapolate In R to Tc from above and below the peak, as illustrated in
Figure 1. The difference in In R between the two extrapolated points gives
the ratio
(3)
close to the BCS value of 1.75, corresponding to 8~j kB - 50°K. The gap
estimated from -d In R/ dT continues to increase slowly with decreasing
temperature below Tn reaching a limiting value of 115°K at low tempera-
tures. This increase may be caused by interchain couplings at lower
temperatures, which further enhance the gap. There is evidence for
2kF one-dimensional anomalies in x-ray scattering extending up to
250°K.
Most complex and mteresting are the phase transitions in TTF-TCNQ,
which have been discussed in great detail in this book. The cation and
anion chams are decoupled more than in other one-dimensional
compounds, with 4kF fluctuations on the TTF chains and 2kF on the
TCNQ chains. While the cause of the 4kF scattering is still uncertain, the
anomalies are probably related to strong Coulomb interactions between
electrons. The TCNQ chains appear to dominate the conductivity above
the 53°K transition and the TTF below. The lower temperature transitions
are driven by the 2kF CDWs on the TCNQ chains. The only transition
that is reasonably simple is the metal-semiconductor transition at
53°K.
The resistivity data for TTF-TCNQ of Horn and Guidotti taken near
the 53°K transition show that it is also mean field, with -d In Rj dT varying
as (Tc - T)I/2 just below Teo The extrapolation of -In R to Tc from lower
temperature is complicated by the presence of the 49°K transition close by,
but a reasonable estimate gives -81n R close to the BCS value, 1.75. It is
helpful in making the e~timate to consider both the plots of -In R and of
d In R/ dT. Data for the b-axis resistivity of TTF- TCNQ show additional
peaks just below the 49°K transition and at the 38°K transition (Reference
1a, Figure 10).
In conclusion, the metal-semiconductor transition associated with a
sharp peak in -d In R/ dT appears to be close to mean field in most cases,
but this does not preclude large one-dimensional fluctuations above Tc
because more than one order parameter may be required to describe the
system. In TTF- TCNQ, x-ray and neutron scattering data show 4kF one-
dimensional fluctuation'> extending to high temperatures, but 2kF fluctua-
tions appear only helow 150 o K. If the 4kF fluctuations are confined to the
TTF chains, conduction on the TCNQ chains may not be strongly affected
by charge-density wave formation, at least ahove 150°K.
382 John Bardeen
2.3. Transitions in TaS1, NbSe3
Other interesting examples of quasi-one-dimensional conductors are

TaS3 (13) with an orthorhombic and NbSe3 (14,15) with a monoclinic structure,
with chains of metal atoms parallel to the direction of high conductivity.
They have a fibrous structure with a morphology similar to that of (SN)x'
A related compound, TaSe3, which has a monoclinic structure, is a super-
conductor below 2, 10 K, as is NbSe3 under hydrostatic pressure, (27) with Tc
reaching about 3.6°K at 6 kbaT. One may regard these compounds as
consisting of metallic chains surrounded by an insulating sheath with a
nearly one-dimensional Fermi surface. The room-temperature resistivity
of TaS3 is ~ 3 X 10- 4 0- 1 cm- 1, decreases to a minimum at T ~ 270oK, and
then increases again, with a metal-semiconductor transition at Tc ~ 245°K
with the unit cell increasing from b to 4b along the axis of high conduc-
tivity. Diffuse scattering lines have been observed by electron diffraction
up to room temperature which depend only on the b* value. Lines occur at
~O.25b* above 245°K. changing to spots below. The low-temperature
slope of In R vs. liT gives a gap 2a~ 1740°, or about 8kBTc.
Although NbSe, remains metallic down to T = OaK, sharp increases in
resistivity appear from two different incommensurate COWs, one below
145°K and the other below 59°K. These open up gaps that affect about
20% of the three-dimensional Fermi surface below 145°K and 50% below
the 59°K transition. Only the COWs associated with the 145°K transition
have so far been confirmed by electron diffraction. They appear as super-
lattice spots at the incommensurate value (a*, 0.244b*, c*) below the
transition and diffuse knes at 0.244b* above the a*, c* plane above the
transition. The remarkable frequency and field dependence of the conduc-
tivity will be discussed in Section 4.
3. Coulomb Interactions and Magnetic Susceptibility
3. 1. Limitations of the Hubbard Model
In this section we consider the arguments concerning the nature of the

ground state and excitation spectrum. Coulomb interactions are often
discussed in terms of the Hubbard model, in which the important
parameters are the on-site Coulomb interaction U and the transfer matrix
element t. The "smaIl- U" limit implies a metallic ground state with a gap
due to charge-density waves opening up at the Fermi surface, ±kF • In the
"Iarge- U" limit there i~ antiferromagnetic ordering in the ground state
with spin waves and a gap for quasiparticle excitationY6) There may be
spin-Peierls distortions at 2kF in the spin-wave spectrum, leading to
charge-density waves at 4kF • These gaps are due to electron-phonon

interactions omitted in the Hubbard model. Associated with variations of
electron density there is a compensating motion of ions, so the COWs
involve both electron and ion motion. The temperature variation of the
magnetic susceptibility is an important factor in deciding which limit is
appropriate for a particular compound.
There is an essential singularity at the point U = 0 in parameter space
in the simple one-dimensional Hubbard model, so that one cannot expand
the Coulomb interaction about the metallic ground state for U = 0 no
matter how small the interaction. The Fermi surface, defined by a dis-
continuity of occupancy in the ground-state wave function in k space, does
not remain at ±kF as U mcreases from zero, but gradually moves to ±2kF
as U -> ex). There is no range of U > 0 for which the usual picture of a metal
with a well-defined Fermi surface remains valid.
In the extended Hubbard model considered by Emery in Chapter 6,
which includes an oft-~Ite Coulomb interaction V, the singular point is
shifted to U - 2 V. As shown in his Figure 5, there are SOWs for U - 2 V >
o and COWs for U -2 V <0 for U +6 V>O.
The anomalous properties of the one-dimensional Hubbard model do
not occur in three dimensions. There is a critical value of U, comparable to
the bandwidth W = 4t, at which a transition from a phase with a metallic to
an antiferromagnetic ground state occurs (the Mott transition). In quasi-
one-dimensional conductors, there is also very likely a critical value of
UI W of order unity below which an approximate description of the system
as a one-dimensional metal with Fermi surface at ±kF is valid. Electron-
phonon interactions may help stabilize the Fermi wave vector at k F • Thus
the simple Hubbard model ~hould be used with caution and only for
qualitative indications of the nature of the states and excitation spectrum
for small and large VI W than for quantitative calculations.
If the transverse transport matrix element t~ is sufficiently large in
highly anisotropic conductor~, the orbits of the electrons may overlap
neighboring chains. The ~ystem is then best described as a semimetal with
pockets of electron~ and hole~ of small effective mass. The large orbits lead
to a large diamagnetic contribution to the susceptibility.
As discussed elsewhere, the 4kF anomalies associated with the TfF
chains in TIF-TCNQ appear to reflect a large Coulomb interaction and an
antiferromagnetic ground state (see Chapters 2, 4, and 6). Although there
is no conclusive evidence for 2kF spin-density waves from neutron scatter-
ing, they may be present along with 4kF charge-density waves, which in
turn would give a gap in the quasiparticle excitation spectrum at 4kF •
Another possibility when U is large is the formation of a Wigner lattice in
which the low-lying excited states of the electrons are bound. Since the
charge density in a Wigner lattice would have Fourier components at 4kF
384 John Bardeen
and higher harmonics, one could account for the observed 4kF lines. The
main differences between the two pictures are in the nature of the excita-
tion spectrum and the coupling between the spin and orbital motion of the
electrons. The Wigner lattice should be considered as consisting of
polarons with as~ociated atomic motions.
In the limit U = 00, the orbital motions and spins are completely
decoupled. The orbital wave function is an antisymmetric function
described by a Slater determinant of individual orbitals for each electron,
which for the lowest state includes those extending up to 4kF in the
tight-binding model. The magnetic susceptibility is that of a system of free
spins.
3.2. Magnetlc Susceptibility
3.2.1. The Large U Limit
The magnetic ~usceptibility has been derived by Shiba(l7) from the

exact solution of Lieb and Wu of the Hubbard model at T = 0 for an
arbitrary ratio of electrons to molecules, p = N INA' In the limit UI t » 1,
the susceptibility may be expressed in the form of a one-dimensional
antiferromagnetic chain of electrons,
2NJ-t~
X= - ,- , U»t (4)
1T- Jet!
where Jet! is an effective nearest-neighbor coupling constant
_~( _Sin21TP) (5)

Jct!-
pU 1 2 1Tp
For a half-filled band (p = 1), Jeff = 2t 2 I U.

The low-lying excitations in the large- U limit are spin waves. Elec-
tron-phonon interactions may give a spin-Peierls transition with a gap in
the spectrum at 2 kF .
For p = 1, the ground state corresponds to each orbital in the band
singly occupied, which i~ equivalent to having one electron on each mole-
cular site. Quasiparticle~ may be described by a band of states correspond-
ing to a site being doubly occupied by electrons of opposite spin and a hole
band in which an electron i~ missing from a site. A minimum energy, the
Hubbard gap, is required to create a pair of excitations corresponding to a
particle and a hole. The gap is quite small for UI W <: 0.5, but increases
rapidly to U -- W for UI W'> I. Such a band picture does not take inter-
actions between quasi particles into account and should be used only when
the number of doubly occupied sites and unoccupied sites is small.
3.2.2. The Small V Limit
As VI W ~ 0, the ~usceptibility approaches the Pauli expression for

noninteracting electrons, which, per molecule, is
X~ = 2N(O)I-t~ (6)
where N(O) is the density of states in energy per molecule for one spin
value:
N(O) = (21ft sin ~1fP r I (7)
A short-range Coulomb repulsion between electrons gives an enhance-
ment to the susceptibility. [n the random phase approximation, the
enhancement factor is II - N(O)Vr I, giving
XI'Ix:: = 1I - N(O)Vr ' = 1 + N(O)V + ... (8)
According to Shiba\ calculations, the correction for small V is approxi-
mately
X"/X;:= 1+~N(O)U+··· (9)
giving a smaller correction factor.
These expressions are valid only when the density of states does not
change appreciably in a range of energies extending ± V about the Fermi
energy. For example, if p = 0.6 as in TTF-TCNQ, and the bandwidth on
the TCNQ chain is O.S e V, the Fermi energy would be only about 0.1 e V
above the bottom of the band, where the density of states is much larger
than at the Fermi level. This means that if V is of the order of 0.1, the
enhancement would he larger than indicated by the expressions given
above.
3.3. Problem of TTF- TCNQ
In TTF-TCNQ there is an enhancement of the susceptibility by a

factor of the order of 2-3, depending on how the total susceptibility,
XT = XO + XF, is divided between the TTF and TCNQ chains and what one
assumes for the bandwidths. Heeger, in Chapter 3, gives arguments for
assuming that XFI XO is greater than 2, uses Equation (8) for the enhance-
ment factor, and estimates that VI W is small. Schultz and Craven in
Chapter 4 find roughly equal contributions from XO and XF. Torrance et
at.,(18) using Shiba's calculation of the enhancement factor, find VIW> 1,
and perhaps as large a~ 2-3. The former find the small-V metallic limit
appropriate, while latter favor the large- V limit. Without attempting to
review the arguments, about all one can say is that the uncertainties are
such that either limit might be valid and that one must rely on experiments
386 John Bardeen
to decide. As discussed earlier, these indicate 4kF anomalies and antifer-

romagnetic behavior on the TTF chain and 2kF anomalies and metallic
behavior on the TCNO chain.
A counterargument comes from the temperature variation of the
magnetic susceptibility of TTF- TCNO, which is similar to that of most
other TCNO compounds. There is roughly a linear decrease in X(T) with
decreasing temperature below 300°K. This means that if the large- V limit
applies, Jet! must be relatively large, approaching thermal energy at room
temperature. The magnitude of the susceptibility from spin waves would
then not be far different from that expected for Pauli susceptibility in the
small- V limit. ]n fact, there may not be much change in susceptibility as
V/ W passes through the transition from metallic to antiferromagnetic-
semiconductor behavior.
Some TeNO compounds exhibit a marked increase in X(T) as T
decreases, characteristic of antiferromagnetic behavior with a small Jeff.
One that has been subject to considerable study is NMP-TCNO. We shall
discuss this compound in greater detail in the following section.
3.4. Properties oj NMP- TCNQ
One of the first organic quasi-one-dimensional conductors whose

physical properties were studied extensively is NMP+TCNO-, which has a
room-temperaturc conductivity of the order of 200-4000- 1 cm- I along
the conducting axis. Epstein et al. (2a) found that the conductivity increases
with decreasing temperature to a peak at around 2000K and then
decreases, with a sharp drop suggestive of a three-dimensional phase
transition occurring at low temperatures. The magnetic susceptibility is
roughly constant down to about 250 0K and below 2000K increases as
X(T) = X(O)18/(T + 8)] (10)
where 8 - 65°K and X(O) == IS x 10- 4 emu/mole. This latter temperature
dependence is expected for an antiferromagnet. The authors assumed that
charge transfer is complete with chains of NMP+ and TCNO- ions. Since
NMP+ has no free spin, it was suggested that in the TCNO chains there is a
broad Mott-Hubbard transition from a metallic state at high temperatures
to an antiferromagnetic semiconductor at low temperature.(2a) There is
nO evidence for a change in magnetic behavior associated with a pos-
sible transition near 1000K suggested by conductivity data of Coleman et
1
a.
(1'))
If the charge transfer is complete, the wave vector 2kF associated with
a possible Peierb distortion would be commensurate and equal to the
reciprocal lattice vector b* = 2 w/ b. Thus a Peierls gap may enhance a
Mott-Hubbard gap.
In later work, Epstein et al. (2b) suggested that NMP-TCNQ is better

