Theory of Superconductivity

From Weak to Strong Coupling

**Series in Condensed Matter Physics
**

Series Editors: J M D Coey, D R Tilley and D R Vij

Other titles in the series include: Nonlinear Dynamics and Chaos in Semiconductors K Aoki Modern Magnetooptics and Magnetooptical Materials A K Zvedin and V A Kotov Permanent Magnetism R Skomski and J M D Coey

Series in Condensed Matter Physics

Theory of Superconductivity

From Weak to Strong Coupling

A S Alexandrov

Loughborough University, UK

Institute of Physics Publishing Bristol and Philadelphia

c IOP Publishing Ltd 2003 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher. Multiple copying is permitted in accordance with the terms of licences issued by the Copyright Licensing Agency under the terms of its agreement with Universities UK (UUK). British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. ISBN 0 7503 0836 2 Library of Congress Cataloging-in-Publication Data are available

Commissioning Editor: Tom Spicer Production Editor: Simon Laurenson Production Control: Sarah Plenty Cover Design: Victoria Le Billon Marketing: Nicola Newey and Verity Cooke Published by Institute of Physics Publishing, wholly owned by The Institute of Physics, London Institute of Physics Publishing, Dirac House, Temple Back, Bristol BS1 6BE, UK US Ofﬁce: Institute of Physics Publishing, The Public Ledger Building, Suite 929, 150 South Independence Mall West, Philadelphia, PA 19106, USA

A Typeset in LTEX 2ε by Text 2 Text, Torquay, Devon Printed in the UK by MPG Books Ltd, Bodmin, Cornwall

Contents

Preface Introduction

ix xi

PART 1 Theory 1 Phenomenology 1.1 Normal-state Boltzmann kinetics 1.2 London theory 1.3 Flux quantization 1.4 Ogg’s pairs 1.5 Pippard and London superconductors 1.6 Ginzburg–Landau theory 1.6.1 Basic equations 1.6.2 Surface energy and thermodynamic critical ﬁeld 1.6.3 Single vortex and lower critical ﬁeld 1.6.4 Upper critical ﬁeld 1.6.5 Vortex lattice 1.6.6 Critical current 1.7 Josephson tunnelling Weak coupling theory 2.1 BCS Hamiltonian 2.2 Ground state and excitations 2.3 Meissner–Ochsenfeld effect 2.4 BCS gap, critical temperature and single-electron tunnelling 2.5 Isotope effect 2.6 Heat capacity 2.7 Sound attenuation 2.8 Nuclear spin relaxation rate 2.9 Thermal conductivity 2.10 Unconventional Cooper pairing

1 3 3 6 10 11 12 13 13 17 20 23 26 28 29 33 33 35 40 41 44 45 46 48 49 50

2

1 Electron–phonon and Coulomb interactions in the Wannier representation 4.5 Coulomb pseudopotential 3.6.6 Magnetic ﬁeld expulsion 4.7 Bipolaronic superconductivity 4.4 Eliashberg equations 3.16 3
Contents
Bogoliubov equations Landau criterion and gapless superconductivity Andreev reﬂection Green’s function formulation of the BCS theory.3 Electrons in metal 3.15 2.3.7 Charged vortex and lower critical ﬁeld 4.5 Critical temperature and isotope effect 4.2 Phonons in metal 3.6.7.8 Upper critical ﬁeld in the strong-coupling regime 4.1 On-site bipolarons and bipolaronic Hamiltonian 4.7.5 Polaronic superconductivity 4.1 Polaron band 4.6 Cooper pairing of repulsive fermions Strong-coupling theory 4.vi 2.3 Excitation spectrum and ground-state energy 4.6 Mobile small bipolarons 4.12 2.2 Damping of the polaron band 4.2 Breakdown of Migdal–Eliashberg theory in the strong-coupling regime 4.2 Inter-site bipolaron in the chain model 4.11 2.7.9 Symmetry of the order parameter
4
.4 Mixture of two Bose condensates 4.3 Small Holstein polaron and small Fr¨ hlich polaron o 4.3 Superlight inter-site bipolarons 4.13 2.6.7.7.1 Electron–phonon interaction 3.4 Polaron-polaron interaction and bipolaron 4.2 Bogoliubov equations in the strong-coupling regime 4.7.7.4 Polaron spectral and Green’s functions 4.3 Polaron dynamics 4.7.7.1 Bipolarons and a charged Bose gas 4. T = 0 Green’s functions of the BCS superconductor at ﬁnite temperatures Microscopic derivation of the Ginzburg–Landau equations 52 55 57 59 62 66 73 73 77 81 85 89 91 95 96 98 102 102 105 107 110 116 119 122 122 126 129 135 135 138 140 142 145 147 148 152 156
Intermediate-coupling theory 3.3.3.3.14 2.

4 Universal upper critical ﬁeld of unconventional superconductors Superconducting state of cuprates 8.4.4 ARPES in Y124 and Y123 8.1 Photocurrent 8.3 Low Fermi energy: pairing is individual in many cuprates 5.2 Isotope effect on Tc and on supercarrier mass 7.1 Parameter-free description of Tc 7.5 Bipolaron model of cuprates Normal state of cuprates 6.1 In-plane resistivity and Hall ratio 6.5 c-axis transport and charge pseudogap 6.4.2 Normal-state resistivity below Tc 6.4 Spin pseudogap in NMR 6.5 Sharp increase of the quasi-particle lifetime below Tc 8.Contents
PART 2 Applications to high-T c superconductors 5 Competing interactions in unconventional superconductors 5.1 Bloch theorem A.1 Low-temperature penetration depth 8.4.6 Symmetry of the order parameter and stripes Conclusion
vii 159 161 161 168 171 174 174 177 180 184 184 188 192 194 195 198 200 200 205 207 209 212 212 216 222 226 227 228 230 231 234 238 243 244 244 247 249
6
7
8
9
A Bloch states A.2 Self-energy of one-dimensional hole in a non-crossing approximation 8.6 Infrared conductivity Superconducting transition 7.3 Speciﬁc heat anomaly 7.2 Effective mass approximation A.4.1 High-Tc superconductors: different concepts 5.1 Apex bipolarons 5.4.3 Exact spectral function of a one-dimensional hole 8.3 Lorenz number: evidence for bipolarons 6.4 Bipolaron bands in high-Tc perovskites 5.4 ARPES 8.4.3 SIS tunnelling 8.2 Band structure and essential interactions in cuprates 5.2 In-plane bipolarons 5.2 SIN tunnelling and Andreev reﬂection 8.3 Tight-binding approximation
.

1 Bose–Einstein condensation temperature B.3.3.1 Grand partition function B.1 Slater determinant C.viii
Contents
252 252 253 254 254 255 257 261 261 262 264 265 265 267 271 274 279 283 294
B Quantum statistics and Boltzmann kinetics B.2 Third-order phase transition B.4 Ideal Bose gas B.2 Speciﬁc heat B.3 Pauli paramagnetism.2 Annihilation and creation operators C.5 Boltzmann equation C Second quantization C.3.4.1 Fermi energy B. Landau diamagnetism and de Haas–van Alphen quantum oscillations B.3 -operators D Analytical properties of one-particle Green’s functions E Canonical transformation References Index
.4.2 Fermi–Dirac and Bose–Einstein distribution functions B.3 Ideal Fermi gas B.

It also sets out an approach to the problem of high-temperature superconductivity. Cooper and Schrieffer (BCS). Ogg Jr and Shafroth-Butler-Blatt. Pippard. A S Alexandrov 1 January 2003
ix
. Green’s functions. it is shown how the BCS theory can be extended to the strong-coupling regime. The book could also be seen as an attempt to bring the level of university training up to the level of modern theoretical condensed matter physics. whereas identifying and understanding the microscopic origin of high-temperature superconductivity stands as one of the greatest theoretical challenges of this century. proceeding from the dynamic properties of a single polaron. The canonical Migdal–Eliashberg extension of the BCS theory to the intermediate coupling strength of electrons and phonons (or any bosons) is also discussed. Superconductivity and. The book starts with the phenomenological ideas by F and H London. where the multi-polaron problem is reduced to a charged Bose gas of bipolarons. applications of the theory to cuprates are presented in greater detail in part II along with a brief discussion of a number of alternative viewpoints. Therefore the book contains introductions to the Bloch states. Finally. high-temperature superconductivity is the topic covered in senior graduate and postgraduate courses virtually in every physics department.Preface
The observation of high-temperature superconductivity in complex layered cuprates by Bednorz and M¨ ller in 1986 should undoubtedly be rated as one u of the greatest experimental discoveries of the last century. based on the extension of the BCS theory to the strong-coupling regime. This book is conceived as a fairly basic introduction to the modern theory of superconductivity. second quantization. Then. quantum statistics and Boltzmann kinetics. in particular. Ginzburg and Landau. and proceeds with the microscopic weak-coupling theory by Bardeen. and canonical transformations in the appendices to make it easy to follow for senior undergraduate and graduate students with a basic knowledge of quantum mechanics.

.

e. some Tc = 135 K at room pressure and approximately 160 K under high (applied) pressure. By that time the frictionless ﬂow (i. 80 K) can be routinely manufactured. Nine years later. with two protons and only one neutron. The bosons in the lowest energy state within the Bose–Einstein condensate thus form a coherent macromolecule. the transition temperature of the alloy Nb3 Ge at 23 K being the highest recorded prior to the discoveries in the high-Tc cuprates. Fermions obey the Pauli exclusion principle and the Fermi–Dirac statistics. no adequate explanation of the superconductivity phenomenon appeared until the work by Bardeen. F London suggested in 1938 that the remarkable superﬂuid properties of 4 He were intimately linked to the Bose–Einstein ‘condensation’ of the entire assembly of Bose particles [3]. An assembly of bosons obey the Bose–Einstein statistics. was a Bose particle (boson). which allows all of them to occupy a single quantum state.Introduction
In 1986 Bednorz and M¨ ller [1] discovered the onset of possible superconductivu ity at exceptionally high temperatures in a black ceramic material comprising four elements: lanthanum. superﬂuidity) of liquid 4 He had been discovered below a temperature of some 2. These discoveries could undoubtedly result in large-scale commercial applications for cheap and efﬁcient electricity production. barium. Cooper and Schrieffer in 1957 [2]. subsequent work revealed that many metals and alloys displayed similar superconducting properties. while its isotope 3 He. two neutrons and two electrons. Within the next decade many more complex copper oxides (cuprates) were synthesized including the mercurycuprate compounds which. As soon as one Bose particle in the Bose liquid meets xi
.17 K. Following Kamerlingh–Onnes’ discovery of superconductivity in elemental mercury in 1911. which dictate that two identical particles must not occupy the same quantum state. It had been known that the helium atom with its two protons. to date. is a fermion. Despite intense efforts worldwide. provided long lengths of superconducting wires operating above the liquid nitrogen temperature (ca. The new phenomenon initiated by Bednorz and M¨ ller broke all constraints on the maximum Tc predicted u by the conventional theory of low-temperature superconducting metals and their alloys. copper and oxygen. have the highest conﬁrmed critical temperature for a superconducting transition. Bogoliubov [4] and Landau [5] explained how Bose statistics can lead to the frictionless ﬂow of a liquid.

m ∗∗ = 2m e 2×10−27 g. The three orders-of-magnitude difference between the critical superﬂuidity temperatures of 4 He and 3 He kindles the view that Bose–Einstein condensation might represent the ‘smoking gun’ of high-temperature superconductivity. involved the pairing of individual electrons.xii
Introduction
an obstacle to its ﬂow. Here we have a superﬂuid formed from pairs of two 3 He fermions below this temperature. in sharp contrast with the model of non-overlapping (local) pairs discussed earlier by Ogg and Schafroth. for example in the form of an impurity. In sharp contrast. in principle. However. This idea was further developed as a natural explanation of superconductivity by Schafroth [7] and Butler and Blatt in 1955 [8]. Cooper then showed that electron pairs were stable only due to their quantum interaction with the other pairs. A crucial demonstration that superﬂuidity was linked to the Bose particles and the Bose– Einstein condensation came after experiments on liquid 3 He. and with the effective mass of each boson twice that of the electron. low Tc ) superconductors nor did it explain the microscopic nature of the attractive force which could overcome the natural Coulomb repulsion between two electrons which constitute a Bose pair. which failed to show the characteristic superﬂuid transition within a reasonable wide temperature interval around the critical temperature for the onset of superﬂuidity in 4 He. The ultimate BCS theory showed that in a small interval round the Fermi energy. The same model which yields a rather precise estimate of the critical temperature of 4 He leads to an utterly unrealistic result for superconductors. with one or two exceptions. which implied that Nb3 Ge already had the highest
. all the others do not allow their condensate partner to be scattered to leave the condensate. The failure of this ‘bosonic’ picture of individual electron pairs became fully transparent when Bardeen. 3 He becomes a superﬂuid only below a very low temperature of some 0.0026 K. ‘Unfortunately’ electrons are fermions. These Cooper pairs would strongly overlap in real space. it is not at all surprising that the ﬁrst proposal for high-temperature superconductivity. the BCS theory led the vast majority of theorists to the conclusion that there could be no superconductivity above 30 K. Cooper and Schrieffer (BCS) [2] proposed that two electrons in a superconductor indeed formed a pair but of a very large (practically macroscopic) dimension—about 104 times the average inter-electron spacing. namely Tc = 104 K with the atomic density of electron pairs of about 1022 cm−3 . If two electrons are chemically coupled together. made by Ogg Jr in 1946 [6]. whose atoms were fermions. Thus. the electrons are correlated into pairs in the momentum space. The BCS theory was derived from an early demonstration by Fr¨ hlich [9] that o conduction electrons in states near the Fermi energy could attract each other on account of their weak interaction with vibrating ions of a crystal lattice. the Ogg–Schafroth picture was condemned and practically forgotten because it neither accounted quantitatively for the critical parameters of the ‘old’ (i. Therefore.e. an ensemble of such two-electron entities can. be condensed into the Bose–Einstein superconducting condensate. Highly successful for metals and alloys with a low Tc . the resulting combination is a boson with total spin S = 0 or 1.

Compared with the early Ogg– Schafroth view. In fact. Such bipolaronic states are ‘dressed’ by the same lattice deformation [13] and. Yu A Firsov. the microscopic BCS theory and its extension to the intermediate-coupling regime at a fairly basic level. J T Devreese. two fermions (now polarons) are bound into a bipolaron by lattice deformation. Chapters 1– 3 of this part are generally accepted themes in the conventional theory of superconductivity. being proportional to the inverse mass of a bipolaron. Further prediction was that high temperature superconductivity could exist in the crossover region of the electron–lattice interaction strength from the BCSlike to bipolaronic superconductivity [11. a Bose liquid of tightly bound electron pairs surrounded by the lattice deformation (i. However. V V Kabanov. D Mihailovic. the BCS theory left perhaps only a limited hope for the discovery of new materials which could be superconducting at room temperatures or. 15] considered small bipolarons as entirely localized objects. In the course of writing the book I have proﬁted from valuable and stimulating discussions with P W Anderson. Part 2 describes key physical properties of high-temperature superconductors. G M Zhao and V N Zavaritsky. Chapter 4 describes what happens to the conventional theory when the electron–phonon coupling becomes strong. L P Gor’kov. which is not yet generally accepted. Chapters 5–8 also present the author’s particular view of cuprates. that in their ground-breaking search for high-Tc superconductivity. earlier studies [14. of so-called small bipolarons) was naturally predicted [10]. K A M¨ ller. it has been shown later that small bipolarons are itinerant quasi-particles existing in the Bloch states at temperatures below the characteristic phonon frequency (chapter 4). J E Hirsch. S A Trugman. u J R Schrieffer. and subsequently in u their Nobel Prize lecture [16]. Their expectation [16] was that if ‘an electron and a surrounding lattice distortion with a high effective mass can travel through the lattice as a whole. they have a mass too large to be mobile. as visiting
. P P Edwards. from Loughborough University. and a strong electron–lattice coupling exists an insulator could be turned into a high temperature superconductor’. was reduced in comparison with an ‘ultra-hot’ local-pair Ogg–Schafroth superconductivity but turned out to be much higher than the BCS theory prediction. at least. A F Andreev. Part 1 describes the phenomenology of superconductivity. As a result. at ﬁrst sight. A R Bishop. A P Levanyuk. By extending the BCS theory towards the strong interaction between electrons and ion vibrations. W Y Liang. Quite remarkably Bednorz and M¨ ller noted in their original publication. they were stimulated and guided by the polaron model. the superconducting critical temperature. P E Kornilovitch. While the Ogg–Schafroth phenomenology led to unrealistically high values of Tc . at liquid nitrogen temperatures. 12].e. The book naturally divides into two parts. It is now clear that the Ogg–Schafroth and BCS descriptions are actually two opposite extremes of the same problem of electron–phonon interaction. Part of the writing was done while I was on leave.Introduction
xiii
Tc .

xiv

Introduction

Professor at the Hewlett-Packard Laboratories in Palo Alto, California. I wish to thank R S Williams and A M Bratkovsky for arranging my visiting professorship and enlightening discussions. I am thankful to my wife Elena and to my son Maxim for their help and understanding.

PART 1 THEORY

Chapter 1 Phenomenology

At temperatures below some critical temperature Tc , many metals, alloys and doped semiconducting inorganic and organic compounds carry the electric current for inﬁnite time without any applied electric ﬁeld. Their thermodynamic and electromagnetic properties in the superconducting state (below Tc ) are dramatically different from the normal state properties above Tc . Quite a few of them might be understood without any insight into the microscopic origin of superconductivity by exploiting the analogy between superconducting electrons and superﬂuid neutral helium atoms (liquid 4 He ﬂows without any friction below Tc 2.17 K). Realizing this analogy, F and H London developed the successful phenomenological approach in 1935 describing the behaviour of superconductors in the external magnetic ﬁeld. Ogg Jr proposed a root to high-temperature superconductivity introducing electron pairs in 1946 and Ginzburg and Landau proposed the phenomenological theory of the superconducting phase transition in 1950 providing a comprehensive understanding of the electromagnetic properties below Tc .

**1.1 Normal-state Boltzmann kinetics
**

If ions in a metal were to be perfectly ordered and were not to vibrate around their equilibrium positions, the electric resistivity would be zero. This is due to the familiar interference of waves scattered off periodically arranged scattering centres. The electron wavefunctions in an ideal periodic potential are the Bloch waves, which obey the Schr¨ dinger equation, o − ∇2 + V (r) ψnk (r) = E nk ψnk (r) 2m e (1.1)

where V (r + l) = V (r) for any lattice vector l connecting two atoms. Here and later, we use the ‘theoretical’ system of units, where the Planck and Boltzmann constants and light velocity are unity ( = kB = c = 1). The momentum and 3

4

Phenomenology

the wavevector are the same in these units ( p = k). Single-particle states are classiﬁed with the wavevector k in the Brillouin zone and with the band index n, so that ψnk (r) = u nk (r) exp(ik · r) where u nk (r) is a periodic function of r and the energy is quantized into bands (appendix A). When a weak electric ﬁeld E is applied, the Bloch states behave like free particles accelerated by the ﬁeld in accordance with the classical Newton law dk = −e E. (1.2) dt Here e is the magnitude on the elementary charge. Impurities, lattice defects and ion vibrations break down the translation symmetry, therefore equation (1.2) should be modiﬁed. When the interaction of electrons with these imperfections is weak, the electric current is found from the kinetic Boltzmann equation for the distribution function f (r, k, t) in the real r and momentum k spaces. This function can be introduced if the characteristic length of ﬁeld variations is fairly long. The number of electrons in an elementary volume of this space at time t is determined by 2 f (r, k, t) dk dr/(2π)3 . In the equilibrium state, the distribution depends only on the energy E ≡ E nk (appendix B): f (r, k, t) = n k ≡ 1 + exp E −µ T

−1

.

(1.3)

The Boltzmann equation for electrons in external electric, E, and magnetic, B, ﬁelds has the following form (appendix B): ∂f ∂f ∂f +v· − e (E + v × B) · = ∂t ∂r ∂k ∂f ∂t (1.4)

c

where v = ∂ E nk /∂ k is the group velocity. Here we consider a single band, so the band quantum number n can be dropped. By applying the Fermi–Dirac golden rule, the collision integral in the right-hand side for any elastic scattering is given by ∂f ∂t = 2πn im

c q 2 Vsc (q)δ(E k − E k+q )[ f (r, k + q, t) − f (r, k, t)] (1.5)

where Vsc (q) is the Fourier component of the scattering potential and n im is the density of scattering centres. The form of this integral does not depend on the particles’ statistics. For weak homogeneous stationary electric and magnetic ﬁelds, the Boltzmann equation is solved by the substitution f (r, k, t) = n k − ∂n k F · v. ∂E (1.6)

**Normal-state Boltzmann kinetics
**

We consider the case when the transport relaxation time deﬁned as 1 ≡ 2πn im τ (E)

2 (1 − cos )Vsc (k − k)δ(E k − E k ) k

5

(1.7)

depends only on energy E. Here is the angle between v = ∂ E k /∂ k and v. Then keeping the terms linear in the electric ﬁeld E, the Boltzmann equation for the function F becomes F · v = −eτ (E)[ E · v − F · (v × B · ∇ k )v]. (1.8)

If the magnetic ﬁeld is sufﬁciently weak (ωτ 1; ω B is the Larmour frequency), one can keep only the terms linear in B with the following result for the non-equilibrium part of the distribution function: F · v = −eτ (E)[ E · v + eτ (E)E · (v × B · ∇ k )v]. Using the current density j = −2e

k

(1.9)

v f (r, k, t) = 2e

k

v

∂n k v·F ∂E

(1.10)

**one obtains the longitudinal conductivity σx x , σx x = −2e2
**

k

∂n k 2 τ (E)vx ∂E

(1.11)

**and the Hall conductivity σ yx , σ yx = −2e3 B
**

k

∂n k 2 τ (E)(v 2 m −1 − v y vx m −1 ). y xx yx ∂E

(1.12)

Here m −1 = ∂ 2 E k /∂kα ∂kβ is the inverse mass tensor (α, β are x, y, z). αβ To calculate the longitudinal conductivities (σx x and σ yy ) and the Hall coefﬁcient RH = −σx y /(Bσx x σ yy ), we apply the effective mass approximation (appendix A), assuming for simplicity that the inverse mass tensor is diagonal (m −1 = δαβ /m α ). If the transport relaxation time is independent of energy αβ (τ (E) = τ ), we obtain, integrating by parts, σx x = and σ yx = ne2 τ mx (1.13)

ne3 Bτ 2 . mymx

(1.14)

which must. In fact. The average of t is the relaxation time τ and j = −nev average = ne2 τ E m (1. there will be no contribution from v 0 to the average electron velocity. the velocity of an electron accelerated by the ﬁeld at time t after a collision is v = v 0 − e Et/m. For example. E(t) = Eeiωt . In our ‘one-band’ approximation.17) Re σ (ν) = m(1 + ω2 τ 2 ) for the energy spectrum E k = k2 /2m. ac conductivity is obtained from dc conductivity. therefore.6
Phenomenology
Then the Hall coefﬁcient is RH = −1/en.18)
where ω p = (4πne2 /m)1/2 is the plasma frequency. It satisﬁes the sum rule
∞ 0
dω Re σ (ω) = ω2 /8 p
(1.2 London theory
The ﬁrst customary macroscopic interpretation of superconductivity as a kind of limiting case of ordinary Drude conductivity met with unresolved difﬁculties. contained a permanent current and this without the
. rather than the free-electron mass m e . As a result. be given entirely by the average of −e Et/m.15)
To calculate the current induced by the time-dependent electric ﬁeld.16)
where F does not depend on time. For example. using the approximation we have to cut the upper limit in the sum rule by a value which is well above the transport relaxation rate 1/τ but still below the interband gap. t) = n k − ∂n k F · veiωt ∂E (1. in the case of an electric ﬁeld alone. One can prove that the band mass in ω p of the sum rule should be replaced by the free-electron mass.
1. which determines the absorption of the electromagnetic radiation. Following this school of thought one was bound to search for a model of a metal which. The Hall coefﬁcient is negative if the carriers are electrons and positive if they are holes. if the upper limit is really inﬁnite. where v 0 is the velocity immediately after that collision. one can solve the Boltzmann equation using the substitution f (r. the real part of ac longitudinal conductivity. which comes into the expression for ω p . which allows for an experimental determination of the carrier density n = 2 k n k . k. These results also follow from the classical Drude model. if we replace 1/τ by −iω + 1/τ . in its most stable state. there is a band mass m. is given by ne2 τ (1. Since the electron emerges from a collision in a random direction.

But disfavouring this concept. in spite of the fact that it shows an electric current without an electric ﬁeld. Meissner’s experiment. A superconductor behaves as an enormous ideal diamagnetic atom of macroscopic dimensions expelling the entire magnetic ﬂux from its volume. however. and particularly the stability of persistent currents. therefore. The idea to replace Fermi statistics by Bose statistics in the theory of metals led F and H London to an equation for the current. The wavefunctions ψ(r) of the electrons with a ﬁnite wavevector are considerably disturbed by the magnetic ﬁeld and. From inﬁnite conductivity it only follows that the magnetic ﬂux in a superconductor must be constant and. the term in brackets
. The analogy with a ferromagnet which in its most stable state contains a permanent magnetization was recalled to support the model. showed that the magnetic ﬂux in a superconductor was always zero if the magnetic ﬁeld was low enough. such as seemed needed for the superﬂuid state of liquid 4 He and for the Meissner superconducting state.20)
where A ≡ A(r) is the vector potential such that ∇ × A = B. The screening of an applied magnetic ﬁeld is affected by volume currents instead of an atomic magnetization with a diamagnetic susceptibility χ =− 1 .London theory
7
assistance of any external ﬁeld.19)
F London [3] noticed that the degenerate Bose–Einstein gas provides a good basis for a phenomenological model. F and H London [18] showed that the Meissner phenomenon could be interpreted very simply by the assumption of a peculiar coupling in the momentum space. Indeed an experiment by Meissner and Ochsenfeld [17] revealed that the phenomenon must not be interpreted as an extreme case of high metallic conductivity. as if there were something like a condensed phase of the Bose gas (appendix B). 4π (1. a general theorem of quantum mechanics was used by Bloch. The thermodynamic stability of the superconducting state. therefore. The density of the electric current is known to be given in quantum mechanics by j(r) = e2 ie (ψ ∗ ∇ψ − ψ∇ψ ∗ ) − Aψ ∗ ψ 2m m (1. dependent upon the way in which the superconductor passes the threshold curve between the normal and superconducting states. For superconductivity. One should sum this expression over all electron states below the Fermi surface of the normal metal. which turned out to be microscopically exact. did not seem to allow for any other conclusion. according to which the most stable state of electrons for general reasons should be without a macroscopic current. So Bloch concluded that the only theorem about superconductivity which can be proved is that any theory of superconductivity considering the phenomenon as an extreme case of ordinary conductivity is refutable. in particular.

so that it vanishes in the Maxwell gauge. in the leading zero and ﬁrst order.23) ∗∗ ∗∗ 2m 2m m
where φ0 (r) is the condensate wavefunction and E 0 is the energy in the absence of the magnetic ﬁeld.24)
where n s = Ns /V is the condensate density. the magnetic ﬁeld is the same for both potentials because ∇ × ∇ f = 0. But suppose the electrons are replaced by charged bosons with mass m ∗∗ and charge e∗ . if ˜ A= A+∇ f (1. Below the Bose–Einstein condensation temperature the number of bosons Ns . The ﬁrst-order correction to the energy E 1 must be proportional to ∇ · A as a scalar.23) vanish. However.21) and the wavefunction in powers of A we obtain.20) does not vanish. respectively. The ﬁrst and third terms on the left-hand side and the ﬁrst term on the right-hand side of equation (1. which occupy the same quantum state. the bracket in equation (1. The non-trivial assumption that carriers in the superconductor are charged bosons leads to the simple London
. while φ1 (r). the unperturbed normalized condensate wavefunction is a constant (φ0 (r) = 1/V 1/2 ). The result is only a very weak Landau diamagnetism (appendix B). so the current density is obtained by multiplying equation (1. expanding equation (1. As a result. In a simply connected superconductor (as a bulk sample with no holes). We conclude that φ1 (r) = 0 and the perturbation correction to the condensate wavefunction is proportional to the square or higher powers of the magnetic ﬁeld. we can choose A.20) (paramagnetic contribution) vanishes in the ﬁrst order of A and only the last (diamagnetic) term remains. free bosons condense into a state with zero momentum. All condensed bosons have the same wavefunction. In particular. φ0 (r).21)
˜ The vector potential A can be replaced by another A without any change in physical observables. is macroscopic. which satisﬁes Maxwell’s gauge ∇ · A = 0 by imposing the ˜ condition f = ∇ · A. f (r) can be always determined for any original choice of vector ˜ potential A. In the absence of the external ﬁeld. − 1 ie∗ 1 φ0 (r) − φ1 (r) − ∗∗ A∇φ0 (r) = E 0 φ1 (r) + E 1 φ0 (r) (1. 2m ∗∗ (1. E 1 are the ﬁrst-order corrections to the wavefunction and to the energy. it turns out to be of the same order of magnitude and of the opposite sign as the last term containing the vector potential. Then. k = 0 and E 0 = 0 (appendix B).20) by Ns as j (r) = − e∗2 n s A m ∗∗ (1. Their wavefunction φ(r) obeys the Schr¨ dinger equation.8
Phenomenology
in equation (1. o − 1 (∇ + ie∗ A)2 φ(r) = Eφ(r). Thus. Hence.22) where f (r) is an arbitrary single-valued function of coordinates.

28)
The ﬁeld. L (1.30) Therefore. when a superconductor occupies half of the space with x ≥ 0.25). The solution is B(x) = H e−x/λL . so that the second term is their kinetic energy and the last term is the energy associated with the magnetic ﬁeld. F = Fs + dr
2 m ∗∗ vs (r) ns + 2
dr
B 2 (r) .London theory
9
relation between the magnetic ﬁeld and the current. we can rewrite the free energy: F = Fs + 1 8π dr(B 2 + λ2 |∇ × B|2 ). L (1. a weak magnetic ﬁeld penetrates only to a very shallow microscopic depth λL .29)
with the boundary condition B(0) = H . satisfy equation (1. therefore. with the atomic density of carriers n s = 1022 cm−3 .
. vs (r) is the velocity of condensed carriers. the free energy changes by δF = 1 4π dr(B + λ2 ∇ × ∇ × B) · δ B(r).26)
Here Fs is the free energy of the unperturbed superﬂuid. must. λ2 L d2 B −B =0 dx 2 (1. m ∗∗ = m e and e∗ = e. (1. the ﬁeld B(x) is parallel to its surface and depends only on x. which minimizes the free energy.25)
where λ2 = m ∗∗ /(4πe∗2 n s ) is known as the London penetration depth. In simple geometry. Indeed. The London equation becomes a simple equation for the magnitude B(x). Replacing vs (r) by vs (r) = j(r)/(e∗ n s ) and the current density by j = ∇ × B/4π. We can combine the London equation with the Maxwell equation ∇ × B = 4π j to obtain an equation for the magnetic ﬁeld in the superconductor: B + λ2 ∇ × ∇ × B = 0 L (1. The London equation explains the Meissner–Ochsenfeld effect. 8π
(1. where H is the external ﬁeld.27)
If B(r) changes by a small vector δ B(r). one obtains the London penetration depth as small as ˚ λL = [m e c2 /(4πn s e2 )]1/2 ≈ 600 A (in ordinary units). this equation is also derived by minimizing the free energy of a charged superﬂuid in the magnetic ﬁeld. On a L more phenomenological level.

1. .07 × 10−7 G cm2 e
(1. we can ﬁnd a contour C surrounding the hole. 2.31)
where the surface integral is taken over the cross section. 2. Let us consider a hole in a bulk superconductor (ﬁgure 1. . The magnetic ﬁeld does not penetrate into the bulk deeper than λL . e∗
(1. Hence. Hence. The wavefunction is single-valued if δ = 2π p where p = 0.10
Phenomenology
Figure 1.
. the ﬂux is quantized ( B = p 0 ) and the ﬂux quantum (in ordinary units)
0
=
π c = 2. . Normalizing the condensate wavefunction as φ(r) = 1/2 n s exp(i ) and taking j = 0 in equation (1. which includes the hole. along which the ﬁeld and the current are zero. e∗
(1. . 3.32)
=
C
dl · A(r) =
δ .33)
Here δ is a change of the phase in the round trip along the contour. we can express the vector potential along the contour as A(r) = − Then the magnetic ﬂux becomes
B
∇ . This constraint leads to a quantization of the magnetic ﬂux. . Flux trapped in the hole is quantized:
B = p 0 . .20). 1.
1.1) with the trapped magnetic ﬂux
B
=
ds · B
(1. p = 1. .34)
for e∗ = 2e as observed experimentally [19]. .3 Flux quantization
The condensate wavefunction φ(r) should be a single-valued function.

This suggested the electron constituent to be almost exclusively in the electron pair conﬁguration. his idea of real-space electron pairing was further developed as a natural explanation of superconductivity by Schafroth. with one or two exceptions. low-Tc ) superconductors nor did it explain the microscopic nature of the attractive force which could overcome the natural Coulomb repulsion between two electrons. In a more concentrated phase. of course. Extremely rapid freezing of such solutions caused an enormous decrease in measured resistance. While independent experiments in metal-ammonia solutions did not conﬁrm Ogg’s claim. The resistance of the liquid sample at −33 ◦ C was some 104 . Ogg proposed that his model could also explain the previously observed superconductivity of quasi-metallic alloys and compounds. the resulting combination is a boson with the total spin S = 0 or 1. Other experimental studies showed the solute to be diamagnetic in the concentration range characterized by liquid–liquid phase separation. The microscopic BCS theory showed that in a small interval round the Fermi energy. and one had essentially a liquid metal with a small temperature-independent Pauli paramagnetism (appendix B). The idea was motivated by his demonstration that electron pairs were a stable constituent of fairly dilute solutions of alkali metals in liquid ammonia. in principle. it appeared that the liquid–liquid phase separation was prevented. No abnormal resistance change accompanying solidiﬁcation was observed. In a more dilute phase. the electrons were still predominantly paired according to Ogg but their Bose–Einstein condensation temperature was low enough due to a low concentration.4 Ogg’s pairs
The London equation successfully explained the Meissner–Ochsenfeld effect but could not. Ogg suggested that an ensemble of such two-electron entities could. By extremely rapid cooling.e. Butler and Blatt [7. If two electrons are chemically coupled together. However. and the solution in the special concentration range was actually a high-temperature superconductor up to its melting point of the order of 190 K. Ogg argued that even such a small residual resistance was due to faulty contact with platinum electrodes. electrons are paired in the momentum space rather than in the real space. and that the system became frozen into the superconducting Bose–Einstein condensate. except for solutions in the concentration range characterized by the phase separation into two dilute liquid phases at sufﬁciently low temperatures.Ogg’s pairs
11
1. which constitute the Bose pair. In 1946 Richard Ogg Jr [6] proposed that the London ‘bosonization’ of electrons could be realized due to their pairing. the electron pairs became unstable. form a superconducting Bose–Einstein condensate. as electrons do not obey Bose statistics. Sufﬁciently rapid cooling of the solutions to temperatures in the range from −90 to −180 ◦C resulted in the production of homogeneous deep-blue solids. The
. while that of the solid at −95 ◦ C was only 16 . explain superconductivity. 8]. the Ogg–Schafroth picture was condemned and practically forgotten because it neither accounted quantitatively for the critical parameters of the ‘old’ (i. All of the solid samples proved to be good electrical conductors.

which couples the current density at one point with vector potential at all neighbouring points: j(r) = − 3 16π 2λ2 ξ L dr (r − r ) A(r ) · (r − r ) −|r−r |/ξ e . ξ m ∗∗ (1. if l ξ . ξ m ∗∗ (1. proposed by Ogg and discussed in more detail by Schafroth.5 Pippard and London superconductors
There is only one characteristic length λL in the London theory.35) by l. the true penetration depth becomes larger than the London value: λH ≈ λL (ξ/λL )1/3 > λL . we need to replace ξ in the exponent of equation (1.12
Phenomenology
Cooper pairs strongly overlap in real space. By doing so. in sharp contrast with the model of non-overlapping (local) pairs. in the opposite case λL the vector potential is essentially non-zero only in a thickness λH smaller than ξ . which couples the current density at some point r with the vector potential in the same point A(r). who introduced the second characteristic length of the superconductor. Pippard
. Pippard noticed that the London equation (1.35)
If the London penetration depth is large. which is the electron mean free path l limited by the presence of impurities. However. equation yields the London equation. Here it is natural to expect that the non-local relation between the current density and the vector potential holds within a distance of the order of l rather than ξ. A substantial step towards the microscopic theory was made by Pippard [20] and by Ginzburg and Landau [21].38) There is another characteristic length in ‘dirty’ superconductors. Thus.37)
In this Pippard limit.
1. and the integral in equation (1. the so-called coherence length ξ .36)
Applying the Maxwell equation. Butler and Blatt. In this case Pippard’s ξ. (1. Pippard proposed a phenomenological non-local relation.24). λL ξ . we obtain the exponential penetration law for the magnetic ﬁeld as in the London case but with the penetration depth λH differing from the London penetration depth λL : 1 λ2 H ≈ 4πλH e∗2 n s . We can roughly estimate the result of integration to be j(r) ≈ − λH e∗2 n s A(r). should be modiﬁed to better ﬁt experimentally observed penetration depths in superconducting metals and alloys.35) is reduced by the factor λH /ξ . the vector potential varies slowly enough to take it out of the integral at point r. |r − r |4 (1.

where λL ξ (type I). and the normal phase.
1.6. Such terms (for example. is much deeper as follows from the Ginzburg–Landau theory. In the vicinity of the transition they expanded the superﬂuid free energy Fs (equation (1.39)
In this limit the penetration depth is proportional to the square root of the impurity concentration in agreement with experimental observations. the supercurrent density is j (r) = ie∗ e∗2 (φ ∗ ∇φ − φ∇φ ∗ ) − ∗∗ A(r)n s (r). we need an equation for the condensate density and the phase (r) of the condensate wavefunction φ(r) ≡ n s (r)1/2 exp[i (r)]. which turned out to be general enough to satisfy the microscopic theory of superconductivity. Ginzburg and Landau (GL) [21] formulated the equation for φ(r). With increasing magnetic ﬁeld. we obtain from Pippard and Maxwell equations in the dirty limit λH . 2m ∗∗ m (1. Actually the difference between the London superconductors with λL ξ (type II) and the Pippard superconductors. the stiffness no longer holds.21)) because pairs are strongly correlated in real superconductors. They applied the Landau theory of the second-order phase transitions assuming that φ(r) is an order parameter which distinguishes the superconducting phase. n s = n s (r). (1. Remarkably. linear
. ∗∗ 2m 8π
Here Fn is the free energy of the normal phase in the absence of a magnetic ﬁeld. If we normalize the condensate wavefunction by the condition |φ(r)|2 = n s (r).26)). where φ(r) =0. in powers of φ(r) keeping the terms up to the fourth power: F − Fn = + dr α|φ(r)|2 + β |φ(r)|4 2 (1.Ginzburg–Landau theory
13
assumed that the contribution to j(r) coming from distances less than the mean free path is not modiﬁed by the impurities.41)
1 B 2 (r) |[∇ + ie∗ A(r)]φ(r)|2 + . As a result.6 Ginzburg–Landau theory
1.1 Basic equations The condensate density n s is homogeneous in the London theory due to a stiffness in the condensate wavefunction to small magnetic perturbations. ξ l: λH ≈ λL (ξ/l)1/2 . This equation might be rather different from the ideal Bose gas (equation (1. where φ(r) = 0.40)
To describe the superconductor in a ﬁnite magnetic ﬁeld. There are no odd terms in this expansion. and the superﬂuid density becomes inhomogeneous.

41).45)
. Varying A(r) by δ A(r) in the free energy. To make sure that the second integral in the variation of F. To determine this new length. but with a nonlinear ‘potential energy’ proportional to |φ(r)|2 and a ﬁxed total energy −α.42)
δφ ∗ (r) ds · [∇ + ie∗ A(r)]φ(r)
where the second integral is over the surface of the superconductor. δ F = 0 and both integrals should vanish for any δφ ∗ (r).40). The GL equation for a dimensionless function f (x) becomes −ξ(T )2 where ξ(T ) = d2 f (x) − f + f3 =0 dx 2 1 [2m ∗∗ |α(T )|]1/2 (1. The master equation and the supercurrent equation (1. As a result.44)
is the natural unit of length for the variation of φ(r). the equations should be supplemented by the boundary condition at the surface: n · [∇ + ie∗ A(r)]φ(r) = 0 (1. which is the case if α(T ) The coherence length is a fundamental feature of the GL theory. φ(r) = n s f (x).21). there are two solutions for the order parameter. o equation (1. Varying φ ∗ (r) by δφ ∗ (r) and integrating by parts.14
Phenomenology
and cubic in φ(r)) would contain an arbitrary phase of the order parameter and should be excluded because F is a physical observable.42). let us consider a situation when the real order parameter changes 1/2 only in one direction. and there is no magnetic ﬁeld. which corresponds to the normal state. transition. equation (1. T = Tc . one obtains δF = + dr δφ ∗ (r) αφ(r) + β|φ(r)|2 φ(r) − 1 2m ∗∗ 1 [∇ + ie∗ A(r)]2 φ(r) 2m ∗∗ (1. In equilibrium. we obtain the master equation of the theory: − 1 [∇ + ie∗ A(r)]2 φ(r) + β|φ(r)|2 φ(r) = −αφ(r). which describes the homogeneous superconducting state.40) provide a complete description of the magnetic properties of inhomogeneous superconductors for any magnetic ﬁeld. the phenomenological coefﬁcient α is (T − Tc ) and β is temperature proportional to the temperature difference α independent. we obtain the Maxwell equation ∇ ×∇ × A = 4π j with the supercurrent density as in equation (1. vanishes. and setting δ F = 0. φ(r) = 0. In the framework of the Landau theory of phase transitions. 2m ∗∗ (1. For homogeneous superconductors without a magnetic ﬁeld. and |φ(r)|2 = n s = −α/β. The superﬂuid density should vanish at the (T − Tc ) and β is a constant.43)
It looks like the Schr¨ dinger equation for the condensate wavefunction. equation (1.

There is also a general constraint on the applicability of the Landau theory of phase transitions. i.49)
because λL (T ) (Tc − T )−1/2 . The expansion of F in powers of the order parameter makes sense if a statistically averaged amplitude of ﬂuctuations |δφ| in the coherence 1/2 ns . the theory is conﬁned to the transition region |Tc − T | Tc 1. potential which requires λL (T ) This inequality leads to another constraint: |Tc − T | Tc κ2 (1. This constraint is more stringent than equation (1. as we know from the Pippard theory. The ratio of the London penetration depth and the coherence length. Using this condition and the current equation (1. respectively.Ginzburg–Landau theory
15
where n is a unit vector normal to the surface.43) linear with respect to the vector potential in Maxwell’s gauge. (1.e. volume Vc remains small compared with the order parameter itself. Substituting λL = [m ∗∗ /(4πn s e∗2 )]1/2. (1. plays a crucial role in superconducting electrodynamics: λL .47) e 2π In 1950 the pairing hypothesis was not yet accepted. Hence. |δφ| The probability of such ﬂuctuations is w exp(−δ F/T ) (1. The London equation follows from the GL theory for a weak magnetic ﬁeld because there is no contribution to φ(r) in equation (1. (1. the normal component of the current should be zero on the surface. However. The GL theory has its own limitations.40). equation (1. we obtain n · j = 0.46) κ= ξ It is known as the Ginzburg–Landau parameter.44).50)
. Hg and Pb. which is important in inhomogeneous superconductors. Sn. ξ = 1/(2m ∗∗ |α|)1/2 and n s = − α/β we obtain the temperature independent m ∗∗ β 1/2 κ= ∗ . therefore Ginzburg and Landau assumed e∗ = e in their pioneering paper [21]. Meanwhile Ogg’s phenomenology and BCS theory predict e∗ = 2e as a result of real or momentum space pairing.48) in superconductors with κ 1 like Al. this boundary condition is applied not only to the interface with vacuum but also to a superconductor–insulator interface. it differs from the London theory as there is another characteristic length ξ(T ). Since the phenomenological parameters α and β were expanded near Tc in powers of (Tc − T ).48)
It predicts the local London relation between the current density and the vector ξ(0).

|α|Vc (1. where E F is the Fermi energy and n ≈ n s (0) is the electron density. However. However. the ﬂuctuations are small. Calculating the second temperature derivative of f cond yields the difference of the speciﬁc heat C = −T ∂ 2 F/∂ T 2 of the superconducting and normal phases which turns out to be ﬁnite at T = Tc : C = (Cs − Cn )T =Tc = α 2 (0) . βTc (1. n s (0) ≈ 10 mass of supercarriers. With these parameters Gi turns out to be larger than unity when Tc ≥ 30 K.16
Phenomenology
which gives an estimate for the ﬂuctuation part of the free energy as δ F ≈ Tc .55)
The ﬁnite jump at Tc of the second derivative of the relevant thermodynamic potential justiﬁes the assumption that the transition is a second-order phase transition. a few ˚ experiments measured an extremely small coherence volume of some 100 A3 . in a homogeneous superconductor with no magnetic ﬁeld. 21 cm−3 . In novel high-temperature superconductors. Gi ≈ (Tc /E F )4 < 10−8 and the ﬂuctuation region is practically absent.53)
In low-temperature superconductors.51)
As a result.
(1. Vc (0) is about (E F /Tc )3 n −1 (chapter 2). GL considered the superconducting transition (T − Tc ) and β as a as the second-order phase transition by taking α(T ) constant near the transition. m ∗∗ ≈ 10m e (part 2). if [22] |Tc − T | Tc where Gi = Gi (1. There might be other reasons which make the expansion in powers of the order parameter impossible or make the temperature dependence of the coefﬁcients different from that in GL theory. which yields |δφ|2 ≈ Tc . Indeed.52)
m ∗∗ Tc n s (0)Vc (0)1/3
2
.52) and (1. the free-energy ﬂuctuation in the coherence volume Vc (T ) ≈ 4πξ(T )3 /3 is some δ F ≈ |α||δφ|2 Vc .54)
near the transition. such deﬁnition of the transition order might depend on the
. Here the GL theory does not apply in its canonical form because equations (1.48) are incompatible. Hence the number Gi is extremely small. fcond = (Fn − Fs )/V ) is f cond ≡ −(αn s + βn 2 /2) = s α 2 (0) 2β Tc − T Tc
2
(1. the difference of free energy densities of two phases (the so-called condensation energy. and an enhanced effective reduced superﬂuid density.

