Professional Documents
Culture Documents
One-Dimensional Solids
PHYSICS OF SOLIDS AND LIQUIDS
SUPERIONIC CONDUCTORS
Edited by Gerald D. Mahan and Walter L. Roth
Roger P. Evrard
University of Liege
Liege, Belgium
and
Victor E. van Doren
State University of Antwerp (RUCA)
Antwerp, Belgium
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
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
References and Notes 219
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
References and Notes 243
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
A. 1. Berlinsky
1 IntroductIOn
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)
3. Excitonic Superconductivity
Tc = 1.148D e- 1/ Ap
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.
4.1. 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).
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-
Introduction to Highly Conducting One-Dimensional Solids 7
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
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-
Introduction to Highly Conducting One-Dimensional Solids 9
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
Introduction to Highly Conducting One-Dimensional Solids 11
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
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
ACKNOWLEDGMENT
References
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
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)
• •
•
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)
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
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)
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.
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
•
•
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
24 R. Comes and G. Shirane
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)]
+ If the expan~lon I~ made to second order. It Just adds the corresponding Debye-Waller factor.
X-Ray and Neutron Scattering 25
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.
26 R. Comes and G. Shirane
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)
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.
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.
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).
30 R. Comes and G. Shirane
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).
X-Ray and Neutron Scattering 31
(ii) The intensity from one sheet to the next increases as the scatter-
ing-vector component 011 increases along the chain direction:
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.
32 R. Comes and G. Shirane
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
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.
X-Ray and Neutron Scattering 33
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)
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/
34 R. Comes and G. Shirane
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).
X-Ray and Neutron Scattering 35
UJ
...J 7000
«
u
~
U'l
a:
.....
6000
.1.
iii N
0::
5000 -,
\.
~
71.000 .3 ~
:I:
~
.2 ;:::.,
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
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)
X-Ray and Neutron Scattering 37
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).
38 R. Comes and G. Shirane
+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)
X-Ray and Neutron Scattering 39
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 - - -
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).
r A z
e
6 RESOLUTION(FWHM)
5
>"
E INTENSITIES
'34 5 .. .. .. .
~
10--
15 - - -
20--
3 30·······
4 0 - - COUNTS
50 - - ~
100 . ... . . .
2
200 - -
300 - - -
400 - -
500 . ..... .
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 ..•.. ..
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)
42 R. Comes and G. Shirane
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
X-Ray and Neutron Scattering 43
, , \ 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).
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
44 R. Comes and G. Shirane
[10 0J
TC NO
TTF
Figure 17. Schematic herringbone stacking of TTF-TCNQ.
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
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*.
48 R. Comes and G. Shirane
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.
: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).
X-Ray and Neutron Scattering 49
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
~
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.
50 R. Comes and G. Shirane
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).
TTF-TCNQ
~1O 1\,
If)
<{
a::
I-
00
a::
<{ 5
o
-0-----£-0
o
-4~
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).
4000 5000
lO.2 +?, \)
...;
.... 3500 4500
'"
g
~
'".....z
:::>
..., 3000
0
4000
>-
I-
V"l
Z
W
I-
z 2500
(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).
X-Ray and Neutron Scattenng 53
ITTF - TeNa I
ONSET OF lD ORDER
AT 54K
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
54 R. Comes and G. Shirane
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).
X-Ray and Neutron Scattering 55
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).
(/)
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.
211'
- 2kF--+
0) 11 I I I 11 I
1l
b) t r
---
J
I
21T
4kF
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).
X-Ray and Neutron Scattering 61
Charge
Compound Measured 2kF transfer Reference
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
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
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
69
70 A. 1. Heeger
~ 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.
(b)
•
4 4
50
0
0
•
.I
3 :3
,
0-
"0-
X ,
><
"...
-'"
) ..><
"
0-
.I
N
2 2
. 0-
j
>< •
•
/
(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
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
Charge-Density Wave Phenomena 75
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):
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.
Charge-Density Wave Phenomena 77
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
Charge-Density Wave Phenomena 79
where
s = gJ-LBh (22)
41TkB
_ f1 + (1 + W2T~ )1 / 2 112
I
• TTF CHAIN PROTON DATA (i . F)
x TCNOCHAIN PROTON DATA ( i- Ol
0 . 12 I- -
• •
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
E
1--
150
100
50 t-(
r--
<xH
112
,
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 -~.