described as a semiconductor over the entire temperature range, with strong
Coulomb interactions. In their theory the maximum in the conductivity
arises from the competition between an activated carrier concentration and
a strongly temperature-dependent mobility. There is some evidence that the
charge transfer is not quite complete.(20)
Alloys in which a fraction of the NMP ions are replaced by phena-
zine have been studied by Miller and Epstein.(21) Since phenazine is
neutral the alloys are expected to have the composition
(NMP+)APhen)l-ATCNO )x(TNCOO)I-D altering the electron concen-
tration on the TCNO chains. Magnetic properties of the alloys are similar
to NMP-TCNO in showing antiferromagnetic behavior at low tempera-
tures. Since the TCNO band is less than half full (p < 1), there should be no
gap for quasiparticle excitations in the Hubbard model. Since an activation
energy for conduction is observed at low temperatures, there must be a
Peierls gap. Transport properties will be discussed in the following section.
A number of other organic conductors,(22) including quino-
linium+(Ot(TCN0 2 ) , are antiferromagnets, but the majority behave
more like ordinary metals or semiconductors, with the susceptibility stay-
ing constant or decreasing with decreasing temperature.
4. Collective Transport
One of the principal motivations for research on quasi-one-dimen-

sional conductors has been the possibility of obtaining high-temperature
superconductivity or at least very high conductivity. One path has followed
from Little's suggestion of designing an excitonic superconductor by sur-
rounding a conducting spine with highly polarizable molecules. As he and
Gutfreund point out in a critical analysis of the problem in Chapter 7, this
has proved to be a very difficult task. The other was stimulated by the
announcement of very high conductivity peaks in TTF-TCNO by Heeger
and co-workers. Although originally thought to be due to superconducting
fluctuations above the transition temperature, the writer suggested that a
more likely explanation was transport by fluctuating charge-density waves,
along the lines of an early suggestion by Frohlich for superconductivity in a
one-dimensional system. Although it was found that the conductivity
values as reported were spurious and caused by the difficulty of making
measurements in highly anisotropic systems, interest in the quasi-one-
dimensional conductors has continued to increase because of the fascinat-
ing problems of chemistry and physics involved.
It was a surprise to find on rereading the early papers of Frohlich(2J)
and Kuper(24) how close the mathematics (although not the physics) of
388 John Bardeen
Frohlich conductivity is to the BCS pairing theory of superconductivity

developed several years later. This analogy has been used by many in
further developments of the Frohlich theory and is the approach used by
Sham in Chapter 5. There are of course very important differences, such as
the role of impurities, commensurability, and interchain coupling in pin-
ning charge-density waves which have no counterpart in pairing super-
conductivity. A major open question is whether or not one can enhance the
conductivity by the Frohlich mechanism over what one would have with
impurity scattering alone. Present evidence is in the negative.
With both mechanisms, one has to contend with the problem of
one-dimensional fluctuations; in an ideal one-dimensional system there are
no phase transitions for T> O. Mattis(25) has shown that with an attractive
interaction (U < () in the Hubbard model) the conductivity goes to infinity
as T ~ 0 even when impurity scattering is present, and others have exten-
ded his work. This conductivity enhancement probably represents incipient
pairing superconductivity. However, as we shall show in the following
section, the effect of fluctuations is so great that one cannot expect to get
conductivity much greater than that of a normal metal except at extremely
low temperatures.
Later we shall discuss evidence for Frohlich conductivity from: (1) the
frequency-dependent dielectric function, e(w), whose imaginary part gives
the conductivity, O'(w); (2) transport properties in the fluctuation regime
above T;; (3) electric-field-dependent conductivity and possible role of
solitons at low temperatures; and (4) frequency and field dependence of
the conductivity of NbSe3 below Te. Since most of these topics have been
discussed in detail elsewhere in the book, we shall confine our remarks to
supplementary information and in a few cases to express new points of
view on controversial subjects.
4.1. Superconductivity and One-Dimensional Fluctuations
There is a difference in effects of fluctuations on so-called one-dimen-

sional superconductors composed of thin wires or whiskers in which the
dimensions transverse to the axis are small compared with the Pippard
coherence distance, go, and those i:1 which conduction is along noninterac-
ting chains of molecule~. In the former, there is little change in transition
temperature from that of bulk material; in the latter there is no phase
transition above T = O°K. True superconductivity requires phase
coherence between chains, which in a pairing superconductor requires at a
minimum the possibility of Josephson tunneling between chains. This
requires that the transverse matrix element, tc, of the tight-binding model
be sufficiently large.
The theory of one-dimensional fluctuations of McCumber and

Halperin(2h) may be extended to individual chains. Their expression for the
ratio of the superfluid conductivity from pair fluctuations to the normal
conductivity is
if.1 = 2 d]
kBT(o ex p [4' 1/2 (11)
(T" 1Tfz 2 nS 31T 2 k B T
where n is the density of conduction electrons, (0 the Pippard coherence
distance, S the cross-sectional area, and d the gap parameter. Inserting the
expreSSIOns
(12)
and defining
(13)
we find
(14 )
To get an enhancement of conductivity requires x > ~4, or T <
~O.l Tc- Thus even if one were able to make an independent chain super-
conductor with a high mean field Te , fluctuations would decrease the
pairing contribution so much that one would get appreciable conductivity
enhancement only at very low temperatures.
Another argument that illustrates that supercurrents would decay
rapidly in independent chains is that individual electrons or phonons have
sufficient momentum to change the collective motion. This may be seen by
considering a chain of length L subject to periodic boundary conditions. If
there are N electrons, the Fermi wave vector is at kF = ±1TN/2L. A change
of wave vector by 2kF changes the momentum by 1TN/ L. This is just the
momentum change that would occur if the quantum number describing the
superflow changes by one unit HN
x (27T/ L)].
If impurity scattering is present, the normal conductivity is also
reduced by one-dimensional fluctuations, so the relative contribution of
the collective motion may be larger than indicated above. However, to get
very high conductivity would require an enhancement over the bulk
conductivity. Thus, it l~ essential to have three-dimensional correlations
that maintain phase coherence between neighboring chains.
While it would be of great scientific interest to find systems that exhibit
fluctuating pairs at high temperatures along the lines suggested by Gut-
freund and Little, it would only be a step along the way to getting true
high-temperature superconductivity; one must also have a reasonably large
t.L'
390 John Bardeen
Similar considerations apply to the effect of fluctuations on Frohlich

conductivity. In addition to the problems associated with pinning, in order
to get high conductivity it would be necessary, as in pairing superconduc-
tivity, to have phase coherence between neighboring chains. This must be
borne in mind considering the implications of theories based on ideal
one-dimensional systems in which interchain interactions are neglected.
In charge transfer systems, it is likely that interchain coupling will lead
to coherence between COWs 180° out of the phase. If these move
together, there would be no net transport of charge in the chain direction
and so the Frohlich mechanism would be ineffective.
Further studies of transport in true metals, such as NbSe3, should be of
great interest, because in this case one does have phase coherence. Since
NbSe3 becomes superconducting under pressure, it should be an excellent
material to study the interplay between Frohlich conductivity and pairing
superconductivity.
4.2. Dielectric Properties
Quasi-one-dimensional conductors that undergo a Peierls transition

generally have a very large static dielectric constant in the semiconducting
phase, values an order of magnitude or more larger than can be accounted
for by the Peierls gap. As described in Chapter 3, Section 4, these large
values are attributed to low-frequency oscillations of the charge-density
waves about their pinning positions. The oscillator strength, IF, of these
Frohlich modes is smalL of the order of the ratio of the electron effective
mass, m* to the Frohlich mass, MF<
(15)
which is typically of order 10- 1 . These modes give a peak in the conduc-
tivity O"«(U), at (U ~ (UF, where (Up is the pinning frequency. The width of the
peak gives the damping constant, r.
From measurements of r, one can estimate what the Frohlich conduc-
tivity would be in the absence of pinning. Since the system is ordered, there
are no one-dimensional fluctuations. The pinning frequencies are so low, in
the very far infrared, that measurements are difficult. It is only for KCP
that fairly complete data are available.
The tirst convincing evidence for moving COWs came from an analysis
of far-infrared reflectivity measurements of Briiesch et al.(28) on KCP. At
temperatures below 1500 K, they found an anomaly in reflectivity which they
interpreted as coming from pinned COWs oscillating at a pinning frequency
(UF of about 15 cm .J .
In KCP, conduction is along chains of Pt atoms. The electrons have a

Fermi energy E F = 3.5 eV and an effective mass about equal to the free-
electron value. The oscillations presumably involve the optically active

(7T/ a, 7T/ a, 2kF ) Peierls distortion, which at low temperatures gives a gap
2d==0.11 eY. From Frohlich's expression based on mean-field theory,
(16)
one finds A = 0.2. From his expression for the effective mass MF of the
COW, one finds
(17)
with an unperturbed phonon energy, Wo == 0.01 e V.
This value for MF is of the same order of magnitude as that required
to fit the far-infrared data. The contribution of the mode to the dielectric
function is
(18)
where Op = (47Tne 2 / MF )1/2 is the plasma frequency of the collective mode,

with n the electron density, and r the damping constant of the wave. Using
the measured static dielectric constant, E 1(0) = 3000 (at 4.2°K), Briiesch et
at.(2S) were able to fit the low-temperature data with Op-
8000 cm ~ 1 MF/ m - 1000, WF - 15 cm ~ I, and r - 8 cm ~ I. In view of the
simple model used, Equation (17) could easily be off by a factor of 2, so
that the agreement is all that one could expect.
The authors also attempted to see what contribution, if any, fluctuat-
ing COWs make to the room-temperature conductivity of KCP, which is
about 300 (0 cm) I. They found that the collective contribution is at most
50% and perhaps a good deal less.
If all of the electrons were carried by unpinned COWs subject to the
damping r, the conductivity would be
IJ"F = ne 2 /(MF J') (19)
Inserting the values for MF and r required to fit the data (at 4.2°K), one
finds IJ"F == 1300 (0 cm) I, several times larger than the room-temperature
value of about 300 (n cmfl. This suggests that if the damping in both cases
is mainly by impurity scattering, the scattering rate per electron is larger for
single-particle motion than it is for collective motion. This is not in accor-
dance with a theory of Fukuyama et at.,(29) who gave an argument to show
that COWs cannot enhance the conductivity over that from single particles
when the conductivity i~ limited by impurity scattering. Collective motion
could of course reduce the ~cattering by 2kF phonons, and it may be that
this scattering is mamly responsible for the resistivity in KCP at room
temperature. A more likely possibility is suppression of the conductivity by
one-dimensional fluctuations at room temperature.
392 John Bardeen
The static dielectric constant £1(0) is much higher than can be account-
ed for by a conventional single-particle picture. The contribution from the
Peierls gap to £ 1 is W ~/ 6t~.z, where W ~ = 41Tne 2 / m * is the electron plasma
frequency and 2Ll the energy gap. This contribution is about 190 for KCP
as compared with the observed value of 3000. One can obtain large values
for £1 from one-dimensional fluctuations, but these should not be present
in KCP at low temperatures.
The oscillator strength of the Frohlich mode, fy = m* / Mp, of the
order of 10- 3 for KCP. is a very small fraction of the total of unity. In a plot
of 0"( w) versus w, the mode appears as a very small peak near w = 0,
centered at w = WF. The peak conductivity is the value calculated above,
1300 (0 cm) I.
Values of the static dielectric constants (at 4.2°K) are 3500 for TTF-
TCNQ and 15,000 for TSeF-TCNQ. From an analysis of infrared data for
TTF-TCNQ, taken at 4.2°K, Coleman et al.(30) estimate MF/m* = 1400,
Op = 120 (0 cm) I, WF ~ 2 cm- I , and r < 3 cm -I. For unpinned modes, the
Frohlich conductivity at zero frequency should then be greater than
~ 100 (0 cm)· I. Since the room-temperature conductivity may be as large
r
as 650 (0 cm I. these data do not provide a critical test of the role of
impurity scattering of Frohlich modes.
Gunning et al.('I) have given evidence for high positive microwave
dielectric constants even at room temperature in a number of metallic
one-dimensional conductors. Unless there is some other very-Iow-
frequency mode. perhaps associated with one-dimensional fluctuations,
this would imply the presence of Peierls fluctuations. In many cases there is
evidence from x-ray or neutron scattering 2kF (or 4k F ) lines from one-
dimensional fluctuations extending to high temperatures. The oscillato'r
strength is reduced by a factor nj n giving the fraction of electrons carried
by collective modes, but pinning friction is no doubt also reduced. The net
result may be a high static dielectric constant even though the contribution
of the collective modes to the dc conductivity is small. We shall discuss the
conductivity in the fluctuation regime above Tc in the following section.
4.3. Transport above Tc
There have been three general approaches to understanding the

nature of the conductivity above the metal-semiconductor transition
temperature, Tc- One starts from the theory of ideal one-dimensional
metals in which fluctuations and associated disorder play the predominant
role. Because of tluctuations, the dc conductivity vanishes at T = O°K. At
low temperature~, there is phonon-assisted hopping conductivity and the
conductivity increases with temperature. At high temperatures, beyond the
conductivity maximum, phonons scatter the electrons and the conductivity

decreases. When T» eD, the phonon Debye temperature, and phonon
scattering dominates impurity scattering, the conductivity decreases at
T- 2 • In these theories, there is no Peierls transition. A second approach
emphasizes transport by charge-density wave fluctuations with an asso-
ciated pseudogap in the energy spectrum of the electrons. The third makes
use of ordinary transport theory based on mean-field theory. It is presumed
that one-dimensional fluctuations are suppressed by interchain coupling. In
cases where there is a three-dimensional Peierls transition, standard
transport theory should be valid below Tv The energy gap at low tempera-
tures then can be estimated from the slope of a In a vs. 1/ T plot.
One of the main applications of the theories has been to account for
the temperature variation of the resistivity of the various TCNQ salts. In
salts where the cations are substituted TTF molecules,
(20)
where in TTF-TCNQ A is of the order of 2.3. (See Chapter 3, Section 5.1.)
This expression applies to temperatures above the conductivity maximum,
which occurs somewhat above Te. Since the conductivity is not greater than
expected from simple transport theory, it is difficult to decide without
detailed arguments whether or not conductivity enhancement by collective
modes occurs.
At higher temperature~, the conductivity is affected by thermal
expansion, particularly along the b axis, that of high conductivity. The
conductivity at constant b differs from that at constant pressure. In the
discussion that follows, we shall be concerned mainly with lower tempera-
tures (below 150 K in TTF-TCNQ), where (20) is approximately valid for
0
specimens of high puritv.
4.3.1. Percolation and Hopping Transport
There is relatively little in this book about hopping transport in dis-