In particular. In particular. there is a critical region close to the transition. where the speciﬁc heat deviates from the GL prediction. If both λL and ξ are taken to be zero. (1. (1. This is possible only in the so-called thermodynamic critical ﬁeld H = Hc . where B = H .57)
(1. In that sense.59)
The thermodynamic critical ﬁeld is linear as a function of temperature near Tc : Hc = 4πα(0)2 β
1/2
Tc − T Tc
. 0) − .2 Surface energy and thermodynamic critical ﬁeld The GL theory allows us to understand the behaviour of superconductors in ﬁnite magnetic ﬁelds. |Tc −T | ≤ Tc Gi . Then the Gibbs energy per unit volume of the superconducting phase. the transition into a superﬂuid state might differ from the second-order phase transition if the order of the transition is deﬁned by the derivative of the thermodynamic potential.2. is H2 H2 H2 − = f n (T. N. While two phases with zero. Landau proposed a more general deﬁnition of the second-order phase transition.60)
. 0) +
2 α2 Hc = fcond = . N.56)
Let us consider the equilibrium in the external ﬁeld between the normal and superconducting phases separated by an inﬁnite plane boundary at x = 0. x ≤ 0. In general. the second-order phase transition is a continuous transition.
(1. the relevant thermodynamic potential describing the equilibrium state is the Gibbs energy G(T. x ≥ 0 and with no order parameter in the normal phase on the lefthand side of the boundary. H ) = F(T.Ginzburg–Landau theory
17
approximation made. 1. N. If the external ﬁeld H is ﬁxed by external currents.and ﬁnite-order parameters coexist at Tc of the ﬁrst-order transition. N. 8π 2β
(1. N. it predicts qualitatively different properties for type I and II superconductors in sufﬁciently strong ﬁelds.6. the transition of an ideal Bose gas into the Bose-condensed state is of the third order with a jump in the third derivative of the free energy (appendix B). is gs = f s (T. This deﬁnition does not depend on our approximations (for the theory of phase transitions see [23]). the ordered phase of the second-order transition should have zero-order parameter at Tc .58) 8π 4π 8π Two phases are in equilibrium if their Gibbs energy densities are equal. For example. ﬁgure 1. the boundary would be sharp with no magnetic ﬁeld penetrating into the superconducting phase on the righthand side. B) − 1 4π dr H · B(r). where B = 0. where gn = f n (T. 0) and the Gibbs energy density in the normal phase.

18
Phenomenology
[.

κ<<1 . φ[.

in any si mply connected superconductor. A y . We can always choose the vector potential in this form. To describe the effect of the ﬁnite boundary ‘thickness’. Also using the gauge transformation (1. Magnetic ﬁeld and order parameter near the boundary between the normal and superconducting phases of type I (a) and type II (b) superconductors.
[ κ>>1 [
Figure 1. It is convenient to introduce the dimensionless coordinate x = x/λL .64) (1.62)
. the boundary is not sharp because both characteristic lengths are ﬁnite.2. The ˜ boundary conditions in the problem are: f =0 a = h = 2−1/2 (1.63) where the double prime means the second derivative with respect to x. we introduce the surface energy of the boundary deﬁned as the difference between the true Gibbs energy and that of a homogeneous sample in the external ﬁeld H = Hc : σs = 1 8π +
∞ −∞ ∞ −∞ 2 dx [B 2(x) − 2Hc B(x) + Hc ]
dx
α|φ(x)|2 +
dφ(x) β 1 |φ(x)|4 + ∗∗ 2 2m dx
2
+
e∗2 A2 |φ(x)|2 . 2m ∗∗ (1. = 0. Then the GL equations describing the surface energy take the following form: f + κ 2 [(1 − a 2 ) f − f 3 ] = 0 and a = af 2 (1. 0}) when the ﬁeld is parallel to the boundary. the order parameter can be made real.22).61)
The term proportional to A · ∇φ(x) vanishes because the vector potential has no x-component ( A = {0. the dimensionless vector potential a = A y /(21/2 Hc λL ) ˜ and the dimensionless magnetic ﬁeld h = B/(21/2 Hc ).
In fact.

the surface energy is zero if a = 2−1/2(1 − f 2 ). ˜ (1. Hence. the contribution of the ﬁrst term is negative and is about −1. we obtain a2 f 2 1 − 1 2κ 2 = 0.70)
. (1.66) allows us to simplify the surface energy as follows: σs = =
2 λL Hc 4π 2 λL Hc 2π ∞ −∞ ∞ −∞
d x {(a − 2−1/2 )2 + (a 2 − 1) f 2 + f 4 /2 + f 2 /κ 2 } ˜ d x {a (a − 2−1/2 ) + f 2 /κ 2 }. the contribution of the second term is positive. This term is non-zero in the region of the order of |x | º 1/κ.63)) yields a = −21/2 f f = a f 2 or f = −a f /21/2. Its value is about −1 in the boundary region of thickness |x | < 1. Integrating the last term f 2 /κ 2 in the ﬁrst integral of equation (1.67)
Let us estimate the contributions to the surface energy of the ﬁrst and second terms in the last brackets. The exact borderline between the Pippard and London superconductors is deﬁned by the condition σs = 0. Substituting this f and a (equation (1. where its value is about 1. In contrast.67) by parts and substituting f from the master equation (1. Equation (1. (1.68)
Hence. because a = h < 2−1/2 in any part of the sample.65)
for x = ∞ (in the superconducting phase). Multiplying equation (1.66)
where the constant (= 1/2) on the right-hand side is found using the boundary conditions.62) by f and integrating it over x yields the ﬁrst integral as ˜ f 2 /κ 2 + (1 − a 2 ) f 2 − f 4 /2 + a 2 = 1/2 (1. The ﬁrst term a (a − 2−1/2) is zero both in the normal and superconducting phases. We conclude that the surface energy is positive in extreme type I superconductors where κ 1 but it is negative in type II superconductors where κ 1. ˜
(1.Ginzburg–Landau theory
for x = −∞ (in the normal phase) and f =1 a =0
19
(1.69)
Now the second equation of GL theory (equation (1. its contribution to the integral is about +1/κ.62) we obtain σs =
2 λL Hc 4π ∞ −∞
d x {(a − 2−1/2)2 − f 4 /2}.69)) into the ﬁrst integral of motion (equation (1.66)). Hence.

Their order parameter is modulated in space so that the normal and superconducting regions are mixed.72)
1 d ρ dh = h.73)
. The properties of a single vortex are well described by the GL equations with proper boundary conditions. . Let us assume that the vortex appears at the origin of the coordinate system and has an axial symmetry.20
Phenomenology
%
ξ
M
Figure 1. Then the GL equations become 1 1 d df ρ − 3 κ 2 ρ dρ dρ f and dh dρ
2
− f3 = 0
(1. 21/2 (1. A generalization to superconductors is due to Abrikosov [24].71)
Due to the negative surface energy.3.3 Single vortex and lower critical ﬁeld With increasing external ﬁeld.3. 1.6. type II superconductors are inhomogeneous in sufﬁciently strong magnetic ﬁelds.
Hence. Single-vortex core screened by supercurrent. Similar vortex lines were found in superﬂuid rotating 4 He and discussed theoretically by Onsager and Feynman. ρ dρ f 2 dρ
(1. the order parameter depends only on r of the cylindrical coordinates {r.e. i. ﬁgure 1. z} with z parallel to the magnetic ﬁeld. the normal region in the bulk type II superconductor appears in the form of a single vortex line with the normal core of the radius about ξ surrounded by a supercurrent. the borderline between type I and II superconductors is found at κ= 1 .

length and magnetic ﬁeld. where p is a positive integer.
f2 dh = −p dρ κρ
(1. + 2 ρ dρ dρ (1. is (1. The total ﬂux carried by the vortex is
B
= 23/2 πλ2 Hc L
∞ 0
dρ ρh(ρ). Three of them are found at ρ = ∞. It behaves as the logarithm for small ρ ( 1): K 0 (ρ) ≈ ln 2 ργ (1.78. h = dh/dρ = 0 for ρ = ∞. the second GL equation is reduced to its London form in the cylindrical coordinates: 1 dh d2 h − h = 0.
(1.
0
ρ dh f 2 dρ
(1. we obtain
B
= −κ Hence. and as the exponent for large ρ ( 1): K 0 (ρ) ≈ π 2ρ
1/2
exp(−ρ).78) h = constant × K 0 (ρ).73).77)
The solution which satisﬁes the boundary conditions.Ginzburg–Landau theory
21
These equations are written in a form which introduces the dimensionless −1/2 quantities f (ρ) = n s φ(r ).74)
Applying equation (1. The fourth boundary condition is derived using the ﬂux quantization.
(1. respectively. = 0 and the paramagnetic term of the current in equation (1.75)
ρ=0
which should be equal to p
0. We choose the gauge where the order parameter is real. The order parameter should be about one in this region. where ρ ² 1/κ. There are four boundary conditions in the single-vortex problem. The second equation is readily derived replacing the current in equation (1.79)
with γ = 1. Then. ρ = r λL and h(ρ) for the order parameter.40) is zero. so that h = dh/ρ = 0 and the dimensionless order parameter is unity ( f = 1). Here the superconductor is not perturbed by the magnetic ﬁeld.76)
for ρ = 0.80)
. Here K 0 (ρ) is the Hankel function of imaginary argument of zero order. Let us ﬁrst consider the region outside the vortex core.40) by j = ∇ × B/4π and taking the curl of both parts of the equation.

76) is applied for the whole interior (0 ≤ ρ º 1).81)
The ﬂux quantization boundary condition allows us to ‘integrate out’ the magnetic ﬁeld in the master equation for the interior of the vortex with the following result (ρ º 1): 1 d df p2 ρ − 2 f − κ 2 f 3 = 0. the boundary condition (1. f ≈1− p2 .78) should be the same. the two regions overlap. The constant c p has to be found by numerical integration of equation (1.4 where c1 1. Vortex core in the BCS superconductor. ρ º 1. The order parameter is signiﬁcantly reduced inside the core (ρ 1/κ) but it becomes almost one in the region ρ 1/κ. which is applied for 1/κ ≤ ρ < ∞.
(1. Therefore.4. Hence. Because the magnetic ﬁeld and its derivative change over the characteristic length ρ 1. dh/dρ calculated using equation (1. Hence.83)
The vortex free energy Þ p is deﬁned as the difference in the free energies of a bulk superconductor with and without a single vortex. If κ 1. However.
The magnetic ﬁeld and current decrease exponentially in the exterior of the vortex. which is the case if the constant = p/κ. we can use the ﬂux quantization condition to determine the constant in equation (1. The GL equations are reduced to the single London equation outside the vortex core where the order parameter is
. κ 2ρ2 (1. the magnetic ﬁeld around the vortex core (r ² ξ in ordinary units) is B(r ) =
0p 2πλ2 L
K 0 (r/λL ). not only at zero. (1.82) ρ dρ dρ ρ This equation is satisﬁed by a regular solution of the form f = c p ρ p for ρ → 0.166.78). the magnetic ﬁeld is almost constant in the interior of the vortex.22
Phenomenology
Figure 1.76) with f = 1 and equation (1. The numerical result for p = 1 is shown in ﬁgure 1.82). where ρ > 1.

Then Þ p is given by the London expression (1.27). f ≈ 1. while the contribution of the core is negligible.4 Upper critical ﬁeld Bulk type I superconductors remain in the Meissner state with no ﬁeld inside the sample if the external ﬁeld is below Hc. is deﬁned by the condition G p = 0. Hc1 = 4π Þ p . 1. where the vortex appears in bulk type II superconductors. When κ 1.
(1.86)
Here the integral is proportional to the ﬂux carried by the vortex. However. 2 κ The ﬁrst (lower) critical ﬁeld appears to be much smaller than the thermodynamic critical ﬁeld in type II superconductors with a large value of κ 1. ln κ (1. If κ is not very large.84)
Here L is the vortex length along the ﬁeld lines. The ﬁrst (lower) critical ﬁeld Hc1. so that G p = Þp − pL 4π
0 H.88)
We see that the lowest ﬁeld corresponds to a vortex carrying one ﬂux quantum ( p = 1).6.89) Hc1 ≈ Hc 1/2 . Integrating the second term under the integral by parts. (1. if the external ﬁeld is weak. if the ﬁeld is strong enough so that G p < 0. It comprises the kinetic energy due to the current and the magnetic energy:
Þp ≈
1 dr (B 2 + λ2 |∇ × B|2 ) L 8π 2 Lλ2 Hc ∞ L dρ ρ[h 2 + (dh/dρ)2].87)
G p > 0 and the vortex state is unfavourable. the vortex state becomes thermodynamically stable. this region yields the main contribution to the vortex energy. we obtain
Þp = −
2 λ2 L Hc dh L h(ρ)ρ 2 dρ
ρ=κ −1
=
2 Lλ2 Hc p2 L ln κ.85)
The single-vortex Gibbs energy is obtained as G p = Þp − 1 4π dr B · H. 2 κ2
(1. They suddenly become normal metals
. = 2 κ −1
(1.Ginzburg–Landau theory
23
almost a constant. p 0 (1. the vortex penetration and the determination of Hc1 become more complicated.

k2 1 m ∗∗ ω2 χ − (x − x k y )2 χ = α + z∗∗ ∗∗ 2m 2 2m χ.93)
and the eigenvalues are (1. n) are the quantum numbers. At a certain ﬁeld. (1. are conserved. so that A(r) = {0. The well-behaved normalized eigenstates are found to be χn (x) = m ∗∗ ω π
1/4
Hn [(x − x k y )(m ∗∗ ω)1/2 ] −m ∗∗ (x−xk )2 ω/2 y e √ 2n n! E n = ω(n + 1/2). The vortex state for Hc1 < H < Hc2 is still superconducting with permanent currents in the sample.24
Phenomenology
when the external ﬁeld is above Hc . k z .90)
With the same accuracy the magnetic ﬁeld inside the sample does not differ from the external ﬁeld. One can measure Hc2 by continuously decreasing the ﬁeld from a high value. The eigenfunctions are found to be φν (r) = ei(k y y+kz z) χn (x) where χ(x) obeys the one-dimensional harmonic oscillator equation. A bulk type II superconductor also exhibits complete ﬂux expulsion in the external ﬁeld H < Hc1 < Hc. Only for a larger magnetic ﬁeld.
(1.92) (1. H = Hc2 . The upper (second) critical ﬁeld Hc2 is one of the fundamental characteristics of type II superconductors. ω = e∗ Hc2/m ∗∗ is the frequency of the oscillator and ν ≡ (k y . where the nucleation occurs. when H > Hc1. However. H B = H .94)
2
Here Hn ξ = (−1)n eξ
2
dn e−ξ dξ n
(1. Hc2 does the sample become entirely normal with no expulsion of the ﬂux. the GL equation for the order parameter can be linearized by neglecting the cubic term: 1 [∇ + ie∗ A(r)]2 φ(r) = αφ(r). Then approaching from the normal phase. k y . k z . superconductivity is just beginning to appear and the density of supercarriers n s = |φ(r)|2 is small. Then the master equation becomes formally identical to the Schr¨ dinger equation for a particle of charge o 2e in a uniform magnetic ﬁeld. The Hamiltonian (1. x Hc2. so that the resistivity and magnetization start to deviate from their normal state values. superconducting regions begin to nucleate spontaneously. whose eigenvalues and eigenfunctions are well known [25].91)
Here x k y = −k y /(m ∗∗ ω) is the equilibrium position. 0}. 2m ∗∗ (1. vortices penetrate inside the sample but the ﬂux passing through the sample remains less than its normal state value.90) does not depend on coordinates y and z and the corresponding momentum components. In the regions.95)
.

3. We are interested in the highest value of H = Hc2. . y and z. Nonetheless. . The number d of allowed values of k y in this range is d= Lx Ly L y ∗∗ m ωL x = eHc2 2π π (1.99) e∗ We see that Hc2(T ) allows for a direct measurement of the superconducting coherence length ξ(T ). − m ∗∗ ω m ∗∗ ω Lx < ky < Lx 2 2 (1. respectively. where α ∝ T − Tc . .
are the Hermite polynomials. .96)
where L x. the upper critical ﬁeld is linear in temperature Hc2(T ) (Tc − T ) in GL theory.84Tc (in tesla. The limit can be exceeded due to the spin–orbit coupling [27] or triplet pairing but in any case Hc2(0) remains ﬁnite in the framework of BCS theory.y. 1. which is also known as the Pauli pair-breaking limit given by H p 1.97)
so that every energy level is d-fold degenerate. The maximum value of k y is determined by the requirement that the equilibrium position of the harmonic oscillator x k y is within the sample. if Tc is in kelvin).z is the length of the sample along x. it might exceed the thermodynamic ﬁeld Hc2 = −
. n = 0. because Hc2 = 0 /2πξ 2 (T ). Hc2(0) at zero temperature is normally below the Clogston–Chandrasekhar [26] limit. Phase diagram of bulk type I (a) and type II (b) superconductors.Ginzburg–Landau theory
25
Figure 1.98)
2m ∗∗ α . which is found from ω(n + 1/2) = − α + with k z = n = 0. (1. Near Tc .5. 2.
2 kz 2m ∗∗
(1.

At the border between type II and I superconductors.102)
where the integral is taken in the plane outside the vortex cores. This happens because vortices repel each other. in agreement with equation (1. h(ρ) = 1 [K 0 (|ρ − ρ 1 |) + K 0 (|ρ − ρ 2 |)].103)
. In particular. where κ = 2−1/2. The free energy is given by the London expression:
Þ≈
1 dr (B 2 + λ2 |∇ × B|2 ) L 8π 2 Lλ2 Hc L = dρ(h2 + |∇ × h|2 ) 4π
(1. y) is represented by a linear combination of degenerate solutions of the linearized GL equation with n = k z = 0. Hc1 = Hc2 = Hc. The magnetic ﬁeld h(ρ) is the superposition of the ﬁelds h 1 (ρ) and h 2 (ρ) of each vortex. the proﬁle of the order parameter can be established analytically [24].101)
We can expect the minimum of the free energy to correspond to a regular arrangement of the vortex lines. even type I superconductors could be in an intermediate state with normal and superconducting domains. y) is periodic in the x and y directions. κ (1. The H –T phase diagram of type I and II bulk superconductors is shown in ﬁgure 1.26
Phenomenology
by several orders of magnitude in superconductors with a large GL parameter.5. if the ﬁeld H < Hc . To see how the repulsion occurs. In the superconducting state close to Hc2(T ) line the order parameter φ(x. If H is only slightly less than Hc2. all critical ﬁelds should be the same. 1. while the superconducting nucleation in type II superconductors could take place at the surface in a ﬁeld Hc2 < H < Hc3 . the lines will take a regular arrangement so that φ(x. The ﬁnite size of the sample and its surface might affect the magnetic ﬁeld penetration and the superconducting nucleation.100). let us consider two p = 1 vortices with their cores at ρ 1 and ρ 2 in the plane perpendicular to B in a type II superconductor with κ 1. equation (1. To minimize their repulsive energy.100) Hc2 = 21/2κ Hc.5 Vortex lattice Vortex conﬁguration in the ﬁeld Hc1 < H < Hc2 can be found by numerical integration of nonlinear GL equations.6. as follows from the relation (1. y) =
ky
c(k y )eik y y exp −
(x − x k y )2 2ξ 2
.
(1.91). φ(x.

. Integrating. the pressure is higher where the velocity of an ideal liquid is lower. creating a repulsive force between them.107)
The magnetic ﬁeld in the core of the ﬁrst vortex. Hence. According to the Bernoulli law. The translation in the y direction by a does not change φ(x.108)
where n is a unit vector parallel to the ﬁeld.105)
where the contour integral is taken over the surface of the second vortex core. one obtains the positive interaction energy U12 =
2 Lλ2 Hc L K 0 (|ρ 1 − ρ 2 |). two vortices repel each other. the pressure is higher in the region between two vortices. (1. ﬁgure 1. Let us now assume that the order parameter in the vicinity of the Hc2 (T ) line is periodic with period a. κ2
(1.
Integrating equation (1.106)
The free energy of the vortex state with two vortex lines is larger than twice the single vortex energy. . Hence.102) by parts and taking the limit κ → ∞. 2ξ 2
(1. created by the second vortex is B12 = 21/2 Hc K 0 (|ρ 1 − ρ 2 |)/κ.101) with the integer l = 0. (1. ±2. . The supercurrents of the two vortices are added to each other in the outer regions but they are subtracted from each other in the region between two vortices. y) if k y = 2πl/a in the sum (1. one obtains in the leading order Þ = 2Þ1 + U12 .Ginzburg–Landau theory
27
1
1
Figure 1. Applying the Maxwell equation. Hence. for H slightly less than Hc2. f 12 = −∇U12 . ±1.104) The interaction energy U12 is given by U12 =
2 Lλ2 Hc L 2π
dl 2 · h1 × ∇ × h2
(1. y) =
l
cl e2πily/a exp −
[x − πl/(ea Hc2)]2 . Two parallel vortex lines repel each other.109)
. ∇ × B12 = 4π j 12 .6. . the repulsion force applied to the ﬁrst vortex is expressed through the current j 12 of the second vortex as f 12 = L
0 j 12
×n
(1.6. we have φ(x.

we can neglect the magnetic energy in the free energy density f s .113) jc = β The Landau criterion of superﬂuidity allows us to understand jc at a more microscopic level. which ﬂows with a constant velocity v. The GL phenomenology allows us to calculate the critical current density jc. the momentum and energy
. With a further lowering of the magnetic ﬁeld.6 Critical current If the current density is too high. The magnetic ﬁeld due to the current in the ﬁlm is proportional to the ﬁlm thickness.
(1. The vortex lattice in type II superconductors is directly observed using ferromagnetic powder sputtered on the sample surface and in neutron scattering experiments. The simplest case is the uniform current distribution across the sample.112)
with a maximum at vs = (2m ∗∗ |α|/3)1/2 . so that their energy k is a function of k. The maximum (i. This condition is found in a thin superconducting ﬁlm. y) is periodic with period a = (2π)1/2ξ . we obtain ns = − 1 β α+
2 vs 2m ∗∗
. This process creates at least one elementary excitation. In a homogeneous system. the elementary excitations of the liquid have well-deﬁned momenta k. the supercurrent density j = e∗ vs n s is a nonlinear function of vs . A triangular lattice is obtained using the condition cl+2 = cl and c1 = ic0 . Let us consider a superﬂuid. when the thickness of the ﬁlm is small compared with the magnetic ﬁeld penetration depth and the coherence length. (1. (1.6. y) also be periodic in x. j =− e∗ vs β α+
2 vs 2m ∗∗
(1. critical) current density is 4e(m ∗∗ )1/2 |α/3|3/2 . In the coordinate frame moving with the liquid. The latter is more stable in a perfect crystal. the superconducting state is destroyed.28
Phenomenology
In order that φ(x. A square lattice is obtained if cl+1 = cl .111)
Hence. Hence. Then φ(x. it is necessary to impose a periodicity on the coefﬁcients cl . if the ﬁlm is sufﬁciently thin. 2 n s vs βn 2 s + . the kinetic energy of the condensate may dissipate and the ﬂow would gradually stop. the density of vortices drops and the period of the vortex lattice becomes larger than the coherence length.e. Because of the friction.110) f s ≈ fn + αn s + ∗∗ 2 2m Minimizing f s as a function of the condensate density n s . 1.

7). That is why the critical current of real type II superconductors depends on the sample quality. In the ideal Bose gas k = k 2 /2m ∗∗ and vc = min[k/2m ∗∗ ] = 0. However. when Hc1 < H < Hc2. Bose–Einstein condensation alone is not sufﬁcient for superﬂuidity. Josephson [29] predicted
. vortices move across the current and the current inevitably ﬂows through their normal cores. According to Bogoliubov [4].
1.108) with j 12 = j. Hence. as follows from equation (1. v vc . Hence.117)
The liquid is superﬂuid. Flow with a higher velocity leads to pair breaking and a loss of superconductivity. the excitation spectrum has a gap near the Fermi surface > 0. In BCS theory (chapter 2). The critical current in the vortex state of bulk type II superconductors does not reach the pair-breaking limit.116)
The critical value of the velocity is obtained for M = ∞ as vc = min
k
k
.
(1. As a result. 2M
(1. different defects of the crystal lattice ‘pin’ vortices preventing their motion.7 Josephson tunnelling
Let us consider a bulk superconductor separated into two parts by a thin contact layer with different properties from those of the bulk. Its thickness is supposed to be small compared with the coherence length (ﬁgure 1. if the external current is not very large. vc = /kF .Josephson tunnelling
29
conservation in a ‘collision’ with an object (such as the retaining wall) takes the following form: −Mv = −Mv + k (1. a normal metal or any weak link with a reduced condensate density or a small cross section. one obtains
k
and
+k·v+
k2 = 0. Disordered samples can carry the critical current density of about 107 A cm−2 or even higher [28].115) = + k. there is a non-zero critical velocity in (k = kF ) and min k = the BCS superconductor. an ideal type II superconductor has zero critical current in the vortex state. the repulsion between bosons modiﬁes their excitation spectrum so that the repulsive Bose gas is a superﬂuid (chapter 4). The layer might be an insulator. 2 2 Here M is the mass of the object and v is the liquid velocity after the collision. if vc = 0 and the ﬂow is slow enough. Hence.114) Mv 2 Mv 2 (1. This vortex motion leads to the energy dissipation. Combining these two equations. The current j in the direction perpendicular to the vortex lines creates the Lorenz force applied to the vortex core.

˙ (1. that there is a supercurrent j = (e∗ /m ∗∗ )n s d /dx along x in the sample. there is a ﬁnite phase shift ϕ = 2 − 1 of the condensate wavefunction on the right.123)
where ϕ0 is the phase difference in the absence of voltage. t) = n s exp[i (x. Hence. x = ±0. the current oscillates with a frequency (in the ordinary units) 2eV ωJ = . Hence. which is a property of the weak link.124)
. if ϕ0 is nonzero. Following Feynman [30].(φ2 ) and left-hand (φ1 ) ‘banks’ of the contact. we obtain I = Ic sin(ϕ0 + e∗ V t) (1.122)
K (n s1 n s2 )1/2 . t)] (x is the direction perpendicular to the contact).121)
The current I through the contact is proportional to the number of carriers tunnelling per second. n s1 = n s2 = n s and ϕ = e∗ V. If we suppose. The carriers change their quantum state from φ1 (t) to φ2 (t) in the contact with a transition amplitude K . I ∝ n s1 .118) iφ1 (t) = K φ2 (t) and e∗ V ˙ iφ2 (t) = K φ1 (t) + e∗ V φ2 (t) (1.119)
where is a change in the electrostatic energy at the transition due to voltage V . If both banks of the Josephson contact are made from where Ic the same superconductor. To keep the same current density along the sample. We conclude that some current I < Ic can ﬂow through the insulating thin layer between two bulk superconductors with no applied voltage. (1. one can write ˙ (1.7).120)
(1. we can understand this effect in the framework of the London description of superconductivity in terms of the condensate wavefunction 1/2 φ(x.30
Phenomenology
that a supercurrent can ﬂow through the weak link without any voltage and it oscillates as a function of time if there is a voltage drop on the contact. When the voltage is applied. where n s is suppressed. so that ˙ I = Ic sin ϕ (1. Taking time derivatives and separating the real and imaginary parts of these equations. ˙ As a result. the condensate phase should depend on x (ﬁgure 1. one obtains n s1 = 2K (n s1 n s2 )1/2 sin ϕ ˙ n s2 = − 2K (n s1 n s2 )1/2 sin ϕ ˙ and ϕ = e∗ V + K [(n s1/n s2 )1/2 − (n s2 /n s1 )1/2] cos ϕ. the phase gradient should be larger in the weak link.

W Φ([.Josephson tunnelling
\ .

31

&

[

Figure 1.7. Josephson’s weak link between two superconductors.

Figure 1.8. Magnetic ﬁeld dependence of the Josephson current.

The Josephson current also oscillates as a function of the applied magnetic ﬁeld. Indeed, let us introduce a magnetic ﬁeld B into the dielectric layer parallel to the contact (ﬁgure 1.7). The magnetic ﬁeld penetrates into the superconductor. The size of the region along x with the magnetic ﬁeld and the supercurrent is about w = t + 2λH , where t is the thickness of the layer. Using the same arguments as in section 1.3 we obtain dl · A(r) =

C

δ e∗

(1.125)

where the contour C crosses the contact twice as shown in ﬁgure 1.7. Here we take into account that the contributions of the transport current ﬂowing along x are mutually cancelled. The contribution of the screening current is negligible if the length of the contour is much larger than w. The left-hand side of equation (1.125) is an elementary ﬂux d B = Bw dy across the area of the contour, while δ is the change of the phase shift along the contact area, so that d

B

=

ϕ(y + dy) − ϕ(y) . e∗

(1.126)

We see that, in the presence of the magnetic ﬁeld, the phase shift and Josephson current density j = jc sin ϕ(y) are not uniform within the contact area because

32

Phenomenology

**(from equation (1.126)) ϕ(y) = e∗ Bwy + ϕ0 . The total current through the contact is I = jc L
**

L/2 −L/2

dy sin(e∗ Bwy + ϕ0 ) = Ic (sin ϕ0 )

sin (π B / 0 ) π B/ 0

(1.127)

where Ic = jc S and S = L 2 is the contact cross section. The Josephson current oscillates as a function of the magnetic ﬂux revealing a characteristic quantum interference pattern (ﬁgure 1.8). The Josephson effect laid the groundwork for designing squids (Superconducting Quantum Interference Devices). Squids can measure magnetic ﬁelds and voltages as low as 10−11 G and 10−15 V, respectively.

Chapter 2 Weak coupling theory

The phenomenological Bose-gas picture rendered no quantitative account for the critical parameters of conventional low-temperature superconductors. Its failure stemmed from the very large diameter ξ of pairs in conventional superconductors which can be estimated using the uncertainty principle, ξ ≈ 1/δk. The uncertainty in the momentum δk is estimated using the uncertainty in the kinetic energy δ E vF δk, which should be of the order of Tc , the only characteristic energy of the superconducting state. Therefore, ξ ≈ vF /Tc turns out to be about 1 µm for simple superconducting metals where Tc 1 K and the Fermi velocity vF 108 cm s−1 . This means that pairs in conventional superconductors strongly overlap and the Ogg–Shafroth model of real space pairs cannot be applied. An ultimately convincing theory of conventional superconductors was formulated by Bardeen, Cooper and Schrieffer [2]. In this and the next chapters we describe the essentials of BCS theory using the Bogoliubov transformation [4] in the weak-coupling regime and Green’s functions for the intermediate coupling (see also the excellent books by Schrieffer [31], Abrikosov et al [32] and De Gennes [33]).

2.1 BCS Hamiltonian

The BCS theory of superconductivity [2] relies on Fr¨ hlich’s observation [9] o that conduction electrons could attract each other due to their interaction with vibrating ions of the crystal lattice. In a more complete analysis by Bardeen and Pines [34] in which Coulomb effects and collective plasma excitations were included, the interaction between electrons and the phonon ﬁeld was shown to dominate over the matrix element of the Coulomb interaction near the Fermi surface. The key point is the fact discovered by Cooper [35] that any attraction between degenerate electrons leads to their pairing no matter how weak the attraction is. One can understand the Cooper phenomenon by transforming the corresponding Schr¨ dinger equation for a pair into the momentum space. Because o of the Pauli principle, a pair of electrons can only ‘move’ along the Fermi surface 33

34

Weak coupling theory

in the momentum space. But it is well known that in two dimensions a bound state exists for any attraction, however weak. The Cooper solution of the twoparticle problem in the presence of the Fermi sea demonstrated the instability of an attractive Fermi liquid versus pair formation with a possible Bose–Einstein condensation in the momentum space. Because of the Pauli principle, two electrons with parallel spins tend to be at larger distances than two electrons with antiparallel spins. Hence, following BCS, one can expect electrons to be paired into singlets rather than into triplets with the zero centre-of-mass momentum of the pairs. For these reasons, BCS introduced a model (truncated) Hamiltonian in which the interaction term contains only the attraction between electrons with the opposite spins and momenta, H=

k,s=↑,↓

ξk c† cks + ks

k

† c† c−k↓ ˆ (k) k↑

(2.1)

where

ˆ (k) =

k

V (k, k )c−k ↓ ck ↑ .

(2.2)

The ﬁrst term describes independent Bloch electrons, where cks is the electron annihilation operator with the momentum k and spin s, ξk = E k − µ is the band energy dispersion referred to the chemical potential µ. It is convenient to consider an open system with a ﬁxed chemical potential rather than with a ﬁxed total number of particles Ne Ó 1 to avoid some artiﬁcial difference between odd and even Ne . H − µNe has a minimum in the ground state for the open system. That is why the electron energy in equation (2.1) is referred to as µ. We recall (appendix C) that annihilation and creation operators anticommute for fermions: {cks , ck s } ≡ cks ck s + ck s cks = 0 and {cks , c† s } ≡ cks c† s + c† s ck s = δkk δss . k k k (2.3) (2.4)

They lower or raise the number of fermions n ks in a single-particle state |k, s by one and multiply a many-particle wavefunction of non-interacting fermions by ±n ks or ±(1 − n ks ) as cks |n k1 s1 , n k2 s2 , . . . , n ks , . . . = ± n ks |n k1 s1 , n k2 s2 , . . . , n ks − 1, . . . c† |n k1 s1 , n k2 s2 , . . . , n ks , . . . ks (2.5)

= ± (1 − n ks )|n k1 s1 , n k2 s2 , . . . , n ks + 1, . . . .

Here + or − depends on the evenness or oddness of the number of occupied states, respectively, which precede the state |k, s in an adopted ordering of singleparticle states. Modelling the attractive potential V (k, k ), BCS took into account the retarded character of the attraction mediated by lattice vibrations. The ions of the lattice must have sufﬁcient time to react, otherwise there would be no over scr eeni ng of the conventional Coulomb repulsion between electrons. In

Ground state and excitations

35

other words, the characteristic time τs for the scattering of one electron with the energy ξk by another one should be longer than the characteristic time of lattice relaxation which is about the inverse (Debye) frequency of vibrations −1 ωD . Using the energy–time uncertainty principle, we estimate the scattering time as τs ≈ |ξk − ξk |−1 . Hence, V (k, k ) could be negative (i.e. attractive), if |ξk − ξk | < ωD . To describe the essential physics of superconductors, BCS introduced a simple approximation for the attractive interaction: V (k, k ) = −2E p if the condition |ξk |, |ξk | < ωD (2.7) is satisﬁed, and zero if otherwise. Here E p is a positive energy depending on the strength of the electron–phonon coupling. Multiplying it by the density of electron states per unit cell at the Fermi level N(E F ) (appendix B), we obtain a dimensionless constant λ = 2E p N(E F ) (2.8) which conveniently characterizes the strength of the coupling. The BCS theory was originally developed for weakly coupled electrons and phonons with λ 1; extended towards the intermediate coupling (λ º 1) by Eliashberg [36]; and to the strong coupling (λ ² 1) by us [10, 11]. (2.6)

**2.2 Ground state and excitations
**

The operators ˆ , ˆ † in the BCS Hamiltonian (2.1) annihilate and create pairs with total momentum K = 0. One can expect that the pairs condense in the ground state with K = 0 like bosons below some critical temperature Tc (appendix B). The number of condensed pairs Ns should be macroscopically large, if T < Tc . The matrix elements of ˆ and ˆ † should be large as well, like the matrix elements of the annihilation and creation boson operators, † Ns − 1|b0 |Ns = Ns + 1|b0 |Ns = (Ns )1/2 Ó 1 (appendix C). That is all commutators of ˆ , ˆ † should be much smaller than the operators themselves † because [b0b0 ] = 1 Ò (Ns )1/2. Hence, we can neglect the fact that ˆ does not commute with ˆ † and c† . It is the same as replacing these operators by their ks expectation values in the open system, ˆ (k) ≈ where

k k

(2.9)

= −2E p (ωD − |ξk |)

k

(ωD − |ξk |) c−k ↓ ck ↑

(2.10)

36

Weak coupling theory

and (x) = 1 for positive x and zero otherwise. The wavefunction of the open system is a superposition of different eigenfunctions of the total number operator and the average equation (2.10) is not zero but macroscopically large in the superconducting state. Here the double brackets mean both quantum and statistical averages (appendix B). The substitution of equation (2.9) transforms the BCS Hamiltonian into a quadratic form with respect to the electron operators, ˜ H=

k † [ξk (c† ck,↑ + c−k↓ c−k↓ ) + k↑ † † k c k↑ c−k↓

+

∗ k c−k↓ c k↑ ] +

| |2 (2.11) 2E p

where the last term is added to make sure that the ground-state energies of the exact BCS Hamiltonian and of the approximation (2.11) are the same. Here = −2E p

k

(ωD − |ξk |) c−k ↓ ck ↑

(2.12)

does not depend on k. Now, following Bogoliubov [4], we can replace the electron operators by new fermion operators:

† ck↑ = u k αk + vk βk

(2.13) (2.14)

c−k↓ = u k βk −

† vk αk .

This transformation reduces the whole problem of correlated electrons to an ideal Fermi gas comprising two types of new non-interacting fermions (i.e. quasiparticles) α and β, if we choose u2 = k ξk 1 1+ 2 k 1 ξk 2 vk = 1− 2 k u k vk = − where

k k

(2.15) (2.16)

and 2 (2.17)

2. k k|

=

2 ξk + |

(2.18)

**Then new annihilation and creation operators anticommute like fermion operators,
**

† † {αk αk } = {βk βk } = δkk † {αk αk } = {βk βk } = {αk βk } = {αk βk } = 0.

**and the transformed Hamiltonian becomes diagonal, ˜ H = E0 +
**

k † k (αk αk † + βk βk )

(2.19)

at T = 0. we obtain E c = 2N(E F )
0 ωD
dξ
ξ−
ξ2 + ξ2 +
2 (0)/2 2 (0)
(2.24)
Using the deﬁnition of E 0 (equation (2. the Debye frequency is small (ωD states (DOS) is practically constant in the narrow energy region ±ωD around the Fermi energy. E 0 in equation (2. 2E p
(2.Ground state and excitations
where E0 = 2
k 2 (ξk vk + k u k vk ) +
37
| |2 . below Tc there are two solutions: = 0 and a real = 0.21) by the integral using the deﬁnition of the density of states N(E) in the Bloch band (appendix A). we obtain √ 2 2 ωD tanh ξ +| | 2T =λ dξ. As a result.
(2. They do not interact.21)
where fk =
† αk αk
=
† βk βk
is the quasi-particle distribution function.23) 0 ξ 2 + | |2 There is a trivial solution of this equation: = 0.12) replacing the electron operators by quasi-particle ones. their chemical potential is zero in the thermal equilibrium. (2. This energy is the ground-state energy referred to the normal state energy.20)). The system prefers to be in the superconducting (condensed) state below Tc because the condensation energy E c is negative. Above some Tc . In µ) and the density of conventional metals. f k = 0. Hence.19) is the ground-state energy of the BCS superconductor. N(E) N(E F ). Unlike the case of bare electrons the total average number of quasi-particles is not ﬁxed. we have Ec = E0 − 2
ξ k <0
ξk . However. At T = 0. (2.25)
.20)
The order parameter is determined from the self-consistent equation (2. One can replace the sum in equation (2. = Ep
k k
(1 − 2 f k )
(2. Therefore.22) fk = exp( k /T ) + 1 We see that there are no quasi-particles in the ground state. this is the only solution corresponding to the normal state. so that (appendix B) 1 .

(2. ˆ † (k)] = 4E 2 (ωD − |ξk |) p
k
(ωD − |ξk |)(1 − 2c† ↑ ck ↑ ). The criterion (chapter 1) of superﬂuidity and the critical velocity vc distribution of ‘bare’ electrons is of the form (T = 0)
2 n k = c† ck↑ = vk k↑
(2. Using
† † † † † † [c1 c2 .9)) replacing the pair operators ˆ (k) by the anomalous averages k is perfectly self-consistent.1.1). c3 c4 ] = δ23 c1 c4 − δ13 c2 c4 + δ24 c3 c1 − δ14 c3 c2
(2. are electrons with spins ‘up’ and ‘down’.29)
we obtain [ ˆ (k). this integral is negative: E c ≈ − 1 N(E F ) 2
2
(0) < 0. Indeed.30) k
.38
Weak coupling theory
(
εN
µ (N
∆ N) N
Figure 2.27)
which has a zero step (Z = 0) which differs from the Fermi distribution at T = 0 with Z = 1 (ﬁgure 2. they are a mixture of both and their energy dispersion k is remarkably different from that of the non-interacting electrons and holes (see ﬁgure 2. respectively. This is a clear manifestation of a breakdown of the Fermi-liquid description of attractive fermions at low temperatures.
(2.26)
Far away from the Fermi surface. In the vicinity of the Fermi surface. The quasi-particle energy spectrum satisﬁes the Landau /kF . ˆ † (k)] (2. α and β. The ‘meanﬁeld’ approximation (equation (2. Indeed. the quasi-particles. if ξk > 0 and holes if ξk < 0.
where
(0) is the order parameter at T = 0.2). Excitation spectrum of the BCS superconductor. let us calculate the commutator [ ˆ (k).28) to show that its expectation value is zero.

34)
A=
k
vk
−1
.33)
In terms of the electron operators. The ground state of the BCS Hamiltonian 0 is a vacuum with respect to quasi-particles. so that applying their annihilation operator to 0 . Distribution of electrons at T = 0.2. we have
† αk = u k ck↑ − vk c−k↓
βk = vk c† + u k c−k↓ k↑ and
0
=A
k
2 † (u k vk − vk c−k↓ c† )|0 k↑
(2.31) because the function under the integral is odd with respect to ξk . every single particle state k in 0 is free from quasi-particles. (2.
(2. The eigenfunction should be normalized. we obtain [ ˆ (k).32) 0 = A
k
Indeed. Quasi-particles are fermions and their vacuum state is obtained from the electron vacuum |0 by applying the quasi-particle annihilation operators for every momentum k: αk βk |0 .
Replacing the electron operators by the quasi-particle ones and taking the average. we obtain zero.35)
.
(2. which is the case if A = 0|
k † † βk αk αk βk |0 −1/2
.Ground state and excitations
39
QN
QRUPDOVWDWH VXSHUFRQGXFWLQJ VWDWH
N)
N
Figure 2. ˆ † (k)] = 4E 2 (ωD − |ξk |) p
k 2 (ωD − |ξk |)(u 2 − vk )(1 − 2 f k ) = 0 k
(2.

The perturbed many-particle state ˜ in the ﬁrst order in A is given by ˜ = |n α . n β
n α .s
(2. respectively. which follows as the symmetrized form of the velocity operator in the second quantization ˆ j j(r) = ˆ p (r) + ˆ d (r) j where ˆ p (r) = − e j 2m c† cks (2k + q)e−iq·r k+qs
k. n β +
n α .q. We apply the perturbation theory for a linear interaction of electrons with the vector potential A(r) taking the BCS Hamiltonian as a zeroorder one.s
where Aq is the Fourier component of A(r).q.3 Meissner–Ochsenfeld effect
BCS reduced the many-body problem to a non-interacting gas of quasi-particles allowing for an analytical description of thermodynamic and kinetic properties of conventional superconductors.n β
|n α . Let us ﬁrst discuss the BCS theory of the magnetic ﬂux expulsion.40
Weak coupling theory
Thus the ground-state wavefunction is
0
=
k
† (u k − vk c−k↓ c† )|0 . k↑
(2. n β E nα nβ − E nα nβ
(2.39)
is the paramagnetic part and ˆ d (r) = − e j m
2 k. The ﬁeld distribution in the sample is determined by the average of the current density operator.36)
ˆ It is a superposition of the eigenfunctions of the total number operator N ≡ † ks c ks c ks . where ∇ · A(r) = 0. Applying the Bogoliubov transformation
.s
c† cks A(r)e−iq·r k+qs
(2. n β |Hint|n α .37) Hint = − k+qs m
k.40)
the diamagnetic part. e (k · Aq )c† cks (2.41)
where |n α .
2.38) (2. This form of the interaction follows from the velocity operator [−i∇ − e A(r)]/m in the effective mass (m) approximation for the band energy dispersion and the gauge.q. n β and E nα nβ are the eigenstates and eigenvalues of the zero˜ order Hamiltonian H .