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).
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)
(27)
the Fermi wave number. Using this and Equation (26) leads to(58)
(28)
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
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)
. 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)
Note that
(t ' )2 . 2 I
NK sm 2q (34)
Ao=2 g2 (d)1 ,
N(0)=(t )2_1_ (35)
liw q F.S. K 21Tt
Charge-Density Wave Phenomena 89
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)
(39)
and
opt
Aph =
("
:-
r;')N
K" (0) =:-""
Aph
(40)
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
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)
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).
~
~
~
'"
~
...
Charge-Density Wave Phenomena 93
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
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
Charge-Density Wave Phenomena 97
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.
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
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)
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).
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)
£1(0)=1+8 1o
00 O'(w)- 0'(0)
w
2 dw
5. Electrical Conductivity
5.1. DC Measurements
(49)
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)
18 .---.---~--~--~----
50 60 70 80 90 100
TE MPERATURE (K)
Charge-Density Wave Phenomena 109
E
<.)
70 r---.---~----~---r---.----.
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 )
---
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
3.00 ,..---- - - - - , - - - - - - - r - - - - - - - - ,
2.8 0
,"
2.60 ....." ,,'
, ' ."41 .,...,,1······ ,"
error comes from the scatter in the data. The variance in A was calculated
using the formula
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).
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).
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
(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
(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)
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
0.4
1 11
38K 49K 54K ~
•______ .
0.3
'-I~
•
----.---
0.2
Figure 24, Temperature dependence of the TTF and TCNQ chain local susceptibilities shown
in detail (Reference 58),
Charge-Density Wave Phenomena 125
(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
(63)
(64)
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).
Charge-Density Wave Phenomena 127
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
ENERGY (eV)
0.04 0.1
~~ T, " , - - - "---,---.--
I
0.61
i
( T TF ) ( TeNQ)
I-
I
300 K
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).
(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)"
Charge-Density Wave Phenomena 129
10 3 ----------------e
38 K
10'
---- _-------e
E
<.>
c:
b
FI(WO)= 1+8 Io
x [u(w)-u(O)]
2
w -Wo
2 dw (69)
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).
Charge-Density Wave Phenomena 131
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.
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
Charge-Density Wave Phenomena 133
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________~____~________~
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.
21T)2 * 2
Vo ( T =MFwF (71 )
(72)
(73)
(p±{X-vt)=4tan
-I{ exp± [(WF) (1-v2/c6)I/2
Co
X-vt ]} (74)
(75)
136 A. 1. Heeger
(76)
(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
Charge-Density Wave Phenomena 137
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)
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
ACKNOWLEDGMENTS
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).
Charge-Density Wave Phenomena 143
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
1. Introduction
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
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
a b c (3
x (A) (A) (A) (deg)
-~.--
1852,-----.---.---.----.------..,
c-LATTICE
CONSTANT (~)
1846
18.40L------'------'---~---'------I
1252,--------,-----,---,----,-------,
1240
12.28't::......--'-----'-----'---~--
3881,----.---.---.---.---
385
~.~~ ,
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
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
154 T. D. Schultz and R. A. Craven
3. Phase Diagram
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 .
156 T. D. Schultz and R. A. Craven
Tel, and finally to the low-temperature phase(s) including the other tran-
sitions.
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.)
25 TTF - TCNQ -
.' .'
;-
/
;"
20
.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.)
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,
160 T. D. Schultz and R. A. Craven
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
(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.
162 T. D. Schultz and R. A. Craven
-60r-------~--------~------__,
-50
-30
.'
..,/ x.<::o~~. - ·
,.
. ~ .... . "
;
;:
o
-20 , .. ./ ,/
.....
~;;;~:::cc~~;~t~
>
-=--10
0..
UJ
I--
o
10
20 X·I ,O
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.
0 . 10
(5)
(6)
4.1.5. Conclusions
.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.)
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
170 T. D. Schultz and R. A. Craven
35
30
., 25
"0
E
"-
::>
E
20
Q)
<t
'0
15
c:
a.
on
)( 10
05
400
TEMPERATURE (OK)
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
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.)