ordered one-dimensional conductors. The theory of phonon-assisted hop-
ping is a difficult one and has been subject to considerable study. There are
excellent reviews by Cohen, Abrikosov, and Ryzhkin, and by Rasha,
Gogolin, and Mel'nikov in the Proceedings of the Siofok Conference on
Organic Conductor~ and Semiconductors. (1 2) Others who have been active
in the field are Brazovskii, Gor'kov, and Dzyaloshinskii.(33,34) We shall give
here only a very brief summary of the nature of the results, since the
theories give an alternative explanation of some of the unusual properties
of quasi-one-dimensional conductors. These include very high static
dielectric constant~ and frequency-dependent conductivity. In systems that
394 John Bardeen
undergo three-dimensional Peierls transItions, theories based on one-

dimensional disordered metals can be expected to apply only at tempera-
tures well above the transition temperature, Te•
As first pointed out by Mott and Twose, (35) all quantum states for
electrons in disordered one-dimensional systems are localized. Bloch et
al. (36) were the first to apply the theory of hopping conductivity in one
dimension to account for the temperature dependence of the conductivity
in NMF-TCNQ and other TCNQ salts. Disorder can be introduced by
impurities, inter- and intramolecular vibrations, random orientation of
asymmetric molecules, and by alloying.
Phonons give a diffusive-type transport with a diffusion coefficient
D ~ 12 /Tp h, where 1 is the hopping distance and Tph the lifetime for inelastic
scattering by phonons. A characteristic feature is the appearance of the
square of a mean free path in the expression for the conductivity. Goglin et
al.(37) were able to fit the temperature dependence of the conductivity and
dielectric constant of quinoline (Qn) and acridizine (Adz) salts of TCNQ,
Qn(TCNQh, and Adz(TCNQh with a theory of this sort. The high dielec-
tric constants (several thousand) are accounted for by a large motion of
charge (~20 A) In the potential wells created by disorder.
The nature of the solutions is illustrated by an exact solution obtained
by Abrikosov and Ryzhkin(38) for a one-dimensional model with impurities
and phonons. At high temperatures, kB T» hwo, they find for the dc
conductivity
2e 2 /* 1* wDI*
0-(0)=---- (21)
rrhS [ph 2VF
r
where S is area per chain and 1* = (C I + I~~ 1. The expression applies in
the limit 1*« 2 Vf) WD' The first term of the product is the Drude conduc-
tivity
(22)
Note that 0-(0) is always less than o-D.

In the pure limit, lph» I" the conductivity is proportional to I~h' or to
r- 2 • Impurities have a much greater effect in reducing the conductivity in
one dimension than in three dimensions, since the conductivity varies in
general as (/*)' / Iph '
Thus theories based on one-dimensional disorder have some of the
properties generally attributed to conduction by charge-density wave
fluctuations, for example large static dielectric constants and extreme
sensitivity of the conductivity to impurities. They no doubt apply to highly
disordered systems. It is difficult to say to what extent they are applicable
to quasi-one-dimensional metals which are relatively pure and in which
Recent Developmer.ts and Comments 395
fluctuations above the Peierls transition temperature Tc are reduced by

interchain coupling or by three-dimensional phonons.
4.3.2. Paraconductivity from Charge-Density Waves
After the suggestion that the conductivity peak observed above Tc in

TIF-TCNQ is due to Frohlich conductivity from CDW fluctuations, a
semiphenomenological theory was derived by Allender et ai. (39) A similar
expression was derived by Strassler and Toombs(40) from an evaluation of a
diagram corresponding to that Aslimosov and Larkin used to describe
paraconductivity from fluctuations above Te in pairing superconductors.
The same result was derived by Rice from a semiphenomenological
theory. (41 )
Patton and Sham(42) showed that this diagram gives other terms, and
also evaluated terms of the same order from additional diagrams cor-
responding to those evaluated by Maki for pairing superconductors. A
further diagram, omitted by Patton and Sham, was evaluated by Takado
and Sakai.(43) All of these calculations are based on the perturbation
approach described by Sham in Chapter 5.
It is interesting to note that the additional terms from these diagrams
give a contribution proportional to l~h in the pure limit, as in one dimen-
sion. The contribution i~ positive (an increase over (Tv) only when the
electron-phonon coupling constant, A, is greater than about 1.2.
The mean-field expression for the contribution from charge-density
wave fluctuations may he written in the following form(41) in the pure limit,
T,» hj(27Tk8T):
(23)
where (ns) is the density of electrons carried by the fluctuations, pro-

portional to (~2), F = (T- TJj T., and
(24)
is the rate at which phonons decay into electron-hole pairs. In this equa-
tion, Wo is the unperturbed phonon frequency and A the electron-phonon
coupling constant. The value of (TF derived from Equation (23) is always
less than (Tv in the applicable range of T, and is far too small to account for
the conductivity peak observed above Te in TIF-TCNQ.
Lee et al.(44) pointed out that fluctuations in quasi-one-dimensional
metals could be much larger than those derived from the mean-field
calculation described above. They suggested that the mean-field transition
temperature might well be three or four times the observed Te. We have
seen that the low-temperature gap is typically the order of (8-1O)kB Tn
396 John Bardeen
much larger than the mean-field value of 3.5kBTc- Fluctuating CDWs give
a pseudogap,
(25)
They find that the reduction in density of states at the Fermi surface from
the pseudogap is
(26)
where ~(T) is the coherence length of the wave. In mean-field theory,
~ = hUFl2"A and N(O) = No.
Rice(4'i) generalized the phenomenological theory by reducing the
scattering rate r by the factor N(O)I No and allowing ~ to take on arbitrary
values. He thus found
(IF = ne2(ii,)24kBn~T) (27)

m* n '/TUFA
This expression is the one Heeger used in Chapter 3, Section 5.1, to

account for the conductivity of TTF-TCNQ below 150°K. He made use of
empirical values of ~(T) derived from diffuse x-ray scattering and assumed
that Tc is well below the mean-field transition temperature so that iis! n
and "AI t:. are both near unity. One may question these latter assumptions
because 2kF CDWs are not observed in x-ray scattering above 150°K.
4.3.3. Use of Conventional Scattering Theory
Fluctuations may be reduced so much by interchain effects that one

may use the Drude formula to determine the conductivity and the usual
expressions for other transport properties. This seems to be the case for the
TCNQ salts with substituted TTF cations studied by Bloch and co-workers.
One finds essentially the same temperature dependence for all such
compounds, even for those with such large interchain coupling that they
remain metallic down to T = OaK.
To account for the T A temperature dependence of conductivity when
one-dimensional fluctuations are suppressed, one must have something
other than acoustic phonon scattering, because the latter would give A = 1
rather than the 2.3-2.4 ohserved. The two suggestions that have received
greatest attention are scattering by low-lying intramolecular vibrations(2c)
and second-order scattering by Iibrational modes(4o) (rotational oscilla-
tions). The latter do not couple in first order. Second-order scattering
requires two quanta whose wave vectors, q] and Q2, add to 2k F.
In their phenomenological approach, Allender et al.(39) pointed out
that decay of 2k f phonons or CDW fluctuations is to electron-hole pairs
with a wave-vector difference 2k F . Thus electrons are transferred from one
side of the Fermi surface to the other with an effect similar to that of an
electric field acting on the system. Weger(47) has used this argument in a
different way to suggest why 2kF phonons are ineffective in scattering and
why a long time constant is involved in the conductivity, as indicated in
Figure 8, Chapter 3. If 2kF acoustic phonon umklapp processes are
neglected, so that the momentum of these phonons reverts to the electrons,
the equations of motion for the electrons and phonons are
d(Pe + Pd Pe
+-=eE (28)
dt Te
and
dPL Pe PL
--- (29)
dt Tph TF
Here Pe and PL represent the crystal momenta in the electrons and

phonons, respectively, 7,. is the electron lifetime for other than 2kF phonon
scattering, Tph is the electron lifetime for scattering by 2kF acoustic
phonons, and TF = r- I the lifetime for decay of phonons or fluctuating
Frohlich modes to electron-hole pairs. The ratio TF/ T ph is a large number,
of the order of MF/ m *.
If an electric field pulse is applied, a large part of the momentum will
go in a short time (~Tph) to the phonons. This momentum will be released
slowly (time ~TF) back to the electrons. Thus there will be an initial
current pulse followed by a long tail that decays slowly. The response to an
alternating field, O'(w), will be similar to that shown schematically in Figure
8, Chapter 3, with Tc = Tr.
In steady state, Pe = eTeE, so that the conductivity is determined by Te'
Further,
(30)
The latter could be taken as the definition of M F .

Intramolecular vibrations and librational modes presumably decay
rapidly to other modes, so that such scattering should be included in Te'
Impurity scattering is also included in T e , although it may be reduced by a
pseudogap. it is not yet certain which of the two mechanisms predominates
in thermal scattering in TTF-TCNQ. Rice and co-workers(48) have shown
how one may determine the coupling between electrons and intramolecu-
lar vibrations from infrared absorption data, so that when sufficient data
are available one may he able to estimate the scattering rate in the substi-
tuted TCNQ salts.
Epstein et ai.(2b) have used a mobility determined by intermolecular
vibrations, together with an activated carrier concentration, to account for
398 John Bardeen
the conductivity in NMP-TCNQ, in which the Coulomb interactions are

strong. A similar model has been applied to alloys in which NMP+ is
replaced by phenazine and to other TCNQ salts.
4.4. Nonlinear Field Dependence of Conductivity
The most direct evidence for Frohlich conductivity comes from

nonlinear electrical conductivity. Such conductivity has been observed at
low temperatures, T« Tn in many quasi-one-dimensional metals that
exhibit an incommensurate Peierls transitions. It is thought to arise from
solitons or other local modes of the pinned COW structure. True Frohlich
conductivity is very likely responsible for the remarkable nonlinear
conduction is observed in NbSe, just below each of the two Peierls tran-
sitions.
4.4.1. Solitons
As discussed by Heeger (Chapter 3, Section 8), a soliton (or phason)

may be regarded as a localized nonlinear excitation in which the phase of
the wave on one of the chains changes by 27T in an otherwise phase-locked
incommensurate Peierls distortion. A phase change of 27T with positive
gradient corresponds to putting an extra oscillation in the wave and thus an
additional quantum state for electrons at or below the energy gap. With
negative gradient. there is a missing state and the charge is of opposite sign.
It has generally been assumed that the charge of a soliton is 2e, the
charge of two electrons of opposite spin in the additional state. If a Peierls
distortion occurs in a material with strong Coulomb repulsions, such that
the Fermi level i~ at ±2kF' each state is at most singly occupied with charge
±e.
While the very-low-temperature data are in general agreement with
the soliton picture, there are some discrepancies. One, pointed out by
Heeger, is that the factor mUltiplying the exponential in the expression for
the conductivity of TTF-TCNQ is several orders of magnitude smaller than
predicted by theory. Another is that magnetic susceptibility data suggest
that the carriers responsible for the conductivity have free spins, contrary
to the soliton picture. It is possible that the nature of the defects respon-
sible for conduction is more complex than given by the simple picture(49)
such that the soliton like excitations may be uncharged or singly charged as
well as doubly charged. There may be bound states of the solitons or
solitons may combine with electrons or holes created by the Peierls gap. In
TJF- TCNQ, the conduction is probably on the TTF chain, which may have
singly charged 2kf solitons.
The energy of a soliton derived from a sinusoidal pinning potential

may be expressed in a simple form. The equation of motion for the phase
is(50)
i4> 1 a2 4> - 2 '
- - - - - d sm4>=O (31)
ax 2 C~ at 2
where Co = VF( m */ MF ) I /2 and d = COWp, where WF is the pinning
frequency. The energy for v = 0 may be written
(32)
where ~o = IiVF/ 7TA is the Pippard coherence distance. Thus if d were equal
to ~o, the energy would be equal to the energy gap, Eg = 2A, required to
create an electron-hole pair.
4.4.2. Nonlinear Conduction in NbSe3
Motions of solitons are motions of kinks in a charge-density wave, and

thus do not correspond to motion of the entire wave as envisaged by
Frohlich. It is very probable that such motion has been observed in the
nonlinear conductivity of NbSel by Monceau et al.(6) and interpreted in
terms of motion of CDWs of Peierls distortions by Ong and MonceauY S )
The Peierls distortions which give peaks in the resistivity below the tran-
sition temperatures are very weakly pinned. As indicated earlier, the peaks
are largely wiped out at microwave frequencies (see Figure 2) and also by
f- 0.6
0::
~
"-
>- 0.5
f-
>
f- 004 -lI
If)
If)
1
w
0:: 0.3f
0
_~_~~, ,,,,,,l
w
N
:::J 0.2.-
<t
::!:
0::
0 O.I~
z
20 40 60 80 100 120 140 160 180 200