In this limit u k+q vk − u k vk+q = 0.21).5 as follows from the master equation (2.1) for a homogeneous system. The exponent in equation (2.4 BCS gap. while the denominator remains ﬁnite: k+q + k > 2 (0).BCS gap.43)
k. This is a remarkable result which demonstrates the instability of the Fermi liquid for any value of the attraction λ in agreement with the Cooper two-particle solution.44)
In the opposite limit (λH < ξ ). which is large compared with the coherence length ξ . At T = 0.42). the paramagnetic contribution vanishes and we obtain the London equation (chapter 1). Actually. the Pippard non-local theory of the ﬂux expulsion follows from equations (2. the limit to which the theory is applied.
.13) and (2.42) and (2.q
× exp(−iq · r)(u k+q vk − u k vk+q ) jd = − ne2 A(r).21)) which is also a gap in the quasi-particle spectrum (ﬁgure 2.
(2.45)
for λ 1. Therefore. critical temperature and single-electron tunnelling
The BCS theory introduces the order parameter (equation (2. j (r) = − ne2 A(r).42) turns out to be zero at the Fermi level and the paramagnetic current appears to be ﬁnite. In this case one can take the limit q → 0 in equation (2. one can show that it cancels the diamagnetic part so that the normal state current is zero in a permanent magnetic ﬁeld. BCS theory predicts a universal ratio 2 (0)/Tc 3. the non-trivial solution is determined from 1 = λ Integration yields (0) 2ωD exp − 1 λ (2. Thus the superconducting ground state cannot be derived by using the perturbation theory with respect to the pairing potential up to any order. m (2.46)
ωD 0
dξ ξ2 +
2 (0)
. critical temperature and single-electron tunnelling
41
(equations (2. In fact.14)) for Hint yields the average current density at T = 0: j p (r) = 2e2 m2 u k+q vk (2k + q)(k · Aq ) k+q + k (2.46) cannot be expanded in a series of λ.43). The value of the gap at T = 0 should be of the order of Tc . m
Let us assume that the magnetic ﬁeld varies over the characteristic length λH .
2. In the normal state the denominator in equation (2.42) (2.

(2. ξ
(2. where (Tc − T )/Tc small compared with the temperature.14.
(2. T
(2.5. a direct expansion in powers of cannot be applied in equation (2.23).42
Weak coupling theory
At T = Tc .
The numerical coefﬁcient in
(2. because every term of such an expansion would render a divergent integral.54)
.50)
Here we have replaced the upper limit in the integral for inﬁnity which is justiﬁed because the quasi-particle distribution function falls down exponentially at low temperatures. T Tc we rewrite the master equation using its zero-temperature form (equation (2.48)
where C 0. the gap should be zero so that Tc is determined by 1 = λ
ωD 0
dξ tanh(ξ/2Tc ) . so that we can replace (T ) by (0) on the right-hand side and expand ln in powers of 1 (T ) = (T ) − (0) on the left-hand side with the following result:
1 (T )
= − 2π T (0) exp −
(0) .
(2.48) is ≈1. In the vicinity of Tc . so that 2 (0) Tc 3. equation (2.53)
The temperature correction to the gap appears to be exponentially small at low 1. the whole gap is temperatures. However. Instead we use tanh x 1 = 2 + [π(n + 1/2)]2 x x n=−∞
∞
(2.49)
To calculate the temperature dependence of (T ) at low temperatures.52)
The remaining integral is exponentially small at low temperatures. ξ 2 + 2 (T ) f (ξ ) ≈ exp − .577 is the Euler constant.47)
Integrating by parts and replacing the upper limit by inﬁnity.45)): ln (T ) = −2 (0)
∞ 0
dξ ξ2 +
2 (T )
f (ξ ).51) T Replacing the integration over ξ by the integration over the energy ξ 2 + 2 (T ) yields ln (T ) = −2 (0)
∞ (T )
=
d
e−
2
/T 2 (T )
−
. we have Tc = 1 2eC ωD exp − π λ (2.

becomes a quasi-particle in the superconductor.58) I (V )
k(ξ k <0). critical temperature and single-electron tunnelling
so that 1 = 2T λ
ωD n 0
43
ξ2
+
2 ωn
dξ +
2 (T )
(2. .
(2. we obtain σ = σN eV (eV )2 −
2 (T )
(2. we obtain the transition rate as 2 Tkk δ( k − ξk − eV ). (2.57)
There is a discontinuity in the temperature derivative of (T ) at Tc which leads to a jump of the speciﬁc heat (section 2.60)
Calculating the remaining integral. Tc Tc
(2. The electron. Replacing the sum by the integral we obtain I (V )
eV −∞ +∞
dξ
−∞
dξ δ
ξ
2
+
2 (T ) − ξ
. tunnelling from the normal metal. ±1.55)
where ωn = π T (2n + 1) are the so-called Matsubara frequencies.6).
(2. .61)
. . equation (B. The δ-function in equation (2.59)
Then the conductance σ = dI /dV is found: σ
+∞ −∞
dξ δ
ξ
2
+
2 (T ) − eV
. The last equation can be expanded in powers of (T ) as ln Tc =2 T
2
(T )Tc
n 0
∞
dξ .k
The matrix element Tkk is almost independent of the momentum k in the normal metal and of k in the superconductor. 2 (ξ 2 + ωn )2
(2. if the voltage is not very high eV ∼ µ. Applying the Fermi–Dirac golden rule. n = 0.06Tc 1 − . ±2. The gap can be measured directly in tunnelling experiments where one applies voltage V to a thin dielectric layer between the normal metal and the superconductor (ﬁgure 2.27)) we obtain (T ) = π Tc 8 7ζ (3)
1/2
1−
T T ≈ 3.BCS gap.58) takes into account the difference eV in the normal and superconducting chemical potentials. The current running through the dielectric is proportional to the number of electrons tunnelling under the barrier per second. .56)
Calculating the integral over ξ and the sum over n (appendix B.3) [37].

the ﬁnite value of α measured experimentally proves that phonons are
.5. There is no current if |eV | < because there are no states inside the gap in the superconductor (ﬁgure 2.3.63) α=− d ln M In fact. (2. ∂|ξ | = √ . The characteristic phonon frequency ωD is proportional to 1/ M as a frequency of any harmonic oscillator.48)) which depend on the mechanism of√ interaction.
where σN is the normal state conductance of the barrier above Tc . However.3).44
Weak coupling theory
(PSW\OHYHOV
2∆ µ . (2. Just above the threshold. the isotope exponent α could be lower than 0. But. 1 6
Figure 2. the isotope exponent is found as d ln Tc = 0.61) rather well in conventional superconductors.5 in a BCS superconductor because of the Coulomb repulsion and the anharmonicity of phonons. eV = . the coupling constant λ is independent of the ion mass (section 3. in any case.
2. Shaded areas correspond to occupied states.62) ρ( ) ≡ 2− 2 ∂ The typical experimental ratio σ/σN as a function of the voltage and the temperature follows the BCS prediction (2. Hence. when an ion mass M is varied without any change of the electronic conﬁguration of the ion. Tunnelling from the normal metal (N) to the superconductor (S) through a dielectric barrier.5 Isotope effect
The origin of the electron–electron attraction in superconductors can be tested by isotope substitution. the conductance has a maximum because the quasi-particle density of states ρ( ) diverges at = .3). There are two parameters in the BCS the expression for Tc (equation (2.

their speciﬁc heat Ce in the normal state is linear as the function of temperature (appendix B).66)
is the quasi-particle grand partition function. we obtain the entropy of the ideal Fermi gas of quasiparticles as S = −2
k
[ f k ln f k + (1 − f k ) ln(1 − f k )]
(2. The isotope effect has been observed in many conventional superconductors and in high-temperature superconductors (Part II).68)
. The temperature dependence of Ce changes drastically in the superconducting state due to the gap.67)
where the distribution function f k = n kα = n kβ is deﬁned in equation (2. The quasi-particle energies depend on the temperature in the self-consistent BCS approximation which should be taken into account in the calculations of temperature derivatives of the thermodynamic potential.64). The number of these electrons is proportional to temperature.65)
where n kα. where the quasi-particle entropy is deﬁned as S = − ln P(U Q . 1 are the quasi-particle occupation numbers. Therefore.Heat capacity
45
involved in the pairing mechanism. (2.
2. We apply the deﬁnition of the speciﬁc heat as Ce = T dS/dT . Here we take into account the fact that the quasi-particle chemical potential is zero.
(2. Calculating the statistical average in equation (2.β = 0.22).64)
Here P(U Q . and Z=
k
[1 + eβ k ]2
(2.6 Heat capacity
Only the electrons near the Fermi surface can absorb heat in a metal because of the Pauli principle. The temperature derivative of the distribution function includes the derivative of k as d fk 1d k k = f k (1 − f k ) − dT T dT T2 so that Ce = 2
k 2 k T2
f k (1 − f k )
−
d k T dT
k
. N) . N) = Z −1 e−βU Q is the statistical probability of ﬁnding the Fermi gas of N quasi-particles with energy U Q (appendix B). UQ =
k k (n kα
+ n kβ )
(2.

74)
where
Mkq = u k u k−q − vk vk−q
(−)
(2.s
where q = ν/s and ν are the wavevector and frequency of the sound.75)
. respectively (s is the sound velocity).
2.70)
and the speciﬁc heat has a discontinuity. k = |ξ | and we obtain Ce = CN . However.c.73).69)
as expected for the ideal Fermi gas (appendix B).43 (2. The sound frequency is low (ν Tc ) and only the scattering terms are relevant: Hint
k (−) † † Mkq (αk αk−q + βk βk−q )
(2. the number of quasi-particles is exponentially small f k e− (0)/T and so is the speciﬁc heat (equation (2. Above Tc . V (q) is proportional to the sound amplitude and H. The relative value of the jump is 1. Ce = CN + if T = Tc − 0.68)). (2. Applying the Bogoliubov transformation.46
Weak coupling theory
At low temperatures. where CN = N(E F ) T2
∞
dξ
0
ξ2 cosh2 (ξ/2T )
=
2π 2 N(E F )T 3
(2.71)
1.202.7 Sound attenuation
The interaction of ultrasound waves with electrons is described by the following Hamiltonian: c† ck−qs + H. Here ζ (3) 8π 2 N(E F )Tc 7ζ(3) (2.72)
12 Ce (Tc − 0) − Ce (Tc + 0) = Ce (Tc + 0) 7ζ(3)
which agrees with the value measured in many conventional superconductors.68) appears to be ﬁnite: d k d π2 = =− T dT T dT 2
k
8 7ζ(3)
(2. Two of them correspond to the annihilation and creation of two different quasi-particles and the other two correspond to their scattering.73) Hint = V (q)eiνt ks
k. The phase transition turns out to be second order as in the GL phenomenology (chapter 1). is the Hermitian conjugate.c. we obtain four terms in the interaction (2. just below Tc . the second term in the brackets of equation (2.

Sound attenuation
1RUPDOL]HG UHVSRQVH
47
.

.

is a so-called coherence factor.
(2. which is the difference between these two rates.76)
and the emission rate is Wemi
k
[Mkq ]2 f k−q (1 − f k )δ(
(−)
k−q
−
k
− ν). .4. ∂
d
∂ξ ∂
2
[u 2 (ξ ) − v 2 (ξ )]2
(2. The rate of the sound absorption is given by the Fermi–Dirac golden rule as Wabs
k (−) [Mkq ]2 f k−q (1 − f k )δ( k−q
−
k
+ ν)
(2. (2. so that −
∞
Only transitions with ξ = ξ ∂f . is given by (−) [Mkq ]2 ( f k − f k−q )δ( k−q − k − ν).81)
. ∂ exp( /T ) + 1
(2.78)
k
If ν T.79)
= ξ 2+ where = ξ 2 + 2 and contribute to the integral over ξ .
77F
Figure 2. Temperature dependence of sound attenuation and thermal conductivity (1) compared with the nuclear spin relaxation rate (2).77)
The sound attenuation .80) as −
∞
d
− 2−
2
2 2
1 ∂f = . we have f k − f k−q ≈ ν∂ f /∂ .80)
Here we can see that the large density of quasi-particle states ∂|ξ |/∂ = √ / 2 − 2 is cancelled by the small coherence factor in the integral in (2. Replacing the integration over k and over the angle between k and q by the integration over ξ = ξk and ξ = ξk−q . respectively. we obtain − dξ dξ [u(ξ )u(ξ ) − v(ξ )v(ξ )]2
2.
∂f δ( − ) ∂
(2.

(+)
(2. the due to the inelastic electron–phonon scattering. Only electrons in a narrow energy region around the Fermi surface can exchange their spin with
.84)
However.87)
In fact. the divergency is cut by some damping of excitations τ −1 . Above Tc . Applying the Fermi–Dirac golden rule.c. ∂
(2.83)
The NMR frequency is very small and the spin-ﬂip scattering is practically elastic.8 Nuclear spin relaxation rate
The measurement of the linewidth of the nuclear magnetic resonance (NMR) is another powerful method of determining (T ).82)
one can measure the temperature dependence of the BCS gap.48
Weak coupling theory
The sound attenuation in the superconducting state depends exponentially on temperature (ﬁgure 2. we obtain 1/T1
k. That is why only the scattering of quasi-particles contribute to 1/T1 as in the case of sound attenuation.
2. The linewidth depends on the inverse time of the relaxation of nuclear magnetic moment due to the spin-ﬂip scattering of carriers off nuclei. where relaxation rate is proportional to T .
s n
=
2 exp( (T )/T ) + 1
(2.85)
is different. Hint
k † † Mkk (βk αk + αk βk ). This scattering is described by the hyperﬁne interaction of the nucleus with the electron spin: Hint
k. This linear temperature dependence of 1/T1 in a normal metal is known as the Korringa law. k
(2. for example = 0. Replacing the sums by the integrals yields a divergent integral. here the coherence factor
(+) Mkk = u k u k + vk vk
(2. Using its ratio to the normal-state attenuation. 1/T1 .86)
ν → 0. 1/T1 −T
∞
d
+ 2−
2
2 2
∂f .4).k
[Mkk ]2 ( f k − f k )δ(
(+)
k
−
k
− ν)
(2. It has the same origin as the linear speciﬁc heat.k
c† ↓ ck↑ + H. Both are due to the Pauli principle.

the NMR relaxation rate is exponentially small because of the gap: 1/T1 e−
/T
ln( τ )
(2. The second term of the coherence factor M on the left-hand side of the Boltzmann equation accounts for the driving force. it has a maximum (see ﬁgure 2. which acts on a quasi-particle due to the temperature dependence of its energy. The solution of equation (2.90)
where v = ∂ k /∂ k is the quasi-particle group velocity.4).91) is
n f˜ = τtr
∂f v · ∇T. The distribution function obeys the Boltzmann equation.
2. Their contribution Q to the heat ﬂow is determined by the deviation f˜ of the distribution function from the equilibrium. T |ξ | ∂
2
(2. τtr τtr In the superconducting state. the transport relaxation rate is diminished by a factor n |ξ |/ < 1 compared with that in the normal state (1/τtr ) because of the square (−) in the probability of scattering. The maximum in 1/T1 is one of the most interesting and important features of BCS theory. This result is distinctly different from the result for sound attenuation.91) v· ∂r ∂r ∂k τtr where the transport relaxation rate for elastic scattering is obtained using the Fermi–Dirac golden rule: 1 |ξ | 1 (2.9 Thermal conductivity
Important information about the excitation spectrum of the superconducting state can also be obtained from thermal conductivity [39]. The observation by Hebel and Slichter [38] of the peak in 1/T1 was one of the ﬁrst in the body of evidence for the detailed nature of the pairing correlations in BCS superconductors.Thermal conductivity
49
nuclei and absorb heat. They differ because the coherence factors are different: M (+.88)
as in sound attenuation. Q=
k
f˜v
k
(2. Quasi-particles contribute to heat transfer in a superconductor.92) s = n. Well below Tc . just below Tc . = ξ 2 + | (T (r)|2 .89)
A simple energy-gap form of a two-ﬂuid model could account for the drop in the sound attenuation but not for the rapid rise of 1/T1 just below Tc .93)
. (2.−) = (uu ± vv ). which for a small deviation from the equilibrium has the form: ∂ k ∂f f˜ ∂f − · =− s (2. However.

97)
according to the Pauli exclusion principle. Applying the Bogoliubov transformation. the states near the Fermi surface contribute to the sum in equation (2. ∞ 2 2 0 dx x / cosh (x)
(2.↓
ξk c† cks + ks
1 2
(
s
ss k
† c† c−ks + H. if the potential allows for such pairing [40].
(2. The simple approximation for the pairing potential (equation (2. which does not contain fermionic operators. k )
ss k
2
(1 − 2 f k ).94) Ks − d 2 .95)) drops exponentially with the temperature lowering below Tc just like sound attenuation (ﬁgure 2.10 Unconventional Cooper pairing
In BCS theory electrons are paired with the opposite momenta on the Fermi surface and the opposite spins.s=↑. k ) takes the following form:
ss k
=−
k
V (k. T ∂ Hence.98)
k
Let us assume that V (k.95)
The normalized thermal conductivity (equation (2. The order parameter should be antisymmetric:
ss k
=−
ss −k
(2. the mean-ﬁeld Hamiltonian is ˜ H=
k. the master equation for any pairing potential V (k. which is uniform along the Fermi surface. Then all q ≈ 21/2kF 1 − cos . the ratio to the normal-state thermal conductivity is Ks = Kn
∞ 2 2 /2T dx x /cosh (x) . It depends only on the distance between two correlated electrons. This is an s-wave order parameter because the pair wavefunction is isotropic in real space. In the general case.50
Weak coupling theory
Substituting it into equation (2. k ) depends on the value of the momentum transfer q = |k − k |. BCS theory describes an anisotropic Cooper pairing with non-zero orbital momentum of pairs l = 0 as well.4).c. In BCS theory.96)
where we omit the c-number term.
2.) ks
(2.90) yields the thermal conductivity K = | Q|/|∇T | as ∂f 1 ∞ (2. where functions in the master equation depend on the angles of k and k with the absolute
.97) and √ is the angle between k and k . which is the case for an isotropic system.6)) leads to the momentum-independent gap k = .

we obtain. Using the orthogonality of the Legendre polynomials and the summation theorem (equation (2.101)
This expansion is instrumental due to the summation theorem: Pl (cos ) = Pl (cos )Pl (cos
l
) ) cos[m(φ − φ )] (2. Substitution of the series into the master equation yields a system of coupled nonlinear equations for the amplitudes (l). (l) = − V (l)N(E F ) (l) 2l + 1
ωD 0
dξ tanh(ξ/2Tc ) .102)
+2
m=1
(l − m)! m P (cos )Plm (cos (l + m)! l
where Plm (x) are associated Legendre functions deﬁned as Plm (cos ) = sinm dm Pl (cos ) (d cos )m
and is the angle between any two directions determined by the spherical angles ( . Pl (x) = 1 dl 2 (x − 1)l 2l l! dx l (2.Unconventional Cooper pairing
51
values of the momenta equal to the Fermi momentum. φ) and ( .103)
Then the critical temperature Tc = 1 2eC ωD exp − π λl ∗ (2. ξ
(2. These equations are linearized and decoupled at T = Tc .100)
which are orthogonal and normalized as
1 −1
dx Pl (x)Pl (x) =
2δll . |k| = |k | = kF . φ ). We can expand the order parameter and the potential in the polynomial series:
∞ ss k
=
l=0 ∞
(l)Pl (cos ) V (l)Pl (cos
l=0
(2.102)).99)
V (q) =
)
where Pl (cos ) are the Legendre polynomials. respectively.104)
. 2l + 1
(2. for an isotropic system with parabolic energy dispersion.

. Hence.) are triplets (S = 1). fully taking into account the interaction of quasi-
. 3. We apply the effective mass (m) approximation for the band dispersion (see appendix A) and drop c-number terms in the total energy. The singleparticle gap.105)
Here and further on. The order parameter. . if the Fourier transform of the pairing potential has an attractive component with l = 0. 5. k . . According to the Pauli exclusion principle.) are singlets (total spin S is zero) and the odd l states ( p. . is proportional to the Fourier transform of the pair wavefunction. while the spin component of even orbital states should be antisymmetric. . .
2. Let us introduce ﬁeld operators (appendix C)
s (r)
=
k
cks exp(ik · r). the spin component of the paired state with an odd orbital momentum should be symmetric under the permutation of the spin coordinates of two particles.107)
is the one-electron Hamiltonian in the external magnetic ( A(r)) and electric (U (r)) ﬁelds. . d. The change in sign of k corresponds to the permutation of real-space coordinates of the two fermions of the pair. has nodes on the Fermi surface if l = 0.
(2. Then the BCS Hamiltonian can be written as follows: ˜ H= where [∇ + ie A(r)] ˆ + U (r) − µ h(r) = − 2m
2
dr
s
† ˆ s (r)h(r)
s (r)
+
(r)
† ↑ (r)
† ↓ (r)
+
∗
(r)
↓ (r)
↑ (r)
(2. It is odd for l = 1. . The solution of equation (2.
(2. . the system condenses into the state with non-zero orbital momentum of the pair wavefunction. −k = (−1)l k . Therefore. the even l states (s.11 Bogoliubov equations
The Bogoliubov transformation of the BCS Hamiltonian can be readily generalized for inhomogeneous superconductors. k Pl (cos ).106) (2. f.103) is trivial: (l) = 0 for all l except l = l ∗ with a maximum value of λl ∗ and negative V (l ∗ ).52
Weak coupling theory
where λl ∗ ≡ |V (l ∗ )|N(E F )/(2l ∗ + 1) for negative V (l ∗ ). The coordinate-dependent order parameter is given by (r) = −2E p
↓ (r) ↑ (r)
. the total wavefunction of the pair should change its sign after the permutation of both the orbital and spin coordinates of two fermions. and even for even l. we take the volume of the system as V = 1.108)
Superﬂuid properties of inhomogeneous superconductors can be studied by the use of the Bogoliubov equations. Hence.

). t) =
d ˆ u(r. we obtain two coupled Schr¨ dinger equations for the wavefunctions o u(r.
↑ (r)] ↑ (r)]
= − δ(r − r ) = − δ(r − r )
† ↓ (r) † ↑ (r). t)αn + vn (r.111). t)
∂t
˜ † = [ H .114)
ψ↓ (r. which annihilate and create quasiparticles in a quantum state n.Bogoliubov equations
53
particles with the condensate. t) = ei Ht
˜ s (r)e ˜ −i H t
. † s (r
)} =
k.112)
The linear Bogoliubov transformation of ψ-operators has the form ψ↑ (r. t) =
n ∗ † [u n (r. t) = h(r)u(r.115)
. ψ↓ (r. dt i
(2. t). ψ↑ (r. t) − (r)v(r.k
{cks . t) + (r)ψ↓ (r.110) and (2. t)
(2. Using this transformation in equations (2.
(2. t)αn ] n † where αn and αn are the fermion operators. t) = h ∗ (r)v(r.
(2.110)
and −i
† ∂ψ↓ (r. t) and v(r. t).113) (2. t):
(2. we introduce the timedependent Heisenberg operators (appendix D) as ψs (r. t)] ∂t † ˆ = h(r)ψ↑ (r. t)]
† ˆ = h ∗ (r)ψ↓ (r.
(2. t). t)βn ] ∗ † [u n (r. t) − ∗
(r)ψ↑ (r. To derive these equations. dt d ˆ −i v(r.111)
Here we have applied commutation relations for the ﬁeld operators: {
s (r). t) + ∗ (r)u(r. c† s } exp[ik · r − ik · r ) k exp[ik · (r − r )] = δss δ(r − r )
k
= δss and [ [
† ↑ (r † ↑ (r
) )
† ↓ (r ↑ (r
).109)
The equations of motion for these operators are readily derived: i ∂ψ↑ (r. t) ˜ = − [ H . t)βn − vn (r.

119) and (2.116)
which retains the Fermi commutation relations for all operators.123)
(2.122)
and.124) (2. We can readily solve the set of Bogoliubov equations (2. t)] = δ(r − r ) n n
(2.
k
(2.117). t)vn (r . these equations are reduced to the steady-state ones:
n u n (r)
=
−
[∇ + ie A(r)]2 + U (r) − µ u n (r) − 2m [∇ − ie A(r)]2 + U (r) − µ vn (r) − 2m
(r)vn (r)
∗
(2. from equation (2. As a result.120) into equations (2.
∗ [u n (r.121)
∗
= ξk u k −
vk uk
= − ξk vk −
(2.108) yield the self-consistent equation for the order parameter: (r) = −2E p
n † † where f n = αn αn = βn βn = (1 + exp n /T )−1 is the equilibrium quasiparticle distribution in the quantum states n with energy n .125) (2.120) Substituting (2.
Using the same transformation. t)u ∗ (r .117) in the homogeneous case. we ﬁnd |u k |2 = 1 ξk 1+ 2 k ξk 1 2 1− |vk | = 2 k 2 . (r) = . When the magnetic and electric ﬁelds are stationary.126)
∗ u k vk = −
.119) (2. equations (2. when A(r) = U (r) = 0. In this case the excitation wavefunctions are plane waves.116). t) + vn (r. we obtain
kuk k vk
(2. |u k |2 + |vk |2 = 1.54
Weak coupling theory
There is also the sum rule.118)
u k (r) = u k eik·r vk (r) = vk eik·r and the order parameter (r) is r-independent. ∗ u n (r)vn (r)(1 − 2 f n )
(2.117)
n vn (r)
= − −
(r)u n (r). the quantum and statistical averages in equations (2.

18)). This state is described by an oscillating complex order parameter (r) = exp(2iq · r) (2.133). the system remains translation invariant and the momentum k is a quantum number.
2 Taking into account the normalization condition u 2 + vk = 1. where the gap is found using the master equation. They are mathematically transparent but the analytical solution is possible only in a few simple cases.133)
If we apply the Landau criterion (1.12 Landau criterion and gapless superconductivity
The Bogoliubov equations are coupled nonlinear integra-differential equations.132)
u k vk = − and the excitation spectrum Ek = ξk+q − ξ k−q 2
+
(ξk+q + ξ k−q )2 4
+
2. If the ﬂow is uniform. and is the real amplitude.127)
as it should be (see equation (2. we consider a uniform ﬂow with the superﬂuid velocity v s . = 2E p
k
2
(1 − 2 f k ).128)
2. we obtain k 2 u 2 . we can keep only the terms of the
.
(2. As an example.Landau criterion and gapless superconductivity
The elementary excitation energy is
k
55
=
2 ξ k + | |2
(2.129)
where 2q = 2mv s is the centre-of-mass momentum of the Cooper pair.
k
(2. Hence. the critical velocity will be vc = /kF . The solution has the form u n (r) = u k exp[i(k + q) · r] vn (r) = vk exp[i(k − q) · r] where the coefﬁcients u k and vk are found from (E k − ξk+q )u k + vk = 0 (E k + ξk−q )vk + u k = 0.131)
ξk+q + ξ k−q 1 1± 2 2 k 2
k
(2.117) to the spectrum equation (2.130)
(2. Let us examine how the ﬂow destroys superconductivity. vk = k
(2.

56
Weak coupling theory
ﬁrst order in q as long as q spectrum becomes
º mvc
Ek ≈
kF in equation (2. However. Then the excitation
k
+ k · vs
(2. we obtain ln (0) = 2 π (0)
kF vs
d
arccos( /kF vs ) . (2. we arrive at ln = ln .136) Here (0) is the BCS gap (equation (2.134)
and the equation for the order parameter is = 2E p
k
2
[1 − 2 f (
k
k
− k · v s )]. we have f ( k −k · v s ) = θ (−
k
+ kvs cos ϕ)
where ϕ is the angle between v s and k. Let us see how it happens by solving the BCS equation (2.
(2. The Landau criterion tells us that superconductivity should disappear at vs > vc . Hence. where e = 2.138)
This equation has only the trivial solution = 0. there are no quasi-particles ( f k = 0) and = (0).140) (0) kF vs kF vs evc /2. which is found from 2 2 kF vs . Calculating the remaining integral.133)) is negative for some directions of k and quasi-particles appear even at zero temperature. integrating over the angle yields kF vs 1 − ( /kF vs ) (0) = d √ . if vs > vc . (2. In this case. The superﬂow in threedimensional superconductors is different [41]. Integrating over the angle.46)) in the absence of the ﬂow.718. when the superﬂuid velocity exceeds vc . √ 2− 2 kF vs
(2. If vs < vc . = 1− ln − ln 1 + 1 − (2. that is superconductivity disappears precisely at vs vc . The excitation spectrum It disappears only at vs (equation (2. For T = 0. where f ( k −k · v s ) = 1 and the gap should be different from (0). there is an interval of |ϕ| arccos[ k /(kF vs )]. the Landau criterion cannot be compromised in two-dimensional s-wave BCS superconductors.137)
in two dimensions.139) ln 2− 2 There is a non-trivial solution = 0 even if vs > vc .135)
The excitation energy (equation (2.135) in two and three dimensions.133)) has no gap but the order parameter is still non-zero if vc < vs <
.133). (2.

The hole
. a new phenomenon is observed.
2.10) and ‘dirty’ superconductors with magnetic impurities [42]. In the absence of the barrier. equation (2. Here the ﬁrst and second terms describe the incident and reﬂected electron plane waves. in which an incoming electron from the normal side of the normal/superconducting contact is reﬂected as a hole along the same trajectory [46]. When quasi-particles are bosons (chapter 4). which can be written in the matrix form: Eψ(x) = (x) −(1/2m) d2/dx 2 − µ(x) (x) (1/2m) d2/dx 2 + µ(x) ψ(x). Normal excitations ﬁll the negative part of their energy spectrum in k-space resulting in two components (normal and superﬂuid). In the normal state (x µ(x) = µn ).141) The gap (x) and the chemical potential µ(x) depend on the coordinate x 0. It shows the gap structure.4). The Andreev reﬂection results in an increase in the tunnelling conductance in the voltage range |eV | º in sharp contrast to its suppression in the case of the barrier. the kinetic energy of a moving condensate entirely dissipates into quasi-particles as soon as vs vc .
(2. There are other examples of the gapless superconductivity like unconventional Cooper pairs (section 2. perpendicular to the contact area. The two-ﬂuid situation at T = 0 is possible due to the Fermi statistics of quasiparticles in the BCS superconductor because the Pauli exclusion principle limits their density.Andreev reﬂection
57
evc /2. The third term describes the reﬂected (Andreev) hole. they renormalize it while vs < evc /2. The quasi-particles do not fully destroy the order parameter.142)
where momenta associated with the energy E are q ± = [2m(µn ± E)]1/2. as at ﬁnite temperatures. NS conductance can be readily calculated using the Fermi–Dirac golden rule (section 2. The superconductivity in this region is gapless. The two-component wavefunction of the normal metal is given by ψn (x < 0) = 1 0 eiq
+x
+b
1 0
e−iq
+x
+a
0 1
e−iq
−x
(2.141) is the free-particle Schr¨ dinger equation (ﬁrst row) o or its time-reversed version (second row). (x) = 0.13 Andreev reﬂection
The Bogoliubov equations are particularly instrumental in handling the interface between normal and superconducting metals (NS interface) and between different superconductors (SS interface). respectively. A simple theory of NS tunnelling in a metallic (no-barrier) regime [47] follows from the one-dimensional Bogoliubov equations. They allow us to calculate the I –V characteristics of tunnelling structures in the same fashion as in the conventional single-particle tunnelling problem in quantum mechanics. If there is a potential barrier due to a dielectric layer.

we obtain 1+b =c+d +d E +ξ E −ξ + + − q (1 − b) = ck − dk and q − a = ck + E +ξ − dk − E −ξ . The coefﬁcients a. which. If the incident energy (or the voltage V = E/e) is larger than the gap (E ). we obtain c = 1 and a= E +ξ (2. which determines the conductance.5). b.144) and (2. c.144) (2. so that ξ is not necessary real.58
Weak coupling theory
appears when the electron with the momentum q ≈ kF tunnels from a normal metal into a superconducting condensate together with its ‘partner’ having the opposite momentum −q.143)
Here the momenta associated with the energy E are k ± = [2m(µs ± ξ )]1/2 .e. √ where ξ = E 2 − 2 . which are the continuity of ψ(x) and its ﬁrst derivative at x = 0. (x) = . µ(x) = µs ) the incoming electron can produce only outgoing particles (i.147) For simplicity. The transmission coefﬁcient. respectively.5). In the superconductor (x 0. means that all reﬂection is the Andreev reﬂection and all transmission occurs without branch crossing.145). is given by T (E) = 1 + |a|2 − |b|2. (2. Then we ﬁnd b = d = 0. The solution in this region is given by ψs (x > 0) = c 1 E +ξ eik
+x
+d
1 E −ξ
e−ik
−x
(2.150) T (E) = √ E + E2 − 2
.146)
from equations (2.149) (2.148) a=c (2. (2. Applying the boundary conditions.145) (2. physically speaking. as in the conventional single-particle tunnelling problem. This simultaneous two-electron tunnelling results in the hole excitation on the hole branch of the excitation energy spectrum near −kF in the normal metal (ﬁgure 2. With b = d = 0. ξ is real and the transmission is obtained as 2E . The energy of the incident electron is deﬁned in the whole positive region (E 0). d are determined from the boundary conditions. with the positive group velocity ∂ k /∂k) (ﬁgure 2. we now take q ± ≈ k ± ≈ kF because the Fermi energy is huge compared with the gap in the BCS superconductors.

T = 0
59
E.Green’s function formulation of the BCS theory.

N)
( N)
.

G.

N) N)
F.

r . 0) −
(−t) ψ↑ (r . 0) − † (−t) ψs (r . 0) are connected by the relation F ∗ (r. (2. the closed circles denote an incident at (1). 0). F(r. t) and ψ↓ (r . t)ψ↑ (r .
It tends to one in the high energy limit (E ). (2. t)
(t) ψ↓ (r. t) . t) one-particle Green’s functions (GF) as iG(r. t)ψs (r . T = 0
There is yet another elegant formalism introduced by Gor’kov [43]. t)ψ↑ (r . t) = iF + (r. T (E) = 1 +
2
2. 0) and F + (r. equation (2. r . r . and transmitted (c). which allows for an economic derivation of BCS results and an extension of the theory to the intermediate coupling regime (chapter 3). Hence. when the energy ). (d) quasi-particles. 0)ψ↓ (r. the transmission is doubled compared with the normal is below the gap (E state limit: = 2. Remarkably. Using the Heisenberg operators. 0)ψ↓ (r. The open circle denotes the Andreev hole. 0) = −F + (r. t) (2.
Here the quantum averages are calculated in the ground state of the system at zero temperature.153)
. r . r . r . r . the Andreev reﬂection is observed as an enhancement of the tunnelling conductance of NS metallic contacts in the gap region V /e. The phenomenon serves as a powerful tool in the gap determination (part 2).151) (E + ξ )(E + ξ ∗ ) Hence. 0) − † † (t) ψ↓ (r. r . t) = iF(r.152) † † (−t) ψ↑ (r . The operators ψ↑ (r. t) and ‘anomalous’ F ± (r. 0)ψs (r.14 Green’s function formulation of the BCS theory. reﬂected (b) electrons.5. t ) anticommute if t = t. t) =
† (t) ψs (r. r .109). Schematic plot of excitation energies versus k at an NS interface. let us deﬁne the time-dependent ‘normal’ G(r. N
QRUPDOPHWDO LQWHUIDFH
VXSHUFRQGXFWRU
Figure 2.

ω) = 0. t) ˆ = −h ∗ (r)F + (r.158)
= . ω) − F + (k. ω) − G(k. The easiest way to apply the condition is to consider the normal state limit of equation (2. ∂t The order parameter is expressed in terms of the anomalous GF as i (r) = −2iE p [F + (r. F + (r.
u2 k ω−
+
k
2 vk ω+
(2. ω) = (ξk ) (−ξk ) + ω − ξk ω − ξk (2. We have to deﬁne a way of bypassing the poles in evaluating the integral. ω) exp(ik · r − iωt)
k
(ω − ξk )G(k. In a homogeneous superconductor. t) = G(r − r . r . ω) = 1 (ω + ξk )F + (k. Applying the Fourier transforms ∞ 1 dω G(k. r.159) with = 0.154)
(2. ω)](0) = 0.154) and (2. ω) exp(ik · r − iωt) (2. t) = F + (r − r . r . t) + ∗ (r)G(r. t). t) (2.110) and (2. The solution of the ﬁrst-order differential equations (2.155) ∂ G(r. r .111). t) = 2π we obtain
∞ −∞
dω F + (k. r . They cannot be used for integration with respect to ω because their simple poles are just on the real axis of ω. t) ˆ = δ(t)δ(r − r ) + h(r)G(r. r . (2.
.159)
k
These expressions are well deﬁned for any ω but not for ω = ± k . G (0) (k. Taking into account that d (t)/dt = δ(t).157) G(r − r . ω) = F + (k. Then the Fourier components are found to be G(k.160)
[F + (k.60
Weak coupling theory
The ﬁrst time derivatives of the GFs are calculated using the equations of motion (2. t) = 2π −∞
k
1 F + (r − r . ω) = ω + ξk ω2 − ω2 −
2 k 2 k
(2. +0)]∗ . t) and (r) is a real constant. r . we obtain a system of two coupled equations: i and ∂ F + (r. G(r. r .155) are not well deﬁned because we have not applied any initial condition so far.156)
We can readily solve these self-consistent equations in the absence of an external ﬁeld. r . t). t) + ∂t (r)F + (r.

As a result. are G(k. t) =
k
(−ξk ) is the Fermi–Dirac distribution at T = 0 and
(ξk ).
(2. if we choose the following way of bypassing the poles: G (0) (k. The pole in the ﬁrst term is below the real axis. ω) = u k vk
ω+
k
− iδ
−
ω−
. while the second one is 2πi (−t) (−ξk ). we recover the ideal Fermi-gas GF (equation (2. t) =
k
cks exp[i(k · r − ξk t)]
so that iG (0) (r − r .Green’s function formulation of the BCS theory. ω) exp(−iωt) =
(ξk ) +
∞ −∞
dω
e−iωt ω − ξk + iδ dω e−iωt .165) + iδ . we have exp[ik · (r − r ) − iξk t]{ (t) (ξk ) − (−t) (−ξk )}.163)
where δ = +0 is an inﬁnitesimal positive constant. ω) = (−ξk ) (ξk ) + ω − ξk + iδ ω − ξk − iδ (2. T = 0
61
Integrating the Fourier transform G (0) (k.162) We obtain the same result using the Fourier transform equation (2. which provide a correct normal state limit. The free-electron Heisenberg operators are found as ψs (r. When t is negative.k
cks c† s exp[i(k · r − k · r ) − iξk t] k (−t) c† s cks exp[i(k · r − k · r ) − i ξk t] k (2. while the pole in the second term is found in the upper half-plane. Indeed let us calculate the integral
∞ −∞
dω G (0)(k. Hence. ω) = u2 k ω−
k
+ iδ
+ 1
2 vk
ω+
k
− iδ 1
k
(2. the superconducting GFs. the ﬁrst integral is zero.161)
− where c† s cks = δkk k cks c† s k = δkk iG (0) (r − r .160). which can be directly calculated using its deﬁnition.164) ω − ξk − iδ
(−ξk )
∞ −∞
When t is positive the contour in both integrals of the right-hand side should be chosen in the lower half-plane of the complex variable. The ﬁrst integral yields −2πi (t) (ξk ) and the second integral is zero. t) = (t)
k.
F + (k.162)). (2. Hence. ω) should yield the normal state GF.

Only the second term of the normal GF. τ1 . τ2 )
(2.15 Green’s functions of the BCS superconductor at ﬁnite temperatures
At ﬁnite temperatures.27)). r . τ1 .167) ne =
k 2 so that the electron distribution function is vk . as it should be (see equation (2. r. equation (2. varying in the interval 0 < τ1 . +0)]∗ = −
2iE p 2π
∞ k −∞
dω eiωt {F + (k.165). τ2 ) = − Tτ ψ↓ (r.166)
Here t is negative and the contour should be taken in the upper half-plane. The electron density is calculated as n e = −iG(r. Calculating the integral with the Fourier transform equation (2. −0) = − i 2π
∞ k −∞
dω e−iωt G(k. r . τ2 ) = − Tτ ψs (r. τ1 . Following Matsubara [44]. Let us show that these GFs also provide the BCS results for the electron distribution function and the gap. τ1 )ψ↑ (r .165).170)
where ψs (r. we replace time t in the deﬁnition of the Heisenberg operators by a ‘thermodynamic time’ τ = it. we obtain the BCS equation at T = 0. ˆ A =
ν
e(
−E ν )/T
ˆ ν| A|ν ≡ Tr{e(
˜ − H)/T
ˆ A}
(2.
Here the thermodynamic ‘times’.171)
. ω)
t →−0
.62
Weak coupling theory
We note that the poles of the Fourier transform of GF yields the excitation spectrum of the superconductor. = −2E p
k
u k vk . τ ) = exp(H τ )
+
s (r) exp(−H τ )
and Gor’kov’s temperature GF is
† † (r.169)
2. The double angular brackets correspond to quantum ˆ as well as statistical averages of any operator A with the Gibbs distribution (appendix B). τ1 )ψs (r . the BCS theory can be formulated with the ‘temperature’ GFs (appendix D). r.
(2. contributes with the following result: 2 vk (2. τ2 ) . Then the temperature GF is deﬁned as
† (r. are real and positive. The gap is found to be
= −2iE p [F + (r. τ2 . ω)}∗ (2.168)
where t = +0.
(2. τ2 < 1/T .

The operation Tτ performs the ‘time’ ordering according to the following deﬁnition:
† Tτ ψs (r. . .179)
. the variable τ is deﬁned in the domain −1/T τ 1/T.172)
Actually the temperature GFs depend on the difference τ = τ1 − τ2 because of the trace in their deﬁnition. r . ±2. τ ) = − exp( /T ) Tr{e(τ −1/T ) H
˜ s (r)e ˜ −τ H † s (r
)}.170). . τ1 .177)
where n = 0. τ2 )
−
† (τ2 − τ1 )ψs (r . τ + 1/T ) = − exp( /T ) Tr{eτ H ˜ ˜ = − (r.
(2. r . ±1.
(2.173)
s (r)}
where we have performed the cyclic permutation under the trace. Following the deﬁnition in equation (2.Green’s functions of the BCS superconductor at ﬁnite temperatures
63
˜ where is the thermodynamic potential and |ν are the eigenstates of H with the eigenvalues E ν . F(τ ) = −F(τ + 1/T ) for a negative τ and f n = 1 (1 − eiπn ) 2
1/T 0
dτ F(τ ) exp(iπnT τ ). so that (r. .178)
In our case. ˜ s (r)e ˜ −(τ +1/T ) H ˜ † s (r
)} (2.176)
The Fourier transform theory states that if a function F(τ ) is deﬁned over the interval −1/T τ 1/T .175)
Here replacing the positive τ by a negative τ as τ = τ + 1/T .
(2. r . τ2 ) = exp( /T ) Tr{e− H /T eτ2 H = exp( /T ) Tr{e
˜ −(τ +1/T ) H ˜ ˜ ˜ † τH s (r )e ˜ † τH s (r )e s (r)e ˜ −τ1 H
} (2.174)
There is a connection between GFs with negative and positive τ . Differing from positive τ1 and τ2 . τ ). τ2 ) ≡ † (τ1 − τ2 )ψs (r. (2. τ1 )ψs (r . for τ > 0. τ1 ). then its Fourier expansion is
∞
F(τ ) = T
n=−∞
f n exp(−iπnT τ )
(2. τ2 )ψs (r. that (r. and fn =
1 2 1/T −1/T
dτ F(τ ) exp(iπnT τ ). τ1 )ψs (r . Indeed let us consider τ1 < τ2 . we ﬁnd. r . we obtain (τ < 0) ˜ ˜ ˜
˜ (r.
(2.

ωn ) = −
. ∂τ
(2. r . r .180) (2.184)
The order parameter is given by
∗
(r) = −2E p
+
(r.186)
= −2E p T
k ωn
+
(2. r . ωn ) = −
+
iωn + ξk 2 2 ωn + k
∗ 2 ωn + 2 k
(2. 0). ±1. + (r.183) (2. τ ) ∂τ (2. τ ) + (r)ψ↓ (r.182)
and the equations for temperature GFs. τ ) † ˆ = h ∗ (r)ψ↓ (r.64
Weak coupling theory
Hence the Fourier components fn of the fermionic GFs are non-zero only for odd n. As a result we can expand the temperature GFs of a homogeneous superconductor into a Fourier series as follows: (r. .
(2. τ ) ˆ = δ(τ )δ(r − r ) + h(r) (r. . ±2. the Fourier transforms of these equations are: (iωn − ξk ) (k. ωn ). r . τ ) − ∗ (r) (r. . . τ ) † ˆ = h(r)ψ↑ (r. we obtain the equations of motion.181)
(r. τ ) = T
ωn +
(k. ωn ) = 0 (k. τ ) ∂ ˆ = h ∗ (r) + (r. Differentiating the Matsubara operators with respect to τ . ωn ) exp[ik · (r − r ) − iωn τ ]
+ ωn
(2. r . τ ).188)
(k. r . τ ) = T
(k. r . ωn ) − (iωn + ξk ) + (k. r. τ ) ∂τ † ∂ψ↓ (r.185)
In the absence of an external ﬁeld. r . ωn ) exp[ik · (r − r ) − iωn τ ]
where the discrete Matsubara frequencies are ωn = π T (2n + 1).187)
The solution is (k. ωn ) − and
∗ + ∗
(k. − ∂ (r. ωn ) = 1 (k. τ ) ∂τ + (r) + (r.
(2. τ ) − ∗ (r)ψ↑ (r. n = 0. τ ). − ∂ψ↑ (r.