174 T D. Schultz and R. A. Craven
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
4.2.3. Conclusions
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
178 T. D. Schultz and R. A. Craven
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
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
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.
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.
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
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 "
/-U
(;() -15 Ii
i
-20 '-
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.)
188 T. D. Schultz and R. A. Craven
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
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.)
190 T. D. Schultz and R. A. Craven
" 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"--- \
' ..... ""--
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
: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
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
196 T. D. Schultz and R. A. Craven
/;
TTF-TeNQ / I
-
,/ '1
/ ,
/
I
;<:
"-
/
E0 / I E "
'"
q-
/
/
'"
q-
Q '0
z I-
a: '0
"-
x'" X
'0
Figure 21. Spin susceptibility and its temperature derivative for TTF-TCNQ in the neigh-
borhood of the metal-semiconductor phase transition. (After Reference 133.)
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.)
I I
800
600
t
.,- f-X · '.O
t:
..,
::::. 400
a::
oS
..,
200
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.
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
204 T. D. Schultz and R. A. Craven
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.
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
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.
206 T. D. Schultz and R. A. Craven
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.
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
(14 )
H (' N S Se
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.
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)'
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
212 T. D. Schultz and R. A. Craven
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).
The Organic Metals (TSeF)x(TTF)I_x-TCNQ 215
""'- ........... ,
(f)
I-
Z
...... ,
:::) \TTF-TCNQ
al
Il:
<I: \\,
r II
I-
Ui
Z
w
till
II
I- II
,,
Z
-t ... ,
7. Summary
ACKNOWLEDGMENTS
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
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).
The Organic Metals (TSeF)x(TTF)I_x-TCNQ 223
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')
224 T. D. Schultz and R. A. Craven
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
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
(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
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
~ phonon
- - el ectron
Figure 1. Electron and phonon self-energy diagrams. ---- cou lomb in teract ion
230 L. f. Sham
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
\t C~-2kFCk)
By Equations (16) and (20b). then
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 .
(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.
Perturbation Approach to Lattice Instabilities 233
(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)
ec -
_[7 (3)J
64
1/3 _
- 0.51 (32)
(34)
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)
(37)
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
(45)
and
(46)
(49)
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:
and
(61)
5. Effects of Impurities
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.
Perturbation Approach to Lattice Instabilities 241
1 liT
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)
Perturbation Approach to Lattice Instabilities 243
(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
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
247
248 V.1. Emery
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
(4)
(5)
Theory of the One-DimensIOnal Electron Gas 251
and
H2 = V L Pmu{Jm+l,o-' (6)
m,CT,O"
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
1.4.1. U<O
(0 ) )( )( )( )( )( )( )( )(
II II II II
(b) )1
II '" '"
II '" '"
II '" '"
II
)E
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) ~
"
~ ~ ~ ~ ~ ~
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.
Theory of the One-DimensIOnal Electron Gas 255
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.
(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)
(13)
256 V.I. Emery
(15)
(16)
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)
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
[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)
[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
(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)
(34)
(36)
and, using Equations (26), (28), and (36), it is straightforward to verify that
[Ho, Pip] = VFPPlp
(37)
[Ho, P2p] = -VFPP2p
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
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
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
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
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 ).
Theory of the One-DimensIOnal Electron Gas 263
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),
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
Al = (27Ta)I/2
,A 2 N,
(59)
A __ e_-;--=
2 - (27Ta )1/2
Theory of the One-Dimensional Electron Gas 265
with
(60)
°
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),
(63)
where
(64)
On the other hand, in the boson picture, Equations (53), (54), (59), and
(61) give
1 21T e -alkl
= - exp( - I
- - { i sin k(x - VFt)
2mx L k>O k
+(2nk + l)[cos k(x -UFt)-In) (67)
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
Theory of the One-Dimensional Electron Gas 267
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.
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),
= {-iO(t)8(X - VFt), i= 1
(72)
-iO(t)8(x + VFt), i= 2
268 V. 1. Emery
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
Theory of the One-Dimensional Electron Gas 269
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
(79)
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
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
Theory of the One-Dimensional Electron Gas 271
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.
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.
(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)
Theory of the One-Dimensional Electron Gas 273
(92)
(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
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.
Pm =S:"
where S!. = s;" + is:''. In this notation, the Hamiltonian may be rewritten
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
(105)
(106)
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.