TEMPERATURE (K)
Figure 2. Relative resistivity of NbSe3 at dc and 9.3 GHz. The reduction in resistivity in
strong electric fields is similar to that at microwave frequencies. (Plot from data of Ong and
Monceau, Reference 15.)
400 John Bardeen
relatively small electric fields (-0.1-1.0 e V / em). The compound NbSe3 is a

metal with nearly planar Fermi surfaces, only parts of which are affected by
the energy gaps opened up by the two Peierls transitions from incom-
mensurate CDWs. One at 145°K affects about 20% of the Fermi surface and
one at 59°K another 50%, leaving only about 30% for normal metallic
conduction when T« 50o K.
Because of the fibrous structure of NbSe3, it is difficult to measure the
transverse conductivity, but it must be in the range of good metals. Thus
NbSe3 is not quasi-one-dimensional in the sense of having a large aniso-
tropy in the conductivity, but it has some of the characteristics because the
Fermi surfaces normal to the axis along which Peierls distortions are found
are nearly planar.
The observed electric field dependence is of the form a =
aa + ab exp( - E/ Eo), where ab and Eo are temperature dependent. The
high field conductivity, aa + ab, is apparently that of the complete Fermi
surface and is ~imilar to that observed at microwave frequencies. It falls
approximately on the (T(T) vs. T plot extrapolated from temperatures
above 145°K and varies approximately as T- '. The temperature at which
the electric field. Eo, is a minimum occurs for each of the transitions at
temperatures somewhat below the transition temperature, Tc-
To get true Frohlich conductivity, the energy gap should follow the
displacement of the Fermi surface associated with a current flow. Well
below Te. in regions of the Fermi surface where there are Peierls gaps, very
little current flows and consequently there is little drag on the CDW. Just
below ~, the current is mostly normal, so again there is little drag. The
problem may be discussed in term~ of the two-fluid models. The mass flow
of the electrons i..,
(33)
The free energy of the electrons associated with the flow is
U
"Fe 1
= '2p"V n2 + 1ZPsV s2 = '2'J2/ P +:2'(p sP" /)(
P v" - Vs )2 (34)
The mutual friction between normal and superflow is a maximum when
p" = p, = !p, so one would expect that the electric field required is least at
the temperature for which this condition is satisfied. This result appears to
be qualitatively in agreement with experiment. There is as yet no under-
standing of the exponential dependence on electric field.
In the limit E ---> IX', one presumably has true Frohlich conductivity;
with the energy gaps at the displaced Fermi surface. Well below Teo where
p, = p, the added resistivity caused by the COWs is eliminated, but there is
no sign of an enhancement of conductivity over that of the normal metal.
Since NbSe, is a true three-dimensional metal, there should be good phase
coherence between neighboring chains. In spite of the gap and
consequently a smaller number of quasiparticles, the effect of scattering

appears to be about the same as if there were no gaps at the Fermi surface.
Because of the inverse T dependence of the conductivity, the scattering
must be mainly by thermal phonons. Scattering is to all parts of the Fermi
surface, of which the 2k r scattering is only a small part. Thus elimination
of effects of 2kF scattering by opening up a gap may not affect the
conductivity appreciably. It is uncertain why other scattering is not reduced
by the gap.
5. Concluding Comments
Although a great deal has been learned during the period of intense
activity in the study of quasi-one-dimensional conductors that started in
1973, many questions remain open. Emphasis has shifted from trying to
find materials of very high conductivity to the many interesting problems of
physics and chemistry involved. In summary remarks made at the Siofok
Conference(32) in the fall of 1976, the writer listed the following as the
major questions that had been discussed:
(1) Nature of the electronic structure and transport in quasi-one-

dimensional systems with strong Coulomb repulsions.
(2) Possibility of enhancement of the conductivity of Frohlich modes.
(3) Role of interchain couplings on phase transitions and on fluctua-
tions above 1;.
(4) The effect of disorder and fluctuations on transport in one dimen-
sIOn.
(5) Nature of the central peak on the structure factor S(2kF' w).
(6) Possibility of superconductive pairing at high temperatures.
Most of these questions are still open and under active investigation.
Much of the recent progress has been reviewed in this book. In particular,
Frohlich conductivity, collective transport by charge-density waves, most
likely has been established in nonlinear transport at very low temperatures
in TTF-TCNQ and below the Peierls transition temperatures in NbSe3, but
not in the fluctuation regime above Teo However, there is no evidence that
the conductivity is enhanced over what it would be in the absence of
charge-density waves, even when pinning is overcome by high electric
fields.
A discussion of the important open questions and suggestions for
future research have been given by Berlinsky in Chapter 1 and elsewhere
in the book. Some additional comments are given below.
402 John Bardeen
To get a better correspondence between theory and experiment and to

investigate other phenomena suggested by theory, such as spin-density
waves or pairing superconductivity, it would be desirable to find systems
that could serve as models other than those reported on in this book. One
would like to find materials that exhibit the phenomena to be investigated
in the simplest possible way. Even those that have been studied most
intensively, KCP, TTF-TCNQ, (TTF)I-x(TSeF)x- TCNQ, and related
TCNQ compounds have limitations as model materials, although the
thorough investigations that have been made are very valuable in showing
the richness of phenomena that can occur. None of these is as simple as
could be desired. A material that has served as a model system for strong
Coulomb interactions, NMP-TCNQ, is complicated by the disorder in the
orientation of the dipole moments on NMP+.
Further study of the inorganic compounds, TaS3, NbSe3, and related
materials should be of great interest. As mentioned earlier, study of NbSe3
under pressure should provide valuable information about the interplay
between Peierls distortions and paring superconductivity.
Aside from intrinsic scientific interest, what is the value of studies of
quasi-one-dimensional materials? Chances of obtaining conductivities in
the range of good metals now appear to be remote. It is possible that
applications will he found that involve unusual properties of these materi-
als other than high anisotropic conductivity. Perhaps the main applications
will be indirect. [n organic materials, the research should aid, through
systematic studies, in a better understanding of how charge transfer and
charge transport take place. For example, coupling between electrons and
intramolecular vibrations can be derived from optical absorption spectra.
The research is directly relevant to charge and energy transport in poly-
mers and biological materials that are composed of one-dimensional
chains. A better understanding of one-dimensional chains should help
elucidate the properties of three-dimensional materials composed of
intersecting chains, such as the A-IS compounds so important for super-
conductivity.
It is hoped that the many controversial questions regarding quasi-one-
dimensional conductors will gradually be resolved and that new questions
will be opened up by future research. As a truly interdisciplinary field, it
should provide continued excitement in the years to come.
References
l. (a) A. N. Bloch, Organic Conductors and Semiconductors, Lecture Notes in Physics, Vol.
6S, pp. 317-348, Springer-Verlag, Berlin (1977). (b) A. N. Bloch, T F. Carruthers, T O.
Poehler, and B O. Cohan, Chemistry and Physics of One-Dimensional Metals, pp. 47-85,
H. 1. Keller (ed.), Plenum Press, New York (1977).
2 (a) A J EpstelO, S Etemad, A F Ganto, and A J Heeger, Phys Rev B 5, 952-977