A (k.
(2. t) = i (−t)
A
(2. ω) ± i coth 2T and between those of G R.195) (2. In fact. which are found on the imaginary frequency axis. iωn ) = (k.
In our case the temperature GF is (k. ωn ) = +
k
(2.
(2.194) G R. for example of the gap (equation (2. ω) =
u2 k ω−
k
+ iδ
+
ω+
k
+ iδ
. −ωn ) = u2 k iωn −
∗
(k. A one-particle real-time GF is deﬁned at ﬁnite temperatures as G(k.189)
The sum over frequencies in equation (2.197)
k
2 2 where u 2 . There is a simple connection between the Fourier components of G and G R. Nonetheless. (2.
2 vk iωn + 2. ω) (2. (2. ω) = Re G(k. vk = ( k ± ξk )/2 k and k = ξk + k of this expression to the upper half-plane yields
The analytical continuation
2 vk
G R (k.A and . t) = −i Tt cks (t)c† ks with the real time t.A (appendix D): ω Im G(k.54)). we obtain the BCS equation for the order parameter: ξ 2 + (T )2 λ dξ (2.191)
(2.192) .198)
. There is no direct physical meaning of poles of the Fourier transforms of temperature GFs. There are also retarded G R and advanced G A GFs: G R (k.196) G R (k. ωn ).190) tanh 1= 2 2T ξ 2 + (T )2 where the integral should be cut at |ξ | ωD . the Matsubara GFs lead directly to real-time GFs. ωn ) for ωn > 0 and (k.193)
{ck (t)c† } k
Their Fourier components are analytical in the upper or lower half-plane of ω. t) = − i (t) {ck (t)c† } k G (k. respectively. The kinetic properties are expressed in terms of real-time GFs. As a result. they allow for a direct calculation of the thermodynamic properties of the system.190)).189) is calculated using the expansion of tanh(x) (equation (2.Green’s functions of the BCS superconductor at ﬁnite temperatures
and the equation for the order parameter takes the following form: = 2E p T
k ωn 2 ωn + 2 k
65
.

r ) in the magnetic ﬁeld. r
) = δ(r − r ) + )= −
∗
(r)
+ ωn (r. r
[∇ + ie A(r)]2 +µ 2m [∇ − ie A(r)]2 +µ 2m
1/T 0
ωn (r. For zero temperature. r . r
). τ ) exp(iπnT τ )
+ ωn (r.183) and (2. r
)
(2. τ ) exp(iπnT τ ) dτ
+
1/T
(r.
(2. r).200)
as it should (see equation (2. ω) = u2 k ω−
k
+ iδ
+
2 vk
ω+
k
− iδ
(2.
2. r . which satisﬁes the following equation: iωn + [∇ + ie A(r)]2 +µ 2m
(n) ωn (r. Then we can expand the GFs in powers of (r).16 Microscopic derivation of the Ginzburg–Landau equations
Using the Green’s function formalism near Tc . Gor’kov [45] derived the Ginzburg–Landau equations from BCS theory. ωn (r.201)
)= )=
0 ∗
dτ (r.202)
. the order parameter is small.184) with respect to the thermodynamic time yields iωn + −iωn + where
ωn (r. Let us consider the BCS superconductor in a stationary magnetic ﬁeld with the vector potential A(r). r + ωn (r.
and
(r) = −2E p T
ωn
If the temperature is close to Tc . r + ωn (r. tanh(ω/2T ) = sign(ω) and G(k. Here the ﬁrst term is understood as the principal value of the integral when integrating with respect to ω.165)). The zero-order GF is a temperature GF of (n) an ideal Fermi gas. The Fourier transform of equations (2. r
) = δ(r − r ).199) where we have applied the relation 1/(ω + iδ) = P(1/ω) − iπδ(ω).66
Weak coupling theory
Then using equation (2. we obtain G(k. ω) = P u2 k ω− +
k 2 vk ω+
− iπ tanh
k
ω [u 2 δ(ω − k 2T
2 k ) + vk δ(ω + k )]
(2.
(r)
ωn (r.194).

209)
If A(r) is a constant then the solution is ψs (r. It satisﬁes the equation of motion ∂ψs (r. GFs in a magnetic ﬁeld are not translation invariant and the exact calculation of the normal (n) ωn (r.Microscopic derivation of the Ginzburg–Landau equations
If there is no magnetic ﬁeld. ωn ) =
1 . iωn − ξk
Transforming back to real space we obtain
(0) ωn (ρ)
=
1 (2π)3
dk
eik·ρ iωn − ξk
(2. Calculating the integral over the angles in the polar spherical coordinates yields
(0) ωn (ρ)
=
1 (2π)2 iρ
∞
k dk
0
eikρ − e−ikρ . r ) using equation (2.207)
The Fourier component of the normal state GF falls exponentially with the characteristic length (2. iωn − ξ
(2.205)
We can replace k by k = kF + ξ/vF because only the states near the Fermi surface EF.205). If we neglect these variations. 2πρ
(2. is found as
(n) ωn (r. τ ) =
k
cks exp[i(k − e A) · r − ξk τ )]
(2. iωn − ξ
(2.204)
where ρ = r − r .210)
. But we can assume that the spatial variations of the vector potential are small over the characteristic distance ξ0 .202) is a challenging problem.203)
(k. τ ). r (0)
67
)=
(0) ωn (r
− r ) and its Fourier transform (2. τ ) = ∂τ [∇ + ie A(r)]2 + µ ψs (r.206)
Using the contour in the upper or lower half-plane.208) ρ ≈ ξ0 where ξ0 = vF /(2π Tc ) is the zero temperature coherence length. when ωn is of the order of Tc The result is
(0) ωn (ρ)
=
m (2π)2 iρ
∞ −∞
dξ
eikF ρ eiξρ/vF − e−ikρ e−iξρ/vF . then the ﬁeld operator in the magnetic ﬁeld can be readily expressed via the free-electron operator. contribute to the integral in equation (2. 2m (2. we obtain
(0) ωn (ρ)
=−
m exp[i sign(ωn )kF ρ − |ωn |ρ/vF ].

r
Now we expand ωn (r. r
) to the
)=
(n) ωn (r.
y) ( y) (2.212) B Hc2 ≈ 2 2πξ0 Tc so that the quasi-classical approximation is applied.209). Hence. r)
(x)
(n) ωn (x.
(2. r
)
(2. r ∗
)
)= − + ×
dx dx
(n) −ωn (x. Using the normal state GF in the magnetic ﬁeld (equation (2.
(z)
(n) ωn (z.215) )
(n) ωn (x.216)
×
(n) −ωn (z. r) ∗
− 2E p T
ωn
dx y)
∗
dy
dz
(n) −ωn (x. (2. the GF differs from the zero-ﬁeld GF only by the phase:
(n) ωn (r. and the phase changes by eBξ0 . we can transform Gor’kov equations (2. | A(r)| varies by Bξ0 over 2 this distance. if the vector potential varies slowly.4) 0 Tc − T (2.
). r)
∗
(x)
(n) ωn (x. r).
Using this expansion.68
Weak coupling theory
which is veriﬁed by direct substitution into equation (2. we obtain the following equation for the order parameter:
∗
(r) = 2E p T
ωn
dx
(n) −ωn (x. r
). The maximum ﬁeld is (see section 1. x) ∗
(x)
+ ωn (x.213)
It is the very same region where the Ginzburg–Landau theory is applied. r) (n) ωn (z.211)).
ωn (r. r)
(x)
ωn (x.
×
+ ωn (r.214)
)=−
dx
(n) −ωn (x. if Tc − T Tc 1. r)
(x)
(n) ωn (x.6.
. r ∗
dy y)
∗
dz (z)
(n) −ωn (x. r
(n) −ωn ( y.201) into the integral form:
ωn (r. r + ωn (r. r
) = e−ie A(r)·ρ
(0) ωn (ρ). r
)=
(n) ωn (r. x)
( y)
(n) ωn ( y. r
)−
dx
∗
dy
(n) ωn (r. x)
(x) (2. r
)+
dx
(n) ωn (r.211)
This quasi-classical approximation with respect to the magnetic ﬁeld is applied if the phase change over the distance ξ0 is small. r
(x)
y) ( y)
(n) −ωn (z. r ) to the terms of the second order and terms of the third order in (r) inclusive:
ωn (r.

Microscopic derivation of the Ginzburg–Landau equations
Let us consider the kernel in the ﬁrst term K (x − r) = T
ωn (n) (n) −ωn (x.221)
We cut the divergent integral in equation (2.
(2. The kernel K (0) (ρ) has a characteristic radius about ξ0 . r)
69
= K (0) (x − r)e−2ie A(r)·(x−r) (2. 3 4(π T )2 sinh x 3 8(π T )2 0 (2.219) Here the second term is zero because the function under the integral is odd. j =1
dρ K (0) (ρ) +
dρ K (0) (ρ)ρ · [∇ r − 2ie A(r)] ∂ − 2ie Ai ∂ri
(r)
∗
1 2
dρ K (0)(ρ)ρi ρ j
∂ − 2ie A j ∂r j
(r). r) ωn (x.218)
and ρ = |x − r|.220) at |ξ | = ωD as usual and introduce the DOS at the Fermi level: N(E F ) = mkF /(2π 2 ). The second term
.217)
where K (0) (x − r) = T = m2 (2πρ)2 exp[−2|ωn |ρ/vF ]
ωn
m2T (2πρ)2 sinh(2π Tρ/vF )
(2. The remaining integrals are: dρ K (0)(ρ) = T
ωn
dρ
(0) (0) −ωn (ρ) ωn (ρ) ωD
=T
k ωn
1 = N(E F ) 2 + ξ2 ωn k
dξ
0
tanh(ξ/2T ) (2. Therefore in the integral over x we can expand all quantities near the point x = r up to the second order in ρ inclusive: dx K (0)(x − r)e−2ie A(r)·(x−r) ≈ +
∗ ∗
(x)
∗
(r)
3 i.220) ξ
and dρ K (0) (ρ)ρi ρ j = δi j dρ K (0)(ρ)ρ 2 3 2 2 ∞ δi j 7ζ(3)vF δi j vF x2 = = N(E F ) dx N(E F ). while (x) changes over a much larger distance of the order of ξ(T ) ξ0 near Tc .

we obtain the phenomenological Ginzburg–Landau equation (section 1.215) in powers of (r). Here we can neglect the space variations of (x) and the phase due to the vector potential under the integrals.1): − 1 [∇ + i2e A(r)]2 φ(r) + β|φ(r)|2 φ(r) = −αφ(r). 4m
Now the coefﬁcients α and β are determined microscopically as α= β= 12π 2 Tc (T − Tc ) 2 7ζ (3)mvF 12π 2 Tc2
2 7ζ (3)mvF n e
(2. (2. ξ
Then close to Tc . Finally let us show that the second GL equation for supercurrent follows from the expansion of GFs (2.
3 Here n e = kF /(3π 2 ) is the electron density.226)
. we can expand it as
ωD
dξ
0
1 Tc − T tanh(ξ/2T ) ≈ + ξ λ Tc
(2. Then applying the Fourier transform of the zero-ﬁeld ideal gas GF.224)
and collecting all integrals in equation (2. equation (B. we obtain the following expression for this term: −2E p T
∗
(r)| (r)|2
k ωn
2 (ωn
1 7ζ(3) = −λ 2 )2 8(π T )2 + ξk
∗
(r)| (r)|2 .216) is cubic in (r).28) from the appendix.225)
.70
Weak coupling theory
on the right-hand side of equation (2.6.
(2.222)
We note that Tc is determined from the linearized BCS equation: 1=λ
0 ωD
dξ
tanh(ξ/2Tc ) . Introducing the ‘condensate wavefunction’ φ(r) as φ(r) = (r) 7ζ(3)n e 8π 2 Tc2 (2.223)
where we use the integral. The current–density operator can be expressed in terms of the ﬁeld operators:
(r) = ˆ
s
ie (∇ r − ∇ r )| r →r 2m
† s (r
)
s (r) −
e2 A(r) m
† s (r)
s (r)
.216).

which oscillates in real space with the electron wavelength (≈1/kF ).229)
Expanding the slowly varying order parameter and the magnetic phase exponent in powers of | y − r| and |x − r|. and the ﬁrst (normal) term in the expansion of GFs (equation (2.231)
. The remaining supercurrent is quadratic with respect to the order parameter: j (r) = ie (∇ r − ∇ r )| r →r T m ×
ωn
dx
dy
(n) ωn (r. m
(2.227)
There is no current in the normal state in the stationary magnetic ﬁeld. Differentiating the oscillating part yields j(r) = ie T m ×[ dx
ωn (0) ωn (|r (n) (0)
d y (x)
(0) ωn (| y
∗
( y)e2i A·(x− y)
(0) ωn (| y
(0) −ωn (| y
− x|)
(0) ωn (|r
− x|)∇ r
− r|) −
− r|)∇ r
− x|)].230)
Here the constant C is given by C= T 3 dx
ωn
dy
(0) −ωn (| y
− x|)
(0) ωn (|r
− x|)[(x − y) · ∇ r ]
(0) ωn (| y
− r|).227). x)
∗
( y)
(n) ωn ( y.215)) cancels the second term on the right-hand side of equation (2.
(2.Microscopic derivation of the Ginzburg–Landau equations
Its expectation value is j(r) = ie (∇ r − ∇ r )| r →r T m
ωn (r. r ) = exp(−ie A · ρ) ωn (ρ) is the product of a slowly varying phase exponent and the zero-ﬁeld GF.
∗
(x) ≈ ( y) ≈
(r) + (x − r) · ∇ (r) (r) + ( y − r) · ∇ ∗ (r)
exp[2ie A · (x − y)] ≈ 1 + 2ie A · (x − y) we obtain j (r) = ie C[ m
∗
(r)∇ (r) −
(r)∇
∗
(r)] −
4e2 C | (r)|2 A(r).
(2. which are of the order of ξ0 . m
(2.228) The normal state GF in the magnetic ﬁeld ωn (r. r
). x)
(x)
(n) −ωn ( y. r ωn
71
)−
e2 n e A(r). (2.

. 3 ωn 16π 2 Tc2
Replacing the order parameter in equation (2. As an example. which is discussed in the next chapter.40)): j(r) = ie∗ e∗2 [φ ∗ (r)∇φ(r) − φ(r)∇φ ∗ (r)] − ∗∗ A(r)|φ(r)|2 2m ∗∗ m (2. At the ﬁrst stage. When the coupling constant λ increases. the coherence length at T = 0) is about ten thousand times larger than the distance between electrons. Here we encounter qualitatively different physics [11]. the Fermi liquid becomes unstable even in the normal state above Tc because of the polaron collapse of the electron band (chapter 4).234)
where e∗ = 2e and m ∗∗ = 2m are the charge and mass of the Cooper pair.233)
2 kF π N(E F )T 6m
ωn
7ζ(3)n e 1 = . in aluminium the size of the pair (i. respectively. Both the retardation effect and the damping are taken into account in the Eliashberg extension of the BCS theory to the intermediate coupling regime [36]. However. (iωn + ξk )(iωn − ξk )3 (2.72
Weak coupling theory
Using the Fourier transform of the ideal gas GF in calculating the integrals over x and y in C leads to C= T 3 1 1 ∂ ·k iωn + ξk ∂ k (iωn − ξk )2 k2 . The BCS theory is a mean-ﬁeld approximation. The Cooper pairs in the BCS superconductor disappear above Tc and the one-particle excitations are fermions. deviations from BCS theory arise in two ways: (1) the BCS approximation for the interaction between electrons does not provide an adequate representation of the retarded nature of the phonon induced attraction. the critical temperature increases and the coherence length becomes smaller. if λ is larger than 1.230) by the ‘condensate wavefunction’ leads to the second GL equation (see equation (1. and (2) the damping rate becomes comparable with the quasiparticle energy.e. one can expect some deviations from the BCS behaviour in the intermediate coupling regime. Hence. which is valid if the volume occupied by the correlated electron pair is large compared with the volume per electron.232)
ωn
k
T = 3m
ωn
k
Integrating with respect to k yields C= and we obtain C=
2 2kF N(E F )T 3m ∞ ωn −∞
dξ
2 ωn − ξ 2 2 + ξ 2 )3 (ωn
(2.

the electron–phonon interaction) is strong. The system is neutral and Ne = Z N. However. The ratio of the characteristic Coulomb energy to the kinetic energy is rs = m e e2 /(4πn e /3)1/3 ≈ 1 for the electron density n e = Z N/V = 1023 cm−3 (further we take the volume of the system as V = 1. R j are the electron and ion coordinates.1 Electron–phonon interaction
The attraction between electrons in BCS theory is the result of an ‘overscreening’ of their Coulomb repulsion by vibrating ions. we can take advantage of the small value 73
. respectively. the ions can be considered as rigid charges. Phonons are also affected by conduction electrons. . the electron Bloch states are affected even in the normal phase. ‘bare’ phonons are well deﬁned in insulating parent compounds but a separation of electron and phonon degrees of freedom might be a problem in a metal. where ∞ is the phenomenological high-frequency dielectric constant. j = 1. . One cannot solve the corresponding Schr¨ dinger equation perturbatively because o the Coulomb interaction is strong. . . The inner electrons are strongly coupled to the nuclei and follow their motion. N.1) by √ e/ ∞ .Chapter 3 Intermediate-coupling theory
3. Here we have to start with the ﬁrst-principle Hamiltonian describing conduction electrons and ions coupled by the Coulomb forces: H= −
i
∇i2 e2 + 2m e 2
i =i
1 − Z e2 |r i − r i | ∇2 j
j
ij
1 |r i − R j | (3. ∇ j = ∂/∂ R j .1)
+
Z 2 e2 2
j=j
1 − |R j − R j |
2M
where r i . In doped insulators (like high-temperature superconductors). . unless speciﬁed otherwise). ∇i = ∂/∂ r i . . Z e is the ion charge and M is the ion mass. When the interaction between ion vibrations and electrons (i. . To account for their high-energy electron degrees of freedom we can replace the elementary charge in equation (3. i = 1. Ne . Hence. .e.

β = x.6) is the vibration energy.5)
is the electron energy in a periodic crystal ﬁeld V (r) = l v(r −l) which is Î (r) calculated at R j = l and with the periodic electron density n (0) (r + l) = n (0) (r). z and Dαβ (l − m) = Z 2 e2 ∂2 ∂lα ∂m β 2 1 + |l − m | dr n (0) (r)V (r) (3. This is called the local density approximation (LDA).7)
l =m
is a dynamic matrix. As a result.β
(3. Here α. Because the vibration amplitudes are small we can expand the Hamiltonian in powers of |u| up to quadratic terms inclusive. the Hamiltonian takes the form in the second s quantization: H = He + Hph + He−ph + He−e (3. y. which replaces the Coulomb electron–electron interaction by an effective onebody potential Î (r):
Î (r) = −Z e2
j
1 + e2 |r − R j |
dr
n(r ) + µex [n(r)] |r − r |
(3. which is to approach the ground state of the many-electron system via a one-electron picture.
(3. Î (r) is the functional of the electron density n(r) = † s (r) s (r) . Any further progress requires a simplifying physical idea.4) where He =
s † s (r)
dr
−
∇2 + V (r) 2m e
s (r)
(3.α.m. The ions are heavy and the amplitudes √ |u| 1/MωD of their vibrations near the equilibrium R 0 ≡ l are much smaller than the lattice constant a = N −1/3 : |u| ≈ a me Mrs
1/4
1. usually calculated numerically or expressed as µex [n(r)] = −βn 1/3 (r) with the constant β in a simple approximation.74
Intermediate-coupling theory
of the adiabatic ratio m e /M < 10−3 . The electron–phonon interaction is given by He−ph =
l
ul ·
∂ ∂l
dr
s
† s (r)
s (r)
− n (0) (r) V (r)
.2)
In this estimate we take the characteristic vibration frequency ωD of the order of the ion plasma frequency ωi = 4π N Z 2 e2 /M.3)
where µex [n(r)] is the exchange interaction. Hph =
l
−
2 ∇u ∂ + ul · 2M ∂l
dr n (0) (r)V (r) +
1 2
u lα u mβ Dαβ (l − m)
l.

13)
† The last equation and the bosonic commutation rules [dqν dq ν ] = δνν δqq follow
from (∂/∂u α )u l − u l (∂/∂u α ) = δαβ .ν
Mωqν dqν exp(iq · l) − H.11)
if the eigenfrequencies ωqν and the eigenstates eqν satisfy
2 α Mωqν eqν = β
Dq eqν
αβ β
(3. Then Hph takes the following form Hph =
q.m.Electron–phonon interaction
+ 1 2 u lα u mβ
l.14)
.β
75 (r) V (r) (3.
β
(3.8)
∂2 ∂lα ∂m β
dr
s
† s (r)
s (r) − n
(0)
and the electron–electron correlations are described by He−e = 1 2 − + dr dr Z 2 e2 2 dr e2 |r − r | dr
† s (r) ss † s (r) s s (r) † s (r
)
s
(r )
s (r)
e2 n (0) (r ) +µex [n (0) (r)] |r − r |
l =m
1 .c.ν ∂ = ∂ ul eqν
q.ν † ωqν (dqν dqν + 1/2)
(3. therefore.α.9)
We include the electrostatic repulsive energy of nuclei in He−e . |l − m|
(3. eqν and ωqν are the unit polarization vector and the phonon frequency. The vibration Hamiltonian Hph is a quadratic form and.c. respectively. Here l l Dq =
m αβ
β
β
exp(iq · m)Dαβ (m)
(3. of the phonon mode ν. can be diagonalized with a linear canonical transformation for the displacement operators eqν ul = dqν exp(iq · l) + H.10) 2N Mωq ν q. so that the average of He−e is zero in the Hartree approximation. 2N
where q is the phonon momentum. (3. dqν is the phonon (Bose) annihilation operator.12)
and
q
∗α eqν eqν = Nδαβ .

16)
One-particle states are sorted with the momentum k in the Brillouin zone.n . which determines the crystal ﬁeld potential V (r). k. the Hamiltonian is written as H = H0 + He−ph + He−e where H0 =
k.19)
(3. Different solutions of equation (3. 3 for a simple lattice and 1. The LDA can explain the shape of the Fermi surface of wide-band metals and gaps in narrow-gap semiconductors.s
(3. 2m e (3. The ﬁrst derivative in equation (3. The part of the electron– phonon interaction.12) are classiﬁed with the phonon branch (mode) quantum number ν. 3k for a lattice with k ions per unit cell. ν) = − N M 1/2 ωqν
3/2 ∗ dr (eqν · ∇v(r))ψnks (r)ψn k−qs (r) † γnn (q. which is linear in phonon operators. which is 1. This is not the case in narrow d.and f-band metals and oxides (and other ionic lattices). . The periodic part of the Hamiltonian He is diagonal in the Bloch representation (appendix A):
s (r)
=
k. ν)ωqν cnks cnk−qs dqν + H.s
ψnks (r)cnks
(3.s † ξnks cnks cnks + q.76
Intermediate-coupling theory
is the Fourier transform of the second derivative of the ion potential energy. Using the phonon and electron annihilation and creation operators. The Bloch function obeys the Schr¨ dinger equation o − ∇2 + V (r) ψnks (r) = E nks ψnks (r). 2.6) is zero in crystals with a centre of symmetry.c.20)
.15)
where cnks are the fermion annihilation operators.18)
describes independent Bloch electrons and phonons.n. The spin-polarized version of LDA can explain a variety of properties of many magnetic materials. can be written as 1 He−ph = √ 2N where γnn (q. ξnks = E nks − µ is the band energy spectrum with respect to the chemical potential. k.17) (3. The solutions of this equation allow us to calculate the periodic electron density n (0) (r).n. k.ν. . . where the electron– phonon interaction and Coulomb correlations are strong.ν † ωqν (dqν dqν + 1/2)
(3.n. band index n and spin s.q. . These materials display much less band dispersion and wider gaps compared with the ﬁrst-principle band structure calculations.

7)) is the Coulomb repulsion between ions.23) ρq = ks
k. Core electrons together with nuclei form compact ions with an effective Z . |l − m | (3. k. which is expanded in the Fourier series as 1 = 4π lim κ→0 r q2 1 eiq·r . Therefore.24)
is a constant. the only relevant and the carrier density interaction in the dynamic matrix (equation (3.Phonons in metal
77
is the dimensionless matrix element. H0 should also include a random potential in doped semiconductors and amorphous metals. They have a role to play only for those phonons which are not coupled with electrons by the linear interaction (3. (3. The carrier wavefunction outside the core can be approximated by a plane wave ψnks (r) n (0) (r) eik·r (3.2 Phonons in metal
In wide-band metals such as Na or K. which yields Dαβ (l − m) = Z 2 e2 ∂ 2 2 ∂lα ∂m β 1 .s
is the density ﬂuctuation operator.20). + κ2 (3.22)
where Vc (q) is a matrix element. which is zero for q = 0 because of electroneutrality and † c† ck+qs (3. The electron–electron correlation energy of a homogeneous electron system is often written as He−e =
1 2 q † Vc (q)ρq ρq
(3. we obtain qα qβ Dαβ (m) = 4π lim cos(q · m) 2 + κ2 κ→0 q q
(3.27)
.25) and (3.
3.25)
l =m
The electron–ion interaction is a pure Coulomb attraction v(r) = −Z e2 /r. k.20).26)
q
Substituting this expansion into equations (3. the correlation energy is relatively small (rs ≤ 1) and carriers are almost free. ν) depending only on the momentum transfer q (the Fr¨ hlich interaction): o γnn (q. Low-energy physics is often described by a single-band approximation with the matrix element γnn (q. ν).21)
The terms of He−ph which are quadratic and higher orders in the phonon operators are small. ν) = γ (q.

core electrons undergo a polarization. The interaction gives rise to a signiﬁcant renormalization of the bare phonon frequency (equation (3.78
Intermediate-coupling theory
+ a
b + c D(q.
and γ (q. We apply the Green’s function (GF) formalism (appendix D) to calculate the renormalized phonon frequency.31) and G(k. ω)]−1 (3. when ions are displaced from their equilibrium positions. In fact.1.27). ω) (3. According to equation (3. The phonon GF in the Migdal approximation (f ). κ→0 q 2 + κ 2 lim
(3.28)
Calculating the Fourier transform of equation (3. (3. the carriers interact only with longitudinal phonons. which is the result of our approximation considering ions as rigid charges. ω)]−1 − [G(k. (c). ω) = [G (0)(k.32)
. ν) = i
4π N Z e2
3 Mωq
eqν · q . one obtains the following equation for phonon frequencies and polarization vectors: eqν · q 2 . (e) corrections to the phonon GF. which yields a ﬁnite shear mode frequency.29) ωq eqν = ωi2 q q2 A longitudinal mode with e q is the ion plasmon ωq = ωi (3.30)
and two shear (transverse) modes with e ⊥ q have zero frequencies. the electron self-energy is introduced as (k. ω) = 1 . (d).28). For interacting electrons. ω − ξk − (k. The Fourier component of the free-electron GF is given by equation (2.30)).163) with ξk = k 2 /2m e −µ.ω) = f d + +
+
e
Figure 3. Second (a) and fourth order (b).

The sum of all dangerous diagrams is given in ﬁgure 3. ω + ξk − ξk+q + iδ sign(ξk+q )
(3.1(b) with two polarization loops is the most ‘dangerous’ one. a vertex (circle) corresponds to the interaction matrix element γ (q) ωq /N and bold lines represent G and D.1(f ) which is (q. Thin straight and dotted lines correspond to G (0) and D (0) . The Fr¨ hlich interaction is the sum of two operators o describing the emission and absorption of a phonon.1.
(3. ω)
=
1 4π 3
dk
(ξk ) − (ξk+q ) . ω)]−1 . 2 ωq D (0) (q.or higher-odd orders corrections to D because the Fr¨ hlich interaction is o off-diagonal with respect to phonon occupation numbers. 2
79
(3. ω)
(3.39)
. which is a (0) .38) ﬁrst with the following result:
(0) e (q. Differing from others. ω)
(3. the singularity of internal vertices is ‘integrated out’ in the diagrams in ﬁgures 3. (3. ω)
|γ (q)|2 ωq N =
e (q.38)
The additional factor 2 in the phonon self-energy is due to a contribution of two electron spin states. ω)
=−
2i (2π)4
dk d G (0) (k + q. ω) = [D (0) (q.1(c)– (d).35)
The Feynman diagram technique is convenient.1(a)) includes the so-called polarization bubble. + ω)G (0) (k. ω)]−1 − [D(q. Therefore wavy lines have no direction. The second-order term (0) in D (ﬁgure 3. t) = −i ωq † † Tt (dq (t)dq + dq (t)dq ) .34) 2 ω − ωq + iδ The phonon self-energy is (q.33)
and its Fourier transform for free phonons is a dimensionless even function of frequency. which is large for small q. e . ω) = 2 . There are no ﬁrst. (3.Phonons in metal
The phonon GF is deﬁned as D(q. see ﬁgure 3. ). It is convenient to integrate over frequency in equation (3. Among different fourth-order diagrams the diagram in convolution of two G ﬁgure 3. Both events are taken into account in the deﬁnition of D.37)
and
(0) e (q.36)
(0) e (q. respectively. it is proportional to 1/q 2. ω) = where
e (q. However.

80
Intermediate-coupling theory
πHT
Figure 3. ω) 0 e (q.1(f ). ω) .28) for ωq = ωi .41)
m2 e |ω| (2kF − q) 2πq
1+x 1 − x2 . ω) = ωi2 . we can take the limit ω → 0 in m e kF h 2π 2 q 2kF
= − = −
(3. ω).42)
As a result. the screened e (q. Screened polarization bubble
e (q. ω) is zero in this limit. (0) While e is ﬁnite at q → 0. we obtain the phonon GF as D(q.40) (3. − ωq ˜2 (3. 1 − (4πe2 /q 2 ) 0 (q.
The phonon frequency ω is small. we obtain ﬁgure 3.2 as in the case of the electron–ion plasmon interaction. ﬁgure 3. ω) e
(3. but with the Coulomb (dashed-dotted) line instead of the dotted phonon line. in the long-wavelength limit q
e (q.40)–(3. Using the RPA expressions (equations (3.43)
=−
where qs = 4m e kF e2 /π is the inverse (Debye) screening radius. In the ‘bubble’ or random phase approximation (RPA).42)) and γq determined in equation (3. Thus. In the analytical form. ω (0) e and obtain Re Im where h(x) = 1 +
0 e (q. Vc (q) = 4πe2 /q 2 . ω)
qs one obtains m e kF q 2 2 π 2 qs (3. ln 2x 1−x
We should also take the Coulomb electron–electron interaction into account because the corresponding vertex is singular in the long-wavelength limit.44)
ω2
. ω) = (0) e (q. we have
e (q. ω)
µ. This leads to a drastic renormalization of the long wavelength behaviour of e .2.

Electrons screen the ˜ small (s/vF bare ion–ion Coulomb repulsion and the residual short-range dynamic matrix has the sound-wave linear dispersion of the eigenfrequencies in the long-wavelength limit. Presented analytically. ω)
(3. the damping is small ( ω). s ω. ω) = 4πe2 q2
(0) e (q. as shown in ﬁgure 3. and the bare electron– phonon interaction γ (q) by a screened one.3 Electrons in metal
The lowest contribution to the electron self-energy is given by two second-order diagrams. which is zero according to equation (3.46)
is the electron dielectric function. ωq )1/2 ˜ where (q. see ﬁgures 3. ω)
81 due to (3.Electrons in metal
The poles of D determine a new phonon dispersion and a damping interaction with electrons: ωi ωq = ˜ (q.49)
Because the ratio of the sound velocity to the Fermi velocity (vF ) is adiabatically √ m e /M).4 corresponds to γsc (q. ω)
(3.28). 0) = 1 +
2 qs q2
(3. ω)γsc (q. ω) = 1 − 4πe2 q2
(0) e (q. Higher-order diagrams are taken into account by replacing the bare ionic plasmon GF by a renormalized one. ˜ vF (3. In the long-wavelength limit. the diagram in ﬁgure 3.3(a) and (b). (q. The imaginary part of ω ˜ determines the damping of the sound.51)
.44).
3.4. ω) = γ (q) + so that γsc (q.45)
(3.48) √ where s = Z kF / 3Mm e is the sound velocity. ω). equation (3.47)
and we obtain the sound wave as the real part of ω: ˜ ωq = sq ˜ (3. The diagram in ﬁgure 3.50)
γ (q) . (q. γsc (q.3(b) is proportional to |γ (q)|2 with q ≡ 0.

47) is appropriate. which is (k.54)
(3.5 for the electron self-energy as a result of the summation of the most divergent diagrams. Second-order electron self-energy. ) = where 2E p = |γ (q)|2 ωq = ˜ ˜ 2i (2π)4 N ˜ dq dω E p G(k − q.44) and (3. one has to consider
.55)
Because Z is of the order of one. where ˜ ˜ γ (q) = ˜ γ (q) (q.52)
and the electron–acoustic phonon vertex γ (q) ωq /N . we introduce the acoustic phonon GF ˜ D(q. respectively. 0) ωi ωq ˜
3/2
.
(3. − ω) D(q.4.51). Screened electron–phonon interaction (dark circle). ω) = ωq ˜2 ω 2 − ωq ˜2 (3.3. ω) 2µ .
Figure 3. the electron–acoustic phonon interaction E p is generally of the order of the Fermi energy.53)
Finally we obtain the diagram in ﬁgure 3.
For low-energy excitations (ω µ) a static approximation of the dielectric function (3. 2 Z (1 + q 2 /qs ) (3.82
Intermediate-coupling theory
Figure 3. Therefore. Instead of the D and γsc given by equations (3.

The main contribution to the integral in equation (3.6.59) To simplify the calculations. The difference between G and G (0) appears to be important only for the damping calculations in a very narrow ˜ region |ξk | ωD s/vF near the Fermi surface. we take E p as a constant and apply the Debye approximation ωq = sq for q < qD where qD ˜ π/a is the Debye momentum. Electron self-energy in the Migdal approximation.and higher-order diagrams with the crossing phonon lines as in ﬁgure 3.58)
and the second order is sufﬁcient in our calculations of .5. 0) is the renormalized Fermi energy.5.57)
and µ = µ + (kF . their contribution is adiabatically small (∼s/vF ) compared with equation (3. Fortunately. The electron energy spectrum. (3. we obtain δ Ek = 2iE p (2π)4 N ˜ ˜ dq dω D(q. ˜ ˜ ξk ≡ E k − µ = vF (k − kF ) + δ E k ˜ where δ Ek = ˜ (k. ξk ) − (kF . We also consider a half-ﬁlled band. kF ≈ π/2a with the energy-independent DOS near the Fermi level N(E F ) = m e a 2 /4π. These diagrams are known as vertex corrections. is determined as the pole of the electron GF. ω)[G (0) (k − q. |k − q| kF . This is appropriate for almost all quasi-particle energies.
fourth. As a result. λ ≈ rs 1 (3. 0) (3. Hence. renormalized by the electron–phonon interaction.60)
.54). we can replace the exact GF in the integral equation (3.59) comes from the momentum region close to the Fermi surface.
(3. as shown by Migdal [48].
Figure 3. ξk − ω) − G (0) (k F − q. The region of large ˜ qs contributes mostly to the integral in equation (3. −ω)].54) by the free GF.Electrons in metal
83
Figure 3.6. Adiabatically small corrections to the electron self-energy.56) (3. The q kF dimensionless coupling constant λ = 2E p N(E F ) is small in this region.54). which are absent in ﬁgure 3. This result is known as Migdal’s ‘theorem’.

and extend the integration to ±∞ for ξ = vF (k − kF ). Then the angular integration in equation (3. according to equation (3. we ﬁnd Re(δ E k ) = −λ ˜ and for low-energy excitations with |ξ | ωD . .64)
˜ dq q ωq sign(ξ ) ˜
˜ ˜ ˜ with qm = |ξ |/s if |ξ | < ωD and qm = qD if |ξ | > ωD .
This means an increase in the effective mass of the electron due to the electron– phonon interaction.67) m where the renormalized mass is m ∗ = (1 + λ)m e .63) (3. corrections are small and the mass is equal to the band mass m e .66)
2 ωD ˜ 2ξ
(3.
(3. kF ˜ ξ = ∗ (k − kF ) (3. The thermodynamic properties of a metal at low temperatures T ωD involve m ∗ but the optical properties in the frequency range ν ωD are determined by high-energy excitations. It is −2πi in the ˜ ﬁrst region and 2πi in the second region. the excitation spectrum of metals has two different regions with two different values of the effective mass. Taking into account that D is an even function of ω.59) yields .61) ˜ ˜ This integral is non-zero only if ξ > ω > 0 or ξ < ω < 0.84
Intermediate-coupling theory
It is convenient to introduce a new variable k = |k − q| instead of the angle between k and q. . ˜ [ξ − ω − ξ + iδ sign(ξ )][ω + ξ − iδ sign(ξ )] (3. For excitations far away ˜ from the Fermi surface (|ξ | ωD ). where.
. we obtain d sin (.62)
The real and imaginary parts of equation (3. respectively: Re(δ E k ) = Im(δ E k ) = Ep 4π 2 vF N Ep 4πvF N
0 qD 0 qm
dq q ωq ln ˜
˜ ωq − ξ ˜ ˜ ωq + ξ ˜
(3.65).68)
Hence. (3.65)
˜ Re(δ E k ) = −λξ . (3.) ∼
−∞ ∞
dξ
˜ ξ
δ Ek =
2E p (2π)2 vF N
qD
˜ |ξ|
dq q
0 0
˜ dω sign(ξ )
ωq ˜2 ω2 − ωq + iδ ˜2
.62) determine the renormalized spectrum and the lifetime of quasi-particles.

69) Im(δ E k ) = sign(ξ ) 3 ˜ if |ξ | > ωD and ˜ Im(δ E k ) = sign(ξ ) ˜ if |ξ | ˜ πλ|ξ |3
2 3ωD
(3. Within the Migdal approximation the electron–phonon interaction does not destroy the Fermi-liquid behaviour of electrons.q. in applying this Hamiltonian to electrons. yields πλωD ˜ (3. From our consideration.71)
˜ where dq is the acoustic-phonon annihilation operator. This
. In the high-energy region ˜ ˜ |ξ | ωD .4 Eliashberg equations
Based on Migdal’s theorem. ˜ ˜ |ξ | ωD . the decay is comparable with the energy and the quasi-particle spectrum loses its meaning. The condensed state is described by a classical ﬁeld. ˜ ˜ ks
k. which is the average of the ψψ or two creation operators product of two annihilation ﬁeld operators + ψ † ψ † . so their frequency already includes the self-energy effect (section 3. The appearance of anomalous averages cannot be seen perturbatively but they should be included in the self-energy diagram.70)
ωD . one should not consider the acoustic phonon self-energy.s
(3.c. In the ˜ intermediate-energy region |ξ | ∼ ωD . These averages are macroscopically large below Tc . In the immediate neighbourhood of the Fermi surface. Acoustic phonons in a metal appear as a result of the electron-plasmon coupling and the Coulomb screening. The Pauli exclusion principle is responsible for the stability of the Fermi liquid. The integral in equation (3. the decay becomes small again in comparison with |ξ | and the quasiparticle concept recovers its meaning. ﬁgure 3. from the very beginning. we have to consider adiabatically small higher-order diagrams.Eliashberg equations
85
Damping shows just the opposite behaviour. it follows that. Eliashberg [36] extended BCS theory towards the intermediate-coupling regime (λ º 1). Going beyond the Migdal approximation.5.64). that is to solve the Hamiltonian of free electrons and acoustic phonons coupled by an interaction: 1 He−ph = √ 2N ˜ γ (q)ωq c† ck−qs dq + H. These expressions describe the rate of decay of quasi-particles due to the emission of phonons. the decay is small compared with the quasi-particle energy |ξ | even for a relatively strong coupling λ ∼ 1 and the concept of well-deﬁned quasi-particles has a deﬁnite meaning.
3.2). applying the Gor’kov formalism.

ωn − ωn ).86
Intermediate-coupling theory
can be done in a compact form by introducing the matrix GF [49]:
s (k.74)
Here the temperature GF of phonons is given by (q. However.7 are expressed in analytical form as a set of Eliashberg equations.2. ωn ) = (iωn τ0 −ξk τ3 )−1 is the normal-state matrix GF.74) as (q. ﬁgure 3. Hence.5) but replacing γ (q) by γ (q)τ3 and the integral over ω by ˜ ˜ the sum over the Matsubara frequencies: ˆ (k.3 are the Pauli matrices.186) but with a microscopically determined pairing potential.7.54). similar to Gor’kov’s equations (2. we do not ﬁnd any anomalous .
0 −i i 0
1 0 0 −1
The generalized equation for the matrix ˆ is given by the same diagram as in the normal state (ﬁgure 3. ωn )τ3 (q. ωn ))−1 −
−1 s (k. Here τ0.
n) = −
ωq ˜2
2 n
+ ωq ˜2
(3. They include all non-crossing (‘ladder’) diagrams in every order of the perturbation theory.74).73)
where (0) (k. let ˜ us consider a momentum-independent γ 2 (q)ωq as in the previous section and ˜ approximate the phonon GF in equation (3.1. The most important difference between equation (3. The diagrams in ﬁgure 3.5.76)
. If we replace s by (0) on the right-hand side of equation (3.74).74) self-consistently. τ )
=−
Tτ ck↑ (τ )c† k↑
† Tτ c−k↓ (τ )c† k↑
Tτ ck↑ (τ )c−k↓
† Tτ c−k↓ (τ )c−k↓
(3. and the normal state equation (3.75)
where n = 2π T n are the Matsubara frequencies for bosons.72)
and the matrix self-energy ˆ (k. ωn ) = (
(0)
(k.
n)
≈ −1
(3. we ﬁnd the ﬁnite anomalous averages. if we solve equation (3. ωn ) = − T (2π)3 N ˜ dq γ 2 (q)ωq τ3 ˜
ωn s (k
− q. τ0 = τ2 = 1 0 0 1 τ1 = τ3 = 0 1 1 0 . ωn )
(3. For simplicity. there are no anomalous averages and no phase transition in the second or in any ﬁnite order of the perturbation theory. ﬁgure 3. is a ﬁnite value of the anomalous GF in the self-consistent solution. (3.