(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
(110)
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.
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
(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
Theory of the One-Dimensional Electron Gas 281
(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
(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)
(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
Theory of the One-Dimensional Electron Gas 283
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.
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[[
(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
Theory of the One-Dimensional Electron Gas 285
XAR -iO(l)
=-- L e- /3£~ +If(£~- £)"(m IOA(x)n)(n
A I I"OA Im)+c.c. (144)
Z m.n
H Bs = sO.L
-2-
27T a
f dx COS[(87TOs) 1/2
<l>s(x)] (146)
(147)
Theory of the One-Dimensional Electron Gas 287
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.
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
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)
(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.
find more accurate solutions, making use of results that have been obtained
in the preceding sections.
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)
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.
II II
(157)
(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
Theory of the One-Dimensional Electron Gas 293
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
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.
Low temperature
T«t:.
Correlation High temperature
function T»t:. UII<IU.c1 UII2:I U.c1
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
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
SP COW
TP sow
-6/5 6/5
- - - - - - - . - - - - - + - - - - - - . - - - - WI!
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.
Theory of the One-Dimensional Electron Gas 297
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.
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
Theory of the One-Dimensional Electron Gas 299
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
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
the coupling constant, the phase factors have been omitted from Equations
(112) and (113) and, by a similar argument, from Equations (115) and
(116).
(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
1. Introduction
305
306 H. Gutfreund and W. A. Little
(1)
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
308 H. Gutfreund and W. A. Little
2.1. Background
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
310 H. Gutfreund and W. A. Little
• ••• •
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. ' '
(2)
2.3. Limitation on Te
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
314 H. Gutfreund and W. A. Little
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).
The Prospects of Excitonic Superconductivity 315
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.
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)
(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
The Prospects of Excitonic Superconductivity 317
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.
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),
318 H. Gutfreund and W. A. Little
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
The Prospects of Excitonic Superconductivity 319
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.
3.1. Exchange
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.
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;
322 H. Gutfreund and W. A. Little
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.
p+q
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.
(10)
1
Go(p, iw n )= iW n -g(P) (11)
p(q, w)= -~
1T
Im[_l_]
c(q,w)
(14 )
. )=
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)
(18)
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('
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)
[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)
328 H. Gutfreund and W. A. Little
(28)
(29)
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
The Prospects of Excitonic Superconductivity 329
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-
330 H. Gutfreund and W. A. Little
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
(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.
(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
The Prospects of Excitonic Superconductivity 335
REAL
8(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,
..
Rn : 05 11
A: 2' 10-"crn'
I - 0 .2 ,.A
10~
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.
338 H. Gutfreund and W. A. Little
: :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.
The Prospects of Excitonic Superconductivity 339
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.
(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
The Prospects of Excitonic Superconductivity 341
TS sow
SS cow
(a) (bJ
Figure 13. Diagrammatic representation of the simplest interaction processes in the particle-
particle channel (a), and in the particle-hole channel (b).
(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.
The Prospects of Excitonic Superconductivity 343
TS
SS
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'))
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.
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
The Prospects of Excitonic Superconductivity 347
(51 )
348 H. Gutfreund and W. A. Little
+--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.
(53)
350 H. Gutfreund and W. A. Little
(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
(56)
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
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.
5. Real Models
(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.
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).
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)
-0.0647
~-0.0274
(61)
>
Q)
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,
f
"/(l dk U(p-k)<fJ(k)
<fJ(p)=- ,,/a 41T [e(k)+<fJ 2 (k)]l/2 (62)
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
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.
The Prospects of Excitonic Superconductivity 363
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.
5.1.3. Results
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
366 H. Gutfreund and W. A. Little
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
References
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)
370 H. Gutfreund and W. A. Little
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).
372 H. Gutfreund and W. A. Little
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
373
374 John Bardeen
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
·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)
Recent Developments and Comments 381
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
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.
2NJ-t~
X= - ,- , U»t (4)
1T- Jet!
where Jet! is an effective nearest-neighbor coupling constant
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.
Recent Developments and Comments 387
4. Collective Transport
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
(18)
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.
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)
(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
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
4.4.1. Solitons
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
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
Recent Developments and Comments 401
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:
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
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).
Recent Developments and Comments 403
: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
413
414 Subject Index