(1972) (b) A J EpstelO, E M Conwell, D J Sandman, and J S Miller, Solid State
Commun 23,355-358 (1977) (e) E M Conwell, Phys Rev Lett 39,777-780 (1977)
(d) A J EpstelO and E M Conwell, Solid State Commun 24,627-630 (1977)
3 L S Jdcob" J W BrdY H R Hart Jr, L V Interrante, J S Kasper, G D WatklOs, D
E Prober, and J C Bonner Phy~ Rev B 14, 1036-3051 (1976)
4 D Jerome and M \\'egu Chemistry and PhysIcs of One-DimensIOnal Metals, pp
341-368, H J Keller (ed ) Plenum Press, New York (1977)
5 T Sambongl, K T~utsuml Y 'ihIOZdkl, M Yamamoto, K Yamaya, and Y Abe, Solid
State Commun 22, 729-711 (1977)
6 P Monceau, N POng A M Portis, A Meer,chant, dnd J Rouxel, Phys Rev Lett 37,
602 (1976)
7 J M Ha~tlOg~, J P Pouget, G Shlrane, A J Heeger, N D Mlro, and A G Mac-
Diarmid, Phys Rev Lett 39 1484-1487 (1977)
8 S Etemad, Phys Ret, B 13 22'54-2261 (1976)
9 D E Moncton, R J BlrgLnedu L V Interrdnte, dnd F Wudl, Phys Rev Lett 39,507
(1977)
10 M Thlemdns, R Deltour [I Jerome dnd J R Cooper, Solid State Commun 19,21-27
(1976)
11 P M Horn, R Hermdn dnd M B Sdlamon, Phys Rev B 12, 5012-5015 (1977)
12 P M Horn and D GUidotti, Phy~ Rev B 16 491-501 (1977)
11 K Tsutuml, T T dgdgdkl M Ydmamoto, Y ShlOZdkl, M Ido, T Sambongl, K Yamaya,
and Y Abe, Phys Rev Lett 39 107'5 (1977)
14 A Meerschdnt and J Rouxel, ] Le\\ Common Metals 39, 197 (1975)
15 N POng dnd Pierre MoncedU Anomdlous tran~port properties of a hnear cham metal
NbSe" Phys Rev B 16 ,443 ,4'5'5 (1977)
16 J B Torrdnle, Chemlsm and PhY~lcs of One-DimensIOnal Metals, pp 137-166, H J
Keller (ed), Plenum PTe'," New York (1977)
17 H Shlba, Phy~ Ret B 6 910-918 (1972) AI,o see Cornehus F Coli III, Phys Rev B 9,
2150-21 '58 (1974)
18 J B [orrance Y TomklewlcL and B D S.lvLrman, Phys Rev B 15, 4718-4749
(1977)
19 L B Lolemdn, J A lohen, A P Gdnto, and A J Heeger, Phys Rev B 7, 2122-2128
(1973 )
20 K Ukel and L Shlrotam Solid State Commun (m press)
21 A J EpstelO and J S Miller Bdnd filhng and disorder m molecular conductors (to be
published)
22 G Mlhdly, K Rltvay-Emandlty, and G Gruner, Organic Conductors and Semiconduc-
tors, Lecture Note~ In PhYSICS Vol 05 pp '53'5-569, Spnnger-Verlag, Berhn (1977)
32 Organic Conductors and Semiconductors Lecture Notes In PhYSICS, Vol 65, Spnnger-
24 C G Kuper Proc R SOt London A227, 214-228 (195'5)
25 D C MdttJ~, Phy~ Rev Lett 32 714-717 (1974)
26 D E McCumber and B I Halperm Phys Rev B 1, 1054-1070 (1970)
27 P Monceau, J Peyard, J Rlchdrd dnd P Mohme, Phys Rev Lett 39, 161-164 (1977)
28 P Brue,ch, S Stra"ler and H R Zeller, Phys Rev B 12, 219-225 (1975)
29 H R Fukuyama T M Rice ,md ( M Varma, Phy~ Rev Lett 33,305-307 (1974)
30 L B Coleman, ( R Fmlher Jr A F Gdnto, and A J Heeger, Phys Status Solidi 75,
219-246 (1976)
31 W J Gunnmg, A J Heeger, I F Shchegolov, and Z I ZolotukhlO, Solid State Commun
25,981-985 (1978)
32 Organic Conductors and Semiconductors Lecture Notes In PhYSICS, Vol 65, Spnnger-
Verldg Berhn (1977)
404 John Bardeen
:no L P. Gor'kov and I. E. Dzyaloshinskii, Zh. Eksp. Teor. Fiz. 67,397-417 (1974) [Sov.
Phys.-JETP46, 198-207 (1975)1.
34. S. A. Brazovskii and L E. Dzyaloshinskii, Zh. Eksp. Teor. Fis. 71, 2338-2348 (1976)
[Sov. Phys.-JETP44, 1233-123K (1976)1.
35. N. F. Molt and W. D. Two;,e, Adv. Phys. 10, 107-164 (1961).
36. A. N. Bloch, R. g. Wiseman, and C. M. Varma, Phys. Rev. Lett. 28, 753-756 (1972).
37. A. A. Gogolin, S. P. Zolotukhin, V. I. Mol'nikov, E. L Rashba, and L F, Shchegolev,
Pis'ma Zh. Eksp. Teor. Fiz. 22, 564-569 (1975) [JETP Lett. 22, 278-280 (1975)J.
38. A. A. Abrikosov and L A. Ryzhkin, Solid State Commun. 24, 317-320 (1977); Adv. Phys.
27,147-230 (1978).
39. D. Allendel, j. W. Bray, and]. Bardeen, Phys. Rev. B 9, 119-129 (1974). ]. W. Bray,
Thesis, University of Illinois, Urbana-Champaign (1974).
40. S. Strassler and (;. A. Toombs, Phys. Lett. 46A, 321-322 (1974).
41. M. 1. Rice, Low-Dimensional Cooperative Phenomena, p. 23, H. j. Keller (ed.), Plenum
Press, New York (I <n 5): M. J. Rice, S. Striissler, and W. R Schneider, One Dimensional
Conductors, Lecture Notes in Physics, Vo!. 34, pp. 282-341, Springer-Verlag, Berlin
(1977).
42. g. R. Patton and L. 1. Sham, Phvs. Rev. Lett. 31, 631-633 (1973): 33, 631'-640 (1974).
43. S. Takado and E. Sakai, Fluctuation conductivity in one-dimensional Peierls system
(private communication).
44. P A. Lee. T. M. RIce, and P. W. Anderson, Solid State Commun. 14, 703~709 (1974).
45. M. j. Rice, Solid State Commun. 16, 1285-12K7 (1975).
46. H. Gutfreund and M. Weger, Phys. Rev. B 16, 1753-1755 (1977).
47. M. Weger (prtvate commUnIcation).
48. M. j. Rice, Phv.\ Rev. Lett. 37, 36-39 (1976); M. j. Rice, C. B. Duke, and N. 0, Lipari,
Solid Stak ((unmun. 17, 1OK9-1 093 ( 1975).
49. B. Horowitz and J A. Krumhansl, Solid State Commun. 26, 81-1'4 (1978).
50. M. j. Rice, A. 9 Bishop. J. A. Krumhansl, and S. E, Trillinger, Phys. Rev. Lett. 36,
432-434 (1976)
Author Index
Abarbanel, A. M., 372 Bardeen, J., 5-6, 14, 118, Bloch, A. N., 13-14, 122,
Abe, Y., 403 145,220,244, 302, 141, 142,144,199,
Abrahams, E., 67,145, 310,367-368,404 202, 220, 222, 224,
212,224,370 Bari, R. A., 328, 369 374,394,396,402
Abramowitz, M., 244 Barisic, S., 57, 65-67, Blount, E. I., 221
Abrikosov, A. A., 244, 143, 145,224-225, Bludeau, W., 164, 221
302,393-394,404 370,378 Blume, M., 302
Afanas'ev, A. M., 64, Barmatz, M., 143 Bogoliubov, N. N., 231,
142,244 Batra, I. P., 302 244, 302
Akhtar, M., 141 Beasley, M. R., 370 Bordas, J., 141
Alizon, J., 142, 222 Bechgaard, K., 142, 145, Borland, R. E., 349, 371
Allen, P. B., 330, 369 223-224 Boudeulle, M., 141
Allender, D., 6,14,118, Beck, A., 14, 141 Bray, J. W., 6, 14, 142,
145,220,244,307, Bennett, B. I., 302 145,220,223,244,
320,355,367, Benson, R. E., 13 367,403-404
395-396,404 Berenblyum, A. S., 371 Brazovskii, S. A., 393,
Ambegaokar, V., 337, Berezinskii, V. L., 11 7, 403
369 120, 144, 371 Breusch, P., 14
Andersen, J. R., 220, 244 Bergersen, B., 16 Bright, A. A., 141,221
Anderson, P. W., 6, 14, Bergmann, G., 330-331, Brockhouse, B. N., 65
64-65,142,220, 369 Brown, I. D., 63, 67
222,244,330,352, Berlinsky, A. J., 14-16, Brown, T. R., 371
368-369,374,404 64, 78-79, 122, 140, Bruesch, P., 7,8,42,65,
Andre, J. J., 140,307, 142-143,145,220, 140,390-391,403
368 401 Bulaevskii, L. N., 64, 65,
Anzai, H., 66, 142, Bernasconi, J., 14, 307,367,371
144-145, 223-224 65-66,223 Buravov, L. I., 121,
Anzer, H., 143 Bernstein, V., 142, 168,222 143-145,172,222,
Appel, J., 330, 369 Bertinotti, A., 65 371
Arthur, P., Jr., 143 Bertinotti, c., 65 Bush, B., 354, 371
Atovmyan, L. 0., 140 Bethe, H. A., 247 Bychkov, Yu. A., 10, 15,
Auderset, H., 65 Bieber, A., 140,368 117,120,144,341,
Auryai, H., 15 Birgeneau, R. J., 403 342,351,370-371
Axe, J. D., 64, 244 Birnboim, A., 371
Bishop, A. B., 145,404
Bak, P., 56, 59, 66-67, Bjelis, A., 59, 65, 67, Cabib, D., 90, 139, 143,
123,125-126,199, 143, 225 164,220
211-212,224,244 Blessing, R. H., 66, 140, Canfield, N. D., 219
Balkarei, Yu. M., 369 223 Carneiro, K., 42, 65-66
405
406 Author Index
Carolan, J. F., 14, Corfield, P., 220 Efetov, K. B., 245, 302
142-143 Cornelius, F., 403 Ehrenfreund, E., 15,
Carruthers, T. F.. 199, Cowan, D.O., 13,66, 141-143,220,222
224,402 140, 141, 144, 220, Ehrenreich, H., 222
Castro, G., 141, 145, 221 224 Eldridge, J. E., 202, 224
Chaikin, P. M., 142, 144, Cowley, R. A., 64, 244 Ellenson, W. D., 14,56,
221-222,328,368 Craven, R. A., 13,66, 61,66,126,145,224
Chambers, 1. Q., 221 144-145,189-190, Emery, V. J., 11, 14-16,
Chaudhari, P., 143 199,210,220,224, 58,59,66,97,123,
Chen, C. H., 141 374, 385 125-126,144,145,
Chesnut, D. B., 368 Cunningham, S. 1., 370 180,199,211-212,
Chiang, C. K., 67, 14 L Currat, R., 64-65 223-224,244,278,
144-145 Cutford, B. D., 62, 67 280,287,302-303,
Chu, C. W., 13, 144, 343,346,370,375,
220-221 Dandman, D. J., 221 383
Chui, S. T., 303, 370 Dartyge, E., 145 Engelsburg, S., 323, 368
Citford, B. D., 67 Dashen, R. F., 303 Engler, E. M., 15,67,
Cobanyan, L. A., 302 Datar, W. R., 67 143-145,202,220,
Cochran, W., 64 Davies, C. G., 67 221-224
Coffen, D. L., 219 Davis, D., 353, 356, 360, Enns, R. H., 244
Cohan, B. 0.,402 368, 371 Epstein, A. J., 142-143,
Cohen, J., 141 Debray, D., 221 221,307,368,
Cohen, J. A., 14,403 Decker, D. L., 221 386-387,397,
Cohen, M. H., 79, 219, Deiseroth, H. J., 65 402-403
221, 393 Delplanque, G., 221 Eremenko, O. N., 143
Cohen, M. J., 13,14,66, Deltour, R., 403 Etemad, S., 10, 14-15,
141,144,201-202, Denenstein, A., 67 66-67,107,122,
220,224,367,393 Denoyer, F., 8, 14, 15, 143-145,151,172,
Cohen, M. L., 368 64,66, 142, 177, 184-185,188,197,
Cohen, M. S., 141 223, 303 201-202,204,
Coleman, L. B., 5, 8, De Pasquali, G., 13, 144, 220-221, 223-224,
13-15,66,141, 144, 145,220,223,224 403
220, 367,386, 392, Deutscher, G., 145
403 Devreux, F., 143 Fabre, J. M., 66, 142,
Coleman, S., 287, 303 Dewames, R. E., 334 145,221-223
Coli, C. F., 83, 142 Dieterich, W., 65, 224, Falicov, L. M., 355, 371
Comes, R., 14-15, 32, 244,370 Ferraris, J. P., 13,66,
54-55,64-67.94, Djurek, D., 66, 145, 190, 141, 144, 220
123, 142, 145, 223 Ferrell, R. A., 6, 10,
223-224, 244, 303, Dolling, G., 64 14-15, 244, 332-333,
373 Dorner, B., 64 369
Conwell, E. M., 194,221, Douglass, D. H., 368 Fetter, A. 1.,355,371
402 Drexel, W., 65, 223, 244 Fields, J., 142
Cook, J. W., 13, 144, Dubois, J. Y., 43, 65 Fincher, C. R., Jr., 15,
220 Duke, C. B., 143,404 144,403
Cooper, J. R., 143, 162, Dunitz, J. D., 140 Finnegan, T. F., 13, 144,
165,202,221,222, Dzyaloshinskii, 1. E., 10, 220
254,403 15,145,244,302, Fischer, J. E., 224
Cooper, L. N., 244, 271, 353, 370-371, 393 Foulkes, 1. F., 371
302,310,368 Fowler, M., 276, 302
Cooper, M. J., 65 Eagens, C. F., 35 Fraenkel, G. K., 142
Coppens, P., 45, 65-66, Eastman, D. E., 143 Franulovic, K., 66, 145,
140, 143 Eckert, J., 65 223
Author Index 407
Friedel, J., 66-67,145, Green, D. c., 143 Hertier, W. R., 13, 219
210, 223-224, 228, Greene, R. L., 141, Hohenberg, P. C., 10, 15,
244 143-144,221 332,334,369
Friend, R. H., 165, 221 Grest, G. S., 370 Hone, D., 65
Friesen, W. I., 16, 143 Grimes, C. C., 371 Honjo, G., 64
Fritchie, C. J., Jr., 143 Grobman, W. D., 143 Hopfield, J. J., 89, 143
Fritzsche, H., 221 Groff, R. P., 113, 144, Horn, P. M., 13, 144-145,
Frohlich, H., 2, 3, 14, 65, 220 191, 193-195, 202,
127,140,159,227, Gruner, G., 224,403 204, 380-381,403
230,231,387,403 Guidotti, D., 191, Horovitz, B., 65-66, 212,
Fukuyama, H., 116, 144 193-195,202,224, 225,370-371,404
342,370,391,403 380-381,403 Hoshino, S., 64
Fulde, P., 142 Guinier, A., 64 Hsu, C. H., 141
Gulley, J. E., 166,222 Huang, K., 242, 245,
Gallice, J., 142,222 Gunning, W. J., 144-145, 369
Garito, A. F., 13-15,64, 202,224,392,403 Hubbard, J., 66, 223
66-67,140-145, Gurgenishvili, G. E., 302 Hufnagel, E. J., 142,219
220-224, 303, Gutfreund, H., 65, 269, Hughes, H. P., 141
367-368,403 302, 346, 356, 368, Hunig, S., 219
Garrett, P. E., 219 370-372,377,389, Hurault, J. P., 369
Gautier, F., 140,368 404
Gavaler, J. R., 368 Gyorffy, B. L., 371 Ibers, J. A., 140
Geballe, T. H., 143, 221 Ido, M., 403
Geilikman, B. T., 320, Haering, R. R., 143,244 !izumi, M., 64, 65, 223,
368 Halperin, B. 1.,234, 244, 244
Gemmer, R., 13, 144, 337,369-370,389, Imry, Y., 16,65,
220 403 144-145,225,
Geserich, H. P., 141 Hansen, H. D., 13 244-245, 370
Gesi, K., 64 Harada, J., 64 Interrante, L. V., 142,
Gibbons, P. C., 142 Harbeke, G., 65 403
Gillespie, R. J., 67 Harder, R. J., 13,219 Ireland, P. R., 67
Gillson, J. L., 13, 144, Hardy, W. N., 9, 15, 122, Ishiguro, T., 15,66-67,
220-221 145,220 142-145,199,
Ginzburg, V. L., 232, Harper, J. M. E., 143, 223-224
244,307,320,329, 221
367, 369 Hart, H. R., Jr., 142,403 Jacobs, L. S., 142,403
Giral, L., 67, 142, 145, Hassing, R. F., 234, 244 Jacobsen, C. S., 15, 141,
221-223 Hasslacher, B., 303 144,221
Girlando, A., 90, 143 Hastings, J. M., 67, 403 Jaklevic, R. C., 141
Glaser, W., 14,65,223, Hawley, M. E., 224 Jakobson, A., 141
244 Heeger, A. J., 2,13-15, Janossy, A., 224-225
Glocker, D. A., 13, 144, 64,66-67,140-145, Jeffrey, L., 65
220 201-202,220-224, Jehanno, G., 67, 145, 223
Gogolin, A. A., 13, 103, 303, 367-368, 374, Jericho, M. H., 143
117,120,144,351, 385,387,398,403 Jerome, D., 66, 82,
371,393,394,404 Heger, G., 65 142-144, 165,
Goldberg, I. B., 369 Hehanno, G., 65 167-169,220-222,
Gorkov, L. P., 10, 15, Heidenreich, R., 287, 303 370,403
65, 116, 142, 145, Herbstein, F. H., 140 Johnson, G. R., 13, 144,
231,244,302,393 Herman, F., 142, 153, 220
Grant, A. J., 141 220,302 Johnson, J. D., 282, 303
Grant, P. M., 141, 145, Herman, R. M., 145, 203, Jones, T. E., 142,221
221 223-224,403 Jose, J. V., 303
408 Author Index
Kadanoff, L. P., 303 Kosterlitz, J. M., 303 Luther, A., 11, 15-16,
Kagan, Yu., 64, 142, 244 Kotor, A. 1., 144 142,267,269-270,
Kagoshima, S., 45, 46, Kramers, H. A., 369 276, 278, 280, 282,
56,66-67,94, Krinsky, S., 303 287,302-303,343,
142-145,177,179, Krogmann, K., 13, 64, 346, 370
182,215,223-224 244 Luttinger,1. M., 258,
Kahanna, S. K., 303 Krumhansl, 1. A., 145, 302,376
Kaiser, S., 65 404 Lynn, J. W., 35, 37, 65,
Kajimura, K., 66, 142, Kukharenko, Yu. A., 371 223, 244
223 Kuper, C. G., 142, Lyubovskii, R. B., 143,
Kamimura, H., 141 230-231,244,352, 222
Kaplunov, M. G., 144 369,387,403 Lyubovskaya, R. N., 222
Karakozov, A. E., 331, Kuse, D., 65, 140-141
369 Kusslich, G., 219 Ma, S. K., 144,225,245
Karpfen, A., 142 Kwak, J. F., 142, 144, Maas, E. T., 143
Kasper, J. S., 403 221 MacDiarmid, A. G., 68,
Kats, E. 1., 353, 371 141,403
Kaufman, F. B., 143 La Placa, S., 220 Madhukar, A., 221
Keldysh, L. V., 307,332, Labbe, J., 67 Mafly, R. L., 65
367 Labes, M. M., 141 Mahler, W., 13, 219
Keller, H. J., 13,64-65, Ladik, J., 142 Maki, K., 244, 330, 369
140-141,143,145, Lambert, M., 14,64-65, Maksimav, E. G., 369
222, 224, 244, 302. 223 Maldague, P. F., 220
307,368, 370, Landau, L. D., 244, 369 Mandelstam, S., 287, 303
402-404 Langer, J. S., 337, 369 Maniv, T., 347,370
Khanna, S. K., 15,50,55, Larkin, A. I., 245, Marcelja, S., 234, 244
66,95,119,141-142, 302-303,352,371, Marchall, J., 144
179,220,223,225 395 Mashkov, S. A., 369
Kharadze, G., 145,302 Launois, H., 64-65 Mataga, N., 359, 372
Khidekel, M. L., 143-144, Lee, P. A., 11, 14, 16, Mathias, B. T., 333, 369
222,307,368.371 65-66,116,119, Mattis, D. c., 15,257,
Khomskii, D. I., 324, 127,142-145,220, 287,302-303,388,
367-369 222, 244, 303, 343, 403
Kimura, M., 303 370,374,395,404 Mavroyannis, C., 211, 225
Kirkpatrick, S., 303 Lehman, G. W., 360 Maximov, E. G., 324,
Kirzhnits, D. A., 307, Levin, K., 342, 370 367-368
324, 358, 367-368 Levy, F. A., 141 McCoy, B. M., 303
Kistenmacher, T. J., 44, Lieb, E. H., 15-16,257, McCumber, D. E., 337,
66,140,220 302, 384 369,389,403
Klemm, R. A., 16,66, Liebert, L., 65, 223, 244 McKenzie, D. R., 64
144,302,303,346, Lifshitz, E. M., 244, 369 McMillan, W. L., 95, 143,
370 Linz, A., 64 316,368
Kleppinger, J., 141 Lipari, N. 0., 66, 143, Meerschant, A., 403
Knop, W., 28, 65 404 Mees, K., 372
Knops, H. 1. F., 303 Lippard, J., 368 Megtert, S., 61, 67
Koch, T. R., 65 Little, W. A., 4,13,302, Melby, L. R., 13, 144,
Kogut, J., 257,302 320,334,337,356, 219-220
Kohn, W., 64, 227, 229, 367-369,371-372, Mel'nikov, V. I., 103,
242,244 377,387 144,221,352,371,
Kondo, J., 66, 223 Logan, B. F., 144 393
Kopaev, Ju. V., 367 Loudon, R., 65 Mengel, P., 145
Author Index 409
Menyhard, N., 289, 303, Nelson, D. R., 303 Pissanetzky, S., 143
342,347,370 Nersesyan, A. A., 302 Pitt, G. D., 141
Merrifield, R. E., 144, Neveu, A., 303 Poehler, T. 0., 13, 144,
220 Newman, P. R., 144 220,224,402
Messmer, R. P., 142, 220, Nicoli, D. F., 143 Pokhodnja, K. I., 144
372 Niedoba, H., 65 Pollak, R. A., 143
Meyhard, N., 15 Nielsen, P., 143 Portis, A. M., 221, 403
Mihaly, G., 403 Nishimoto, K., 359, 372 Pouget, J. P., 15,48, 63,
Mihaly, 1., 347, 370 Nozieres, P., 368 66-67,142,179,
Mikulski, C. M., 141 223,225,303,403
Miles, M. G., 13, 144, Ong, N. P., 13, 221, 399, Prester, M., 66, 145, 223
219-220 403 Prober, D. E., 403
Miljak, M., 221 Otnes, K" 64 Pyle, R. E., 13, 144, 220
Miller, I. S., 221, 307, Ovchinnikov, A. A., 16, Pytte, E., 223
368,387,402-403 66,303
Millet, R., 221 Overhauser, A. W., 140 Quasi, H., 219
Milliken, J., 141 Queisser, H. J., 244
Mills, D. 1., 370 Pal, 1., 224
Minot, M. I., 141 Pan, F. P., 145 Raaen, V. F., 13, 144,
Minyhard, N., 370 Parks, R. D., 224, 244, 220
Miro, N. D., 67,403 368-369 Rainer, D., 330-331, 369
Mochel, W. E., B, 219 Patel, V. V., 14, 150, 220 Rashba, E. I., 144, 221,
Molinie, P., 403 Patton, B. R., 65, 118, 393,404
Moller, W., 141 145,220,225,228, Reiger, P. N., 142
Mol'nikov, V. 1.,404 234, 243, 244-245, Renker, B., 14, 33-34,
Molulevitch, G. P., 367 371,395,404 65,223,244
Monceau, P., 13, 399,403 Pecile, C., 90, 143 Rice, M. J., 14,65-66.
Moncton, D. E., 379,403 Peebles, D. 1., 141 118,130,135,137,
Montano, P. A., 65 Peierls, R. E., 2, 13, 21, 143,223,396-397,
Montgomery, H. c., 105, 65,140,159,213, 404
144 227,230,243 Rice, S. 0., 144
Mook, H. A., 46,58.66, Penn, D., 144 Rice, T. M., 6, 10, 14-16,
143,179,223 Penney, T., 15, 143,220, 65-66,142-144,
Moran, M. J., 141 222,224 220,222,244,332,
Morawitz, H., 210,224 Perlstein, J. H., 66, 141, 334,369-370,374,
Morel, P., 368 220 377,403-404
Mori, H., 193,224 Perov, P. 1.,224 Richard, I., 403
Motegi, H., 64 Peschel, I., 15, 267 Richards, P., 122, 145
Mott, N. F., 371, 394, 269-270,276,287, Rickayzen, G., 368
404 302-303 Rietschel, H., 14, 65,
Miihlschlegel, B., 244 Petersen, J. 1., 65 223,244
Mukamel, D., 66, 212, Peyard, J., 403 Rimai, D., 13, 144-145,
225 Phillips, J. C., 65, 221, 220,224,191,204
Muller, W., 66,143-144 368 Riste, T., 64
Murgich, J., 143 Phillips, T. E., 66, 140, Ritsko, J. J., 142
220 Ritvay-Emandity, K., 403
Nakao, K., 141 Pincus, P., 16, 142-143, Robert, H., 142,222
Nathans, R., 65 145,222,244,370 Roberts, B. W., 368
Natsume, Y., 141 Pines, D., 368 Roger, W. A., 143
Nawbower, R. S., 337, Pintschovius, L., 14,65, Rosenberg, 1. P., 143
370 141,223,244 Rouxel, J., 403
410 Author Index
Rupp,1. W., Jr., 222 Schultz, T. D., 66-67, Smith, G. M., 142,219
Russel, A. A., 142,221 144-145,188,220, Smith, 1. S., 15,66, 142,
Russo, P. 1., 141 222-224, 374, 385 144, 222
Rybaczewski, E. P., 15, Schumaker, R., 222 Soda, G., 78, 82, 83, 142,
66,142-143,171. Schuster, H. G., 225, 244 172,175,222
222 Scott, B. A., 143, 145, Solyom, J., 10, 15,67,
Ryzkin, 1. A., 393-394, 220,222,224 224, 289, 294, 303,
404 Scott, J. C., 15, 142-143, 342, 347, 370
166-167,171-172, Spal, R., 67
Saillant, R. B., 65, 223, 174,203,222 Springer, T" 64
244 Sears, M., 6, 14, 244 Stanley, H. E., 223
Saitoh, M., 6, 14 Seiden, P. E., 15,90, Stegun,1. A., 244
Sak, 1., 303 139, 143, 164, Steigmeyer, E. F., 42, 65
Sakai, E., 395,404 220-222 Stephen, Ed., 368
Sakurai, J., 64 Seiler, R., 303 Stern, F., 371
Salahab, D. R., 142,220, Seitz, F., 222, 369 Stollhoff, G., 142
372 Sen, P. N., 65 Striissler, S., 14, 65-66,
Salamon, M. B., 13,66, Shacklette, L. W., 224 118, 140, 142, 145,
139,144-145,190, Shafer, D. E., 106 223,395,403-404
220-221,223-224, Sham, L. 1., 65, 118, Street, G. B., 141, 145
403 145,180,220,223, Strongin, M., 367
Salem, 1., 371 225, 234, 243-245, Stucky, G. D., 65-66,
Sambongi, T., 403 351, 371, 374, 395, 140, 145, 221,
Samuelsen, E. J., 64 404 223-224
Sandman, D. J., 13, 66, Shapiro, S. M., 14-15, Suezaki, Y., 193, 224
142-143,220,367 64,66, 142, 145, Suna, A., 145,220
Saran, M. S., 141 223-224 Susskind, 1., 257, 302
Saub, K., 65-66, 145, Shchegolev,1. F., 86, Suter,1. J., 141
224,370 121,143-145,307, Swendsen, R. H., 303
Savit, R., 303 368,371,403-404 Sy, H. K., 211, 225
Scalapino, D. J., 6, It, Shiba, H., 384,403
14,16,145,237, Shibaeva, R. P., 140, 143 Tagagaki, T., 403
244, 369-370 Shibuya, I., 64 Takado, S., 395, 404
Schafer, D. E., 13, 144, Shimaoka, K., 64 Takenaka, T., 90,143
220,222 Shiozaki, Y., 64,403 Tanner, D. B., 15, 141,
Schaffman, M. J., 66, 221 Shirane, G., 14-15,47, 145, 221
Scheiber, G., 65 59,64-67,94-95, Tao, 1. J., 145,222
Scheutzow, D., 219 142, 145, 179, Taranko, A. R., 15,66,
Schick, M., 269, 302 223-224, 244, 303, 144-145,220,
Schnatterly, S. E., 142 373,403 222-223
Schneider, W. R., 66, Shirotani, 1., 387,403 Testardi,1. R., 143, 368
223,404 Silbernagel, B., 15,66, Thiemans, M., 403
Schotte, K. D., 302 142, 222 Thomas, G. A., 13, 144,
Schottc, U., 302 Silcox, J., 141 220,222
Schrieffer, J. R., 142, Silverman, B. D., 143, Thouless, D. J., 371
302, 310, 323, 222,403 Tiablikov, S. V., 368
368-369 Simpson, A., 143 Tiedje, J. T., 5, 14, 143
Schroer, B., 303 Singer, K., 67, 172,222 Tinkham, M., 337,
Schultz, A. J., 65-66, Skanavi, G. 1.,64 369-370
140,145,221,223 Skove, M. J., 13, 144, Title, R. S., 145,222
Schultz, H., 65 220 Tolmachev, V. V" 368
Author Index 411
Tomic, S., 66, 145, 223 Walsh, M. W., Jr., Williams, D. R., 219
Tomkiewicz, Y., 15,66-67, 169-170,222 Williams, J. M., 65
144-145,166-167, Wannier, G. H., 369 Wilson, J. D., 13, 144,
169, 172-1'74, 185, Warmack, R. J., 13, 144, 219-220
188,199,203-206, 220 Wiseman, R. B., 404
212,220,222-224, Watkins, G. D., 403 Wolfram, T., 369
403 Watson, CR., 66, 143, Woynarovich, F., 67, 224
Tonewaga, T., 65 179,223 Wright on, G. C., 141,221
Toombs, G. A., 65,118, Weeks, J. D., 303 Wu, F. Y., 16,302, 384
145,395,404 Weger, M., 58, 65-66 Wudl, F., 13, 144,
Torrance, J. B., 58-59, 82,107,142-144, 219-220,403
66, 84-85, 143-145, 180,210,221-223,
180,222-223,302, 225,347,369,370, Yafet, Y., 222
385,403 397,403-404 Yagubskii, E. B., 143-144,
Trillinger, S. E., 145,404 Wei, T., 67,143,222 307,368
Tsutsumi, K., 403 Weiderbonner, J. E., 220 Yakimov, E. B., 371
Tucker, J. R., 234, 244 Weiher, J. F., 166, 222 Yamada, Y., 64
Turnbull, D., 222, 369 Weiler, L., 14-16, Yamagiski, F. G., 13, 66,
Twose, W. D., 371,394, 142-143,220 220,367
404 Weissman, R. B., 14 Yamaji, K., 66, 223
Welber, B., 221-222 Yamamoto, M., 403
Uhlenbrock, D., 303 Werner, S. A., 65, 223, Yamaya, K., 403
Ukei, K., 387,403 244 Yoffe, A. D., 141
Werthamer, N. R., 244
Valatin, J. G., 231, 244 Weyl, C., 67, 145, 182, Zarkov, G. F., 367
Van Hove, L., 15,369 216,223 Zavadovskii, A., 351,
Van Schyndel, A., 67 Wheatley, J. C., 302 370-371
Varma, C M., 14,65, Wheland, R. C, 13, 144, Zeller, H. R., 14,64-65,
144,221,370, 220-221 140-141,223,244,
403-404 Whitham, G. B., 145 307,368,403
Vekris, J. E., 67 Whitmore, D. M., 372 Zia, A., 64
Vischer, P. B., 355 Wick, R. F., 144 Zittartz, J., 244
Wiesendanger, E., 64 Zolotukhin, S. P., 144,
Walatka, V. V., 13, 66, Wilkins, J. W., 234, 244, 403-404
141,144,220 369 Zumsteg, F. C., 224
Subject Index
AC conductivity, sliding mode, 118 Bubble, 238, 242