81) with Z = 1 and χ = 0.78)
Transforming the inverse matrix (equation (3.7. and zero otherwise.1. Normal Eliashberg theory. In the last equation.74) leads to the following simpliﬁed Eliashberg equations: [1 − Z (ωn )]iωn = − λT χ(ωn ) = − λT
ωn ∞ −∞
˜ dξ
ωn ∞ −∞
(ωD − |ωn − ωn |)iωn Z (3.Eliashberg equations
*
87
)
Figure 3. ωn )
=−
iZ ωn +
˜ τ1 + ξ k τ3 2 ˜ Z 2 ωn + ξ 2 + | |2
k
(3.80) ˜ Z 2 ω 2 + ξ 2 + | |2
n
˜ dξ
˜ (ωD − |ωn − ωn |)ξ ˜ Z 2 ω 2 + ξ 2 + | |2
n
(3.
and anomalous
+ GFs of the BCS superconductor in the
if | n | < ωD .77)
= iZ ωn τ0 −
˜ τ1 − ξ k τ3 . Substituting equations (3.81)
and (ωn ) = λT
ωn
∞ −∞
˜ dξ
(ωD − |ωn − ωn |) (ωn ) . ˆ is the sum of three Pauli matrices τ0.80) and (3. ˆ (k. respectively. we extend the summation over frequencies to inﬁnity but cut the integral
. ωn ) = i(1 − Z )ωn τ0 + and
−1 s (k.76) and (3.78)) back into the original one yields
s (k.
(3.82)
We can satisfy equations (3.79) into the master equation (3.79)
˜ where ξk = ξk + χ. ωn )
τ1 + χτ3
(3. which in this case are functions of frequency alone. and χ.3 with the coefﬁcients (1 − Z )iωn . 2 ˜ Z 2 ωn + ξ 2 + | |2
(3.

k
(3. the vertex corrections are small and Migdal’s theorem holds in the BCS state as well. As the gap depends on the phonon spectrum.ν )δ(ξk )δ(ξk ) ˜ ˜ k−k ˜
k.
.86) which is an average over the Fermi surface of the interaction matrix element squared multiplied by the phonon spectral density. The critical temperature and the BCS gap are adiabatically small (º ωD ) compared with the Fermi energy and we could worry about the adiabatically small crossing diagrams in ﬁgure 3.
(3. However.84)) can be measured in the tunnelling experiments (section 2.6. ν)ω2 .54) for tanh yields the familiar BCS equation for the order parameter.74) takes properly into account the phonon spectrum. phonons affect the tunnelling I –V characteristics. 1=λ
0 ωD
˜ dξ ˜ ξ 2 + | |2
tanh
˜ ξ 2 + | |2 2T
. Let us deﬁne the phonon density of states F(ω) = 1 N δ(ω − ωqν ) ˜
qν
(3.84)) can be obtained in a direct fashion using the tunnelling conductivity of the superconducting tunnel junctions. Then applying equation (2. which are neglected in the master equation. (3.84)
The DOS (equation (3. which makes them small as in the normal state.4).83)
Hence the Migdal–Eliashberg theory reproduces the BCS results. As a result. The quasi-particle DOS (equation (3. Then the function α 2 F can be calculated to ﬁt the experimental energy dependence of the gap ( ) [54] and compared with the phonon density of states measured independently. which is transformed into the energy-dependent gap in the quasi-particle DOS as ρ( ) =
2
− | ( )|2
. retardation and realistic matrix elements of the electron–phonon interaction in metals [50–52].
(3.85)
and the so-called Eliashberg function α 2 F. α 2 (ω)F(ω) = 1 N(E F )N γ 2 (k − k . Within a more general consideration the master equation (3.88
Intermediate-coupling theory
˜ over |ξ | at ωD .74). The outside regions mainly contribute to the integrals of the crossing diagrams.ν δ(ω − ωk−k . The important feature of the intermediate-coupling theory is the explicit frequency dependence of the order parameter (ωn ). if a similar approximation for the attraction between electrons is adopted. the BCS state is essentially the same as the normal state outside the narrow momentum region around the Fermi surface.

for example in lead. which might be much stronger than the attraction induced by phonons. Nevertheless.89) by the integral: π Tc because Tc ωD
1
≈
∞ π Tc
dω
(3. Then the solution is found in the form (2ωD − |ω|) +
2
(ω) =
(2ωe − |ω|) (|ω| − 2ωD )
(3.87)
where µc = Vc N(E F ). (ωn ) = T can be parametrized as K (ωn − ωn ) = λ (ωD − |ωn − ωn |) − µc (ωe − |ωn − ωn |) (3.
3.74) by the Fourier component ˜ of the Coulomb potential Vc . The Coulomb interaction is non-retarded for frequencies less than ωe .91)
. At T = Tc . we neglect second and higher powers of the order parameter and integrating over ξ obtain (ωn ) = π Tc
ωn
K (ωn − ωn )
(ωn ) .5 Coulomb pseudopotential
The theory of superconductivity has to take account of the Coulomb repulsive correlations between electrons.89)
Let us adopt the BCS-like approximation of the kernel: K (ωn − ωn ) λ (2ωD − |ωn |) (2ωD − |ωn |) − µc (2ωe − |ωn |) (2ωe − |ωn |) and replace the summation in equation (3. |ωn |
(3. ωn − ωn ) in equation (3. we can account for the Coulomb repulsion in the same fashion as for the electron–phonon interaction but ˜ replacing γ 2 (q)ωq D(q.Coulomb pseudopotential
89
for example.90)
ωe . The comparison shows that frequency dependence of the Eliashberg function and the phonon density of states are similar in many low-temperature superconductors with a weak or intermediate electron–phonon coupling. The observation of the characteristic phonon frequencies in tunnelling (usually in the second derivative of the current versus voltage) and the isotope effect are used to verify the phonon-mediated pairing mechanism.88) dξ
ωn
K (ωn − ωn )
(ωn )
2 ωn
+ ξ2
+ | (ωn )|
2
(3. Then the kernel K (ωn − ωn ) in the BCS equation. There is no adiabatic parameter for this interaction because the electron plasma frequency ωe = 4πn e e2 /m e has about the same order of magnitude as the Fermi energy in metals. in neutron scattering experiments.

which works well for simple metals and their alloys.94)
µc (3. if Tc = where 2ωD 1 exp − π λ − µ∗ c (3.62λ) c However. McMillan’s formula predicts Tc ≈ 2 K.97)
.93)
A non-trivial solution of these coupled equations is found. The retarded attraction mediated by phonons acts well after two electrons meet each other.45 λ − µ∗ (1 + 0. estimated from this equation and from the ﬁrst-principle band-structure calculations. clearly too low to explain high Tc values in novel superconductors.04(1 + λ) ωD exp − Tc = . One can formally compute Tc and the gap using the Eliashberg equations (3.74) also in the strong-coupling regime λ > 1.89) yields the following equations for 1. V3 Si and in other A-15 compounds). (3.95) 1 + µc ln(ωe /ωD ) is the so-called Coulomb pseudopotential [53].91) into equation (3. In particular. the critical temperature may rise as µ∗ = c Tc ≈ ωD λ1/2 . Allen and Dynes [56] found that in the extreme strong-coupling limit (λ 1).14 and with the Debye temperature c as high as ωD = 400 K. √ |ξ | º ωD m e /M. This is a remarkable result. In the original papers. ωD
2 µc ln
(3. the discrepancy between the value of λ. This time delay is sufﬁcient for two electrons to be separated by a relative distance.2 :
1
1 − (λ − µc ) ln
1 µc ln
2ωD + π Tc
2
2ωD + π Tc
ωe =0 ωD ωe 1 + µc ln = 0.90
Intermediate-coupling theory
with constant but different values of the order parameter 1 and 2 below and above the cut-off energy 2ωD . The damping due to the Fr¨ hlich interaction dominates outside o this region. It shows that even a large Coulomb repulsion µc > λ does not destroy the Cooper pairs because its contribution is suppressed down to the value about ln−1 (ωe /ωD ) 1. respectively. at which the Coulomb repulsion is small. Computational analysis of the Eliashberg equations led McMillan [54] to suggest an empirical formula for Tc . Substituting equation (3.92) (3.96) 1. Migdal [48] and Eliashberg [36] restricted the applicability of their approach to the intermediate region of coupling λ < 1. 1. The Coulomb correlations also lead to a damping of excitations of the order which is relevant only in a narrow region around the Fermi surface of ξ 2 /µ. With the typical values of λ = 0. 2π (3.5 and µ∗ = 0. exceeds the limit allowed by the experimental and computation accuracy by several times [55]. in materials with a moderate Tc 20 K (like Nb3 Sn.

They found that a dilute Fermi gas cannot remain normal down to absolute zero of temperature even in the case of purely repulsive short-range interaction between the particles.100)
. unconventional Cooper pairs can form with non-zero orbital momentum (section 2. As we shall later discuss in chapter 4. great attention has been paid to a possibility of superconductivity mediated by strong electron–electron correlations without any involvement of phonons.10). the Migdal–Eliashberg theory is based on the assumption that the Fermi liquid is stable and the adiabatic condition µ ωD is satisﬁed. (3.e. k ) < 0 in equation (2. Following Kohn and Luttinger. We also assume that the gas is diluted. and V (r − r ) = 0.6 Cooper pairing of repulsive fermions
The phenomenon of superconductivity is due to the interaction of electrons with vibrating ions. kFr0 1. Using these regions of screened interaction. To understand what is involved. The coupling constant λl is expressed in terms of spherical harmonics V (l) of the Fourier transform of the interaction potential. if |r − r | > r0 . let us consider a simpliﬁed model of spin. V (l)mkF (3. which mediates an effective attractive potential V (k. one should consider the screening of a charge placed in a metal. The critical temperature of unconventional Cooper pairing with the orbital momentum l is found as (see equation (2.98).1 fermions with a weak short-range repulsive interaction 2 between them. In recent years.
3. this assumption cannot be applied in the strong-coupling regime and the proper extension of the BCS theory to λ > 1 inevitably involves small polarons and bipolarons. Kohn and Luttinger [57] pointed out that such possibility exists at least theoretically. For simplicity.99) λl ≡ − 2π(2l + 1) on condition that V (l) is negative.Cooper pairing of repulsive fermions
91
However. their Fermi surface is well deﬁned). It has long been known [58] that if fermions are degenerate (i. the screening produces an oscillatory potential of the form cos(2kFr + ϕ)/r 3 at the distance r from the charge (here ϕ is a constant) which has attractive regions. we choose the repulsive bare potential as V (r − r ) = U with a positive U. Some time ago. if |r − r | r0 .98)
Here we replaced the Debye temperature by the Fermi energy µ because the fermion–fermion interaction is non-retarded.104)) Tc ≈ µ exp − 1 λl . that is the radius of the potential is small compared with the characteristic wave-length of fermions. (3. A system of fermions with purely repulsive short-range forces will inevitably be superﬂuid at zero temperature.

92

Intermediate-coupling theory

**It is easy to estimate V (l) for such a potential, V (l) = 2l + 1 2
**

π

d

0

sin

Pl (cos )V (q)

(3.101)

where q 2 = 2kF (1 − cos ) is about 2kF or less and (qr0 )2 (qr0 )4 4π 3 Ur0 1 − + + Ç[(kFr0 )6 ] . 3 10 40 r a (3.102) Substituting the Fourier transform (3.102) into equation (3.101), we obtain a positive V (l) of the order of V (q) = U dr exp(iq ·r) ≈ V (l)

3 Ur0 (kFr0 )2l .

(3.103)

We conclude that the repulsive bare potential does not produce any pairing. Let us now take screening into account. It is sufﬁcient to consider it perturbatively, if the potential is weak, U µ. The diagrams contributing to the effective interaction up to the second order in U are shown in ﬁgure 3.8. In the analytical form, the Fourier transform of the effective interaction of two electrons on the Fermi surface is given by ˜ V (k, k ) = V (q) − × + × T (2π)3 d p[2V 2 (q) − 2V (q)V (k − p)]

ωn (0) ωn ( p, ωn )

(0) ωn ( p

+ q, ωn ) d p V (k − p)V (k − p)

ωn (0) ωn ( p, ωn )

T (2π)3

(0) ωn ( p

− k − k, ωn ).

(3.104)

Because the density is low (equation (3.100)) we can neglect the q3 dependence of V (q) and take V (q) = 4πUr0 /3 ≡ v in the second-order terms of equation (3.104). Then the diagrams (b) and (c) + (d) cancel each other and the second term in the right-hand side of equation (3.104) vanishes. The remaining last term yields the familiar polarization bubble (equation (3.40)) which ﬁnally leads to ˜ λl ≡ − ˜ V (l)mkF 2π(2l + 1)

π

= λl − λ(−1)l

0

d

sin

Pl (cos ) 1 +

2 4kF − q 2 2kF + q ln 4kF q 2kF − q

(3.105) where λ = (akF /π)2 1. Here the ﬁrst repulsive term is about (akF )(kFr0 )2l , where a = vm e /(2π) is the s-wave scattering amplitude in

Cooper pairing of repulsive fermions

93

D E

G F

I

Figure 3.8. The ﬁrst (a) and second-order diagrams (b–e) which contribute to the screened electron–electron interaction. Note that the non-crossing second-order diagram (f) as well as all higher-order non-crossing diagrams are fully taken into account in the BCS equation (section 3.4) and should not be included in the irreducible scattering vertex.

the Born approximation. The second ‘polarization’ contribution to effective interaction is of the order of (akF )2 . Hence, if the gas parameter is sufﬁciently small, U (3.106) (kFr0 )2l−3 µ the polarization contribution overcomes direct repulsion for all l 2 and it is attractive. For example, for l = 2 (d-wave pairing), the integral in equation (3.105) yields ˜ λ2 = λ 4 (8 − 11 ln 2) ≈ 0.015λ. 105 1 0.015λ (3.107)

The corresponding critical temperature is Tc ≈ µ exp − (3.108)

which is practically zero at any λ < 1. The situation is slightly better when the repulsive potential U is strong (U µ) (i.e. for hard-core spheres) [59]. Partial scattering amplitudes fl in a vacuum are of the order of r0 (kFr0 )2l in this limit. They are small compared with the attractive polarization contribution to the amplitudes ( r0 (kFr0 )) starting from l = 1. Hence, at T = 0, hard-core fermions with repulsive scattering in a vacuum are necessarily in a superﬂuid p-wave state. Calculating the integral in equation (3.105) for l = 1, we obtain 2 ˜ λ1 = λ (2 ln 2 − 1) ≈ 0.15λ 5 (3.109)

H

94 and

**Intermediate-coupling theory
**

1 . 0.15λ

Tc ≈ µ exp −

(3.110)

The Fermi energy could be as large as µ = 104 K but Tc still remains very low because λ 1 in the dilute limit.

Chapter 4 Strong-coupling theory

The electron–phonon coupling constant λ is about the ratio of the electron– phonon interaction energy E p to the half-bandwidth D N(E F )−1 (appendix A). We expect [11] that when the coupling is strong (λ > 1), all electrons in the Bloch band are ‘dressed’ by phonons because their kinetic energy (<D) is small compared with the potential energy due to a local lattice deformation, E p , caused by an electron. If phonon frequencies are very low, the local lattice deformation traps the electron. This self-trapping phenomenon was predicted by Landau [60]. It has been studied in greater detail by Pekar [61], Fr¨ hlich [62], Feynman [63], o Devreese [64] and other authors in the effective mass approximation, which leads to the so-called large polaron. The large polaron propagates through the lattice like a free electron but with the enhanced effective mass. In the strong-coupling regime (λ > 1), the ﬁnite bandwidth becomes important, so that the effective mass approximation cannot be applied. The electron is called a small polaron in this regime. The self-trapping is never ‘complete’, that is any polaron can tunnel through the lattice. Only in the extreme adiabatic limit, when the phonon frequencies tend to zero, is the self-trapping complete and the polaron motion no longer translationally continuous (section 4.2). The main features of the small polaron were understood by Tjablikov [65], Yamashita and Kurosava [66], Sewell [67], Holstein [68] and his school [69, 70], Lang and Firsov [71], Eagles [72] and others and described in several review papers and textbooks [13, 64, 73–76]. The exponential reduction of the bandwidth at large values of λ is one of those features (section 4.3). The small polaron bandwidth decreases with increasing temperature up to a crossover region from the coherent small polaron tunnelling to a thermally activated hopping. The crossover from the polaron Bloch states to the incoherent hopping takes place at temperatures T ≈ ω0 /2 or higher, where ω0 is the characteristic phonon frequency. In this chapter, we extend BCS theory to the strong-coupling regime (λ > 1) with the itinerant (Bloch) states of polarons and bipolarons. 95

96

Strong-coupling theory

**4.1 Electron–phonon and Coulomb interactions in the Wannier representation
**

For doped semiconductors and metals with a strong electron–phonon (e–ph) interaction, it is convenient to transform the Bloch states to the site (Wannier) states using the canonical linear transformation of the electron operators (appendix C): 1 ci = √ eik·m cks (4.1) N k where i ≡ (m, s) includes both site m and spin s quantum numbers. In the new representation, the periodic part of the Hamiltonian (3.17) takes the following form: He = [T (m − m )δss − µδi j ]ci† c j , (4.2)

i, j

where T (m) = 1 N E nk eik·m

k

is the ‘bare’ hopping integral (appendix A). Here i = (m, s) and j = (n, s ). The electron–phonon interaction and the Coulomb correlations acquire simple forms in the Wannier representation, if their matrix elements in the momentum representation depend only on the momentum transfer q, equation (3.21): He−ph =

q,ν,i

ωqν n i [u i (q, ν)dqν + H.c.] ˆ

1 2

(4.3) (4.4)

He−e = Here

Vc (m − n)n i n j . ˆ ˆ

i= j

**1 u i (q, ν) = √ γ (q, ν)eiq·m 2N and Vc (m) = 1 N Vc (q)eiq·m
**

q

(4.5)

(4.6)

are the matrix elements of the electron–phonon and Coulomb interactions, respectively, in the Wannier representation for electrons, and n i = ci† ci is the ˆ density operator. Assuming the interaction matrix elements depend only on the momentum transfer, we neglect the terms in the electron–phonon and Coulomb interactions which are proportional to the overlap integrals of the Wannier orbitals on different sites. This approximation is justiﬁed for narrow-band materials,

Electron–phonon and Coulomb interactions

97

whose bandwidth 2D is less than the characteristic value of the crystal ﬁeld. As a result, in the Wannier representation, the Hamiltonian is H=

i, j

**[T (m − m )δss − µδi j ]ci†c j +
**

q,ν,i 1 2 i= j

ωqν n i [u i (q, ν)dqν + H.c.] ˆ (4.7)

+

Vc (m − n)n i n j + ˆ ˆ

q

† ωqν (dqν dqν + 1/2).

This Hamiltonian should be treated as a ‘bare’ one for metals, where the matrix elements and phonon frequencies are ill deﬁned. In contrast, the bare phonons ωqν and the electron band structure E nk are well deﬁned in doped semiconductors, which have their ‘parent’ dielectric compounds. Here, the effect of carriers on the crystal ﬁeld and on the dynamic matrix is small while the carrier density is much less than the atomic one. Therefore, we can use the band structure and the crystal ﬁeld of the parent insulators to calculate the parameters of the Hamiltonian (4.7). Depending on the particular phonon branch, the interaction constant γ (q, ν) has a different q-dependence. For example, in the long-wavelength limit (q π/a), γ (q, ν) 1 q constant 1 √ q

for optical, molecular (ωq constant) and acoustic (ωq q) phonons, respectively. We can transform the e–ph interaction further using the siterepresentation also for phonons. Replacing the Bloch functions in the deﬁnition of γ (q, ν) (equation (3.20)) by their Wannier series yields γ (q, ν) = − 1 M 1/2 ωqν

3/2 n

e−iq·n eqν · ∇n v(n).

(4.8)

This result is obtained by neglecting the overlap integrals of the Wannier orbitals on different sites and by assuming that the single-ion potential v(r) varies over the distance, which is much larger than the radius of the orbital. After substituting equation (4.8) into equation (4.5) and using the displacement operators (equation (3.10)), one arrives at the following expression: He−ph =

m,n,s

n ms u n · ∇n v(m − n) ˆ

(4.9)

which can also be derived by replacing the ﬁeld operators by the Wannier series in equation (3.8). The site representation of He−ph (equation (4.9)) is particularly convenient for the interaction with dispersionless local modes whose ωqν = ων and eqν = eν are q independent. Introducing the phonon site operators 1 dnν = √ N eiq·n dqν

k

(4.10)

it was then shown that there was no lattice instability but only a small renormalization of the phonon frequencies of the order of the adiabatic ratio.3) o correctly describes the electron self-energy for any value of λ but it should not be applied to further renormalize phonons (see also chapter 3). the electron does not interact with the displacement of the ion.ν † ων gν (m − n)(eν · em−n )n ms (dnν + dnν ) ˆ
(4. Because of this lattice instability. that is gν (0) = 0. ω/µ 1. ˜ The frequency turns out to be zero at λ = 0. The conclusion was that the Fr¨ hlich Hamiltonian (4. the Migdal–Eliashberg theory cannot be applied at λ > 1 for a reason. The inverse (1/λ) expansion technique [11] showed that the many-electron system collapses into a small polaron regime at λ ≈ 1 for any adiabatic ratio. However. But one might expect that the self-consistent Migdal–Eliashberg (ME) theory is still valid in the strong-coupling regime because it sums the inﬁnite set of particular (noncrossing) diagrams in the electron self-energy (chapter 3).5.ν
and He−ph =
n.11) is convenient in modelling the electron–phonon interaction in complex lattices.2 Breakdown of Migdal–Eliashberg theory in the strong-coupling regime
The perturbative approach to the e–ph interaction fails when λ > 1. Migdal [48] and Eliashberg [36] restricted the applicability of their approach to λ < 1. The ‘real space’ representation (4. if the adiabatic Born–Oppenheimer approach was properly applied [77]. which has nothing to do with lattice instability. many authors used ME theory with λ much larger than 1 [50].98
Strong-coupling theory
we obtain in this case un =
ν
eν † (dnν + dnν ) √ 2Mων
† ων (dnν dnν + 1/2)
Hph =
n. Atomic orbitals of an ion adiabatically follow its motion. Therefore. whose orbital it occupies.m. One of the problems with such an extension of ME theory is lattice instability.
4. This regime is beyond
. In fact. for any value of λ. and e m−n ≡ (m − n)/|m − n| is the unit vector in the direction from the electron m to the ion n. The same theory applied to phonons yields the renormalized phonon frequency ω = ω(1 − 2λ)1/2 [48]. As a result.11)
where gν (m) =
dv(m) 1 √ ων 2Mων dm
is a dimensionless force acting between the electron on site m and the displacement of ion n.

The transformation X = (x 1 + x 2 ). which is coupled only with the total density (n 1 + n 2 = 1).4). in the framework of ME theory. 4
(4. ω/µ → 0.3). Setting the potential equal to zero in the solution of the equations of motion restores the translation symmetry but in a new polaron band (section 4. the exact ground state is a self-trapped insulator at any ﬁlling of the band if λ ² 1. This assumption excludes the possibility of the breakdown of the local translation symmetry due to lattice deformation. 2 with one electron and then generalize the result for an inﬁnite lattice with many electrons. To enable the electron relaxation into the lowest polaron states. That leaves the following Hamiltonian to be solved in the extreme adiabatic limit M → ∞:
† † † † H = −t (c1 c2 + c2 c1 ) + (λkt)1/2 ξ(c1 c1 − c2 c2 ) +
kξ 2 .12)
+
i
−
1 + 2M ∂ x i2 2
∂2
kx i2
where t is the nearest-neighbour hopping integral.c.14)
. The major problem with the extension of ME theory to strong coupling originates from its basic assumption that the electron and phonon GFs are translation invariants. τ ) = G(r − r . x i is the normal coordinate of the molecule (site) i and k = Mω2 . i. Hence. The Hamiltonian of the model is H = −t
ij
ci† c j + H. In fact. one has to introduce an inﬁnitesimal translation-non-invariant potential in the Hamiltonian which should be set equal to zero only in the ﬁnal solution for the GFs [78]. The Migdal theorem is exact in this limit. because of the broken translation symmetry in the strong-coupling limit. if the e–ph coupling is sufﬁciently strong (λ ² 1). ξ = x 1 − x 2 allows us to eliminate the coordinate X. To illustrate the point let us compare the Migdal solution of the molecularchain (Holstein) model of the e–ph interaction [68] with the exact solution in the adiabatic limit. j = 1.Breakdown of Migdal–Eliashberg theory
99
ME theory. It cannot be described by a summation of the standard Feynman– Dyson perturbation diagrams even including the vertex corrections (section 3. we consider a two-site case (zero-dimensional limit). One assumes that G(r. + 2(λkt)1/2
i
x i ci† ci (4. First. As in the case of the off-diagonal superconducting order parameter. similar to the absence of the anomalous Bogoliubov averages in any order of the perturbation theory (section 3.3) rather than in the bare electron band. r . τ ) prior to solving the self-energy equation. a small translation symmetry-breaking potential drives the system into a new ground state. one would expect Fermi-liquid behaviour above Tc and the BCS ground state below Tc at any value of λ.13)
The solution is
† † ψ = (αc1 + βc2 )|0
(4.

−0) = i † † (c − c2 )(c1 + c2 ) 2 1 (4. k .23)
which should vanish in the Migdal theory. the Fourier transform of the GF should be diagonal in the Migdal approximation. we obtain G(k. τ )δk. τ ) = G(k.22)
at τ = −0.
(4.5.17) with respect to ξ . the ground state is translationally invariant. (4. Indeed.18)
.21)
where k = 2πn/Na. the theory provides only a symmetric (translation invariant) solution with |α| = |β|.19)
The symmetry-breaking ‘order’ parameter is ≡ β 2 − α2 = [2λ + (4λ2 − 1)1/2]2 − 1 . Hence. so the groundstate energy E 0 and the ground-state displacement ξ0 are obtained by minimizing equation (4.5. β = −α. When λ > 0.17) 4 In the extreme adiabatic limit the displacement ξ is classical. k . ξ = 0.20)
If λ < 0. The site operators are transformed into momentum space as ck = N −1/2
j
c j exp(ika j )
(4. k .100 where
Strong-coupling theory
t [t 2 + ((λkt)1/2 ξ + (t 2 + λktξ 2 )1/2)2 ]1/2 (λkt)1/2 ξ + (t 2 + λktξ 2 )1/2 β= − 2 [t + ((λkt)1/2 ξ + (t 2 + λktξ 2 )1/2 )2 ]1/2 α= and the energy is E=
(4. Precisely this state is the ‘Migdal’ solution of the Holstein model. Then the off-diagonal GF with k = 0 and k = π/a of the two-site chain (N = 2) is given by G(k.5. this
.16)
kξ 2 − (t 2 + λktξ 2 )1/2 . −N/2 < n ≤ N/2. we obtain E 0 = −t λ + and ξ0 = t (4λ2 − 1) λk 1 4λ
1/2
(4. [2λ + (4λ2 − 1)1/2]2 + 1 (4.k . = 0 and E 0 = −t. Calculating this average. G(k.15) (4. If λ ≥ 0. −0) = i 2 (α − β 2 ) 2 (4.

81].1. The only instability which might occur in this regime is the formation of on-site selftrapped bipolarons (section 4. ω/zt [10.24) It becomes smaller than the lattice constant (r p = a/λ) for λ ≥ 1. On-site self-trapped bipolarons form a chargeordered state due to a weak repulsion between them [10].1 1. According to [80]. The asymptotically exact many-particle ground state of the half-ﬁlled Holstein model in the strongcoupling limit (λ → ∞) is ψ=
j ∈B
c†↑ c†↓ |0 j j
(4. That is why a multi-polaron system remains in a self-trapped insulating state in the strong-coupling adiabatic regime. The radius of the self-trapped adiabatic polaron.5. The system collapses into a localized adiabatic polaron trapped on the ‘right-hand’ or ‘left-hand’ site due to the ﬁnite lattice deformation ξ0 .875 and at λ ≥ 0. is readily derived from its continuous wavefunction [79] ψ(x) 1/ cosh(λx/a). if the on-site attractive interaction. Here the j s are B-sites of a
. The latter authors also studied the Holstein two-dimensional and three-dimensional lattices in the adiabatic limit. (4. 2λzt.
solution is not the ground state of the system (ﬁgure 4. the self-trapping of a single electron occurs at λ ≥ 0.4 0.e. The adiabatic solution of the inﬁnite one-dimensional chain with one electron was obtained by Rashba [79] in a continuous (i.3
101
Figure 4. r p .7 0.4 0.92 in two and three dimensions.1).9 Λ 1. The ground-state energy (in units of t.0 0.8 0. no matter how many polarons it has.6).5 0.Breakdown of Migdal–Eliashberg theory
E t.2 1. is larger than the repulsive Hubbard U [14]. The Migdal solution is shown as the broken line. respectively. full line) and the order parameter (thin full line) of the adiabatic Holstein model.8 1. order parameter 0. The generalization to a multi-polaron system on an inﬁnite lattice of any dimension is straightforward in the extreme adiabatic regime. when λ < 0. effective mass) approximation and by Holstein [68] and Kabanov and Mashtakov [80] for a discrete lattice.25)
for any value of the adiabatic ratio. Alternatively.6 0. the Migdal solution is the only solution.

Benedetti and Zeyher [91] applied the dynamical mean-ﬁeld theory in inﬁnite dimensions.6).3. quite a few numerical and analytical studies have conﬁrmed these conclusions (see. Hiramoto and Toyozawa [97] calculated the strength of the deformation potential. Thus. It is an insulating state rather than a Fermi liquid. a self-consistent approach to the many-polaron problem is possible with the ‘1/λ’
. The radius of the acoustic polaron and bipolaron is about the lattice constant. [80–96]). so that the critical value of λ does not very much depend on the number of electrons in this case either. The effect acoustic polaron takes place in the extreme adiabatic limit. For example. Hence. As in the 1/λ expansion technique. the bipolaron superconducting state is essentially different from the BCS superconductor (section 4. 88] applied the gauge-invariant self-consistent method neglecting the momentum dependence of the vertex. sqD of the adiabatic ratio sqD /zt on the critical value of λ was found to be negligible.3 and 4. ω → 0. for example. Hence.102
Strong-coupling theory
bipartite lattice A + B. The non-adiabatic corrections (phonons) allow polarons and bipolarons to propagate as Bloch states in a new narrow band (sections 4.5 < λ < 1. There is a qualitative difference between the ordinary Fermi liquid and the polaronic one. In particular. They found that the transition of two electrons into a self-trapped small bipolaron occurs at the electron–acoustic phonon coupling λ 0.1 Polaron band The kinetic energy is smaller than the interaction energy as long as λ > 1. In contrast. they arrived at the same conclusion about the applicability of the Migdal approach (in [91] the critical value of λ was found to be 1.7.5.5). which transforms electrons into small polarons and bipolarons. Takada [86. there is a large isotope effect on the carrier mass in polaronic metals [82] (section 4. that is half of the critical value of λ at which the transition of the electron into the small zt. In particular. both approaches avoided the problem of broken translation symmetry by using the non-dispersive vertex and GFs as the starting point. under certain conditions. The transition into the self-trapped state due to the broken translational symmetry is expected at 0.3 in the adiabatic limit).3). In the last years. the multi-polaron system is metallic with polaronic or bipolaronic carriers rather than with bare electrons. As a result.3 (depending on the lattice dimensionality) for any electron–phonon interaction conserving the on-site electron occupation numbers. which is expected in the Migdal approximation at any value of λ in the adiabatic limit. the polaron effective mass m ∗ will depend on M (section 4.68)) because λ does not depend on the isotope mass. Likewise.
4. the renormalized (effective) mass of electrons is independent of the ion mass M in ordinary metals (equation (3.5) while there is no carrier mass isotope effect in ordinary metals.3 Polaron dynamics
4.

(4. The electron and phonon operators are transformed as ci = ci exp ˜
q. ν)]dqν − H. The eigenstates ˜ | N = |n i .ν
u i (q. It follows from equation (4. ν)dqν − H. we apply the canonical transformation e S to diagonalize the Hamiltonian.c. ν)|2 ωqν cos[q · (m − n)]
q.i
n i [u i (q. j
[σi j − µδi j ]ci† c j − E p ˆ
i
ni + ˆ
q.ν. that there is an analytical exact solution of a single polaron problem in the strong-coupling limit λ → ∞ [71].33)
.
(4.c. The diagonalization is exact if T (m) = 0 (or λ = ∞): ˜ (4. In the extreme inﬁnite-coupling limit (λ → ∞) we can neglect the hopping term of the transformed Hamiltonian.26) H = e S H e−S where S=−
q.Polaron dynamics
103
expansion technique [98].30)
where σi j = T (m − n)δss exp ˆ
q.ν
† ωqν (dqν dqν + 1/2) +
1 2 i= j
vi j n i n j ˆ ˆ (4.31)
is the renormalized hopping integral depending on the phonon operators and vi j ≡ v(m − n) = Vc (m − n) − 1 N |γ (q. known for a long time.29)
respectively (appendix E).32)
is the interaction of polarons comprising their Coulomb repulsion and the interaction via a local lattice deformation.27)
is such that S † = −S.ν
(4. it changes the boson vacuum. ν)dqν − H.28)
and
˜ dqν = dqν −
i
n i u ∗ (q. which treats the kinetic energy as a perturbation.ν
[u j (q. As a result.] ˆ
(4.29) that the Lang–Firsov canonical transformation shifts the ions to new equilibrium positions. Following Lang and Firsov. The technique is based on the fact.c. n qν are sorted by the polaron n ms and phonon n qν occupation numbers. ν) ˆ i
(4. The energy levels are E = −(µ + E p )
i
ni +
1 2
vi j n i n j +
i= j q
ωqν (n qν + 1/2)
(4. ν) − u i (q. In a more general sense. the transformed Hamiltonian takes the following form: ˜ H=
i. The rest has analytically determined eigenstates and eigenvalues.

yields a negative k-independent correction
.36)
where |0 is the vacuum. n qν |2 ˆ ωqν n qν
k
(4. Up to the second order in the hopping integral. To see how the tunnelling occurs. 0 = √ N c† exp(ik · m)|0 ms
m
(4.37)
q. which depend on the polaron occupation numbers. First we restrict the discussion to a single-polaron problem with no polaron–polaron interaction and µ = 0. we apply the perturbation theory using 1/λ as a small parameter. The second term in equation (4. The middle of the electron band falls by the polaron level-shift E p due to a potential well created by lattice deformation. The ﬁnite hopping term leads to the polaron tunnelling because of degeneracy of the zero-order Hamiltonian with respect to the site position of the polaron.35) D Here D = zT (a). which is linear with respect to the bare hopping T (m).104
Strong-coupling theory
where n i = 0. 1. 2.30). j
σi j ci† c j |k . The phonon frequencies remain unchanged in this limit.38)
where g 2 (m) =
1 2N
|γ (q. ν)|2 [1 − cos(q · m)]
q. quadratic in T (m). one readily calculates the perturbed energy levels. n qν are the excited states of the unperturbed Hamiltonian with one electron and at least one real phonon. describes a polaron-band dispersion. in zero order with respect to the hopping describes localized polarons and independent phonons.39)
is the band-narrowing factor at zero temperature. . 1 and n qν = 0. The third term in equation (4. 0|
i.34) Ep = 2N q. where Ep λ≡ . ν)|2 ωqν . .37).37). 3. (4. ∞. By applying the textbook perturbation theory. The Hamiltonian (4. which are vibrations of ions relative to new equilibrium positions.ν
where |k . 1 |γ (q. . they are given by E(k) = −E p + −
k .ν
(4.n qν
| k. The proper Bloch set of N-degenerate zero-order eigenstates with the lowest energy (−E p ) of the unperturbed Hamiltonian is 1 |k.ν Now let us discuss the 1/λ expansion.
k
=
m
T (m)e−g
2 (m)
exp(−ik · m)
(4. (4. z is the coordination lattice number and T (a) is the nearestneighbour hopping integral. .

As in any second-order correction. the simultaneous emission and absorption of phonons (ﬁgure 4. The polaron half-bandwidth is exponentially reduced. 2 w ≈ De−g (4.3.3) become possible. These transitions shift the middle of the band down without any real charge delocalization. 0|
i.40)
if k = k .2. The electron hops 2 back and forth many times (about eg ) waiting for a sufﬁcient lattice deformation to appear around the site n. The polaron localized in the potential well of depth E p on site m hops onto a neighbouring site n with no deformation around and comes back. It has little to do with the polaron effective mass and the polaron tunnelling mobility because the lattice deformation around m does not follow the electron. this transition shifts the energy down by an amount of about −T 2 (m)/E p . one can raise a concern about its existence in real solids. The emission of a single high-frequency phonon is impossible for any k because of the energy conservation. is understood in ﬁgure 4. Only after the deformation around n is created does the polaron tunnel onto the next site together with the deformation.
. no matter how strong the interaction is. Hence.2 Damping of the polaron band The polaron band is exponentially narrow. Hence.38)) (containing a small exponent). However. These phonons ‘dress’ the electron and coherently follow its motion. j
σi. At zero temperature the perturbation term of the transformed Hamiltonian conserves the momentum because all off-diagonal matrix elements vanish: k. there is no damping of the polaron band at T = 0 caused by optical phonons.41) and it is usually less than the optical phonon energy ω0 (g 2 is of the order of Dλ/ω0 ). 4. see equation (4. 0 = 0 ˆ
(4. ‘Back and forth’ virtual transitions of the polaron without any transfer of the lattice deformation from one site to another.Polaron dynamics
Q
105
(S P
Figure 4.
to the polaron level-shift of the order of 1/λ2 . The origin of this correction. which could be much larger than the ﬁrst-order contribution (equation (4. at ﬁnite temperatures. j ci† c j |k .2.38).

3) becomes more important with increasing temperature. we estimate the matrix element of the two-phonon scattering as Mq q ≈ 1 2 wγ0 n q (n q + 1).43)
where the corresponding matrix element is Mq q =
i.ν
However.106
Strong-coupling theory
∼γZ T TC
N
NTTC
Figure 4. n q . ˆ
For simplicity we drop the phonon branch index ν and consider the momentum independent γ (q) = γ0 and ωq = ω0 . n q − 1. j
k + q − q .q k
−
k+q−q
)
ph
(4.44)
Using this estimate and the polaron density of states. j ci† c j |k. Two-phonon scattering responsible for the damping of the polaron band. N (4. ν)|2 ωqν [1 − cos(q · m)].42)
where T0−1 =
1 N
−1 |γ (q. Expanding σi j -operators in powers of the ˆ phonon creation and annihilation operators. 2T
For high temperatures (T
ω0 /2) the band narrows exponentially: w ≈ De−T /T0 (4.
Moreover. We can estimate the scattering rate by applying the Fermi–Dirac golden rule: 1 = 2π τ |Mqq |2 δ(
q. ν)|2 [1 − cos(q · m)] coth
q.45)
. the two-phonon scattering of polarons (ﬁgure 4. N p (ξ ) ≡ 1 N δ(ξ −
k k)
≈
1 2w
(4. the polaron bandwidth shrinks with increasing temperature because the phonon-averaged hopping integrals depend on temperature (appendix E): σi j ˆ
ph
= T (m − n)δss exp
−
1 2N
|γ (q.3. n q .ν
ωqν . n q + 1|σi. q.

Introducing a normal coordinate at site n as o ξn =
q
(2N Mωq )−1/2 eiq·n dq + H. which we refer as the o small Fr¨ hlich polaron (SFP). Phonons dominate in the scattering at ﬁnite temperatures if the number of impurities is sufﬁciently low.47) τ which is satisﬁed for a wide temperature range T ≤ Tmin ≈ ω0 4 ln γ0 (4.48)
2 about half of the characteristic phonon frequency for relevant values of γ0 .c.50)
we rewrite the e–ph interaction.
4. The polaron transport theory requires a special diagrammatic technique in this region [74.51)
.i
f (m − n)ξn n i .3). the polaron mobility decreases when the temperature increases from zero to Tmin due to an increasing number of phonons. Hence. ˆ
(4. The polaron band is well deﬁned if 1 <w (4. f (m) = N −1
q 3 γ (q)(Mωq )1/2eiq·m
(4. as He−ph =
n. 75]. There is a temperature range around Tmin where the thermal activated hopping still makes a small contribution to the conductivity but the uncertainty in the polaron band is already signiﬁcant [71]. the incoherent thermal activated hopping dominates in the polaron dynamics [66–68. At higher temperatures.3. where it is at minimum.46) τ where n ω = [exp(ω0 /T ) − 1]−1 is the phonon distribution function. due to thermal activated hopping. The optical phonon frequencies are exceptionally high. Let us compare the small Holstein polaron (SHP) formed by the short-range e–ph interaction and a small polaron formed by the long-range (Fr¨ hlich) interaction.3 Small Holstein polaron and small Fr¨ hlich polaron o The narrowing of the band and the polaron effective mass strongly depend on the radius of the electron–phonon interaction [100]. 71] and the polaron states are no longer the Bloch states. about 1000 K or even higher and polarons are in Bloch states in the whole relevant range of temperatures in novel superconductors. the mobility increases above Tmin .Polaron dynamics
we obtain
107
1 4 ≈ wγ0 n ω (1 + n ω ) (4. equation (4.49)
and a ‘force’ between the electron at site m and the normal coordinate ξn . Therefore.
(4.

(4. polarized in a direction perpendicular to the chains [94] (ﬁgure 4. the inter-chain distance is also a and we take a = 1.39E p /ω. The corresponding force is given by f (m − n) = κ . The exact polaron mass in a o
. for dispersionless phonons. using equations (4. is measured in units of the lattice constant. Indeed. (|m − n|2 + 1)3/2 (4.49κ 2/(2Mω3 ) and g 2 = 0. then g 2 = E p /ω. For this longrange interaction.52)
1
3 2Mω0 m
[ f 2 (m) − f (m) f (m + a)]
(4. In general. we obtain E p = 1.34) and (4. a. g 2 = 0. In the nearest-neighbour approximation. To estimate γ . we have g 2 = γ E p /ω with the numerical coefﬁcient γ = 1−
m
f (m) f (m + a) 2 n f (n)
(4. The theory appears almost exact in a wide region of parameters for the Fr¨ hlich interaction. there is no simple relation between the polaron level shift E p and the exponent g 2 of the mass enhancement. This relation depends on the form of the electron–phonon interaction.56)
Thus the effective mass renormalization is much smaller than in the Holstein model.54)
which might be less than 1. |m − n|. roughly m ∗ ∝ (m ∗ )1/2 . f (m) = κδm. Here κ is a constant. let us consider a one-dimensional chain model with a long-range Coulomb interaction between the electron on the chain (×) and ion vibrations of the chain (◦).4).53)
where a is the primitive lattice vector. If the interaction is short range. ωq = ω0 .27κ 2/(2Mω2 ). the effective mass renormalization is given by m ∗ /m = eg
2
where m is the bare band mass and 1/m ∗ = ∂ 2 k /∂k 2 at k → 0 is the inverse polaron mass.39) we obtain Ep = and g2 = 1 2 2Mω0 f 2 (m)
m
(4.55)
Here the distance along the chains.108
Strong-coupling theory
For simplicity we consider the interaction with a single phonon branch and γ (−q) = γ (q). SFP SHP Not only does the small polaron mass strongly depend on the radius of the electron–phonon interaction but the range of the applicability of the analytical 1/λ expansion theory also does.0 (the Holstein model). In general.

. 2
7D.Polaron dynamics
3
109
.

56λ . Such a polaron tunnels with a larger probability than the Holstein polaron due to a smaller relative lattice distortion around two neighbouring sites. the ion displacements. This is not the case for the Holstein polaron. the polaron is almost localized on one site m. ﬁnite-time-step and ﬁnitetemperature errors and allows for an exact (in the QMC sense) calculation of the ground-state energy and the effective mass of the lattice polaron for any electron– phonon interaction described by the Hamiltonian (4. the SFP was found to be much lighter than the SHP.78λ and e1. respectively.5).51). 100].
. In the strong-coupling limit.e. The exponents are remarkably close to those obtained with the Lang–Firsov transformation. Including all
. if the interaction is longranged.
wide region of the adiabatic parameter ω/T (a) and coupling was calculated with the continuous-time path-integral quantum Monte Carlo (QMC) algorithm [94]. in the case of the Fr¨ hlich interaction the transformation is perfectly accurate even in the o moderate adiabatic regime. allows us to generalize this result.
. only a fraction of the total deformation changes every time the polaron tunnels from one site to its neighbour and γ is smaller than 1. while the large Fr¨ hlich polaron (i. This method is free from any systematic ﬁnite-size. If the interaction is short range. proportional to the displacement force f (m − n). At large λ (>1. e0. For a short-range e–ph interaction. the entire lattice deformation disappears at one site and then forms at its neighbour.4λ for ω = 0. spread over a large distance. the size of its wavefunction is the atomic size. A one-dimensional model of a small polaron on the chain interacting with the ion displacements of another chain. The mass ratio m ∗ /m ∗ is FP HP a non-monotonic function of λ. γ = 1 and the polaron is very heavy already at λ ≈ 1. The polaron cloud (i.
Figure 4. Hence. In contrast. Another interesting point is that the size of the SFP and the length over which the distortion spreads are different. m ∗ (λ) o FP is well ﬁtted by a single exponent. the same analytical technique is applied only in the non-adiabatic regime ω/T (a) > 1.e.5).
. which is e0. Hence. 72.4. The fact that the Lang–Firsov transformation is rather accurate for the long-range interaction in a wide region of parameters. at λ < 1) was heavier than the large Holstein o polaron with the same binding energy (ﬁgure 4. Therefore.5T (a). lattice distortion) is more extended than the polaron itself [66. when the polaron tunnels from site to site. Their amplitude at a site n falls with distance as |m − n|−3 in our one-dimensional model. The effective mass of Fr¨ hlich polarons. In contrast.73λ for ω0 = T (a) and e1. ω/T (a) ≤ 1 for any coupling strength.

3)). SFP SHP where the masses are measured in units of the band mass. γ (q) o 1/q).4 Polaron spectral and Green’s functions The multi-polaron problem has an exact solution in the extreme inﬁnite-coupling limit (λ = ∞) for any type of e–ph interaction conserving the on-site occupation numbers of electrons (equation (4.3.57)
For the Fr¨ hlich interaction (i.
phonon branches in the three-dimensional lattice. ν)
. 74. ωq ∼ q) also leads to a lighter polaron in the strong-coupling regime compared with the SHP. the 1/λ perturbation expansion is applied. ν)[1 − cos(q · m)]
q. a short-range interaction with dispersive acoustic phonons (γ (q) ∼ 1/q 1/2. Actually. However. Inverse effective polaron mass in units of 1/m = 2T (a)a 2 . 98.
(4. As an example. 99] 1 2zλ2 1
so that the analytical perturbation theory has a wider region of applicability than one can expect using a semiclassical estimate E p > D. we obtain m ∗ = (m ∗ )γ .
. A lighter mass of SFP compared with the non-dispersive SHP is a generic feature of any dispersive electron–phonon interaction.57 in a cubic lattice and γ = 0. we ﬁnd γ = 0.e. The expansion parameter is actually [72. and γ =
q.255 for the in-plane oxygen hole in cuprates (Part II).ν
γ 2 (q. Holstein [68] pointed out in his original paper that the dispersion is a vital ingredient of the polaron theory. For ﬁnite coupling. 4.ν
γ 2 (q.5.110
Strong-coupling theory
Figure 4.

dynamical mean-ﬁeld approach in inﬁnite dimensions [91]. if the interaction is short-ranged [83]. However.60)
is the free phonon part. E(k) = m T (m) exp(−ik · m) is the bare dispersion in the rigid lattice and Z =
m
T (m)e−g (m) exp(−ik · m) m T (m) exp(−ik · m)
2
(4. The interaction term Hint comprises the polaron–polaron interaction.61)
is the band-narrowing factor (Z = exp(−γ E p /ω)) as discussed earlier). The chemical potential µ includes the polaron level shift −E p .62)
. the bandwidth and the polaron density of states strongly depend on the adiabatic ratio ω/T (a) and on the radius of the interaction. 92.Polaron dynamics
111
the expansion convergency is different for different e–ph interactions. 89. 93]. Applying the canonical transformation we write the transformed Hamiltonian as ˜ (4. It might also include all higher orders in 1/λ corrections to the spectrum independent of k. 85. Keeping this in mind. the analytical theory is practically exact in a wider range of the adiabatic parameter and of the coupling constant for the long-range Fr¨ hlich o interaction. variational calculations [83.58) H = H p + Hph + Hint where Hp =
k
ξ(k)c† ck k
(4. Exact numerical diagonalizations of vibrating clusters. 87. the effective mass. it overestimates the polaron mass by a few orders of magnitude in the adiabatic case (ω/T (a) < 1). equation (4. and the residual polaron–phonon interaction Hp−ph ≡
i= j
[σi j − σi j ˆ ˆ
† ph ]ci c j . However. For simplicity.59)
is the ‘free’ polaron contribution and Hph =
q † ωq (dq dq + 1/2)
(4. and quantum Monte Carlo simulations [94] revealed that the ground state energy (≈−E p ) is not very sensitive to the parameters. we drop the spin and phonon branch indexes. The tunnelling probability is extremely sensitive to the shape of this potential. Here ξ(k) = Z E(k) − µ is the renormalized polaron band dispersion.32). The ﬁrst order in 1/λ perturbation theory is practically exact in the non-adiabatic regime (ω/T (a) > 1) for any value of the coupling constant and any type of e–ph interaction.
(4. let us calculate the one-particle GF in the ﬁrst order in 1/λ. In contrast. A much lower effective mass of the adiabatic Holstein polaron compared with that estimated using ﬁrst-order perturbation theory is the result of the poor convergency of the perturbation expansion owing to the double-well potential [68] in the adiabatic limit.

∞. let us also neglect the polaron–polaron interaction. . n q = 0.112
Strong-coupling theory
We can neglect Hp−ph in the ﬁrst order in 1/λ 1.68) (4.64)
by the use of the Wannier representation and the Lang–Firsov transformation. .67)
ck (t)c† k (4. j (4. If Hint = 0. X i = exp
q
Now.66)
The Heisenberg free-polaron operator evolves with time as ck (t) = ck e−iξk t and ci (t)ci† = 1 N ei(k ·m−k
k .65)
Here the quantum and statistical averages are performed for independent polarons and phonons. It can be written as 1 n|ck |m = √ N e−ik·m n|ci X i |m ˜ ˆ ˜
m
(4. 1. n k = 0. therefore ˆ ˆ ci (t) X i (t) X † ci† j = ci (t)c† j ˆ ˆ X i (t) X † .n ˆ ˆ ci (t) X i (t)c† X † j j
1 N
eik·(n−m) ˆ ˆ + c† X † ci (t) X i (t) }. ω) =
1 2 ∞ −∞
1 2π
∞ −∞
dt ei(ωnm +ω)t
dt eiωt
×{
m.69)
1 = N ci† ci (t) = 1 N
[1 − n(k )]eik ·(m−n)−iξk t ¯
k
n(k )eik ·(m−n)−iξk t ¯
k
. δ(ωnm + ω) = the spectral function is expressed as A(k. The spectral function (equation (D. . . . j j (4.k ·n)
(4.10). and the phonon occupation numbers. the energy levels are E {m} = ˜
k
ξk n k +
q
ωq [n q + 1/2]
(4. 1. 2.10)) is deﬁned by the matrix element n|ck |m .c. To better understand the spectral properties of a single polaron. applying the Fourier transform of the δ-function in equation (D. Here ˆ u i (q)dq − H.63)
and the transformed eigenstates |m are sorted by the polaron Bloch-state ˜ occupation numbers.

dq ] = 1.c.] e[−u j (q)dq −H. Collecting all multiplies in equation (4. hence. t) − u i (q).t )u j (q)]/2 j i
e−αdq e−|α|
2 /2
.c. t) = [1 − cos(q · m) cos(ωq t)] coth ωq + i cos(q · m) sin(ωq t). is obtained by u j (q) in the previous expressions. and. j (q)e−iωq t .] e[−u j (q)dq −H.73) where n ω = [exp(ωq /T ) − 1]−1 is the Bose–Einstein distribution function of phonons. dq (t) = dq e−iωq t and ˆ ˆ X i (t) X † j =
q
exp[u i (q.c.]
(4.72)
Now the quantum and statistical averages are calculated by the use of (appendix E) ∗ † 2 eα dq e−αdq = e−|α| nω (4.t )dq −H. whose commutator [ A. B]/2 (4. Applying once again the same identity yields e[u i (q.75) 2T = exp − 1 2N |γ (q)|2 f q (m − n.
(4. j (q.72). we can apply this identity in equation (4.] = eα
∗d † q ∗ †
×e
[u i (q. The Heisenberg free-phonon operator evolves in a similar way.76)
. The result is replacing u i (q.74)
Here we have used the symmetry of γ (−q) = γ (q). The average X † X i (t) .t )u j (q)−u i (q. the terms ˆ ˆ containing sin(q · m) have disappeared.70) to obtain e[u i (q.] = e(α dq −αdq ) ∗ ∗ × e[u i (q.t )dq −H.c. (4. t)
q
(4.
(4.70)
where u i. t) ˆ ˆ X † X i (t) j = ˆ ˆ X i (t) X † j
∗
.c. we arrive at ˆ ˆ X i (t) X † j where f q (m. which is a j multiplier in the second term in the brackets of equation (4. This average is calculated using the operator identity (appendix E) ˆ ˆ ˆ ˆ ˆ ˆ e A+ B = e A e B e−[ A. t)dq − H.Polaron dynamics
113
where n(k) = [1 + exp ξk /T ]−1 is the Fermi–Dirac distribution function of ¯ polarons.t )u j (q)]/2 where α ≡ u j (q.65).t )u ∗ (q)−u ∗ (q.] exp[−u j (q)dq − H.c. Because [dq . t) = u i.71) ˆ ˆ ˆ ˆ which is applied for any two operators A and B. B] is a † number.

ω) =
l=0
[ Al(−) (k.. it is also applied in the weak-coupling and intermediate regimes (section 4.
q ∞
=Z
l=0
{
q
|γ (q)|2 ei[q·(m−n)−ωq t ] }l (2N)ll!
(4..79)
where Al(−) (k. when coth(ωq /2T ) ≈ 1. ω)]
(4..77) and (4.79) is in the form of a perturbative multi-phonon expansion. Each contribution Al(±) (k..65).68) into equation (4. k and integration with respect to time in equation (4.q l
|γ (q r )|2 (2N)l l!
l l
× 1−n k− ¯
r=1
qr
δ ω−
r=1
ωq r −ξ k−
l r=1
qr
(4.74) yields ˆ ˆ X i (t) X † j where Z = exp − 1 2N |γ (q)|2 .q l
×n k+ ¯
r=1
q r δ ω+
r=1
ωq r −ξ k+
l r=1
qr
.. equation (4.. T ωq .3. n. In the case of the long-range Fr¨ hlich interaction with higho frequency phonons..
(4. ω) = π Z
l 2 r=1 |γ (q r )| (2N)l l! l l
q 1 .66) and performing summation with respect to m. we arrive at [101]
∞
A(k.80)
and Al(+) (k.3).81)
Obviously.79)) is applied to low-energy polaron excitations in the strongcoupling limit. ω) = π Z
l r=1 q 1 . ω) to the spectral function describes the l transition from the initial state k of the polaron band to the ﬁnal state k ± r=1 q r with the emission (or absorption) of l phonons.77)
(4. ω) + Al(+) (k.78)
Substituting equations (4. Differing from the canonical Migdal GF (chapter 3). Then expanding the exponent in equation (4.114
Strong-coupling theory
To proceed with the analytical results.. we consider low temperatures.79). Instead the e–ph coupling leads to the coherent dressing of electrons by phonons because of the
. there is no damping of polaronic excitations in equation (4. The 1/λ expansion result (equation (4.

We note that its spectral density spreads over a wide energy range of about twice the polaron level shift E p . because Z part Aincoh(k. which is beyond the ﬁrst-order 1/λ expansion. It is interesting that there is some k dependence of the incoherent background as well. The ﬁrst (l = 0) k-dependent coherent term arises from the polaron band tunnelling.82)
the higher-momentum integrals..2).80) and (4. The dressing can be seen as the phonon ‘sidebands’ with l ≥ 1. −∞ dω ωp A(k. 2w = 2Z D.3.q l ∞
l 2 r=1 |γ (q r )| (2N)l l! l
=Z
l=0
1 1 l! 2N 1 2N
|γ (q)|2
q
= Z exp
∞
|γ (q)|2 = 1
q
(4.3). which might be larger than the unrenormalized bandwidth 2D in the rigid lattice without phonons. 1 π
∞ −∞ ∞
dω A(k. However. the coherent part shows a dispersion only in the energy window of the order of the polaron bandwidth. The spectral function of the polaronic carriers comprises two different parts. In contrast. ω) with p > 0. in the case of the Fr¨ hlich interaction. the effective mass is less enhanced.81).11)) is satisﬁed. ω) = π Z N
∞ l=1 2l γ0 2l l!
×
k
{[1 − n(k )]δ(ω − lω0 − ξk ) + n(k )δ(ω + lω0 − ξk )} ¯ ¯ (4. ω)] = π Z δ(ω − ξ k ). ω) = Z
l=0 q 1 . The second incoherent ξk = Z E k − µ. we obtain the following expression in this case: Aincoh(k.3.... differ from the exact value by an amount proportional to 1/λ. ω) comprises all the terms with l ≥ 1. Only in the Holstein model with the short-range dispersionless e–ph interaction (γ (q) = γ0 and ωq = ω0 ) is the incoherent part momentum l independent. calculated using equation (4. Acoh (k. ω) = [ A(−) (k.79). 0 0 (4. ω) + A(+) (k. if the matrix element of the e–ph interaction and/or phonon frequencies depend on q. Replacing k ± r=1 q r by k in equations (4. o Z < 1 (section 4. While the major sum rule (equation (D.Polaron dynamics
115
energy conservation (section 4. The difference is due to a partial ‘undressing’ of high-energy excitations in the side-bands. It describes the excitations accompanied by emission and absorption of phonons.84)
.83)
The spectral weight of the coherent part is suppressed as Z 1.

32)). is 4πe2 − |γ (q)|2 ωq .26).4 Polaron-polaron interaction and bipolaron
Polarons interact with each other (equation (4. we have to take into account the dynamic properties of the polaron response function. but cannot overscreen it at large distances. respectively. v(m − n) = e2 > 0.86) yields the repulsive interaction in real space. v(q). polaron GFs are easily obtained using their analytical properties (appendix D). the residual long-range interaction turns out to be rather weak and repulsive in ionic crystals [13].85) Here the ﬁrst term describes the coherent tunnelling in the narrow polaron band while the second k-independent sum is due to the phonon cloud ‘dressing’ the electron. which is described by [62] |γ (q)|2 ω0 = 4πe2 (
−1 −1 ∞ − 0 ) .86) v(q) = q2 ∞ In the long-wave limit (q π/a).79). For example.
(4. where 0 1. comprising the direct Coulomb repulsion and the attraction mediated by phonons. 0 |m − n| (4. ωn ) = Z Z + iωn −ξ k N
∞ l=1 2l γ0 2l l!
k
n(k ) ¯ 1 − n(k ) ¯ + iωn − lω0 − ξk iωn + lω0 − ξk
. (4. The Fourier component of the polaron–polaron interaction. Taking into account both the long-range attraction of polarons owing to the lattice deformations and their direct Coulomb repulsion. Calculating the integral with the spectral density equation (4. Fourier transforming equation (4. we ﬁnd. One can erroneously believe that the long-range Fr¨ hlich interaction becomes o a short-range (Holstein) one due to the screening of ions by heavy polaronic
. As soon as we know the spectral function. that (k. The range of the deformation surrounding the Fr¨ hlich polarons is quite large and their deformation ﬁelds o overlap at ﬁnite density. Considering the polaron–phonon interaction in the multi-polaron system. of the host ionic insulator which are usually well known from experiment. q2
(4.
4. in the Holstein model [102].116
Strong-coupling theory
which has no k-dependence. the temperature polaron GF is given by the integral (D. the Fr¨ hlich interaction dominates in the o attractive part.87)
Here ∞ and 0 are the high-frequency and static dielectric constants.88)
We see that optical phonons nearly nullify the bare Coulomb repulsion in ionic solids.

ω0 ). which is about
2 ω2 (q) º ωe p
me ∞ m∗ 0
2 ωe . In fact. as shown in ﬁgure 3.93)
is the temperature-dependent polaron plasma frequency squared. − k + k+q
(4. we obtain γsc (q. 0) = 1 + where qs = 2πe2 n(2 − n) 3 0T a (2 − n) n 1− 3 2T 2a
2 qs q2 1/2 k
(4. ω0 ) = γ (q) ≈ γ (q) (q. the polaron response becomes dynamic: frequency (ω0 ² (q. we can approximate the polaron distribution function as n(k) ≈ ¯ and obtain (q. For a ﬁnite but rather low w).94)
. we can apply the ordinary bubble approximation (chapter 3) to calculate the dielectric response function of polarons at the frequency : (q. In the absence of bipolarons (see later). when phonons in the polaron cloud follow the polaron motion. we obtain the usual Debye screening at large distances (q → 0). ) = 1 − where ω2 (q) = 2v(q) p
k
ω2 (q) p
2
(4.89)
This expression describes the response of small polarons to any external ﬁeld of the frequency º ω0 . γsc (q.
The polaron plasma frequency is very low due to the large static dielectric constant 1) and the enhanced polaron mass m ∗ . ) = 1 − 2v(q)
k
n(k + q) − n(k) ¯ ¯ .90)
(4.Polaron-polaron interaction and bipolaron
117
carriers. For a temperature larger than the polaron half-bandwidth (T > w).91)
and n is the number of polarons per a unit cell. small polarons cannot screen high-frequency optical vibrations because their renormalized plasma frequency is comparable with or even less than the phonon frequency.4. ( 0 Now replacing the bare electron–phonon interaction vertex γ (q) by a screened one. ω 0 ) (4. In the static limit.92) −
k)
n(k)(
k+q
(4.

they overcome the Coulomb repulsion at a short distance of about the lattice constant. Then.118
Strong-coupling theory
because ω0 > ω p . i. It means that they tunnel together with the entire phonon cloud no matter how ‘thin’ the cloud is.5 and the ﬁrst-order 1/λ expansion is accurate in the whole region of coupling. Let us estimate the coupling constant λ and the adiabatic ratio ω0 /T (a).6. owing to a narrow band.96)
The corresponding ‘phase’ diagram is shown in ﬁgure 4. As a result. where t ≡ T (a) and ω ≡ ω0 . The characteristic attractive potential is V = D/(λ − µc ). At large distance. at which the small ‘bipolaronic’ instability occurs.5). Therefore.95)
Substituting the polaron mass. 8m ∗ a 2
V ≥
(4. The radius of the potential is about a. as one can see in ﬁgure 4. Polarons are too slow to screen high-frequency crystal ﬁeld oscillations. 103]. Nevertheless two large polarons can be bound into a large bipolaron by an exchange interaction even with no additional e–ph interaction other than the Fr¨ hlich one [64.88)). In the Fr¨ hlich interaction.5). m ∗ = [2a 2 T (a)]−1 exp(γ λD/ω0 ). and λ includes the interaction with all phonon branches. ≤ (γ zλ)−1 ln ω0 4z(λ − µc )
(4. two heavy polarons easily form a bound state. Small bipolarons form at λ ≥ µc + π 2 /4z almost independently of the adiabatic ratio. o When a short-range deformation potential and molecular-type (i. Another important point is the possibility of the Wigner crystallization of the polaronic liquid. a bound state of two attractive particles appears. we can say that the carriers are small polarons i ndependent of the value of λ in this case. The Wigner crystallization appears around rs 100 or larger.e. if π2 . there is no sharp transition between o small and large polarons. where µc is the dimensionless Coulomb repulsion (section 3. Hence.
. we ﬁnd T (a) π2 . a small bipolaron. the strong interaction with high-frequency optical phonons in ionic solids remains unscreened at any density of small polarons. Jahn–Teller [104]) e–ph interactions are taken into account together with the Fr¨ hlich o interaction.e. Because the net long-range repulsion is relatively weak. In three dimensions. if the adiabatic parameter is not very small (down to ω/T (a) ≈ 0. polarons repel each other (equation (4. the singular behaviour of γ (q) ∼ 1/q is unaffected by screening. which corresponds to the atomic density of polarons n ≤ 10−6 with 0 = 30 and m ∗ = 5m e . the relevant dimensionless parameter rs = m ∗ e2 / 0 (4πn/3)1/3 is not very large in doped cuprates. This estimate tells us that polaronic carriers are usually in the liquid state.

a superconducting state of small polarons is possible only if λ > µc .32)). (4.5 Polaronic superconductivity
The polaron–polaron interaction is the sum of two large contributions of the opposite sign (equation (4. γ = 0.5) in the case of non-adiabatic carriers. However. The ‘t/ω–λ’ diagram with a small bipolaron domain and a region of unbound small polarons for z = 6. where the Fermi energy is the largest. so that bipolarons are the Cooper pairs formed by two small polarons [11].5.97)
where is the binding energy of the order of an attraction potential |V |. The BCS approach is applied if rb n −1/3 . Here the BCS approach is applied to non-adiabatic carriers with a nonretarded attraction.6. There is no ‘Tolmachev’ logarithm (section 3.61)). where one can still apply the BCS mean-ﬁeld approach.99) 0 < λ − µc
. This condition is opposite to the weak-coupling BCS regime.98) |V | F.Polaronic superconductivity
119
Figure 4. Z < 1.
4.4 and Coulomb potential µc = 0. which puts a severe constraint on the value of the attraction (4. It is generally larger than the polaron bandwidth and the polaron Fermi energy. where bipolarons are ‘extended’ enough and pairs of two small polarons overlap similarly to Cooper pairs. This consideration leaves a rather narrow crossover region from the normal polaron Fermi liquid to a superconductor. because the attraction is non-retarded if F º ω0 . The size of the bipolaron is estimated as rb ≈ 1 (m ∗ )1/2 (4. there is still a narrow window of parameters. F = Z E F (equation (4. Hence.

(4. j
[( σi j ˆ
ph
− µδi j )ci† c j + 1 vi j ci† c† c j ci ] j 2
(4. Hp =
k.5).107)
.100)
written in the Wannier representation.↑ cm.↓ =
tanh
2T
(4.1–0. At least one of them should be attractive to ensure that the ground state is superconducting. one obtains the following expressions:
k 2 2 ξk + 2 k 2 ξk + 2 k
ck.6. As long as λ 1/ 2z.c.99) is satisﬁed.120
Strong-coupling theory
In the case of the Fr¨ hlich interaction.104)
is the order parameter.99). Introducing two order parameters
0 1
= − v0 cm.↓ = − v1 cm.102)
and transforming to the k-space results in the usual BCS Hamiltonian. Hence.↑ c−k. For simplicity.105)
and = − v0 N v1 Nw
k k 2 2 ξk + k (ξk k 2 k 2 ξk + 2 k
0
tanh
2T
2 ξk + 2 k
(4.↓
(4.103)
where ξk =
k
− µ is the renormalized kinetic energy and
k
=
0
−
1
ξk + µ w
(4. Z is about 0.3 for typical values of λ o (ﬁgure 4.]
(4. equation (4.101) (4.106)
1
= −
+ µ)
2 k
tanh
2 2 ξk +
2T
. the region. Applying the Bogoliubov diagonalization procedure (chapter 2).↑ cm+a. is on the border-line in ﬁgure 4. let us keep only the on-site v0 and the nearestneighbour v1 interactions. we can treat the polaron–polaron interaction approximately by the use of BCS theory (chapter 2). we can drop their residual interaction with phonons in the transformed Hamiltonian. ˜ H≈
i. In the crossover region polarons behave like fermions in a narrow band with √ a weak non-retarded attraction.s
ξk c† cks + ks
k
[
† † k c k↑ c−k↓
+ H. If the condition (4.

where the nearest-neighbour contributions to pairing cancel each other.110)
−
1 µ/w
and zv1 2w
w−µ −w−µ w−µ −w−µ
A=
dη dη
tanh(η/2Tc ) η η tanh(η/2Tc ) .106) and (4. rather than the Debye temperature: (ξ ) =
w−µ −w−µ
dη N p (η)V (ξ. With a constant density of states in the polaron band. η) = −v0 − zv1 (ξ + µ)(η + µ)/w2 . if this repulsion is less than the total inter-site attraction. so that superconductivity exists even in the case of onsite repulsion (v0 > 0).14w 1 −
2
µ2 2w exp 2 w v0 + zv1 µ2 /w2
(4. The critical temperature Tc of the polaronic superconductor is determined by two linearized equations (4. w2
zv1 B= 2w
These equations are applied only if the polaron–polaron coupling is small (|v0. (4.14w n(2 − n) exp 2w v0 + zv1 [n − 1]2 .112) so that Tc 1.107) in the limit 0. There is a non-trivial dependence of Tc on doping.
.111)
if v0 + zv1 µ2 /w < 0.109) (4. the Fermi energy F ≈ µ is expressed via the number of polarons per atom n as µ = w(n − 1) (4.1 | < w).1 → 0: 1+ A − where =
0
v0 µ2 + 2 zv1 w Aµ w + (1 + B)
−
1
Bµ w =0
1
=0
(4. η)
(η) 2 η2 +
2 (η)
tanh
η2 + 2T
2 (η)
. (4.108)
Here V (ξ. A non-trivial solution is found at Tc ≈ 1.Polaronic superconductivity
121
These equations are equivalent to a single BCS equation for k = (ξk ) but with the half polaron bandwidth w cutting the integral.113)
Tc have two maxima as a function of n separated by a deep minimum in a halfﬁlled band (n = 1). z|v1 |.

Hpert (equation (4.116)
Let us ﬁrst discuss the dynamics of on-site bipolarons. which are the ground state of the system with the Holstein non-dispersive e–ph interaction. qualitatively different from the normal Fermi liquid and from the BCS superﬂuid. that there are no unbound polarons in the ground state. it has no diagonal matrix elements. 11]. Consideration of particular lattice structures shows that small bipolarons are mobile even when the electron–phonon coupling is strong and the bipolaron binding energy is large [100]. bipolarons repel single polarons and each other.114)
vi j ci† c† c j ci + j
q. small bipolarons are not overlapped. Then the bipolaron dynamics. Hence.ν
† ωqν [dqν dqν + 1 ] 2
(4.122
Strong-coupling theory
4.117) where U is the on-site Coulomb correlation energy (the so-called Hubbard U ).32) is purely the Coulomb repulsion because the phonon-mediated attraction between two polarons on different sites is zero in the Holstein model. The on-site bipolaron is formed if 2E p > U (4. λ − µc Z . a charged Bose liquid.1 On-site bipolarons and bipolaronic Hamiltonian 1 allows for a consistent treatment The small parameter Z /(λ − µc ) of bipolaronic systems [10. Here we encounter a novel electronic state of matter. so energy. rather than into the BCS-like ground state of the previous section.115)
σi j ci† c j . = 2E p − U. Their binding w. j 1 2 i.6. Two or more on-site bipolarons as well as three or more polarons cannot occupy the same site because of the Pauli exclusion principle. When this condition is fulﬁlled. Hence. is larger than the polaron half-bandwidth.116)) destroys bipolarons in the ﬁrst order. The Bose liquid is stable because bipolarons repel each other (see later).6 Mobile small bipolarons
The attractive energy of two small polarons is generally larger than the polaron bandwidth. The inter-site polaron–polaron interaction (4. including superconductivity. j
(4. is described by the use
. ˆ
(4. Hence the polaronic Fermi liquid transforms into a Bose liquid of double-charged carriers in the strong-coupling regime. 4. Under this condition the hopping term in ˜ the transformed Hamiltonian H is a small perturbation of the ground state of immobile bipolarons and free phonons: ˜ H = H0 + Hpert where H0 = and Hpert =
i.

The expression (4.119) refer to conﬁgurations involving two unpaired polarons and any number of phonons. There are no real phonons emitted or absorbed because the bipolaron band is narrow (see later). we have Ep − Ef =
f.121)
.Mobile small bipolarons
123
of a new canonical transformation.118) Ef − Ep
i.119) f |σii ci† ci ˆ . . j
Here E f. 2. On-site bipolaron tunnelling is a twoms ms step transition. The subsequent tunnelling of its ‘partner’ to the same site restores the initial energy state of the system. |p p|σ j j c† c j ˆ j |f
1 1 + Ep − Ef Ep − Ef
S2 couples a localized on-site bipolaron and a state of two unbound polarons on different sites. which has only doubly occupied c† c† |0 or empty |0 sites. exp(S2 ) [10]. j = j
(4.ν
ωqν (n qν − n f ) qν
p
(4. 1 =i Ep − Ef we obtain Hb = H0 − i
m=m .p and | f . . This equation is an explicit deﬁnition of the bipolaron binding energy which takes into account the residual inter-site repulsion between bipolarons and between two unpaired polarons. 3. we obtain ˜ (Hb)ff ≡ (e S2 H e−S2 )ff (Hb)ff ≈ (H0 )ff − ×
1 2 ν i =i . | p are the energy levels and the eigenstates of H0. Hence. It takes place via a single polaron tunnelling to a neighbouring site. . 1.120)
where n qν are phonon occupation numbers (0.p
+
q.119) determines the matrix elements of the transformed bipolaronic Hamiltonian Hb in the subspace | f . However. we can average Hb with respect to phonons. (4.
(4. Replacing the energy denominators in the second term in equation (4.119) by the integrals with respect to time.s n=n . | f with no single (unbound) polarons. the intermediate bra p| and ket | p in equation (4. . ∞). The lowest eigenstates of Hb are in the subspace. Hence. Neglecting the terms of the orders higher than (w/ )2 .s † × c† cm s cns cn s ms ∞ 0 ∞ 0
dt ei(Ef −Ep +iδ)t
T (m − m )T (n − n ) dt e−i
t nn mm
(t). which eliminates the ﬁrst order of Hpert: f |σi j ci† c j | p ˆ (S2 )fp = .

ν
cosh[ωqν ((1/2T ) − it)] sinh[ωqν /2T ] (4. t) − u i (q.c.71) twice yields ˆ ˆ ˆ ˆ X i† (t) X i (t) X † X j = j
q
eβ ×e
∗d † q
e−βdq e−|β|
2 /2
[u i (q. ν)|2 Fq (m. X † and X j commute ˆ ˆ X j also. we ﬁnd
nn mm
(t) = e−g
2 (m−m
) −g 2 (n−n )
e
× exp
1 2N
|γ (q. ν) = γ (−q. m .127) Taking into account that there are only bipolarons in the subspace where Hb operates. m .c. Applying the identity (4.122)
ˆ which is calculated using the commutation relations in section 4. n. t) + u j (q) − u j (q). n )
q.t )−u i (q. n ) = cos[q · (n − m)] + cos[q · (n − m )] − cos[q · (n − m )] − cos[q · (n − m)].c.] (q)−u j (q)]dq −H.t )]dq −H. X i† (t) and ˆ i (t) commute for any γ (q.128)
+
m=m
[t (m − m )b† bm + 1 v(m − m )n m n m ]. (4.124)
ˆ ˆ X† X j = j
where the phonon branch index ν is dropped for simplicity. we ﬁnally rewrite the Hamiltonian in terms of the creation b† = m c† c† and annihilation bm = cm↓ cm↑ operators of singlet pairs as m↑ m↓ Hb = −
m
+
1 2 m
v (2) (m − m ) n m (4.126)
where Fq (m.73).]
(4. so that we can write ˆ ˆ X i† (t) X i (t) =
q
e[u i e[u j
q
(q.t )−u i (q. n. j j
(4.123) (4.124 Here
nn mm
Strong-coupling theory
(t) is a multi-phonon correlator.
nn mm
(t) ≡
ˆ ˆ ˆ ˆ X i† (t) X i (t) X † X j j
(4. m 2¯
. ν) (appendix E). Finally using the average equation (4.4.3.t )][u ∗ (q)−u ∗ (q)]/2−H.125)
where β = u i (q.

Mobile small bipolarons
125
There are no triplet pairs in the Holstein model because the Pauli exclusion principle does not allow two electrons with the same spin to occupy the same site. equation (4.
(4. to calculate t (m) and v (2) (m).133)
When ω0 . m v(m − m ) = 4v(m − m ) + v (2) (m − m ) ¯ (4. we can keep the ﬁrst term only with l = 0 in the bipolaron hopping integral in equation (4.130) yield [105] t (m) = − and v (2) (m) = 2T 2 (m) e−2g
2 (m)
∞ l=0
[−2g 2(m)]l l!(1 + lω0 / ) [2g 2 (m)]l . The bipolaron transfer integral is of the second order in T (m): t (m − m ) = −2iT 2 (m − m )
0 ∞
dt e−i
t
mm mm
(t). if the bipolaron binding energy is large ( ω0 ). Expanding the time-dependent exponents in the Fourier series and calculating the integrals in equations (4. However.131) and (4.128) describes the low-energy physics of strongly coupled electrons and phonons.129)
is the bipolaron–bipolaron interaction including a direct polaron–polaron interaction v(m − m ) and a second order in T (m) repulsive correction: v (2) (m − m ) = 2iT 2 (m − m )
0 ∞
dt e−i
t
mm mm (t).130)
This additional repulsion appears because a virtual hop of one of two polarons of the pair is forbidden if the neighbouring site is occupied by another pair. However.131)
The bipolaronic Hamiltonian (4. If the phonon frequency is dispersionless. the bipolaron half-bandwidth zt (a) is of the order of 2w2 /(z ).134) exp −2g 2 − ω0 ω0
. l!(1 + lω0 / )
(4. = 2E p − Vc < 2g 2 ω0 .132)
2T 2 (m)
e−2g
2 (m)
∞ l=0
(4. we obtain
mm mm (t) mm mm (t)
= e−2g = e−2g
2 (m−m 2 (m−m
) )
exp[−2g 2 (m − m )e−iω0 t ] exp[2g 2 (m − m )e−iω0 t ]
at T ω0 .132).132) yields √ 2g 2(m)ω0 2 2π T 2 (m) √ 1 + ln t (m) ≈ .
(4. Here n m = b† bm is the bipolaron site-occupation operator. In this case. (4. ω0 < equation (4. this limit is not realistic because < 2g 2 ω0 .126). the bipolaron bandwidth dramatically decreases proportionally 2 to e−4g Ò 1 in the limit → ∞. We use the explicit form of the multi-phonon correlator. In a more realistic regime.

in a long-range interaction.25 (part 2).2 eV in novel superconductors like oxides and doped fullerenes. There are two reasons for the lowering of its mass with an increasing radius for the e–ph interaction. In contrast. and the renormalization of its mass. where the numeric coefﬁcient γ depends on the radius of the interaction.42)). Thus the small polaron mass could be ≤10m e where a Holstein-like estimate would yield a huge mass of 10 000m e. Optical phonon frequencies are about 0. the entire lattice deformation disappears and then forms at another site. the mass enhancement exponent of on-site bipolarons in equation (4.3). E p . This longer-range interaction leads to a lighter polaron in the strongcoupling regime (section 4. in the framework of the Holstein model. this results in a dramatic lightening of the polaron since γ enters the exponent.135) t (m) ∝ √ exp − 2T T for T > ω0 . when the polaron moves between the nearest lattices sites. v (2) D 2 / . which is typically of the order of 1 eV or higher. Therefore. Together with the direct Coulomb (2) ensures the stability of the bipolaronic liquid repulsion. m ∝ exp (γ E p /ω0 ). which is unscreened in polaronic systems (section 4.134) is rather large (² exp(2E p /ω0 ) > 150). Therefore. Indeed. The Fr¨ hlich interaction with o optical phonons. Therefore.126
Strong-coupling theory
In contrast. For a shortrange e–ph interaction. there exists a generic relation between the polaron stabilization energy.6. Indeed. only a fraction of the total deformation changes every time the polaron moves and γ could be as small as 0. The bipolaron also becomes much lighter. is much stronger. The lower mass has important consequences because lighter polarons are more likely to remain mobile and less likely to trap on impurities. if the e–ph interaction is long range. While the polaron band collapses with increasing temperature (equation (4. so that on-site bipolarons could hardly account for high values of the superconducting critical temperature [100].1–0. the bipolaron band becomes wider [106]: Ep + 1 (4. The ﬁrst one is the same as in the case of the
. γ = 1 and polarons and on-site bipolarons are very heavy for the characteristic values of E p and ω0 . 4.2 Inter-site bipolaron in the chain model On-site bipolarons are very heavy for realistic values of the on-site attractive energy 2E p and phonon frequencies. The high-temperature behaviour of the bipolaron bandwidth is just the opposite to that of the small polaron bandwidth.4).3. the polaron is heavy because it has to carry the lattice deformation with it. Clearly. the same deformation that forms the polaron itself. But the Holstein model is not a typical model.133)) has no small exponent in the limit → ∞. to bind two polarons on a single site 2E p should overcome the on-site Coulomb energy. the second order v against clustering. the bipolaron–bipolaron repulsion (equation (4.

Second-order inter-site bipolaron tunnelling is shown by arrows. respectively.3). j ) = g0 [δi. In the momentum representation.134). j.3 (ﬁgure 4.3.137)
.3.7.136)
+ ω0
i
[di† di
+ 1/2] + U
j
n j ↑n j ↓ ˆ ˆ
in the site representation for electrons and phonons (section 4. One can further simplify the chain model by placing ions in the interstitial sites located between the Wannier orbitals of one chain and allowing for the e–ph interaction only with the nearest neighbours of another chain. in contrast with on-site bipolarons.3. j )n j s (di† + di ) ˆ (4.s c j s + H. where g(i. equation (4.s
g(i. j + δi. We can view the EHHM as the simplest model with a longer range than the Holstein interaction.c. Following Bonˇ a and Trugman [107]. The Coulomb interaction is represented by the on-site Hubbard U term.s
[c†+1. we ﬁrst consider a single bipolaron in the chain model c of section 4.1).7.] + ω0 j
i. This model is referred to as the extended Holstein–Hubbard model (EHHM) [107].7). with √ γ (q) = g0 2(1 + eiqa ) (4. The model Hamiltonian is H = T (a)
j. j are integers sorting the ions and the Wannier sites. To illustrate the essential dynamic properties of bipolarons formed by the longer-range e–ph interaction let us discuss a few simpliﬁed models. j +1 ] and i .6. the model is a one-dimensional case of the generic Hamiltonian (4.
single polaron discussed earlier. the o EHHM lacks a long-range tail in the e–ph interaction but reveals similar physical properties. Simpliﬁed chain model with two electrons on the chain interacting with nearest-neighbour ions of another chain. as shown in ﬁgure 4. already tunnel coherently in the ﬁrst-order in T (m) [100] (section 4.Mobile small bipolarons
127
Figure 4. The second reason is the possibility to form inter-site bipolarons which. In comparison with the Fr¨ hlich model of section 4. which tunnel only in the second order.7). in certain lattice structures.

In the case when U < 2g0 ω0 . m ∗ SHP 2 exp(E p /ω0 ). when U > 2g0 ω0 . j +1. g2 =
2 g0 a π π/a −π/a 2 dq [1 − cos2 qa] = g0
(4. j
. Its binding energy 2 = 2g0 ω0 (4. the inter-site bipolaron has four spin states. j +1 follows. one singlet S = 0 and three triplet states. Hence. Differing from the on-site singlet bipolaron. the inter-site bipolaron has the lowest energy. j (equation (4.
(4. the EHHM has the numerical coefﬁcient γ = 1/2. S = 1.138)
for the polaron level shift. when one of the electrons within the pair hops to the left (right) and then the other j +2. with different z-components of the total spin. This tunnelling involves the multi-phonon correlation function j +1. the inter-site bipolaron also tunnels only in the second order in T (a).32).140)
π/a −π/a 2 dq [1 + cos qa] cos qa = −2g0 ω0
v(a) = −
2 2g0 ω0 a
π
for the on-site and inter-site polaron–polaron interactions.141)
scales as the square root of the small Holstein polaron mass. and the polaron mass m∗ EHP exp Ep 2ω0 (4.126)): 2 j +2. we obtain the bipolaron mass as 2E p exp m ∗∗ (4.139)
for the mass enhancement exponent and
2 v(0) = U − 4g0 ω0
(4. It scales as (m ∗ /m)4 but is much smaller than the on-site bipolaron EHP mass in the Holstein model.144) does not depend on U . we obtain Ep =
2 g0 ω0 a π π/a −π/a 2 dq [1 + cos qa] = 2g0 ω0
(4. respectively.142)
Using expression (4. ±1. (4. In this regime the bipolaron binding energy is
2 = 4g0 ω0 − U. which scales as (m ∗ /m)4 . the on-site bipolaron has the lowest energy because |v(0)| > |v(a)|. Using equations (4.128
Strong-coupling theory
and ω(q) = ω0 . Sz = 0. j +1 = e−2g0 . In the chain model (ﬁgure 4.34). SHB SHP 2 In the opposite regime.132) for the bipolaron hopping integral.39) and (4.143) EHB ω0 if ω0 . m ∗∗ exp(4E p /ω0 ).7).

The implicitly present inﬁnite Hubbard U prohibits double occupancy and removes the need to distinguish the fermionic spin.ν † (dmν dmν + 1 ).146)
The e–ph term is written in real space. in order to form stable bipolarons. the e–ph interaction has to be strong enough to overcome the Coulomb repulsion at short distances. In general.146) is of considerable complexity.ν † † Vc (n − n )cn cn cn cn
+ ω0 + ω0
(4. the kinetic energy is a perturbation and the model can be grossly
. which o explicitly includes the inﬁnite-range Coulomb and electron–phonon interactions. We see that the inter-site bipolaron in the chain model is lighter than the on-site bipolaron but still remains much heavier than the polaron. the system is a charge-segregated insulator and at a stronger Coulomb repulsion. In this case. bipolarons repel each other and propagate in a narrow band. there is a potential possibility for clustering. the many-body model equation (4. 2 m. which is more convenient when working with complex lattices. the inter-site bipolaron mass enhancement is m ∗∗ EHB T −2 (a) exp Ep ω0 m∗ EHP m
2
129
(4. the Hamiltonian is written as H=
n=n † T (n − n )cn cn + n=n † † gν (m − n)(eν · em−n )cn cn (dmν + d mν ) n=m.Mobile small bipolarons
Hence. More precisely. if it is onedimensional. Introducing spinless fermion operators cn and phonon operators dmν .3 Superlight inter-site bipolarons Any realistic theory of doped ionic insulators must include both the longrange Coulomb repulsion between carriers and the strong long-range electron– phonon interaction. In this parameter window. we are interested in the limit of the strong e–ph interaction.6. At a weaker Coulomb interaction. Let us now apply a generic ‘Fr¨ hlich–Coulomb’ Hamiltonian. 4. there is a certain window of Vc /E p inside which the clusters are unstable but bipolarons form nonetheless. the system is the Fermi liquid or the Luttinger liquid. the inclusion of the longrange Coulomb repulsion is critical in ensuring that the carriers would not form clusters. to a particular lattice structure [108].145)
in the inﬁnite Hubbard U limit (U → ∞). The inclusion of the Coulomb repulsion Vc makes the clusters unstable. However. Indeed. From a theoretical standpoint. Since the e–ph interaction is long range.

147) is the polaron–polaron interaction: v(n − n ) = Vc (n − n ) − Vph (n − n ) where Vph (n − n ) = 2ω0
m. the result is ˆ t (n − n ) ≡ g 2 (n − n ) =
m. where bipolarons already tunnel in the ﬁrst order in T (n). When T º ω0 . is due to displacements of common ions by two electrons.151)
gν (m − n)(eν · em−n ) × [gν (m − n)(eν · e m−n ) − gν (m − n )(eν · e m−n )].
.149)
gν (m − n)gν (m − n )(eν · em−n )(eν · em−n ).ν
σnn ˆ
ph
= T (n − n ) exp[−g 2 (n − n )]
(4. Finally.147)
The last term describes the energy gained by polarons due to the e–ph interaction. Vph . so that σnn can be averaged over phonons.ν † † gν (m − n)(eν · e m−n )cn cn (dmν − d mν ).150)
This term is a perturbation at large λ. E p is the familiar polaron level shift: E p = ω0
mν 2 gν (m − n)(eν · em−n )2
(4. n
+
n=n
v(n − n
− Ep
(4. the transformed hopping operator σnn in the ﬁrst term in ˆ equation (4.130
Strong-coupling theory
simpliﬁed using the canonical transformation of section 4.
The phonon-induced interaction.148)
which is independent of n.
[gν (m − n)(eν · e m−n ) (4.ν
(4.147) is given by σnn = T (n − n ) exp ˆ
m. The third term on the right-hand side of equation (4.ν † − gν (m − n )(eν · e m−n )](dmα − d mα ) .
The transformed Hamiltonian is ˜ H = e−S H e S =
n=n † σnn cn cn + ω0 ˆ mα † † )cn cn cn cn † (dmν dmν + 1 ) 2 † cn cn . Here we consider a particular lattice structure (ladder).3 in the Wannier representation for electrons and phonons: S=
m=n.

consequently. assuming that all sites are isotropic twodimensional harmonic oscillators. if otherwise it will be repulsive. If the ion is an isotropic harmonic oscillator. it generates an additional inter-polaron attraction because the distant ions have small angle φ.149).149) more fully. The overall sign and magnitude of the interaction is given by the lattice sum in equation (4. And. the full interaction becomes negative and the polarons form pairs. If the angle φ between f 1 and f 2 is less than π/2 the polaron–polaron interaction will be attractive. Here s is the ion’s force constant. Here the last term should be 1 2 interpreted as an ion-mediated interaction between the two carriers. the long-range nature of the e–ph interaction serves a double purpose. The total energy of the carrier–ion pair is − f 2 /(2s). When a carrier (electron or hole) acts on an ion with a force f . This is precisely the summand in equation (4. It depends on the scalar product of f 1 and f 2 and. the carrier density and the lattice geometry. Thus.148) and (4. on the relative positions of the carriers with respect to the ion. One should also note that. First. The real-space representation allows us to elaborate the physics behind the lattice sums in equations (4. as we assume here. consequently. Vc (n) ≡ Vc . and for
.8(b)). In general. then the following simple rule applies.8(a). some ions will generate attraction and some repulsion between polarons (ﬁgure 4. The many-particle ground state of H0 depends on the sign of the polaron– polaron interaction. the evaluation of which is elementary. see ﬁgure 4.151) and (4. we also adopt the nearestneighbour approximation for both interactions. while repulsive interaction enhances the mass.
When Vph exceeds Vc . This additional attraction helps to overcome the direct Coulomb repulsion between the polarons. the bipolaron mass). ω0 2 (4. the mass renormalization exponent can be expressed via E p and Vph as follows: g 2 (n − n ) = 1 1 E p − Vph (n − n ) .153) v(n − n
† † )cn cn cn cn
Hpert =
n=n
† t (n − n )cn cn . Now consider two carriers interacting with the same ion. For simplicity.152)
Now phonons are ‘integrated out’ and the polaronic Hamiltonian is H p = H0 + Hpert H0 = − E p
n † cn cn + n=n
(4.149).148) expressed via dimensionless coupling constants. an attractive interaction reduces the polaron mass (and.9(a). according to equation (4. gν (l) ≡ g. the Fr¨ hlich o interaction makes the bipolarons lighter.Mobile small bipolarons
131
By comparing equations (4. second. Here we consider the zig-zag ladder in ﬁgure 4. it displaces the ion by some vector x = f /s. The ion displacement is x = ( f 1 + f 2 )/s and the energy is − f 2 /(2s) − f 2 /(2s) − ( f 1 · f 2 )/s.152).

z = 4. we ﬁnd E p = 4g 2ω0 t = TNN exp − 7E p 8ω0 3E p t = TNN exp − 4ω0 (4.154)
where cn and pn are polaron annihilation operators on the lower and upper sites of the ladder.]. (4.c.
the hopping integrals. so that the one-particle energy spectrum E 1 (k) is found from det 2t cos k − E p − E 1 (k) t (1 + e−ik ) t (1 + eik ) = 0. 2t cos k − E p − E 1 (k) (4. (b) A mixed situation.132
Strong-coupling theory
Figure 4.148). T (m) = TNN for l = n = m = a. (4.8. the two polarons (ﬁlled circles) deform the lattice more effectively than separately.154) into momentum space yields Hp =
k † † (2t cos k − E p )(ck ck + pk pk ) + t k † [(1 + eik ) pk ck + H.149) and (4. Hereafter we set the lattice period a = 1. respectively (ﬁgure 4.156)
A linear transformation of ck and pk diagonalizes the Hamiltonian. and zero otherwise. Then.155)
.c. There are four nearest neighbours in the ladder. Using equations (4. An effective attraction occurs when the angle φ between x1 and x2 is less than π/2. Ion 1 results in repulsion between two polarons while ion 2 results in attraction. the one-particle polaronic Hamiltonian takes the form Hp = − E p
n † † (cn cn + pn pn ) † † † † [t (cn+1 cn + pn+1 pn ) + t ( pn cn + pn−1 cn ) + H. (a) Together.152).
The Fourier transform of equation (4.] n
+
(4.157)
. The mechanism for the polaron–polaron interaction.9(b)).

when Vc < E p /2 and. The attractive interaction is provided via the displacement of the lattice sites. There are two such nearest neighbours for the inter-site bipolaron of type A or B (ﬁgure 4. the two polarons are bound into an inter-site bipolaron of types A or B.9(c)) but there is only one common nearest neighbour for bipolaron C (ﬁgure 4. (c) two degenerate bipolaron conﬁgurations A and B. and v = Vc − E p /4.
There are two overlapping polaronic bands. When Vc > E p /2. It is quite remarkable that bipolaron tunnelling in the ladder already appears
.9(d)). The nearest-neighbour interaction (equation (4. if two polarons are on different sides of the ladder.Mobile small bipolarons
133
Figure 4. and (d) a different bipolaron conﬁguration. (b) two types of polarons with their respective deformations. v < 0. consequently. One-dimensional zig-zag ladder: (a) initial ladder with the bare hopping amplitude T (a). However. Let us now place two polarons on the ladder. C. which are the common nearest neighbours to both polarons.149)) is v = Vc − E p /2. whose energy is higher than that of A and B. E 1 (k) = −E p + 2t cos(k) ± 2t cos(k/2) with effective mass m ∗ = 2/|4t ± t| near their edges. there are no bound states and the multi-polaron system is a onedimensional Luttinger liquid.9. if both polarons are on the same side.

there is no ﬁrst-order kinetic energy contribution to their energy.1 and from the inter-site chain bipolaron of section 4. there is a possibility of clusterization. In models with strong inter-site attraction. t is Ep ≡ 2E 1 (0) − E 2 (0) = − Vc .c. is m ∗∗ = 4m ∗ 1 + 0. m ∗∗ = 2/t .162)
for any n ≥ 3. 6n − 12 (4. the lowest energy of n polarons placed on the nearest neighbours of the ladder is found as E n = (2n − 3)Vc − 6n − 1 Ep 4 (4. The result of such a projection is a bipolaronic Hamiltonian: Hb = (Vc − 5 E p ) 2
n † [ A † A n + Bn Bn ] − t n n † † [Bn An + Bn−1 An + H.6. Indeed. (4. providing a superlight bipolaron.158)
where An = cn pn and Bn = pn cn+1 are inter-site bipolaron annihilation operators and the bipolaron–bipolaron interaction is dropped (see later). E 2 (k) = Vc − 5 E p ± 2t cos(k/2) 2 (4. This fact combines with a weaker renormalization of m ∗ .161)
The numerical coefﬁcient 1/8 ensures that m ∗∗ remains of the order of m ∗ even at large E p .159)
with a combined width 4|t |. The bipolaron binding energy in zero order with respect to t. There are no resonating states for an n-polaron conﬁguration if n ≥ 3. (4.6. equation (4.163)
. ‘Odd’ clusters are stable if Vc < and ‘even’ clusters are stable if Vc < n−1 E p. Similar to the two-particle case described earlier.] (4. we can project the Hamiltonian onto the subspace containing A.25 exp Ep 8ω0 .155). This case is different from both the on-site bipolarons discussed in section 4.2. Therefore. B and empty sites. They are coupled by Hpert.164) n Ep 6n − 10 (4. the lowest-energy conﬁgurations A and B are degenerate. E n should be compared with the energy E 1 +(n −1)E 2/2 of far separated (n − 1)/2 bipolarons and a single polaron for odd n ≥ 3.160) 2 The bipolaron mass near the bottom of the lowest band. Its Fourier transform yields two bipolaron bands.134
Strong-coupling theory
in the ﬁrst order with respect to a single-electron tunnelling. where the bipolaron tunnelling was of the second order in T (a). Neglecting all higher-energy conﬁgurations. or with the energy of far separated n bipolarons for even n ≥ 4.

bm .7.9) results in a bipolaron bandwidth of the same order as the polaron one. and annihilation. then immobile bound clusters of three and 8 8 more polarons could form. their operators commute as bm b† + b† bm = 1 m m bm b† m − b† m bm = 0
y
(4. the polaron–polaron interaction is always repulsive (section 4. One should note that at distances much larger than the lattice constant. Sz = 0. 4. stripes or strings is impossible. inter-site bipolarons tunnel via a one-particle hopping rather than via simultaneous two-particle tunnelling of onsite bipolarons. In the subspace of pairs and empty sites. Keeping this in mind. ±1. bipolaron operators as m m↑ m↓ Hb =
m=m
[t (m − m )b† bm + 1 v(m − m )n m n m ] m 2¯
(4.3 and part 2). we obtain a window of parameters
3 8 Ep
< Vc < 1 E p 2
(4. Outside the window. If Vc is less than 3 E p .167) (4. the ladder is either charge-segregated into ﬁnite-size clusters (small Vc ) or it is a liquid of repulsive polarons (large Vc ).168)
for m = m .166)) to both on-site and inter-site bipolarons. There are additional (spin) quantum numbers S = 0.166)
where v(m − m ) is the bipolaron–bipolaron interaction. b† = c† c† .165)
where the ladder is a bipolaronic conductor.
x b † = Sm − iSm m
(4. implying that the site index m is the position of the centre of mass of a pair. the Hamiltonian of electrons strongly coupled with phonons is reduced to the bipolaronic Hamiltonian written in terms of creation. which should be added to the deﬁnition of bm for the case of inter-site bipolarons.7 Bipolaronic superconductivity
In the subspace with no single polarons.
4. This ‘crab-like’ tunnelling (ﬁgure 4. n m = b† bm .Bipolaronic superconductivity
135
As a result. and even to more extended non-overlapping pairs. we ﬁnd that bipolarons repel each other and single polarons at Vc > 3 E p . Combining the condition of bipolaron formation and that of the instability of larger clusters.1 Bipolarons and a charged Bose gas Bipolarons are not perfect bosons.6. 1.169)
. and ¯ m the position of the middle of the bipolaron band is taken as zero. Also in some lattice structures (section 4. we can apply Hb (equation (4.4) and the formation of inﬁnite clusters. This makes useful the pseudospin analogy [10].

we transform the bipolaronic Hamiltonian into the anisotropic Heisenberg Hamiltonian:
Hb =
m=m
z z x x [ 1 vmm Sm Sm + tmm (Sm Sm + Sm Sm )]. 109]).172)
if the bipolaron density n b is conserved. Replacing bipolarons by spin operators. in our case the total ‘magnetization’ is ﬁxed.174) into equations (4.1 Heisenberg Hamiltonian 2 (4.171)
This Hamiltonian has been investigated in detail as a relevant form for magnetism and also for quantum solids like a lattice model of 4 He.173) and (4.168): β0 = 1 β1 = −1 β2 = √ 3 1 + .171) cannot be solved analytically. when they are on different sites and their fermionic internal structure. Sm = 1 corresponds 2 2 2 z = − 1 to a site occupied by the bipolaron. However.y.174)
where
† † am am − a m am = δm. 2¯
y
y
(4. 1 N
z Sm m
=
1 − nb 2
(4.
(4.176)
.3 .167) and (4.167) and (4.168)) make the problem hard but 1 (for a complete phase not in the limit of low atomic density of bipolarons. n b diagram of bipolarons on a lattice see [10.136 and
Strong-coupling theory
b† bm = m
1 2 z − Sm
(4.m .z = 1 τ1. In this limit we can reduce the problem to a charged Bose gas on a lattice [110]. to an empty site m and Sm 2 Spin operators preserve the bosonic nature of bipolarons. 2 6 (4.173)
b† = m
k=0
† k βk (am )k+1 am
(4. The complicated commutation rules for bipolaron operators (equations (4.170)
z with the pseudospin 1 operators S x. while in those cases the magnetic ﬁeld is an independent thermodynamic variable. The spin.2.175)
The ﬁrst few coefﬁcients βk are found by substituting equations (4. Let us transform the bipolaronic † Hamiltonian to a representation containing only the Bose operators am and am deﬁned as
∞
bm =
k=0 ∞
† k+1 βk (am )k am
(4.