Accumulation layer inversion, 355 Bulk superconductivity, 310, 364
Acridizine, 394
Activation energy, 203 CDW: see Charge-density wave
magnetic, 185 Central peak, 20, 39, 239, 242, 243
measurement, 201 Channel, Cooper, 341
Adz(TCNQh,394 Charge-density wave (CDW), 3,6,70,127,
Alloys, transition temperature of super- 154,180,239,242,253,254,259,
conducting, 309 270,321,328,339,363
Amplitude phase mode, 42 correlations, 301
Analysis of TCNQ salts, electron- gap, 301
microprobe, 150 interchain, 238
Anderson localization, 79 modes, 281
Anisotropy, 69 response, 284
Anomalous Green's function, 324 static, 249
Anomalous modes, 179 susceptibility for, 339, 342
Antiferromagnetic interaction, 180,254 Charge transfer, 8, 45, 154, 160, 178
ATTF-TCNQ, 86 in TSeF-TCNQ, 182
A15 compounds, 62 [C(NH 2h 12Pt(CN)4Bro,23-xH20, 42
Coherence distance, Pippard, 388
Backscattering, interchain, 346 Coherence length, 76, 86, 95, 235, 237
Backward scattering, 278 Coherent neutron scattering, 27
Balance, Faraday, 166 Commensurability, 76, 212
Band structure pinning, 136
of fulvalene, 164 Compounds
of TCNQ, 164 Adz (TCNQh, 394
ofTTF-TCNQ,77 ATTF-TCNQ, 86
BCS gap equation, 232 A15,62
BCS mean-field theory, 378 beta-«(3-) tungsten, 332
Beta-tungsten compounds, 332 Br (KCP), 1
Binding energy, polaron, 92 [C(NH2h12Pt(CN)4Bro.23 . xH 20, 42
Bogoliubov-Valatin transformation, 231 CS1.7SPt(CN)4 . xH 2 0, 42
Bond, polaron polarization, 87 CS 2TCNQ3, 93
Bonding, interstack, 176 deuterated TCNQ salts, 80
Bose-Einstein condensate, 311 diethylTSeF-TCNQ, 150
Boson free energy, 258, 261 diethylTTF-TCNQ,150
Boson representation, 258, 260, 263, 267, DSeDTF-TCNQ
279,299 cis-, 150
Bragg diffraction, 25 trans-, 150
413
414 Subject Index
Compounds (cant.) Condensate