181)
is the dynamic part.180)
dr
¯ dr v(r − r )
†
(r)
†
(r ) (r ) (r)
(4.129)).Bipolaronic superconductivity
We also introduce bipolaron and boson 1 (r) = √ N 1 (r) = √ N -operators as δ(r − m)bm
m
137
(4. Here t (r − r ) =
k ∗∗ ik·(r−r ) k e
1 v(r − r ) = ¯ N vk = ¯ and
vk eik·(r−r ) ¯
k
¯ m=0 v(m) exp(ik·m) is the Fourier component of the dynamic interaction
∗∗ k
=
m=0
t (m) exp(−ik · m)
(4.and higher-body collisions. (4.178)
δ(r − m)am .177) (4.179)
Then we write the bipolaronic Hamiltonian as Hb = where Hd =
1 2
dr
dr
†
(r)t (r − r ) (r ) + Hd + HK + H (3)
(4. HK = 2 dr dr t (r − r ) N × [ † (r) † (r ) (r ) (r ) +
†
(r)
†
(r) (r) (r )].182)
is the kinematic (hard-core) part due to the ‘imperfect’ commutation rules. Because v already has a short-range part v (2) (equation (4. (4. the kinematic ¯
.
m
The transformation of the ﬁeld operators takes the form: (r) = 1 −
† (r)
(r)
N
+
(1/2 +
√
3/6) † (r) (r) (r) + ··· N2
(r). and H (3) includes three.183)
is the bipolaron band dispersion. In the dilute limit (n b 1) only two-particle interactions are essential which include the short-range kinematic and direct density–density repulsions. H (3) contains powers of the ﬁeld operator higher than four.

ψ(r. As a result Hb is reduced to ¯ the Hamiltonian of interacting hard-core charged bosons tunnelling in a narrow band.2 Bogoliubov equations in the strong-coupling regime Let us derive equations describing the order parameter and excitations of CBG by the use of the Bogoliubov transformation similar to that in chapter 2. is derived using the Hamiltonian (4.138
Strong-coupling theory
contribution can be included in the deﬁnition of v. we can expand k = 0 to obtain [∇ + 2ie A(r)]2 t (r.186)
is the inverse bipolaron mass. r ) =
k ∗∗ ik·(r−r ) k−2e A e ∗∗ k
(4. Here we assume a parabolic dispersion near the ∗∗ k 2 .184) in the vicinity of (4.187) describes a charged Bose gas (CBG) with the effective boson mass m ∗∗ and charge 2e. t)ψ(r. t). equation (4. t). Hence.187)
¯ dr dr v(r − r )
(r)
†
(r ) (r) (r )
where we include the bipolaron chemical potential µ. The equation of motion for the ﬁeld operator. eH a 2 t (r − r ) → t (r. It has the following form: atomic ﬁeld.188)
. we arrive at bottom of the band. 4. v(r) ∼ 1/( 0r ) at large distances ¯ (section 4. We note that the bipolaron– bipolaron interaction is the Coulomb repulsion. k Hb ≈ − +
1 2
dr
†
(r)
[∇ + 2ie A(r)]2 +µ 2m ∗∗
†
(r) (4. t)ψ(r .185)
in real space. r ) ≈ − δ(r − r ) 2m ∗∗ where ∗∗ d2 k 1 = m ∗∗ dk 2
k→0
(4. ψ(r. (4.187) as i (∇ − i2e A)2 d ψ(r.16). To describe the electrodynamics of bipolarons. t) dt 2m ∗∗ + ¯ dr v(r − r )ψ † (r . t)] = − − µ ψ(r.4) and the hard-core repulsion is irrelevant in the dilute limit. If the magnetic ﬁeld is weak. t) = [H. we introduce the vector potential A(r) using the so-called Peierls substitution [111]: t (m − m ) → t (m − m )ei2e A(m)·(m−m ) which is a fair approximation when the magnetic ﬁeld is weak compared with the 1 (see also section 2. Finally.7.

The weight of each state in the superposition is a smooth function of √ ¯ N which is practically unchanged in the window ± N near the average number ¯ N .Bipolaronic superconductivity
139
1) in ionic solids makes the A large value for the static dielectric constant ( 0 Coulomb repulsion between bipolarons rather weak. t) 2m ∗∗
∗ ˜ dr v(r − r )[n s (r. calculated for states ¯ ¯ with ﬁxed N = N + 1 and N = N . Hence. ¯
(4. t). t) + ψs (r . we expect that some properties of CBG to be similar to the properties of an ideal Bose gas (appendix B). The macroscopic condensate wavefunction ψs (r. obeys the following equation: i d (∇ + i2e A)2 ψs (r.190) which is a generalization of the so-called Gross–Pitaevskii (GP) [113] equation applied to neutral bosons. t)|2 4m ∗∗ e2 1/3 0 (4πn b /3) (4. t).189)
Substituting equation (4.192) is the condensate density. In particular. If the interaction is weak. t)ψs (r.188) we also obtain i d ˜ ψ(r. If the Coulomb repulsion of bosons is not very large. t) + dt 2m ∗∗ dr v(r − r )n s (r .191)
º1
(4. which plays the role of an order parameter. Introducing the chemical potential µ we consider each quantum state as a superposition of states with a slightly different total number of bosons. because ψ changes the number of particles only by one. t )ψs (r. and neglecting the interaction between the supra-condensate bosons. t) = ψs (r. its diagonal matrix element is practically the same as the off-diagonal one. t) = |ψs (r. Following Bogoliubov [4] we separate the ˜ large diagonal matrix element ψs from ψ treating ψs as a number and the rest. as a small ﬂuctuation ˜ ψ(r. (4. t)ψs (r. t) = − − µ ψs (r.188) and collecting c-number terms of ψs we obtain a set of Bogoliubov-type equations for the CBG [112]. t) ¯
˜ dr v(r − r )ψs (r .189) into equation (4. Taking into account the interaction ˜ of ‘supracondensate’ bosons (described by ψ(r. ψ. t)ψ † (r . t) = dt − + + Here n s (r. t) + ψ(r. rs = (∇ + i2e A)2 ˜ − µ ψ(r. from equation (4. t). t)) with the condensate. t) has an anomalously large matrix element between the states with N + 1 and N bosons. t) ¯
(4.193)
. t)]ψ(r . the state with zero momentum (k = 0) remains macroscopically occupied and the corresponding Fourier component of the ﬁeld operator ψ(r.

194)
i
d u(r. the contribution of the terms nonlinear in ψ is negligible in equation (4. t)ψs (r. t) + ψs (r . t)αn ]
(4. the quasi-particle wavefunctions are plane waves. t)ψs (r. t)] = δ(r − r ) n n
(4. Applying the linear Bogoliubov transformation ˜ of ψ (chapter 2). ¯
(4. t) + ψs (r .195) and (4. (4. t) =
n † where now αn and αn are Bose operators. ˜ ψ(r. t)|2 v(r. t) = dt − + + (∇ − i2e A)2 − µ v(r.3 Excitation spectrum and ground-state energy For a homogeneous case with A(r) = 0. t)ψ0 (r. t). t)|2 u(r. t)] ¯
dr v(r − r )ψ0 (r . t) = dt
− + +
(∇ + i2e A)2 − µ u(r.195)
and −i d v(r.196)
There is also a sum rule.
∗ [u n (r. u(r.140
Strong-coupling theory
the number of bosons n pushed up from the condensate by the repulsion is ˜ ˜ small (see later).197)
which retains the Bose commutation relations for new operators.191).7. t) ¯ ∗ ∗ dr v(r − r )ψs (r . t)u(r .196) plays the same role in the strong-coupling theory as the Bogoliubov equations for the BCS superconductors (section 2. u k (r. t)v(r . t) 2m ∗∗
∗ dr v(r − r )[|ψs (r . t) = vk e
kt
(4. t)ψs (r. t). 4. t)vn (r . t) 2m ∗∗
∗ dr v(r − r )[|ψs (r . we obtain two coupled Schr¨ dinger o equations for the quasi-particle wavefunctions. as ∗ † [u n (r. t) and v(r. The set of equations (4. t) − vn (r. t)]v(r . Therefore.190). t)αn + vn (r. t)u(r . t) = u k eik·r−i vk (r. t) ¯
(4. t)u ∗ (r .198)
ik·r−i k t
.11).

∗∗ )2 4(m m ∗∗ (4. The quasi-particle spectrum (equation (4. because of the global electroneutrality). t) = constant.208)
which is the plasma frequency. ψs (r. Then a solution of equation (4.206) (4. k .207)) differs qualitatively from the BCS excitation spectrum (section 2. ωe .203) (4. we ﬁnd ¯ ¯ 1 ξk 1+ 2 k ξk 1 2 1− vk = − 2 k vk n s ¯ u k vk = − 2 k u2 = k where ξk = k 2 /2m ∗∗ + vk n s . Then the Bogoliubov equations yield k2 u k + n s vk [u k + vk ] ¯ 2m ∗∗ k2 vk + n s vk [u k + vk ] ¯ − k vk = 2m ∗∗
kuk
=
(4.202)
k
Using the Fourier component of the Coulomb interaction yields [114] = k4 2 + ωps 4(m ∗∗ )2 (4.199)
because the volume average of v(r) is zero due to the interaction with the ion ¯ background (i. The BCS quasi-particles are fermions and their energy is of the order of the BCS gap.201)
and |u k |2 − |vk |2 = 1. The quasi-particles which is well below the electron plasma frequency.204) (4.205) (4.200) (4.e. t)-independent.2).190) is µ=0 (4. k in CBG are bosons and their energy is about the (renormalized) plasma frequency
. Here vk is the Fourier transform of v(r). The quasi-particle energy is ¯ = k4 ¯ k 2 vk n s + .207)
k
with a gap ωps = 16πe2n s ∗∗ 0m (4. As a result.Bipolaronic superconductivity
141
√ ns = and the condensate wavefunction is (r.

The groundstate energy E 0 is obtained by substituting equation (4. αk |0 = 0. E 0 ≡ 0|H |0 = This can be written per particle as 23/2 E0 3/4 = 1/4 ωpsrs nb 3 π
∞ 0 1 2 k
(
k
− ξk ). therefore the CBG is a superconductor. The actual
. n s is the number of condensed bosons per unit area.
3/4
(4.
4. |E 0 | plays the same role as the condensation energy in the BCS superconductor. k 4 = E s k/qs + k 4 /qs . the quasi-particle spectrum is given by
k
=
(
∗∗ 2 k )
+2
∗∗ ¯ k vk n s .2rs . vk = ¯ 8πe2 /( 0 k).209) n = ψ † (r)ψ(r) = ˜
k
which is small compared with the total density n b as n ˜ 3/4 ≈ 0. The density of bosons pushed up from the condensate by the Coulomb repulsion at T = 0 is 2 ˜ ˜ vk (4.207)) satisﬁes the Landau criterion of superﬂuidity (chapter 1). if the free-boson dispersion k is anisotropic. Therefore. the plasma gap is anisotropic as well.194) into the Hamiltonian (4.211)
dk k 2
k 4 + 1 − k 2 − 1/(2k 2 ) ≈ −0.187) and neglecting terms of higher order than quadratic in αk .4 Mixture of two Bose condensates The quasi-particle spectrum (equation (4. In an extreme case of (quasi-)two-dimensional bosons with a parabolic dispersion and a two-dimensional Coulomb repulsion.210)
The ground state |0 is a vacuum of quasi-particles.
(4. The value of |E 0 | is considered as the gain in the total energy due to the condensation of interacting bosons with respect to the ground-state energy (= 0) of an ideal Bose gas. the Bogoliubov spectrum is gapless.
(4.142
Strong-coupling theory
ωps .212) The negative value of the ground-state energy is due to the opposite charge background.
(4.7.23ωpsrs .213)
∗∗ Hence.214)
2 Here E s = qs /2m ∗∗ and qs = (32πe2 n s / 0 )1/3 is a two-dimensional screening wave-number. nb (4. These results for three-dimensional charged bosons are readily generalized for any bosons on a lattice beyond the effective mass approximation. For example.

t) + 2m j dr V j j (r − r )
j
∗ × [|ψs j (r .189)) and the equations of motion for the Heisenberg ψ-operators. the condensate wavefunctions are found from two coupled GP equations: i ∂ ψs j (r. t)ψs j (r. t) = ∂t − +
j
∇2 − µ j ψs j (r. Using the displacement transformation (equation (4. t). because inter-site bipolarons have ‘internal’ quantum numbers like the spin (section 4. t)v j (r . t)ψs j (r. t) + ψs j (r . These internal degrees of freedom can affect the collective excitations of a bipolaronic superconductor. t) ∗ ∗ + ψs j (r . t)u j (r . t) = ∂t − ∇2 − µ j v j (r. orbital momentum and different symmetries of the one-particle Wannier orbitals bound into the bipolaron.
(4. t) + 2m j dr V j j (r − r )
j
(4. t)(αn + βn ) + vn j (r.217)
˜ Here we applied the linear transformation of ψ. ˜ ψ j (r. t)u j (r .6). t) =
n ∗ † † u n j (r. j
† j (r)
−
∇2 − µj 2m j
j (r)
dr
¯ dr v j j (r − r )
† j (r)
j (r)
† j (r
)
j
(r )
where m j is the mass of the boson j . t)]. t) 2m j dr V j j (r − r )|ψs j (r .Bipolaronic superconductivity
143
spectrum of bipolarons on a lattice is more complicated. t)|2 u j (r. t)|2 ψs j (r.2
dr +
1 2 j. t)ψs j (r. t) + ψs j (r . t)v j (r . t) = ∂t − ∇2 − µ j u j (r. t)
+ ψs j (r . t)(αn + βn )
. t)]. t)|2 v j (r.216)
∗ × [|ψs j (r . (4.215)
The supracondensate wavefunctions satisfy four Bogoliubov-type equations: i ∂ u j (r. and −i ∂ v j (r. Here we extend the Bogoliubov-type equations of the previous subsection to the multi-component Bose condensate of non-converting charged bosons [115]. t)ψs j (r. Let us consider a two-component (1 and 2) mixture of bosons described by H=
j =1.

It corresponds to a low-frequency oscillation in which two condensates move in anti-phase with one another. u ≡ u 0 . we determine the excitation spectrum. Solving two GP equations (4. ¯ w ≡ w0 and n j = |ψs j |2 are the condensate densities.220)
.2 = 2−1/2 1 (k) + 2 (k) ± [ 1 (k) − 2 (k)]2 + m 1m 2 k (4.219) ¯ where 1 (k) = [k 4 /(4m 2 ) + k 2 vk n 1 /m 1 ]1/2 and 2 (k) = [k 4 /(4m 2 ) + 1 2 ¯ k 2 u k n 2 /m 2 ]1/2 are Bogoliubov modes of two components. differing from these normal-state APs with a linear dispersion.215). in contrast to the usual optical high-frequency plasmon (equation (4. ξ2 (k) = k 2 /2m 2 + u k n 2 and vk . respectively.
(4. u k = 4πq2 /k 2 and wk = ¯ 2 .218) Here ξ1 (k) = k 2 /2m 1 + vk n 1 . v22 (r) and v12 (r). E 2 (k) = k2 2(m 1 m 2 )1/2
2 2 q1 n 1 m 1 + q2 n 2 m 2 2 2 q1 n 1 m 2 +q2 n 2 m 1 2 2 4πq2 n 2 4πq1 n 1 + m1 m2
wk ψs1 ψs2 ¯ ∗ ∗
s1
∗ wk ψs1 ψs2 ¯
u k ψs2 ¯ ∗2
ξ2 (k) + E(k)
s2
(4. wk are the ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Fourier components of v11 (r). if m 1 = m 2 .144
Strong-coupling theory
where αn . Det ∗ψ wk ψ s2 ¯ wk ψs1 ψs2 ¯ ξ2 (k) − E(k) ukψ2 ¯ (4. The mode is similar to the acoustic plasmon (AP) mode in the electron–ion [116] and electron–hole [117] plasmas. from ∗ ξ1 (k) − E(k) vk ψs1 ¯ 2 wk ψs1 ψs2 ¯ wk ψs1 ψs2 ¯ vk ψs1 ¯ ∗2 ξ1 (k) + E(k) wk ψs1 ψs2 ¯ ∗ ∗ wk ψs1 ψs2 ¯ ∗ = 0. However. E(k). vk = 4πq1 /k 2 . the upper branch is the geometric sum of familiar plasmon modes 4πq1 q2 /k for k → 0.220)) in which the components oscillate inphase. E 1 (k) = while the lowest branch is gapless.221)
Remarkably. this mode does not depend on the interaction at any charges of the components. v ≡ v0 . If the interaction 2 2 ¯ ¯ is purely the Coulomb repulsion. We conclude that while the
. the AP of Bose mixtures has a quadratic dispersion in the long-wavelength limit. There are two branches of ¯ ¯ excitations with dispersion 1/2 4k 4 2 2 2 2 2 w n1n2 ¯ E(k)1. we obtain the chemical potentials of a homogeneous system as µ1 = vn 1 + wn 2 ¯ ¯ µ2 = un 2 + wn 1 ¯ ¯ and solving the four Bogoliubov equations. u k . βn are bosonic operators annihilating quasi-particles in the quantum state n.

the two-component condensate of bipolarons is a superconductor.7. 4. m 1 = m 2 and u = v = ω p . transforms the lowest AP mode into the ¯ ¯ ¯ Bogoliubov sound with a linear dispersion at k → 0. wk ∝ constant + 1/k 2 . u k . i.0
k/qs
Figure 4. spins and between themselves.
CBG is a superﬂuid (according to the Landau criterion). Combining both interactions. 0.Bipolaronic superconductivity
145
2
E(k)/ ωp
1
0 0. where the Migdal adiabatic approximation is believed to be valid (chapter 3).5 Critical temperature and isotope effect Polarons and bipolarons differ from electrons in many ways.5 (dots).10. Electrons change their mass in solids due to interactions with ions. Hence. taking vk . A is a constant independent of M and ω0 is the characteristic
.7.5
1. Excitation ¯ ¯ energy is measured in units of the plasma frequency ω p = 4πne2 /m and momentum k is measured in inverse screening length qs = (16πe2 nm)1/4 .3. their mixture is not. electrons form polarons and their effective mass m ∗ depends on M as m ∗ = m exp(A/ω0 ) [82] (section 4. Excitation energy spectrum of the two-component Bose–Einstein condensate with the long-range repulsive and hard-core interactions. ﬁgure 4. In the case of bipolarons.1).0
0. In this plot n 1 = n 2 = n. and 0. q1 = q2 = e. However.10. if the interaction between electrons and ion vibrations is strong and/or the adiabatic approximation is not applied. respectively.e.95 (full line). the renormalized mass of electrons is independent of the ion mass M in ordinary metals. where m is the band mass in the absence of the electron– phonon interaction. the interaction includes both the long-range repulsion and the hard-core interaction (section 4.1). Because λ does not depend on the ion mass.1 (dashes). One of the differences is their effective mass. Different curves correspond to the different ¯ ¯ values of w/v = 0.

interpolating between BCS. there is a substantial isotope effect on the carrier mass in polaronic and bipolaronic systems. α≡− d ln Tc = αm ∗∗ .6. (4. caused by the Coulomb repulsion.225) λ − µc √ 2 M depends on M.3).225) with respect to M yields α = αm ∗ 1 − Z λ − µc . The isotope exponent in m ∗ is deﬁned as αm ∗ = i d ln m ∗ /d ln Mi .146
Strong-coupling theory
phonon frequency. In this case we can apply equation (4. The isotope effect on the polaron/bipolaron mass leads to an anomalous isotope effect on the superconducting critical temperature of polaronic and bipolaronic superconductors [82].224)
In the crossover region from the normal polaron Fermi liquid to bipolarons. when their mass m ∗∗ is proportional to the polaron mass (section 4.225).119]. in contrast to the zero isotope effect in ordinary metals. Hence. Tc of bipolarons is the Bose–Einstein condensation (BEC) temperature of CBG. It is approximately the Bose–Einstein condensation temperature of ideal bosons (appendix B) as long as the Coulomb repulsion is weak (rs º 1): 2/3 3. Equation (4.222)
The same isotope exponent is predicted for inter-site bipolarons. allows us to understand the origin of high values of Tc in comparison with the
. the isotope exponent α in Tc of inter-site bipolarons is the same as αm ∗∗ .223).223) Tc ≈ m ∗∗ where n b is the volume density of bipolarons. Hence. are numerically small even at rs = 1 [118. one can apply the BCS-like expression (4.225) to the whole region of the phase diagram (ﬁgure 4. Differentiating equation (4. which depends on M as ω0 M −1/2 . Corrections to equation (4. one readily ﬁnds αm ∗ =
1 2
ln(m ∗ /m).and BEC-type superconductivity. d ln M (4.223)) if their effective mass m ∗∗ m ∗ .113) for Tc of non-adiabatic carriers which we write as Z Tc D exp −g 2 − . where λ − µc > Z . When λ − µc Z = e−g . (4.226)
We see that the isotope exponent is negative in polaronic superconductors. where λ − µc < Z but it is positive in bipolaronic superconductors. where Mi is the mass of the i th ion in a unit cell. this Here only g 2 expression also describes the isotope exponent in Tc of inter-site bipolarons (equation (4.6) including the BCS-like crossover regime and the Bose–Einstein condensation.31n b (4. Using this deﬁnition and the expression for the polaron mass m ∗ mentioned above.
(4.

t)]2 d ψs (r. t)∇ψs (r. so that the maximum value of Tc is estimated to be Tc º ω0 /3 [13]. −ω)} − ¯ q · a(q.231) 2 q 2 ω2 − q The expectation value of the current is given by j(r.227)
where jα (q. respectively. ω) and aβ (q. ω) (4. The exponentially enhanced DOS of the narrow polaronic band effectively eliminates the small exponential factor in equation (4. −ω) = − q · a(q. ¯
Using a perturbed wavefunction.y. ω). t)∇ψs (r.z
K αβ (q. n s + φ(r. we need to solve equation (4. Therefore. ω) + φ ∗ (−q. 4. ω).225). ω) =
β=x. t)+ dt 2m ∗∗ ψs (r. t) − ψs (r. In novel oxygen and carbon-based superconductors (part 2). (4. ω) − φ ∗ (−q. t)|2 ψs (r. t)] − ∗∗ A(r. ∗∗ m m (4.229) The solution is √ 2e n s ω φ(q. But at a very strong e–ph coupling. t) = √
∗ dr v(r −r )|ψs (r . t) = − ψs (r. 2m ∗∗ 2m ∗∗ (4. we obtain for the Fourier transform φ(q. ω)
(4.232)
. −ω) = − q · a(q. i [∇ + i2e A(r. Tc drops again because the carriers become very heavy. ω) + n s vq {φ(q. ω) + φ ∗ (−q. To calculate the response. t).Bipolaronic superconductivity
147
weak-coupling BCS theory (equation (2. ω): √ 2e n s q2 ωφ(q. ω) are the Fourier transforms of the current and of the vector potential.7. High Tc originates in the polaron narrowing of the band.48)). ω) = φ(q.6 Magnetic ﬁeld expulsion The linear response function is deﬁned as jα (q.228)
and keeping only the terms linear in A(r. The polaron half-bandwidth Z D is normally less or about the phonon frequency ω0 . That is why Tc is remarkably high in these compounds.190) with µ = 0 in the ﬁrst order with respect to A(r. the characteristic optical phonon frequency is about 500 to 2000 K. we conclude that the highest Tc is in the crossover region from polaronic to bipolaronic superconductivity where λ − µc ≈ Z . t). t) = ie 4e2n s ∗ ∗ [ψs (r. t). ω)aβ (q. t) (4.230) ∗∗ ω2 − 2 m q √ 2 4e n s q φ(q. t).

(4.234)
This response function has been split into a longitudinal K l ( δ αβ ) and a transverse K t ( (q α q β − δ αβ q 2 )) part. we ﬁnd σex (q. t)|2 − n b ψs (r.238)
m ∗∗ 16πe2n s
. we can analyse a single vortex in the CBG and calculate the lower (ﬁrst) critical ﬁeld Hc1 by solving a stationary equation for the macroscopic condensate wavefunction [120]: − 4e2 [∇ + 2ie A(r)]2 −µ+ 2m ∗∗ 0 dr |ψs (r . Hence. ω) = σex (q.236)
The conductivity sum rule (1. ω) = ∗∗ q[φ(q.18) is satisﬁed as
∞ 0
dω Re σex (ω) =
2 0 ωps
8
.
(4. t) = 0.237)
The conductivity in the transverse electromagnetic ﬁeld ( D ⊥ q) is given by σt = where λH = i 4πλ2 ω H
1/2
(4.7 Charged vortex and lower critical ﬁeld The CBG is an extreme type II superconductor.148
Strong-coupling theory
Using the perturbed wavefunction. ω) (4. as shown later.233) m m and K αβ (q.235)
(4. 2 iω 4π(ω2 − q )
(4. ω) − φ ∗ (−q. 4.234). ω) D(q. we obtain the Fourier transform of the current as √ e ns 4e2 n s j (q.240) |r − r |
. (4. (4.7. The longitudinal response to the ﬁeld ( D q) is expressed in terms of the so-called external conductivity σex : j l (q.239)
This expression combined with the Maxwell equation describes the Meissner– Ochsenfeld effect in the CBG with the magnetic ﬁeld penetration depth λH . −ω)] − ∗∗ a(q. ω) = 4e2n s m ∗∗ δ αβ ω2 + (q α q β − δ αβ q 2 ) 2 2 − ω2 q ( q
2 q 2 q
− ω2 )
. ω) where D is the external electric ﬁeld. ω) =
2 i 0 ωωps Kl = . Using equation (4.

1/2 Indeed. While the CBG shares quantum coherence with the BCS superconductors owing to the Bose–Einstein condensate (BEC). the long-range (non-local) interaction leads to some peculiarities. f = 1 for ρ = ∞ and the ﬂux quantization condition.242) (4. the vortex is charged in the CBG and the coherence length occurs just the same as the screening radius.245) ξ(0) = (21/2 m ∗∗ ωps )−1/2 and the London penetration depth is λ(0) = m ∗∗ 16πn B e2
1/2
. φ = 0 for ρ = ∞ and φ(0) =
0 ∞
ρ ln(ρ)(1 − f 2 ) dρ
(4.240) and the Maxwell equations take the following form: 1 dh 2 1 d df ρ − 3 − φf = 0 dρ κ 2 ρ dρ dρ f 1 d dφ ρ =1− f2 κ 2 ρ dρ dρ 1 d ρ dh = h.6. ρ = 0.43).
(4.240) is quite different from the Ginsburg–Landau equation (1. equation (4. introducing dimensionless quantities f = |ψs |/n b .246)
There are now six boundary conditions in a single-vortex problem.244)
with a new fundamental unit φc = em ∗∗ ξ(0)2 . ρ dρ f 2 dρ (4.Bipolaronic superconductivity
149
Subtracting n b in the integral of equation (4.247)
for the electric ﬁeld at the origin.242).240). Four of them are the same as in the BCS superconductor (section 1.3). ρ = r/λ(0) and h = 2eξ(0)λ(0)∇ × A for the order parameter. The integra-differential equation (4. In particular. where p is an integer. we explicitly take into account the Coulomb interaction with a homogeneous charge background with the same density as the density of charged bosons. We note that the chemical potential is zero at any point in the thermal equilibrium. the coherence length is the same as the screening radius. respectively. length and magnetic ﬁeld. h = dh/ρ = 0. (4. The remaining two conditions are derived from the global charge neutrality.
. At T = 0.243)
The new feature compared with the GL equations of chapter 1 is the electric ﬁeld potential determined as φ= 1 2eφc ¯ dr v(r − r )[|ψs (r )|2 − n B ] (4.241) (4. dh/dρ = − p f 2 /κρ for ρ = 0. The potential is calculated using the Poisson equation (4.

where c1 1. The breakdown of the local charge neutrality is due to the absence of any equilibrium normal-state solution in the CBG below the Hc2(T ) line. we obtain ξ(0) 0. rs London equation with the familiar solution h = pK 0 (ρ)/κ. such as m ∗∗ = 10m e . Indeed. equation (4. Owing to a large dielectric constant.1664. Then the system decides which of two phases (or their mixture) is energetically favourable but the local charge neutrality is respected.12.11 and 4. This leads to an increase of the order parameter compared with the homogeneous case ( f = 1) in the region close to the vortex core. If κ 1. f differs qualitatively from the BCS order parameter. In this region.6.3). there is no equilibrium normal-state solution (with ψs = 0) in CBG below the Hc2(T )-line because it does not respect the density sum rule. The numerical 1. the solutions are f = 1 + (4 p2 /r 4 ) and 2 /r 2 .249) r dr dr They are satisﬁed by regular solutions of the form f = c p r p and φ = φ(0) + (r 2 /4).46. In the region p r < pκ. λ(0) 265 nm. we can use the ﬂux quantization condition to ‘integrate out’ the magnetic ﬁeld in equation (4. where K 0 (ρ) is the Hankel function of imaginary argument of zero order (section 1. Being quite different from the BCS superconductor. there are no different phases to mix and the only
. φ = −p f BCS = 1 − ( p2 /r 2 ) (see also ﬁgure 4.241). when r → 0.249). where c1 φ(0) −1.11).12) (together with the magnetic ﬁeld) acts as an additional centrifugal force increasing the steepness (c p ) of the order parameter compared with the BCS superﬂuid. n b = 1021 cm−3 and 0 = 103 .0515. with the material parameters typical for oxides. where the total density of electrons remains constant across the sample. the CBG allows for ﬂux penetration by redistributing the density of bosons within the coherence volume. The difference is due to a local charge redistribution caused by the magnetic ﬁeld in the CBG. The constants c p and φ(0) are determined by complete numerical integration of equations (4. For the region ρ ≤ p. respectively. The resulting electric ﬁeld (ﬁgure 4. This leaves us with two parameterfree equations written for r = κρ: 1 d df p2 f r − 2 − φf = 0 r dr dr r and (4.248)
1 d dφ r = 1 − f 2. In contrast. Hence. Inside the core the order parameter is suppressed (ﬁgure 4. where the order parameter and the electric ﬁeld differ from unity and zero. and the Ginsburg–Landau ratio κ 552.11) as in the BCS superconductor. (4.243) is reduced to the for heavy bipolarons. the Coulomb repulsion remains weak even 0.248) and (4. Both superconducting ( k = 0) and normal ( k = 0) solutions are allowed at any temperature in the BCS superconductors (chapter 2).150
Strong-coupling theory
CBG is an extreme type II superconductor with a very large Ginsburg– Landau parameter (κ = λ(0)/ξ(0) 1). ﬁgure 4.11(a).5188 and results for p = 1 are shown in ﬁgures 4.48 nm.

This energy determines the vortex free energy = E v − E 0 .3). h c1 ≈ ln κ/(2κ). this can be written in dimensionless form: = dr F = 2π
0 ∞
[h 2 − 1 φ(1 + f 2 )]ρ dρ. where f 1 and φ − p2 /(κ 2 ρ 2 ).11. However. thus. 2
(4.243). f BEC . |[∇ + 2ie A(r)]ψs (r)|2 + eφc φ[|ψs (r)|2 − n b ] + 2m ∗ 8π (4. the CBG vortex structure is derived here in the
. which is the difference between the energy of the CBG with.251)
In the large κ limit.6. f BCS . and without. p = 1. the main contribution comes from the region p/κ < ρ < p. the same as that in the BCS superconductor. (4. E 0 .250) Using equations (4. F 2π p2 ln(κ)/κ 2 (section 1.241).16).
way to acquire a ﬂux in the thermal equilibrium is to redistribute the local density of bosons at the expense of their Coulomb energy. (b). It can be seen that the most stable solution is the formation of the vortex with one ﬂux quantum. The energy is. and the lower critical ﬁeld is the same as in the BCS superconductor. differing from the BCS superconductor.242) and (4.Bipolaronic superconductivity
151
Figure 4. Vortex core proﬁle in the charged Bose gas. E v . (a) compared with the vortex in the BCS superconductor. a magnetic ﬂux: 1 (∇ × A)2 . where the Ginsburg–Landau phenomenology is microscopically justiﬁed in the temperature region close to Tc (section 2.

4). the upper critical ﬁeld of the CBG differs signiﬁcantly from the Hc2 (T ) of BCS superconductors (section 1.6.4). where the depletion of the condensate remains small. The actual size of the charged core is about 4ξ (ﬁgure 4. and eliminate the one-dimensional singularity of the density of states.11) is applied in a wide temperature region well below the Bose–Einstein condensation temperature. As we show later.12).
low-temperature region. It has an unusual positive curvature near Tc (Hc2(T ) (Tc − T )3/2 ) and diverges at T → 0. In line with the conventional deﬁnition (section 1. Collisions between bosons and/or with impurities and phonons make the motion three dimensional. Hc2(T ) is a ﬁeld. The localization can drastically change the low-temperature behaviour of Hc2(T ) (part 2).7. an interacting CBG condenses in a ﬁeld lower than a certain critical value Hc2(T ) [121]. Electric ﬁeld potential φ as a function of the distance (measured in units of ξ(0)) (lower curve) together with the CBG (upper curve) and BCS order parameters (a).152
Strong-coupling theory
Figure 4. the zero-temperature solution (ﬁgure 4. its proﬁle is shown in (b). an ideal CBG cannot be bose-condensed at ﬁnite temperatures in a homogeneous magnetic ﬁeld because of one-dimensional particle motion at the lowest Landau level [7].6.8 Upper critical ﬁeld in the strong-coupling regime If we ‘switch off’ the Coulomb repulsion between bosons. In fact. where the ﬁrst non-zero solution of the linearized stationary equation for the
. if there is no localization due to a random potential. which prevents BEC of the ideal gas in the ﬁeld. However.12. 4.

∞.252)
Here we include the ‘scattering’ potential Vscat (r) caused. quite different from the GL equation (1.255) into equation (4. we have µ = E c = ω/2 and √ ∞ 2(m ∗∗ )3/2 ω 1 N( . n(T ) = n b (T /Tc )3/2 (see equation (B. 2. 2m ∗∗
153
(4. i. This number is practically the same as in zero ﬁeld. .254)
where N( .e. For ideal bosons.Bipolaronic superconductivity
macroscopic condensate wavefunction occurs: − 1 [∇ − 2ie A(r)]2 + Vscat (r) ψs (r) = µψs (r). ˜ if ω Tc . . by impurities. µ = ω/2 (appendix B).4) εn = ω(n + 1/2) +
2 kz 2m ∗∗
(4. it determines the value of the chemical potential. f ( ) = [exp( −µ)/T − 1]−1 is the Bose–Einstein distribution function and E c is the lowest band edge.253)
where ω = 2eHc2/m ∗∗ and n = 0. Hence the only solution to equation (4.6.258) 0 (Hc2 )
.255) Re √ 2 4π − ω(n + 1/2) n=0 Substituting equation (4. Hc2 ) = . Hc2 ) ∝ √ . Their energy spectrum in a homogeneous magnetic ﬁeld is (section 1.257)
is the number of bosons occupying the levels from n = 1 to n = ∞.61)).252).
∞ Ec
f ( )N( . Let us ﬁrst discuss non-interacting bosons. In contrast. Then using this value. Hence.256) is Hc2 (T ) = 0.256)
dx exp(x/T ) − 1
∞
√
n=1
1 x − ωn
(4. is given by a divergent integral on the left-hand side of equation (4. (4. Hc2 N0 ( . the number of bosons on the lowest level.252) does not allow for a direct determination of Hc2 . Hc2 ) ω( − ω/2)−1/2 leading to a ﬁnite DOS near the bottom of the lowest level. 1. . Hc2 ) is the DOS of the Hamiltonian (4. the Schr¨ dinger equation o (4. Vscat (r) = 0. Hc2 ) d = n b
(4.256). In fact. the upper critical ﬁeld is found from the density sum rule. BEC occurs when the chemical potential ‘touches’ the lowest band edge from below.254) yields √ 2(m ∗∗ )3/2ω ∞ dx 1 = n b − n(T ) ˜ 4π 2 x 1/2 exp(x/T ) − 1 0 where n(T ) = ˜ √ 2(m ∗∗ )3/2 ω 4π 2
∞ 0
(4. for example. (4.90). n = 0. The scattering of bosons effectively removes the one-dimensional singularity in N0 ( . .

154
Strong-coupling theory
Using the Fermi–Dirac golden rule. Here we apply the one-dimensional DOS but cut the integral at 0 from below. For impurities. Then the DOS is found to be ˜ N ( . Hc2 ) and their lowest energy E c (the so-called mobility edge.e.
These simple scaling arguments are fully conﬁrmed by DOS calculations with impurity [121] and boson–boson [122] scattering. then the characteristic length is ξ0 ≈ l nb
1/4
(4. k x . The density of extended ˜ states N ( . depends on the scattering mechanism. Here ν ≡ (n. If we write H0 = 2 0 /(2πξ0 ). k z ) are the quantum numbers of a charge particle in the magnetic ﬁeld. the energy spectrum of the Hamiltonian (4. H0. The scaling constant.264)
. where ˜ (E c . Hc2) = 0) can be found.260) 4π 2 x 1/2 exp(x/T ) − 1 0 as long as T 0 . The upper critical ﬁeld has a nonlinear behaviour: Hc2(T ) ∝ (Tc − T )3/2 in the vicinity of Tc and diverges at low temperatures as Hc2(T ) T −3/2 . Hc2 ) = 1 πV 2 Im
n n(
). the collision broadening of the lowest level 0 (Hc2 ) is proportional to the same DOS:
0 (Hc2 )
∝ N0 ( .263)
To obtain analytical results.
(4.
(4. Finally we arrive at Hc2 (T ) = H0(t −1 − t 1/2 )3/2 (4. we choose the matrix elements of the scattering potential as Vνν = V δnn .252) consists of discrete levels (i.262)
where l is the zero-ﬁeld mean-free path of low-energy bosons. if the electron self-energy ( ) is known.3 considering the impurity scattering as the elastic limit of the phonon scattering:
ν(
)=
ν
2 Vνν − ν −
ν
( )
. localized states) and a continuous spectrum (extended states).259)
so that 0 scales with the ﬁeld as 0 (Hc2) Then the number of bosons at the lowest level is estimated to be √ 2(m ∗∗ )3/2 ω ∞ dx 1 2/3 n0 = T Hc2 (4. Hc2 )
2/3 Hc2 .261) where t = T /Tc and H0 is a temperature-independent constant. We N calculate ( ) in the non-crossing approximation of section 3.
(4.