hexamethyITSeF-TCNQ,150 Bose-Einstein, 311
hexamethylTTF-TCNQ, 150 pair, 364
Hg 2. 88 AsF 6 ,62 Condensation, phonon, 75, 230
Hg 3 _x AsF 6 ,375 Condensed phonons, 55
HMTSeF-TCNQ, 61, 85 Conductance in TTF-TCNQ, 107
HMTTF-TCNQ,61 Continuum limit, 255, 276
KCP,2,6, 177,239,242, 358,373 Cooper channel, 341
Krogman salt, 30 Cooper pairs, 271
of the Krogman type, 353 Correction, vertex, 234, 241, 322
K 1.62 [Pt (C 2 0 4 h 1 . xH 2 0, 42 Correlation
K1.7SPt(CN)4· I.5H 20, 42 CDW, 301
K2Pt(CN)4Clo.32 . xH 2 0, 42 electron, 176
K2Pt(CN)4XO.3 . xH 2 0, I function, 265, 267, 274, 283, 295
X = CI, I lattice displacement, 241
K2Se04,20 order-parameter-order-parameter, 193
Magnus-Green salt, 365 Ornstein-Zernicke, 26, 31
Mg O• 82 [Pt(C 20 4 h] . xH 20, 42 length, 167, 183
NaN0 2 ,20 longitudinal, 35
Na2Pt(CN)4Bro.23 . xH 20, 42 transverse, 34
NbSe3, 2, 375 Coulomb coupling
Nb 3Ge,314 interchain, 125
Nb 3Sn,242 Coulomb enhancement, 168, 174
(NMP+)x (Phen) 1- x (TCNQ-)x- Coulomb gas, 287
(TCNQo)I_X' 387 Coulomb interaction
NMP-TCNQ, 5, 93 interchain, 238, 239
(NMP)+ (TCNQ)-. 374 intrachain, 238
platinum cyanides, 42 Coulomb pinning, 136
platinum oxalates. 42 Coupling
Q-TCNQ,104 electron-phonon, 74
quinolinium+ (Q)+ (TCNQ2)-, 387 constant, 230
quinolinium-TCNQ,4 interchain, 76, 78, 125, 236, 243, 345
Rb1.7sPt(CN)4· xH 2 0, 42 Coupling strength, electron-exciton, 320
(SN)x, 12, 70, 382 Critical resistivity
(SNBro.4)x,70 of TSeF-TCNQ, 191
superconducting, 309 ofTTF-TCNQ,191
TaS3,375 Crystal structure
TCNQ salts, 1 of KCP, 28
TMTTF-TCNQ,86 of TSeF-TCNQ, 149
TSeF-TCNQ, 9,61,86,148,373 of (TSeFh (TTFh_x-TCNQ, 149
(TSeF)x (TTF) 1- x-TCNQ, 148 ofTTF-TCNQ, 44,149
(TSeThBr, 103 CS1.7S Pt(CN)4-xH20 ,42
TTF-AuS4C4(CF3)4,375 Cs 2TCNQ3,93
TH-CUS4C4(CF3)4,375 Curie-Weiss law, 18
TTF(D 4 )-TCNQ,80
TTF-TCNQ, 5,70,148,373
TTF-TCNQ(D 4 ),80 Damping, phonon, 230
(TTF)x (TSeF) 1 _ X- TCNQ, 9 DC conductivity, 392
(TTTh-I3' 104 of DSeDTF-TCNQ, 156
(TTT) (TCNQ)2, 86 in NBS, 399
Compressibility in NMP-TCNQ, 397
of TSeF-TCNQ, 164 ofTSeF-TCNQ, 156, 164
ofTTF-TCNQ,164 of (TSeF)x (TTF)I_x-TCNQ, 155
Subject Index 415
DC conductivity (cont. ) Dyson-Bloembergen line shape, 122

ofTTF-TCNQ, 105, 109,111,156,164, Dyson equation, 234
398
DC resistivity, 337 Elastic neutron scattering, 34, 208, 379
of DSeDTF-TCNQ, 157 of (TSeF)x (TTF)l_x-TCNQ, 155,213
ofTSeF-TCNQ,157 in TTF-TCNQ, 70, 94, 208
ofTTF-TCNQ, 109, 157 Electron correlation, 176
Density operators, 259 Electron-electron scattering, 159
Deuterated TCNQ salts, 80 Electron energy loss in TTF-TCNQ, 77,104
Dielectric constant, microwave Electron-exciton coupling strength, 320
of DSeDTF-TCNQ, 130 Electron-exciton interaction, 327, 359
ofTMTTF-TCNQ,130 Electron Green's function, 229, 231, 234,
ofTSeF-TCNQ, 130,392 240
ofTTF-TCNQ, 127, 130, 392 Electron-hole
Dielectric function, 128, 164 pair-breaking parameter, 241
Dielectric properties, 390 relaxation rate, 118
Dielectric response function, 72 Electron interaction, interchain, 238
DiethylTSeF-TCNQ, 150 Electron-libron scattering, 159
DiethylTTF-TCNQ, 150 Electron-microprobe analysis of TCNQ
Diffuse structure factor, 25, 26 salts, 150
Diffuse x-ray diffraction Electron-phonon coupling, 74
pattern, 206 constant, 230
of (TSeF)x (TTF)l_x-TCNQ, 154 Electron-phonon vertex, 322
in TTF-TCNQ, 177, 208 Electron self-energy, 233, 234
Diffuse x-ray scattering, 22, 25, 26, 31 Electron-two-phonon scattering, 159
in Hg 2 . ss AsF 6 , 61 Eliashberg equation, 322, 330
in KCP, 31 Elliott relation, 172
in platinum cyanides, 43 Emission
in platinum oxalates, 43 of one libron, 180
in TSeF-TCNQ, 61, 216 of two magnons, 180
in (TSeF)x (TTFh_ x -TCNQ, 155, of two softened librons, 180
213 of two softened phonons, 180
in TTF-TCNQ, 45,70,94 Energy gap, 231,237,254,277
Diffusionlike mode, 230 fluctuation, 233, 235, 242
Diffusive hopping, 70 hybridization, 78, 86, 188
Peierls, 201
Diffusivity, thermal, 190
Energy loss in TTF-TCNQ, electron, 77,
Displacement correlation function, lattice, 104
241 Enhancement, Coulomb, 168, 174
Displacive phase transitions. 17, 242 EPR linewidth
Distortion ofTSeF-TCNQ,174
fluctuation lattice, 233, 236 ofTTF-TCNQ,I72
incommensurate lattice, 155 EPR measurements
Peierls, 93, 206 in TSeF-TCNQ, 173
spin-Peierls, 206 in TTF-TCNQ, 166
DSeDTF-TCNQ Exchange interaction, antiferromagnetic,
cis-, 150 254
dc conductivity of, 156 Exciton
dc resistivity of, 157 exchange interaction, 327
microwave dielectric constant of, 130 mechanism, 317, 320
trans-, 150 propagator, 327
transport properties of, 156 Excitonic insulator, 240
Dynamical modulation, 25 Excitonic polaron, 87
416 Subject Index
Excitonic superconductivity, 3, 317 Hg 2 . 88 AsF 6 ,62

Excitonic superconductor, 357 diffuse x-ray scattering in, 61
filamentary, 357 Hg 3 - x AsF 6 ,375
transition temperature of, 323, 358, 361 High-temperature superconductivity, 4,
Extended Hubbard model, 251, 280 319,330,363
HMTSeF-TCNQ, 61,85
Faraday balance, 166 HMTTF-TCNQ,61
Fermions, spinless, 255 Hopping, 236, 251
F erroelectrics, 242 diffusive, 70
Filamentary excitonic superconductor, 357 interchain, 79
First-order phase transition, 212 transport, 393
Fixed point, 292 Hubbard gap, 384
Fluctuation Hubbard model, 251, 280, 338, 382
energy gap, 233, 235, 242 extended, 251, 280
lattice distortion, 233, 236 Hybridization energy gap, 78, 86, 188
in one dimension, 76,193,232,236, Hysteresis,S 7, 61
388
order, 333 Impurities, 239, 243, 351
Peieris, 167, 171, 176, 178 magnetic, 241
three-dimensional, 193 Impurity scattering, 240, 388
Wigner, 182 Incoherent x-ray scattering, 208
Forward scattering, interchain, 346 Incommensurate lattice distortion, 155
Free energy Inelastic neutron scattering, 28, 33
boson, 258, 261 in KCP, 38
Ginzburg-Landau, 233, 243 in TTF-TCNQ, 46, 58, 71, 95,177,179
Friedel oscillation, 242 Instability
Frohlich conductivity, 388,400 lattice, 17, 20, 230, 239
Frohlich mode, 127, 390 Peierls, 70, 71, 76, 93
Frohlich-Peierls phase transition, 230 SDW,254
Frohlich state, 98 structural, 332
Fulvalene, band structure of, 164 Interchain
backscattering, 346
Gap function, 232, 378 charge-density waves, 238
BCS, 232 Coulomb coupling, 125
singlet, 362 Coulomb interaction, 238, 239
triplet, 362 coupling, 76, 78,125,236,243,345
Ginzburg criterion, 233 electron interaction, 238
Ginzburg-Landau equation, 334 forward scattering, 346
Ginzburg-Landau free energy, 233, 243 hopping, 79
Ginzburg-Landau theory, 2,86 interaction, 242
Gorkov's method, 231 Interface roughening, 289
Green's function Intermolecular phonon modulation, 88
anomalous, 324 Intersite interaction, 251
electron, 229, 231, 234, 240, 241 Interstack bonding, 176
phonon, 229,241 Intrachain Coulomb interaction, 238
g-ology, 239, 278, 280, 344,363 Intramolecular phonon modulation, 89
g-shift, 171, 200 Intramolecular vibrations, 179
Inversion accumulation, 355
Hartree approximation, 234 Isotope effect, 315
Hartree-Fock approximation, 234
HexamethylTSeF-TCNQ, ISO Jordon-Wigner transformation, 282
HexamethylTTF-TCNQ, 150 Josephson tunneling, 352, 388
Subject Index 417
K1. 62 (Pt(C 20 4 hl-xH20, 42 Mean-field theory, 230, 233, 324, 378

K1.7s Pt (CN)4-1.5H 20,42 Metal-insulator transition, 5, 70
K2Pt(CN)4Clo.32-xH20,42 Mott-Hubbard, 5
K2Se04,20 Metal-semico'1ductor transition, 154, 183
KCP, 6, 358 Metal trichalcogenides, 2
crystal structure of, 28 MethylTCNQ,199
diffuse x-ray scattering in, 31 Metallic phase of (TSeF)x-
inelastic neutron scattering in, 38 (TTFh_x-TCNQ,156
phase transitions in, 35, 379 MgO,82 (Pt (C 2 0 4hl-xH 20, 42
screening parameter for, 353,360 Microwave
Knight shift, 81,124,171,203 conductivity, 121
Kohn anomaly, 21, 71, 75, 95,177,242 ofTTF-TCNQ,122
Kohn effect, 229, 233, 236 dielectric constant
Korringa product, 80 of DSeDTF-TCNQ, 130
Korringa relaxation, 85 properties of TTF-TCNQ, 120
Krogrnann salts, 30, 353 ofTMTTF-TCNQ,130
ofTSeF-TCNQ,130
Lattice displacement, correlation function, ofTTF-TCNQ,130
241 Migdal approximation, 357
Lattice distortion Migdal theorem, 322
fluctuation, 233, 236 Modulated structure, 20
incommensurate, 155 Modulation
Lattice instability, 17, 20, 230, 239 dynamical, 25
Laue technique, monochromatic, 26 intermolecular phonon, 88
Libration, molecular, 58 intramolecular phonon, 89
Libron softening, 181 static, 32
Librons, 179 Molecular libration, 58
Localization, 349 Monochromatic Laue technique, 26
Anderson, 79
MOO.38Reo.62,315
Longitudinal correlation length, 35 Mott-Hubbard transition, 5, 386
Long-range order, 275 Matt transition, 383
in one dimension, 249
Lyddane-Sachs-Teller relation, 18
NaN0 2 ,20
Magnetic activation energy, 185 Na2Pt(CN)4Bro.23 . xH 2 0, 42
Magnetic impurities, 241 NbSe3
Magnetic properties compounds, 375
of (TSeF)x (TTF) 1- [TCNQ, 172, 202, dc conductivity in, 399
204 phase transitions in, 382
ofTTF-TCNQ,166 Nb 3Ge compounds, 314
Magnetic susceptibility, 384 Nb 3Sn compounds, 242
ofNMP-TCNQ,386 Neutron scattering, 241
of (NMP+)x (Phenh - x (TCNQ-h- coheren t, 27
(TCNQO)I_x,387 elastic, 34, 208, 379
of (Q)+ (TCNQ2)-, 387 inelastic, 28, 33
of TSeF-TCNQ, 174, 205 in TTF-TCNQ, 71
of (TSeFh (TTFh _x-TCNQ, 154 NMP-TCNQ, 5, 93
ofTTF-TCNQ, 80, 83,166,195,202, conductivity of, 5
385 dc conductivity in, 397
Magnons, 180 magnetic susceptibility of, 386
Magnus-Green salt, 365 (NMP+h (Phen)l_x (TCNQ-), magnetic
Mathiessen's rule, 116, 120 susceptibility of, 387
418 Subject Index
(NMP+)x (Ph en) 1- x (TCNQ-)x- Phase pinning, 213

(TCNQoh -x, 387 Phase transitions, 378
(NMPt (TCNQ)-, 374 displacive, 17
Nonlinear conductivity, 398 first-order, 212
Nonlinear transport, 131 F rohlich-Peierls, 230
Non-Ohmic conductivity, 134, 197,202 in KCP, 35, 359
Nuclear relaxation rate, 80, 82 metal-semiconductor, 184
Nuclear spin-lattice relaxation, 197 in NbSe3, 382
in one dimension, 232, 237, 248, 332
Off-diagonal long-range order, 333 second-order, 211, 230
One-dimensionality, 69 in TaS3, 382
One-libron emission, 180 in TCNQ compounds, 380
On-site interaction, 251, 257 in TSeF-TCNQ, 199, 380
Operators in (TSeF)x (TTF)I_x-TCNQ, 155
density, 259 in TTF-CUS4C4(CF3)4, 379
spin-density, 277 in TTF-TCNQ, 8,45,54,109,197,214,
Optical infrared properties of TTF-TCNQ, 381
103 Phason modes, 13 7
Optical properties Phi (<p) particles, 13 5
ofTSeF-TCNQ, 10, 163 Phonon
ofTTF-TCNQ, 8, 70, 163 anomalies, 17,21,71,177,206
Order fluctuations, 333 condensation, 75, 230
Order parameter, 249, 335 damping, 230
Order-parameter-order-parameter correla- frequency, 230
tion function, 193 renormalized, 234
Ornstein-Zernicke approximation, 196, Green's function, 229, 241
347 modulation
Ornstein-Zernicke correlation function, intermolecular, 88
26, 31 intramolecular, 89
self-energy, 74, 229, 234, 238
Pair-breaking parameter, electron-hole,
softening, 93, 172, 230, 232, 234
241
Photoconductivity of TTF -TCNQ, 202
Pair condensate, 364
Pinned regime, 127
Pairing
Pinning, 6,76
singlet, 284, 339
commensurability, 136
triplet, 284, 339
Coulomb, 136
Paraconductivity, 395
frequency, 127, 129,390,399
Parquet approximation, 341
phase, 213
Peierls distortion, 93, 206
potential, 399
Peierls energy gap, 76, 201
Pippard coherence distance, 388
Peierls fluctuations, 167, 171, 176, 178
Platinum cyanides, 42
Peierls-Frohlich sliding mode, 159
Platinum oxalates, 42
Peierls instability, 70, 71, 76,93
Polaron, 384
Peierls transition, 54,168,347,348,351,
binding energy, 92
379,380
excitonic, 87
temperature, 75
Percolation, 393 polarization bond, 87
small, 95
Phase diagram of (TSeF)x-
(TTFh_ x -TCNQ,154 Potassium cyano-platinide, 239
Phase kinks, 135 Precession techniques, Weissenberg, 26
soliton, 13 7 Precursor scattering, 46
Phase mode, amplitude, 42 Propagator, exciton, 327
Subject Index 419
Pseudogap, 76, 96 Scattering (cont.)

Pseudointeraction, 316 neutron, 241
precursor, 46
Q-TCNQ,104 processes, 338, 341, 396
(Q+) (TCNQ2) -, magnetic susceptibility spin-flip, 172
of, 387 Screening, 352, 360
Quinoline, 394 parameter for KCP, 353, 360
Quinolinium+ (Qt (TCNQ2)-' 387 SDW: see Spin-density waves
Quinolinium-TCNQ, 4 Second-order phase transition, 230
Self-energy
Rb1.7sPt(CN)4-xH20,42 electron, 233, 234
Reflectance in TTF-TCNQ, 98, 104, 129 phonon, 74,229,234,238
Reflectivity Semiconducting phase of (TSeF)x
ofTSeF-TCNQ, 163, 164 (TTFh _x-TNCQ, 197
ofTTF-TCNQ, 70,163, 1-64 Shear softness, 90
Relaxation Sine-Gordon equation, 135, 286
Korringa, 85 Singlet
nuclear spin-lattice, 197 gap function, 362
rate pairing, 284, 339
electron-hole, 118 superconductivity, 249, 253, 345, 361
nuclear, 80, 82 susceptibility for, 339, 342
Renormalization group method, 289, 342 Sliding mode
Renormalized phonon frequency, 342 ac conductivity, 118
Representation, boson, 258, 260, 263, Peieris-Frohlich, 159
267, 279, 299 (SN)x, 1, 12,70,382
Residual resistivity of TTF-TCNQ, 115 (SNBro.4h, 70
Resistivity Soft mode, 22, 39, 242
critical transition, 71, 76
of TSeF-TCNQ, 191 Softphonon,154,177,206,230,232,
ofTTF-TCNQ,191 242
dc, 337 Softened librons, emission of two, 180
ofTTF-TCNQ, 109,112 Softened phonons, emission of two, 180
Response Softening
charge-density wave, 284 libron,181
function, dielectric, 72 phonon, 93, 172,230,232, 234
spin-density wave, 284 Softness, shear, 90
Roughening, interface, 289 Solitary wave, 135
Scaling equations, 290 Soliton, 202, 237, 398
Scaling trajectories, 291 phase-kink, 137
Scattering Specific heat
backward, 278 measurements
coherent neutron, 27 in TSeF-TCNQ, 184, 189
diffuse x-ray, 25,31 in TTF-TCNQ, 184, 189
elastic neutron, 34, 207,379 of (TSeFh (TTFh_x-TCNQ, 155
electron-electron, 159 Spin
electron-libron, 159 excitations, 281
electron-two-phonon, 159 susceptibility, 80, 166, 169
impurity, 240, 388 of (TSeF)x (TTFh _x-TCNQ, 155
incoherent x-ray, 208 waves, 83
inelastic neutron, 28, 33 in TTF-TCNQ, 58
inter chain forwarp, 346 Spin-chain model, 282
420 Subject Index
Spin-density operators, 277 TaS3,375

Spin-density wave response, 284 phase transitions of, 382
Spin-density waves, 253, 339 TCNQ
instability, 254 band structure of, 164
states, 254 compounds
static, 249 electron-microprobe analysis of, 150
susceptibility for, 339, 342 phase transitions in, 380
Spin-flip scattering, 172 spin-resonance measurements of, 150
Spin-lattice relaxation, nuclear, 197 transport measurements of, 150
Spin-Peierls distortion, 206 Tetraselenafulvalene, 148
Spin-Peierls transition, 379, 384 Tetrathiafulvalene-tetracyanoquino-
Spin-phonon interaction, 174 dimethane, 147
Spin-resonance measurements of TCNQ Tetrathiotetracene, 86
salts, 150 Thermal conductivity of TTF-TCNQ, 190,
Spinless fermions, 255 197
Static charge-density waves, 249 Thermal diffusivity of TTF-TCNQ, 190
Static modulation, 32 Thermoelectric power
Static spin-density waves, 249 of (TSeF)x (TTFh _x-TCNQ, 160
Structural instability, 332 ofTTF-TCNQ, 153, 197
Superconducting alloys, transition temper- Thermoreflectance spectra, 197
ature of, 309 Three-dimensional fluctuations, 193
Superconducting compounds, 309 TMTTF-TCNQ, 86
Superconducting transition, 348 microwave dielectric constant of, 130
temperature, 232 Trajectories, scaling, 291
Superconductivity Transformation
bulk, 310, 364 Bogoliubov-Valatin, 231
excitonic, 3, 317 Jordon-Wigner, 282
high-temperature, 319, 330, 363 Transition
singlet, 249,253.345,362,363 first-order phase, 212
susceptibility Frohlich-Peierls phase, 230
for singlet, 339, 342 metal-insulator, 5, 70
for triplet, 339, 342 metal-semiconductor, 154
transition, 348 phase, 183
triplet, 249, 253, 345, 362 Mott, 383
Superconductor Mott-Hubbard, 386
excitonic, 357 metal-insulator, 5
filamentary excitonic, 357 in one dimension, 237
transition temperature of, 312 Peierls, 54, 168, 347-348, 351, 379-380
excitonic, 323, 358, 361 region ofTTF-TCNQ, 122
Superfluid conductivity, 389 second-order phase, 211, 230
Superlattice, 70, 206 spin-Peierls, 379, 384
vector, 206 superconducting, 348
Superstructure, 19 superconductivity, 348
Surface electrons, 355 temperature
Susceptibility of excitonic superconductor, 358, 361
for charge-density waves, 339, 342 in one dimension, 237, 343
2kF,269 Peierls,75
magnetic, 384 superconducting, 232
for singlet superconductivity, 339, superconducting alloys, 309
342 of superconductor, 312
spin, 80, 166, 169 in TSeF-TCNQ, 184
for spin-density wave, 339, 342 in (TSeF)x (TTFh_x-TCNQ, 155
for triplet superconductivity, 339, 342 in TTF-TCNQ, 184
Subject Index 421
Transport (TSeF)x (TTFh _x-TCNQ (cont.)

hopping, 393 semiconducting phase of, 197
measurements specific heat of, 155
of TCNQ salts, 150 spin susceptibility of, 155
ofTSeF-TCNQ,165 thermoelectric power in, 160
ofTTF-TCNQ, 165 transport properties of, 156, 197
nonlinear, 131 (TSeThBr, 103
properties TTF-AuS4C4(CF 3)4,375
of DSeDTF-TCNQ, 156 TTF-CUS4C4(CF 3)4,375
ofTSeF-TCNQ, 156, 197 phase transitions in, 379
of (TSeF)x (TTF)l_x-TCNQ, 156, TTF(D 4 )-TCNQ,80
197 TTF-TCNQ, 5, 70,148,373
ofTTF-TCNQ, 70, 107, 156, 197 band structure of, 77
Transverse conductivity, 79 compressibility of, 164
Transverse correlation, 34 conductance in, 107
Transverse transfer integral, 78 conductivity of, 5
Trichalcogenides, metal, 2 cri tical resistivity of, 191
Triplet crystal structure of, 44, 149
gap function, 362 dc conductivity in, 105, 109, 111, 156,
pairing, 284, 339 164,398
superconductivity, 249,253,345, 362 dc resistivity of, 109, 157
susceptibility for, 339, 342 dielectric constant of, 127, 392
TSeF-TCNQ, 9, 61,86,148,373 diffuse x-ray scattering in, 45, 70, 94,
charge transfer in, 182 177, 208
compressibility of, 164 elastic neutron scattering in, 70, 208
critical resistivity of, 191 electron energy loss in, 77,104
crystal structure of, 149 EPR linewidth, 172
dc conductivity, 156, 164 EPR measurements, 166
dc resistivity, 157 inelastic neutron scattering in, 46, 58,
dielectric constant of, 392 70,95, 177,179
diffuse x-ray scattering In, 61, 216 magnetic properties of, 166
EPR linewidth, 174 magnetic susceptibility of, 80, 83, 166,
EPR measurements, 173 195,202,385
magnetic susceptibility of, 174,205 microwave conductivity of, 122
microwave dielectric constant of, 130 microwave dielectric constant of, 130
optical properties of, 10, 163 microwave properties of, 120
phase transitions in, 199, 380 neutron elastic scattering in, 94
reflectivity, 163, 164 optical infrared properties of, 103
specific heat measurements in, 184, 189 optical properties of, 8, 70, 163
transition temperature in, 184 phase transitions in, 8,45,54,109,197,
transport measurements of, 165 214, 381
transport properties of, 156, 197 photoconductivity of, 202
(TSeF)x (TTFh -x- TCNQ, 148 reflectance in, 98, 104, 129
crystal structure of, 149 reflectivity of, 70,163-164
dc conductivity of, 155 residual resistivity of, 115
diffuse x-ray diffraction of, 154 resistivity of, 112
diffuse x-ray scattering of, 155,213 specific heat measurements in, 184, 189
elastic neutron scattering of, 155,213 spin waves in, 58
magnetic properties of, 172, 202, 204 thermal conductivity of, 191, 197
magnetic susceptibility 0[, 154 thermal diffusivity of, 191
metallic phase of, 156 thermoelectric power of, 153, 197
phase diagrams of, 154 transition region of, 122
phase transitions in, 155 transition temperature in, 184
422 Subject Index
TTF-TCNQ (cant.) Wigner fluctuations, 182

transport measurements of, 165 Wigner lattice, 180, 206, 383
transport properties of, 70,107,156,
197 x-y model, 289
x-ray scattering In, 70 X-ray diffraction pattern, 206
TTF-TCNQ(D4),80 in (TSeF)x (TTFh_x-TCNQ, 154
(TTFh (TSeF)l_X-TCNQ, 9 in TTF-TCNQ, 177,208
(TTTh-I 3,104 X-ray scattering, 22, 25-26, 31
(TTT) (TCNQh, 86 in Hg 2 . 88 AsF 6 , 61
incoherent, 208
Vertex
in KCP, 31
correction, 234, 241, 322
in platinum cyanides, 43
electron-phonon, 322
in platinum oxalates, 43
functions, 341
in TSeF-TCNQ, 61, 216
Weissenberg precession techniques, 26 in (TSeF)x (TTF)l_x-TCNQ, 155,213
Wiedemann-Franz law, 191 in TTF-TCNQ, 45,70,94

Highly Conducting One-Dimensional Solids PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Highly Conducting One-Dimensional Solids PDF

Uploaded by

Copyright:

Available Formats

Highly Conducting

Editorial Board: Jozef T. Devreese. University of Antwerp, Belgium

HIGHLY CONDUCTING ONE-DIMENSIONAL SOliDS

ELECTRON SPECTROSCOPY OF CRYSTALS

PLENUM PRESS • NEW YORK AND LONDON

Main entry under title:

© 1979 Plenum Press, New York

No part of this book may be reproduced, stored in a retrieval system, or transmitted,

John Bardeen, Department of Physics, University of Illinois at Urbana-

3 4 SPin Waves In ITF-TCNQ? 58

4.3. Transport above Tc 392

Author Index 405

In the past several years a number of materIals, such as

2 Some Prellmmary Thoughts

The problem of a metal 10 one dimensIOn has long been known to

A ] Berlmsky • Department of PhYSICS University of Bntlsh Columbia Vancouver B C

lic properties." However, he immediately proceeded to qualify this remark

Very little progress was made on the subject of one-dimensional

organic semiconductors were beginning to appear, Little(5) proposed a

where 8D is the Debye temperature (k8 D is a typical phonon energy), and

4. TCNQ Salts and KCP

An important breakthrough in the manufacture of highly conducting

4.2. TTF- TCNQ

esting questions. Was the conductivity above 60 0 K enhanced by some

Bardeen's proposal, and later detailed calculations by Allender et

CDW mechanism. Soon afterwards, Lee, Rice, and Anderson (LRA)(18)

cations of these theories were to the quasi-one-dimensional square planar

ture of about 1000oK. At low temperatures they observe a well-defined

5. TTF- TCNQ and TSeF- TCNQ

5.1. Structural Transitions in TTF - TCNQ

In 1975, the long-awaited observation of the Peierls distortion in

5.2. Electromagnetic Properties of TTF- TeNQ

infrared spectrometers and Michelson interferometers are extremely

5.3. ESR and Alloys of TTF- TCNQ and TSeF- TCNQ

In charge transfer compounds such as TTF-TCNQ where two distinct,

The synthesis of the TSeF molecule by Engler and Patel at IBM(43)

4kF instability at high temperature, as it apparently does in TTF-TCNQ, it

7. Some Concluding Thoughts

This IS particularly relevant in the case of the orgamc conductors, where

The author gratefully acknowledges the hospltahty of the Natuur-

1 R E Pelerls Quantum Theory of Solzds, Oxford Umverslty Press, London (1955)

9 Marshall1 Cohen, L B Coleman, A F Ganto, and A 1 Heeger, Electromc propertIes

30 J P Pouget, S K Khanna, F Denoyer, R Comes, A F Ganto, and A J Heeger, X-ray

R. Comes and G. Shirane

1.1. Lattice Instabilities and Phonon Anomalies

1.1.1. The General Case of Displacive Phase Transitions

Since the original formulation in the early sixties(4) of a dynamical

R. Comes • Laboratoire de Physique des Solides, Associe au Centre National de la

assocIated precursor phonon anomahes (soft mode) Hlstoncally, the soft

The relatIOn to a polar opttc phonon mode IS then gIven by the

where eoo IS the hIgh-frequency dielectnc constant. ThIS leads to a Bnlloum-

At the phase transition, the eigenvectors of this mode condense into a

strong anisotropies,(7·8) highly damped or even overdamped modes,(8) and

1.1.2. The Case of One-Dimensional Metals

x(q) = L f(E k )- f(E k+ q )

where Ek is the energy of the electronic state of momentum k and f(Ed is

(ii) The motions of ions in metals are screened by the conduction

1.2. X-Ray Diffuse Scattering

1.2.1. The Origin of Diffuse Scattering

If k. and kF are, respectively, the wave vectors of the incoming and

Bragg diffrac lion

arbitrary origin is then

In the case of a perfect three-dimensional lattice, with one atom per