266)
and the mobility edge Ec = Here
0
3 0 ω − 2/3 .268)) −1/3 depends on the mean free path l and the inter-particle distance n b .
Solving this equation. The curvature is a universal feature of the CBG. With this DOS in the sum rule (equation (4. Thus Hc2(T ) of strongly-coupled superconductors has a ‘3/2’ curvature near Tc which differs from the linear weak-coupling Hc2(T ) of section 1.4). H) = 4π 2
∞
6m ∗∗ 3 1 ˜n + + 27 2
0
×
n=0
3 1 ˜n + 27 4
1/3
−
3 ˜n
27
+
1 − 2
1/3 3 1 ˜n + 27 4 (4.261).
(4.6. we obtain the same Hc2(T ) for T 0 as through the scaling. which does not depend on a particular scattering mechanism and on any approximations.268)
The zero-ﬁeld mean free path l is expressed via microscopic parameters as l = π/(V m ∗∗ )−2 .13).265)
√ π 2
)
.263) yields a cubic algebraic equation for the self-energy n ( ) of the nth level:
n(
)= =
dV 2 2π eHc2
∞ −∞ √ V 2 m ∗∗
dk z
1 2 − ω(n + 1/2) − k z /(2m ∗∗ ) − √ i − ω(n + 1/2) −
n(
n(
) (4. The ‘coherence’ length ξ0 of the CBG (equation (4.6 (ﬁgure 4. and ξ0 ≈ 0. The Hc2(T ) of bipolarons exceeds this limit because the singlet
. It has little to do with the size of the bipolaron and could be as large as the coherence length of weak-coupling BCS superconductors. 2 2
2/3
(4. we obtain eHc2 ˜ N( . Another interesting feature of strongly-coupled superconductors is a breakdown of the 1.8 l nb
1/4
.267)
1 = 2
√ 2V 2 eH m ∗∗ π
is the collision broadening of levels and ˜n = [ − ω(n + 1/2]/ 0 .84Tc in the weak-coupling theory Pauli paramagnetic limit given by H p (section 1.Bipolaronic superconductivity
155
Integrating over k z in equation (4. equation (4.254)).

156
Strong-coupling theory
+F7.

Therefore.10). depends on the relative momentum k and on the centre-of-mass momentum. K . the symmetry of the order parameter and the symmetry of the gap are the same in the weak-coupling regime. Bosons are condensed at T = 0 no matter what their energy spectrum is.7.13). Hence.2). Hc2(0) = ∞ (ﬁgure 4. in the CBG model. so that Hc2(0) Hp.
bipolaron binding energy is much larger than their Tc .and strong-coupling regimes. In the BCS theory. 4. DUELWUDU\ XQLWV %(&
%&6
77F
Figure 4. Under the rotation of the coordinate system. f (k. BEC critical ﬁeld of a CBG compared with the Hc2 (T ) of BCS superconductors. k changes its sign if the Cooper pairing is unconventional (section 2. In this case. let us consider a tight-binding bipolaron spectrum comprising two bands on a square
. the former (‘external’) symmetry is related to the rotation of K . K ) with respect to the rotation of k. For composed bosons. r 2 ) =
s (r 1 ) s
(r 2 )
is the superconducting order parameter both in the weak. While the latter describes the transformation of f (k. the BCS quasi-particle spectrum is gapless. like bipolarons. it depends on the bipolaron band dispersion but not on the symmetry of the bound state. f (k. As an example. Hence. its Fourier transform.13. In the bipolaron theory. It depends on the relative coordinate ρ = r 1 − r 2 of two electrons of the pair and on the centre-of-mass coordinate R = (r 1 + r 2 )/2. K ). the pairbreaking limit is given by µB Hc2(0) ≈ .9 Symmetry of the order parameter The anomalous Bogoliubov average Fss (r 1 . K ) k (section 2. K = 0 (in a homogeneous superconductor) and the Fourier transform of the order parameter is proportional to the gap in the quasi-particle excitation spectrum. Hence. the symmetry of the BEC is not necessarily the same as the ‘internal’ symmetry of a pair [123].

The one-particle excitation gap is half of the
. The order parameter has different signs in the shaded cells and is zero in the blank cells.269)
b K e−i K ·m . m y + 1).270) respects the time-reversal and parity symmetries. Both have a d-wave symmetry-changing sign. the condensate wavefunction is s-wave with respect to the centre-of-mass coordinate. in the Wannier representation. ‘x’ bipolaron band has its minima at K = (±π. 0) and the y-band at K = (0.14).0).14. is given by ψs (m) = N −1/2
K =(±π. m y ) = ψs (m x . ψs (m x . The two order parameters (equation (4.±π) y
(4.
lattice with the period a = 1: E x = t cos(K x ) − t cos(K y ) K E K = − t cos(K x ) + t cos(K y ). m y ).271) ψs (m) = n s [cos(πm x ) ± cos(πm y )]. The d-wave symmetry is entirely due to the bipolaron energy dispersion with four minima at K = 0. These four states are degenerate. d-wave condensate wavefunction.270)
√ where b K = ± n s are c-numbers at T = 0. if √ ± (4. The symmetry of the gap has little to do with the symmetry of the order parameter in the strong-coupling regime. so that the condensate wavefunction ψs (m) in the site space. When the bipolaron spectrum is not degenerate and its minimum is located at the point of the Brillouin zone. If t. t > 0. The superposition (4.Bipolaronic superconductivity
157
Figure 4.271)) are physically identical because + − they are related by the translation transformation. when the lattice is rotated by π/2 (ﬁgure 4.(0. ±π).
(4. They transform into one another under π/2 rotation. m = (m x .

due to a ﬁnite dispersion of polaron and bipolaron bands (sections 4.158
Strong-coupling theory
bipolaron binding energy. and does not depend on any momentum in zero order of the polaron bandwidth.3 and 4. A multi-band electron structure can include bands only weakly coupled with phonons which could overlap with the bipolaronic band (see also part 2). In this case.
. it has an ‘s’-wave symmetry. the one-particle gap is an anisotropic s-wave.e. In fact. /2. The one-particle excitation spectrum of such mixtures is gapless. the CBG coexists with the Fermi gas just as 4 He bosons coexist with 3 He fermions in the mixture of He-4 and He-3.6). i.

PART 2 APPLICATIONS TO HIGH-T c SUPERCONDUCTORS
.

.

The main motivation behind these concepts 161
. as in BCS and bipolaron theory. λ ² 1. ﬁgure 5. which ‘glues’ two carriers together. but also ‘excitonic’ [133.1. as was noted by Ginzburg [133] in 1968 (chapter 1). so that it describes the cooperative quantum phenomenon of superconductivity well even with these non-phononic attraction mechanisms. the bosonic ﬁeld. magnetic interaction could result in the spin–bipolaron formation as suggested by Mott [139]. As highlighted by Simon [132]. if the coupling is weak (see left-hand side of ﬁgure 5. ‘plasmonic’ [135. 127] and a few other oxides. Nevertheless. At any value of the e–ph interaction superconductivity is a correlated many-body state of pairs which is well described by BCS theory in the weak-coupling regime λ 1. 138]. If the coupling is strong. where pairing is individual (upper half-plane in ﬁgure 5. These discoveries have broken all constraints on the maximum Tc predicted by conventional theory of low-temperature superconducting metals and their alloys. Hence. can not only be ‘phononic’. 129] (M is an alkali metal) and the more recently discovered magnesium diborade MgB2 [130] with a critical temperature above 30 K. the canonical theory has not provided a ‘materials’ aspect in the search for new high-Tc superconductors. 136] and/or of magnetic origin [137. in a phenomenological sense.2).Chapter 5 Competing interactions in unconventional superconductors
5. where pairing takes place in the momentum space due to the Pauli exclusion principle (collective pairing) and by the bipolaron theory in the strong coupling regime. any superconductivity can just be treated as a consequence of the formation of pairs and their condensation—even within the BCS framework.2). 134].1 High-T c superconductors: different concepts
Nowadays there are many complex high-Tc superconductors including copper [1. BCS theory like any mean-ﬁeld theory is rather universal. In general. knowledge of the carrier-pairing mechanism and of the nature of the normal state is central to an understanding of high-Tc supeconductivity. doped fullerenes Mx C60 [128.

We also include the characteristic temperatures of a variety of cryogenic liquids. Towards a century of superconductivity.162
Competing interactions in unconventional superconductors
Figure 5. from 1911. A plot of the evolution of the superconducting transition temperature. to the present situation. as well as the lowest recorded ground temperature on Earth (−89.
.1.2 ◦ C) [131].

if the normal state is not the Fermi liquid. Indeed all undoped parent compounds are insulators with an insulating gap of about 2 eV or so. A motivation for this concept can be found in the earlier work by Kohn and Luttinger [57]. With the increasing energy scale of the bosonic ﬁeld. 141] dogmatized that the interaction in novel superconductors is essentially repulsive and unretarded and that it also provides high Tc without any phonons. who showed that the Cooper pairing of repulsive fermions is possible. There are no obvious a priori reasons for discarding the dogma.2. due to charge and/or spin density waves [142]). the Coulomb pseudopotential creates a serious problem for any of these non-phononic mechanisms. it is the strong on-site repulsion of d-electrons in cuprates which results in their parent ‘Mott’ insulating state (section 5. Nevertheless.6). where the carrier density is low. However.High-Tc superconductors: different concepts
163
Figure 5. There is little doubt that strong on-site repulsive correlations (Hubbard U ) are an essential feature of the cuprates. Therefore. a much smaller gap caused by lattice and spin distortions (i. Right-hand side: non-Fermi liquid theories with electron–phonon (top) and electron–electron (bottom) correlations. But if the repulsive correlations are weak.
is that high Tc might be achieved by replacing low-frequency acoustic phonons in conventional BCS theory by higher-frequency bosonic modes.
. Therefore. such as the electron plasmon (chapter 3) or by spin waves (pseudomagnons). BCS and BCS-like theories (including the Kohn– Luttinger consideration) rely heavily on the Fermi-liquid model of the normal state. The problem is particularly grave in novel superconductors. some authors [140. BSC-like theories of high-Tc superconductors with the electron–phonon (top) and electron–electron (bottom) interactions (left-hand side).5) cannot be applied. This model fails in many high-temperature superconductors (chapter 6). But the same work clearly showed that the Tc of repulsive fermions is extremely low. one would expect a metallic behaviour for the half-ﬁlled d-band of copper in cuprates or. and the screening of the direct Coulomb repulsion is poor.2). well below the mK scale (section 3. at most. the retardation of the effective attraction vanishes and the Tolmachev logarithm (section 3. Then the direct Coulomb repulsion would deﬁnitely prevent any pairing in these models because the exchange electron–electron interaction is always weaker than direct repulsion.e.

conversely. the doubly occupied sites take a large Coulomb energy and the Hubbard Hamiltonian can be reduced to the so-called t– J model [148] ˜ H = T (a)
mn . While the interlayer RVB model was found to be incompatible with experiments [145]. Bose quasi-particles imply a condensate. In the strong correlation limit. Here t and t⊥ are the in-plane and out-of-plane renormalized hopping integrals. we have to realize that the insulating properties of the Mott insulator do not depend on the ordering of the spins. There is no single-particle coherent tunnelling between two spinon–holon planes above Tc . When on-site correlations are strong and dimensionality is low. such as occur in onedimensional models like the one-dimensional Hubbard model (see later). The microscopic Hubbard Hamiltonian [147] H = T (a)
mn . The insulator is antiferromagnetic with one hole and spin. This is the case for one-dimensional repulsive electrons.c.1 component (spinon) 2 and. In Anderson’s resonating-valence-bond (RVB) theory. respectively. must form again in order to come out.1)
was proposed to justify the RVB concept. there is an alternative to the usual Fermi-liquid description proposed by Anderson [140]. the Mott insulator arises from a potentially metallic half-ﬁlled band due to the Coulomb blockade of electron tunnelling to neighbouring sites. it was suggested that the superconductivity of copper-based high-Tc materials is due to holonpair tunnelling between the copper–oxygen planes.2)
. there is no experimental evidence for a charge e superﬂuid. However.2) would necessarily lead to the Luttinger liquid even in two dimensions. observed in the caxis conductivity. in the so-called interlayer RVB extension of the model [144].] + U ms
m
n m↑ n m↓ ˆ ˆ
(5. there is a coherent two-particle tunnelling between them below Tc . Anderson argued that the existence of the upper Hubbard band (section 5.s
[c† cns + H. U T (a). the ground state supports ‘topological solitons’ (the so-called spinons and holons). The Hamiltonian describes the antiferromagnetic Mott insulator at half ﬁlling when U > D.164
Competing interactions in unconventional superconductors
Differing from conventional band-structure insulators with completely ﬁlled or empty Bloch bands. the basic idea of spin and charge separation had been worked out in great detail [146]. In using this model. Therefore.s
[c† cns + H.1 2 per site. Then the corresponding kinetic energy should be responsible for the BCS-like pairing at 2 temperatures below Tc ≈ t⊥ /t and for the plasma-like gap. The main idea is that an electron injected into a two-dimensional layer decays into a singlet charge e component (holon) and a spin. as opposed to the Fermi liquid. which form the so-called Luttinger liquid in one dimension.] + J ˜ms ˜
mn
ˆ ˆ ( S m Sn − 1 n m n n ) 4ˆ ˆ
(5. where mn are the nearest neigbours.c. they persist above the N´ el temperature and arise because the on-site Coulomb repulsion is larger e than the half-bandwidth. if U > zT (a) [143]. However.

which support the unconventional symmetry of the order parameter in some cuprates.1 2 Heisenberg antiferromagnet with the exchange energy J= 4T 2 (a) U (5. In our view [12]. ˆ ˆ ˆ ˆα Sm =
1 2 ss
c† (τα )ss cms ˜ms ˜
are three components of the on-site spin. 3) and τα are the 2 ˜ Pauli matrices as in equation (3. where the direct Coulomb repulsion is diminished due to the d-wave symmetry of the BCS order parameter k . However.e. the diffractive scattering between holes of cuprate superconductors needs to be absent. While some groups claimed that they describe high-Tc superconductors at ﬁnite doping. Therefore. But in one-dimensional interacting fermion systems the scattering is non-diffractive due to topological reasons. 2. As a result an analytical solution of the t– J model is a very hard problem but not in the dilute limit. which dress a single hole. one
.1 eV. The second term in H describes the spin. the Hubbard U and t– J models share an inherent difﬁculty in determining the order.73). In this limit a doped hole dressed by antiferromagnetic spin ﬂuctuations propagates coherently in a narrow band of the order of J T (a) like a small polaron. The exchange energy leads to spin polarons similar to lattice polarons but dressed with magnetic ﬂuctuations rather than with phonons. One could believe that the same spin ﬂuctuations. The possibility of d-wave BCS-like superconductivity in the two-dimensional Hubbard model near half-ﬁlling was suggested by Scalapino et al [149] concurrently with the discovery of novel superconductors.10). the problem with the RVB-like concepts of high-Tc is that for them to be valid. J should be of the order of 0. The magnetic singlet pairing might be effective in the d-channel (section 2. mediate a superconducting pairing of spin polarons due to an effective attractive interaction of the order of J . To be consistent with the N´ el temperature of parent undoped cuprates and with their e spin-wave excitation spectrum. other authors could not ﬁnd any superconducting instabilty without an additonal (i.2).1 operator (α = 1. the non-fermionic commutation relations for the projected electron operators lead to an additional kinematic interaction between carriers doped into the Mott insulator. it has been concluded that models of this kind are highly conﬂicting and confuse the issue by exaggerating the magnetism rather than clarifying it [151]. while other experiments appear to contradict this symmetry (see section 8. e–ph) interaction [150].3)
for the nearest neigbours.High-Tc superconductors: different concepts
165
ˆ where the projected electron operators cm↑ = cm↑ (1 − n m↓ ) act in the subspace ˜ without double occupancy. A known case where diffractive scattering is absent is that of one-dimensional interacting fermion systems. There is now a variety of phase-sensitive experiments. n m = n m↑ + n m↓ . Independent of any experimental evidence. In two and higher dimensions.

some exact solutions of various limits and numerical studies on t– J and Hubbard models led to the conclusion that the electronic structure in cuprates is much more prone to inhomogeneity and intermediate-scale structures such as stripes of hole-rich domains separated by insulating antiferromagnetic domains. there are measurements of out-of-plane resistivity ρc [152. In this scenario. especially in the underdoped regime.6. this ‘stripe’ scenario is incompatible with Coulomb’s law: there are no stripes if the long-range Coulomb repulsion is properly taken into account. The dressing causes an increase in the electron’s effective mass and when the dressing is large.166
Competing interactions in unconventional superconductors
does not have such a topological constraint.2). is that of small polarons with a nonlinear interaction with a background bosonic degree of freedom which gives rise to an effective mass enhancement that depends on the local charge occupation. Moreover. As we discuss in section 8. Strong repulsive correlations between holes could provide another novel mechanism for Cooper pairing. i.e. the kinetic-energy-driven mechanism of superconductivity proposed by Hirsch [156] (see ﬁgure 5. where their interaction is allegedly attractive due to spin ﬂuctuations. doped fullerenes and MgB2 . If. The undressing process can only occur if the carriers are ‘holes’ rather than electrons and when two hole
. Such a mechanism does not require any dynamic attraction between holes. electrons in the metal pair even if the interaction between them is repulsive.2)). it has been proposed that the stripes are essential to a high-Tc mechanism [141].2).1) and (5. which deﬁnitely fail to account for dρc /dT > 0 and high-Tc of highly oxygenated cuprates. As a result. recent neutron [154] and X-ray [155] spectroscopic studies did not ﬁnd any bulk charge/spin segregation in the normal state of a few cuprates suggesting an absence of inplane carrier density modulations in these materials above Tc . It is thus favourable for the holes to pair so that the pairs can spread out somewhat into the antiferromagnetic neighbourhood of the stripe. stripe formation permits hole delocalization in one direction. On the experimental side. The qualitative explanation is as follows: electrons in metals are ‘dressed’ by a cloud of other electrons with which they interact and form the so-called electronic polarons. The proximity effect in conventional superconductor–normal metal structures is a prototypical example of such a mechanism of pairing: when the BCS superconductor and a normal metal are placed in contact with each other.2). A model Hamiltonian. which describes the ‘undressing’. the electrons manage to ‘undress’ when the temperature is lowered. however. the metal is a poor conductor. which are behind the RVB concept. 153] which show the metalliclike temperature dependence of ρc and the mean-free path in the c-direction comparable with the interplane spacing in a number of high-Tc cuprates. On the experimental side. their effective mass will be reduced and electricity will ﬂow easily. but hole motion transverse to the stripe is still restricted. Another problem of the microscopic Hamiltonians (equations (5. is that they neglect the longrange Coulomb and electon-phonon interactions. Welldeveloped interlayer tunnelling invalidates these and any other low-dimensional concepts (see ﬁgure 5. which are essential in novel superconductors (section 5.

High-Tc superconductors: different concepts
167
carriers form a pair. It follows from these data that the local magnetic moments are practically absent from optimally doped cuprates. The quasi-elastic peak and diffuse magnetic scattering are virtually absent in the metallic phase of YBa2 Cu3 O7−δ in a wide energy range up to 30 meV. respectively. There are several semi-phenomelogical concepts for the behaviour of a non-Fermi liquid of cuprates. where x is the doping level like in La2−x Srx CuO4 . It possesses a well-deﬁned Fermi surface but it is not the ordinary Landau–Fermi liquid. According to [157]. As an alternative to a threedimensional Bose–Einstein condensation of bipolarons (chapter 4). these authors
.164] dismiss any real-space pairing claiming that pairing is collective even in underdoped cuprates. In particular. A linear frequency dependence of the imaginary part of the electron self-energy does not follow from the e–ph or Coulomb interactions (see chapter 3). this observation implies a change in the kinetic energy anticipated in the theory of ‘hole’ superconductivity [156] and in the bipolaron theory [13]. In particular. Less exotic concepts are based on the Fermi-liquid and BCS-like approach but take into account the van Hove singularities of the electron density of states [158]. obtained on the high ﬂux reactor at Grenoble. 142. Some authors attribute high Tc and non-Fermi liquid features of the cuprates to the proximity of the Fermi level to a van Hove singularity in the density of states. 159. Pines and his co-workers argued that the magnetic interaction between planar quasi-particles in cuprate superconductors leads to a new quantum state of matter. it is conceivable that novel superconductors are marginal Fermi liquids (MFL) [163] in the sense that the energy interval near the Fermi level in which the Landau quasi-particles exist is small compared with Tc . 160]. because of a singular interaction between electrons with their Fourier component peaked at some momentum caused by soft magnetic ﬂuctuations. an MFL needs a singular interaction like NAFL. that this model yields Tc 10 K for both La2−x Srx CuO4 and YBa2 Cu3 O7−δ . Therefore. High-precision optical conductivity measurements [157] observed a temperature-dependent change in the conductivity sum rule (1. Other authors [141. The model clearly contradicts the neutron data by Smith et al [161]. where it is cubic or quadratic. They believe in a large Fermi surface with the number of holes (1 + x) rather than x in superconducting cuprates.18) in the optimally doped Bi-cuprate integrated up frequencies a few orders of magnitude greater than Tc . Hence. Outside the interval a linear energy dependence of the quasi-particle scattering rate has been postulated. However. which allows for a description of optical responce in some cuprates. when its parameters are constrained by neutron-scattering and transport measurements. the van Hove singularity scenario enhances Tc much less effectively than weak-coupling ad hoc calculations would suggest. which might be caused by a closeness to a quantum phase transition of any type. the nearly antiferromagnetic Fermi liquid (NAFL). other authors [162] conclude in the framework of the Eliashberg theory (chapter 3). and/or a singular interaction between holes due to a closeness to a quantum phase transition of any type [138.

In this concept. 166–169]. and to a bandnarrowing effect (section 4.168
Competing interactions in unconventional superconductors
suggest a collective pairing (i. is that the extension of BCS theory towards the strong interaction between electrons and ion vibrations (chapter 4) describes the phenomenon naturally and that high-temperature superconductivity exists in the crossover region of the electron–phonon interaction strength from the BCSlike to bipolaronic superconductivity as was predicted before [11]. However. unambiguously established for many oxides. 64. magnetic and kinetic measurements. the Cooper pairs in the momentum space) at some temperature T ∗ > Tc but without phase ordering. should lead to a high-energy spectral weight in the spectral function of an oxygen hole at energies about twice the Franck–Condon (polaron) level shift. Here a d-band of a transition metal is split into the lower and upper Hubbard bands due to on-site repulsive interaction (the Hubbard U ).3 and 7. E p . Our view. and explored in greater detail after discovery [13. For although high-temperature superconductivity has not yet been targeted as ‘the shame and despair of theoretical physics’—a label attributed to superconductivity during the ﬁrst half-century after its discovery—the parlous state of current theoretical constructions has led to a current consensus that there is no consensus on the theory of high-Tc superconductivity. rather than to the total density of carriers (1 + x). Experimental [170–177] and theoretical [64. 178–182] evidence for an exceptionally strong electron– phonon interaction in high temperature superconductors is now so overwhelming.
. which show that only holes doped into a parent insulator are carriers in the normal state (chapter 6). A strong coupling with phonons. On theoretical grounds this preformed Cooper-pair (or phase-ﬂuctuation) scenario contradicts a theorem [165]. 84. as explained in sections 5. that even advocates of a non-phononic mechanism [141] accept this fact.
5. their one-particle DOS is schematically represented by ﬁgure 5. while the ﬁrst band to be doped is the oxygen band lying within the Hubbard gap. It also contradicts a parameter-free estimate of the Fermi energy and Tc in the cuprates.2 Band structure and essential interactions in cuprates
In this book we take the view that cuprates and related transition metal oxides are charge transfer Mott insulators at any level of doping [183]. a large Fermi surface is clearly incompatible with a great number of thermodynamic.3). the phase coherence and superconducting critical temperature Tc are determined by the superﬂuid density. a single oxygen hole has well-deﬁned quasi-particle properties in the absence of interactions with phonons and ﬂuctuations of d-band spins. which proves that the number of supercarriers (at T = 0) and normal-state carriers is the same in any clean superﬂuid. which we discuss in the rest of the book. respectively.3. As a result. The oxygen band is less correlated and completely ﬁlled in parent insulators.e. As established in a few site-selective experiments [184] and in ﬁrst-principle (the so-called ‘LDA+U’) band-structure calculations. which is proportional to doping x due to a low dimensionality.1.

It might enter the oxygen band in overdoped cuprates (b). ﬁgure 5. One of their roles in the one-particle DOS is to pin the chemical potential inside the charge transfer gap. Then a Fermi-level crossing could be seen in the angle-resolved photoemission spectra (ARPES) (section 8. c. which is the Brillouin zone (BZ) is determined by the lattice constants a. In particular.5).4). we apply o an expression for E p .Band structure and essential interactions in cuprates
169
Figure 5.4 and 6. which depends only on the experimentally measured highfrequency. half the bipolaron binding energy above the oxygen band edge. To assess quantitatively the role of the Fr¨ hlich e–ph interaction.7. b. ∞ .3.4)
BZ
−1 −1 where κ −1 = ∞ − 0 . ﬁgure 5. dielectric constants of the host insulator as [187] Ep = 1 2κ d3 q 4πe2 (2π)3 q 2 (5. This binding energy as well as the singlet– triplet bipolaron exchange energy are thought to be the origin of normal-state pseudogaps. Density of states in cuprates.3.
The e–ph interaction also binds holes into inter-site oxygen bipolarons the size of a lattice constant (section 4. as ﬁrst proposed by us [185] (sections 6.4). In overdoped samples. Here the size of the integration region. the binding energy and the hole spectral function depend on doping. the bipolaron and polaron bands could overlap because the bipolaron binding energy becomes smaller [186].15 per cell) at any relevant level of doping and the residual repulsive interaction of bipolarons is strongly suppressed by the lattice polarization owing to a large static dielectric constant (section 4. 0 . The bipolaron density remains relatively low (below 0.3(b). and static.4). The chemical potential is inside the charge transfer gap due to bipolaron formation (a) and bipolarons coexist with thermally excited non-degenerate polarons. where bipolarons may coexist with unpaired degenerate polarons. As shown in
. carriers screen part of the e–ph interaction with low-frequency phonons. Hence. shifted by E p . so the chemical potential could enter the oxygen band in the overdoped cuprates. That is why bipolarons in oxides are fairly well described by the charged Bose-gas model of section 4.

The polaron shift about 1 eV results in the formation of small polarons no matter which criterion for their formation is applied. Polaron shift E p due to the Fr¨ hlich interaction.21473 4.266
5. equation (5.4
a Distance between CuO planes.643 0.982 0. System SrTiO3 BaTiO3 BaBiO3 La2 CuO4 LaMnO3 La2−2x Sr1+2x -Mn2 O7 NiO TiO2 MgO CdO WO3 NaCl EuS EuSe
∞ 0
˚ a × b × c ( A3 ) 3.9683 6.2 5. the variational criterion for small polaron formation fails.445 0.032 4.7 5.429 0. Exact Monte Carlo calculations [94] show that the polaron theory based on
. Hence.183 4.32 3.7 5.9c 4.863 3. fullerenes and MgB2 .4 6–7.4).842 0.807 0.9922 × 4.170
Competing interactions in unconventional superconductors
Table 5.749 0.31 × 7.0 5.19363
E p (eV) 0.1. c Distance between MnO planes. The correct criterion for small polaron formation is determined by the convergence of the 1/λ perturbation expansion as discussed in chapter 4. ω0 ² 0. In this non-adabatic or intermediate regime. the Fr¨ hlich interaction alone provides the binding energy of two holes o (≈2E p ) almost by one order of magnitude larger than the magnetic interaction J ∼ 0.1–5.0
310 1499 30.342 × 4.4 5 2.9053 3.592 × 2. 2 b Ishikawa T private communication. Moreover.9 100–300 5.852 0.54 × 7.1 9. There is virtually no screening of c-axis polarized optical phonons in cuprates because an upper limit for the out-of-plane plasmon frequency (º200 cm−1 [188]) is well below the phonon frequency (section 4. The data in the Table represent lower boundaries for the polaron shift.647 0.73 7.0 3.3 5.324 0. Data taken from the 1997 o Handbook of Optical Constants of Solids ed E D Palik (New York: Academic). so even a naive variational criterion of the small polaron formation (E p > D) could be satisﬁed. 2
table 5.6433 5. since the deformation potential and/or the molecular-type (Jahn–Teller) e–ph interaction are not included in equation (5.522 0.2 2.1 eV of the t– J model.4) is perfectly applied at any doping.884 0.1 eV.816 21.96 4.9b 5. The bare half-bandwidth is about 1 eV or less in cuprates.964 5.1.82 × 6a 3.579 0.862 × 3.44 5.4). optical phonon frequencies are high in oxides. The value ∞ = 5 for WO3 is an estimate.90 11.9b 4.4 30 16b 38b 12 89–173 9.

i. The only exception is the minimum at k = (π.7) does Hence. g may depend on doping in cuprates. m ∗ are the density of holes and their i effective mass in each of the hole sub-bands i renormalized by the electron– phonon (and by any other) interaction. and n i . all their wavevectors would belong to the stars with two or more prongs. Because equation (5. In cuprates and MgB2 . It means that the spectrum will be degenerate with respect to the number of prongs which the star has.3 Low Fermi energy: pairing is individual in many cuprates
First-principle band-structure calculations show that copper.5) F = m∗ i where d is the interplane distance. and in MgB2 .
5. in cuprates the degeneracy is 1 not contain any other band-structure parameters. Applying the parabolic approximation for the band dispersion we obtain the renormalized Fermi energy as πn i d (5. which is g = 2 in MgB2 . in particular if a weaker interaction with the deformation potential is taken into account. the band structure is quasi-two-dimensional with a few degenerate hole pockets. π) with one prong and g = 1. It is sufﬁcient to overcome the inter-site Coulomb repulsion at short distances. doped fullerenes. The effective polaron–polaron attraction due to the overlap of the deformation ﬁelds is about 2E p (section 4. One can express the renormalized bandstructure parameters through the in-plane magnetic-ﬁeld penetration depth at T 0.7)
where g is the degeneracy of the spectrum. C60 .4). g 2. g 4. We show later that the adiabatic ratio in all novel superconductors is of the order or even larger than one for most essential optical phonons. alkali metals and magnesium donate their outer electrons to oxygen. respectively. and boron in cuprates. measured experimentally: λ−2 = 4πe2 H
i
ni . If the hole band minima are shifted with doping to BZ boundaries.6)
As a result. One expects four hole pockets inside the Brillouin zone (BZ) due to the Mott–Hubbard gap in underdoped cuprates. The groups of wavevectors for these stars have only one-dimensional representations.e. the estimate of F using this
. we obtain a parameter-free expression for the ‘true’ Fermi energy as
F
=
d 4ge2λ2 H
(5. m∗ i
(5.Low Fermi energy: pairing is individual in many cuprates
171
this expansion describes quantitatively both strong-coupling (small polaron) and weak-coupling (large polaron) regimes without any restriction on the value of E p for a long-range electron–phonon interaction with high-frequency phonons.

6) might be different from the total density of delocalized carriers in equation (5. It is about 10 A or even less. if the degeneracy g ≥ 2 is taken into account. As established experimentally in cuprates (section 5. when ω0 / F ² 1. The parameter-free expression (5. it is rather probable that all novel superconductors. at λ ≤ 0. the mean free path was ˚ found to be surprisingly large at low temperatures. so that the parameter-free expression for F . However.2. The estimate agrees with the measured phonon spectra. c µ∗ µc 1 because the Tolmachev logarithm is ineffective. obtain a reduced value for the zero-temperature superﬂuid density in the dirty limit. l ∼ 100–1000 A. Since the Fermi energy is small and phonon frequencies are high. M = 10 au.5) and (5.172
Competing interactions in unconventional superconductors
equation does not depend very much on the parabolic approximation for the band dispersion. In contrast. λ > µc .e. the Coulomb pseudopotential µ∗ is of the order of the bare Coulomb repulsion. F º ω0 . M (5. The renormalized Fermi energy in magnesium diboride and in more than 30 cuprates is less than 100 meV. it is well established that they are also in the nonadiabatic regime [190]. however. the non-crossing diagrams cannot be treated as vertex corrections as in [191].1).6) are not necessarily the same. the ratios n/m ∗ in equations (5. the renormalized Fermi energy is so small that
. Hence. where ξ(0) is the zero-temperature coherence length. A low Fermi energy is a serious problem for the Migdal–Eliashberg approach (chapter 3). one has to have a strong coupling. one obtains ω0 69 meV for MgB2 . l ξ(0).7). This should be compared with the characteristic phonon frequency.7) does not apply to doped fullerenes with their three-dimensional energy spectrum. M = 16 au for YBa2 Cu3 O6 .8)
With Z = 1. interaction with phonons must be treated within the multi-polaron theory (chapter 4) at any value of λ. Here Vcell is the volume of the (chemical) unit cell. N = 2/Vcell . l ξ(0). including MgB2 are in the clean limit. The latter was measured directly in cuprates as the size of the ˚ vortex core. Because novel superconductors are in the non-adiabatic regime. However. one cannot increase λ without accounting for the polaron collapse of the band (chapter 4). However. is perfectly applicable. since they are comparable to the standard terms.5). the high-frequency phonons are strongly coupled with carriers. Hence. One can. equation (5. where quantum ﬂuctuations are important. to get an c experimental Tc .2). and ω0 = 84 meV with Z = 2. A parameter-free estimate of the Fermi energy obtained by using equation (5. The ‘superﬂuid’ density in equation (5. Even in the region of the applicability of the Eliashberg theory (i.5). Generally. In many cases (table 5. N = 6/Vcell . ω0 = 4π Z 2 e2 N . This is also true even in the extreme case of a pure two-dimensional superﬂuid. which is estimated as the plasma frequency of boron or oxygen ions.7) is presented in table 5. in a translationally invariant system they must be the same [165].

9 Sr0.066 HgBa2 CuO4.0 37.9 83.e.79 YBa2 Cu3 O6.76 Sr0.29 4.6 95. i.Low Fermi energy: pairing is individual in many cuprates
173
Table 5.15 CuO4 La1.22 CuO4 La1.055 HgBa2 CuO4.8 Sr0.2.29 4.2 78.9 89.8 60.5 9.29 4.097 HgBa2 CuO4.5 90.85 Sr0.4 66.5 9.5 3.6 HgBa2 CuO4.1 HgBa2 CuO4.
(5.2 55.5 9.25 CuO4 YBa2 Cu3 O7 YBaCuO(2%Zn) YBaCuO(3%Zn) YBaCuO(5%Zn) YBa2 Cu3 O6. Indeed.29 4.4 67.5 9.7 73.ab (A) 2000 1980 2050 2400 3200 2800 1400 2600 3000 3700 2100 2900 1861 1864 1771 2156 2150 2022 2096 2035 2285 2160 1610 2000 1530 1450 1650 1580 1560 1390 1560 1770 1400 ˚ d (A) 6.6 6.108 MgB2 Tc (K) 36.6 4.9) is well satisﬁed in underdoped and even
.4 92.29 4.5 20.5 95.0 78.7 YBa2 Cu3 O6. ² 100 meV. is smaller than the inter-carrier distance.9 63.5 9.2 CuO4 La1.2 27.7 YBa2 Cu3 O6.77 YBa2 Cu3 O6.101 HgBa2 CuO4.29 4.6 6.97)).055 HgBa2 CuO4.78 Sr0.29 9.92 YBa2 Cu3 O6. the bipolaron radius (equation (4.29 4.1 CuO4 La1.6 6.105 HgBa2 CuO4.88 YBa2 Cu3 O6.29 4.75 Sr0.1 39 ˚ λH.0 92.1 93.5 88.5 9. The Fermi energy (multiplied by the degeneracy) of cuprates and MgB2 [189]. this is the case. if
F
ºπ
.5 9.57 YBa2 Cu3 O6.9)
The bipolaron binding energy is thought to be twice the so-called pseudogap experimentally measured in the normal state of many cuprates (section 6.29 4.3 94.29 4.5 9.101 HgBa2 CuO4.5).096 HgBa2 CuO4.29 4.0 56.84 YBa2 Cu3 O6.0 56. so that equation (5.52 g F (meV) 112 114 106 77 44 57 148 43 32 21 66 34 84 84 92 62 63 71 66 70 56 138 248 161 274 305 236 257 264 332 264 205 122
pairing is certainly individual.5 9.049 HgBa2 CuO4.6 6.24 CuO4 La1.5 9.29 4.0 58.0 46.29 4.29 4.0 30.65 YBa2 Cu3 O6.0 24. Compound La1.5 68.0 70.5 87.6 6.74 YBa2 Cu3 O6.0 91.

9). The correct criterion is given by equation (5.4. the lattice deformations and the Coulomb repulsion fully taken into account but neglecting the kinetic energy terms. One should note that the coherence length in the charged Bose gas has nothing to do with the size of the boson.4. so that we can apply the Fr¨ hlich interaction with the coupling constant o γ (q) 1 q
to calculate the bipolaron mass. in La2 CuO4 [192]. Using these techniques net inter-site interactions of an in-plane oxygen hole with an apex hole. and of two in-plane oxygen holes were found to be atractive (the binding energies = 119 meV and = 60 meV.1 Apex bipolarons An apex bipolaron can tunnel from one unit cell to another via a direct singlepolaron tunnelling from one apex oxygen to its apex neighbour as in a ladder (section 4. Hence.1) eliminates an essential part of the electron–phonon interaction. Lattice polarization is coupled with the electron density. The canonical displacement transformation (section 4.174
Competing interactions in unconventional superconductors
in a few optimally and overdoped cuprates.6. It depends on the inter-particle distance and the mean-free path. respectively).3). 5.6). the interaction between holes can be analysed using computer simulation techniques based on a minimization of the ground-state energy of an ionic insulator with two holes.4 Bipolaron bands in high-T c perovskites
Consideration of particular lattice structures shows that small inter-site bipolarons are perfectly mobile even when the electron–phonon coupling is strong and the bipolaron binding energy is large (section 4. All other interactions were found to be repulsive. therefore the interaction is diagonal in the site representation and the coupling constant does not depend on a particular orbital state of the hole. They are doped charged-transfer ionic insulators with narrow electron bands. Therefore. Oxides are highly polarizable materials. it would be incorrect to apply the ratio of the coherence length to the inter-carrier distance as a criterion of the BCS–Bose liquid crossover. ﬁgure 5. The Bloch bands of those bipolarons in the perovskites are obtained using the canonical transformations of chapter 4 [100. and might be as large as in the BCS superconductor. Let us analyse the important case of copper-based high-Tc oxides. equation (4.262). 108]. The transformed Hamiltonian is given by ˜ H = e S H e−S = (T p − E p )
i(p)
n i(p) + (Td − E d )
i(d)
n i(d)
+
i= j
σi j ci† c j + ˆ
q
† ωq (dq dq + 1/2)
.
5.3.

13)
ν
0|σ pd |ν ν|σd p |0 ˆ ˆ ≈− e E0 − Eν Eg
where |ν . ˆ ˆ j
(5.10) without the third hopping term. |0 is the phonon vacuum and the reduction factor is 1 |γ (q)|2 (1 − cos[q · (m p − m p )]). t)dq − H.15)
u p (q.c. j
(2ωq u i (q)u ∗ (q) − Vi j )n i n j .11)
ω0 . which is the same for oxygen and copper ions E p = Ed =
q
|u j (q)|2 ωq .c. (5. E ν are eigenstates and eigenvalues of the transformed Hamiltonian (5.4. (5. . σ pp ≡ 0|σ pp |0 = T pp e ˆ σ pp ≡
(2) −g 2 pp 2 T pd −g 2 pp
(5. the bandwidth narrowing If the charge-transfer gap is large enough. the second-order indirect hopping (equation (5. Apex bipolaron tunnelling in perovskites.i.13)) is written as σ pp = −i where σ pd (t) = T pd exp ˆ
q † u ∗ (q.
−
1 2 q. t)dq − H.Bipolaron bands in high-Tc perovskites
175
9
Figure 5.14) g2 = pp 2N q These expressions are the result of straightforward calculations described later.16)
.
(5. Taking into account that E ν − E 0 = E g + q ωq n q . E g (2) factors are the same for the direct σ pp and the second-order (via copper σ pp ) oxygen–oxygen polaron hopping integrals (T = 0). exp d q (2) ∞ 0
dt e−iEg t 0|σ pd (t)σd p |0 ˆ ˆ
(5.12) (5.10)
The oxygen ( p) and the copper (d) diagonal terms include the polaron level shift.

let us say. both band-narrowing factors are much less than one can expect in the Holstein model. Here x sin t dt Si(x) = t 0 √ m = a/ 2 and m = a for the in-plane oxygen–oxygen.13) is obtained in the limit E g g2 = pp Ep ω0 1−
(5. Hence. .y orbitals and spins of the four plane oxygen ions in the † cell (ﬁgure 5. t. px3 and p y4 are involved or by |Ai | = 1/ 2 if only two of them are relevant. The probability amplitudes Ai are normalized by the condition |Ai | = 1/2. which is the same or opposite of the spin of the in-plane hole depending on the total spin of the bipolaron.4) and capex is the creation operator for the hole on one of the three apex oxygen orbitals with the spin. 2.14) yields (5. respectively.2E /ω for an in-plane hopping. p y2 . The bipolaron hopping integral. Calculating the bracket in equation (5.8 A 2 ≈ 0.20)
where i denotes the px. . The wavefunction of an apex bipolaron localized.10) of ﬁrst order with respect to the transfer integral responsible for the bipolaron tunnelling to the nearest-neighbour cell m + a is 2 apex ˜ (5. The matrix element of the Hamiltonian (5. is obtained by projecting the Hamiltonian (5.3E /ω for one obtains g pp p 0 p 0 an apex–apex one. . = T pd e
2 −g 2 −gd p pd
e
exp
−
q
[u d (q) − u p (q)][u ∗ (q) − u ∗ (q)]e−iωq t . Equation (5. and g 2 ≈ 0. one obtains
2 .18) ω0 .10) onto a reduced Hilbert space containing only empty or doubly occupied elementary cells and averaging the result with respect to phonons.17) If ωq is q-independent the integral in equation (5. and for apex–apex ˚ ˚ narrowing factors.21) t = m| H |m + a = |Ai |2 T pp e−g
. the phonon occupation numbers. in the cell m is written as
4
|m =
i=1
† Ai ci† capex |0
(5.176
Competing interactions in unconventional superconductors
Here u j (q.15). qd m
if we approximate the Brillouin zone by a sphere of the radius qd (the Debye approximation).19)
Si(qd m) . In cuprates with qd ≈ 0. t) ≡ u j (q) exp(−iωq t) and n q = 0. d p
(5.7 A−1 and a 3.17) as σ pp = −
(2) 2 T pd
Eg
e
2 −g 2 −gd p pd
∞ k=0
e
(−1)k (
q [u d (q)
− u p (q)][u ∗ (q) − u ∗ (q)])k d p
k!(1 + kω0 /E g )
. 1. . if four plane orbitals √ px1. .
Then equation (5.15) is calculated by the expansion of the exponent in equation (5.

The apex z bipolaron. are static but apical oxygens are independent three-dimensional isotropic harmonic oscillators. The bipolaron bandwidth is of the same order as the polaron one. m ∗∗ = x y 4m ∗∗ . 0) for ‘y’ bipolarons. z. as in equation (4. the next nearest
. we choose κx = κ y = κz / 2. Hence. We also include the bare NN hopping TNN . In this equation t is the renormalized hopping integral between p orbitals of the same symmetry elongated in the direction of the hopping ( ppσ ) and t is the renormalized hopping integral in the perpendicular direction ( ppπ).y (K ). which is a speciﬁc feature of the inter-site bipolarons (section 4.x
(5.2 In-plane bipolarons Here we consider a two-dimensional lattice of ideal octahedra that can be regarded as a simpliﬁed model of the copper–oxygen perovskite layer as shown in ﬁgure 5. For mathematical transparency. Two different bands are not mixed because T px . apex apex Their ratio t/t = T pp /T pp = 4 as follows from the tables of hopping apex integrals in solids. For a large part of the Brillouin zone near (0.3 is applied.y y. Due to poor screening. both copper and oxygen. the hole–apical interaction is purely Coulombic: gν (m − n) = κν |m − n|2
where ν = x. one can adopt the effective mass approximation EK =
x.269). The random potential does not mix them either if it varies smoothly on the lattice scale.5. As a result. respectively. the hole bipolaron energy spectrum consists of two bands E x. x 5. p = 0 y for the nearest neighbours. The lattice period is a = 1 and the distance between the apical sites and the central plane is h = a/2 = 0. we assume that all in-plane atoms.Bipolaron bands in high-Tc perovskites
apex
2
177
where T pp e−g is a single-polaron hopping integral between two apex ions. y.y 2 2 Ky Kx + 2m ∗∗ 2m ∗∗ x.22)
with K x. π) for ‘x’ and (π. formed by an overlap of px and p y apex polaron orbitals. we can distinguish ‘x’ and ‘y’ bipolarons with a lighter effective mass in the x or y direction.6. To account for the experimental fact that z-polarized√ phonons couple to holes stronger than others [172].y taken relative to the band bottom positions and m ∗∗ = 1/t. The direct hole–hole repulsion is Vc Vc (n − n ) = √ 2|n − n | so that the repulsion between two holes in the nearest-neighbour (NN) conﬁguration is Vc .4. if formed is approximately four times less mobile than x and y bipolarons.6). respectively.4. so that the model Hamilonian of section 4.

80E p and 0.82E p .c. vNNN = √c − 0. the Cartesian coordinates are n = (n x + 1 .23E p then inter-site NN bipolarons form.82E p and the mass renormalization exponents as gNN =
E 1 (k) = −E p − 2t [cos k x + cos k y ] − 4t cos(k x /2) cos(k y /2).23)
where the factor two accounts for two layers of apical sites. m = (m x . (Since gNNN ≈ (gNNN)2 one can neglect the difference between NNN hoppings within and between the octahedra. The result of such a projection is a bipolaron Hamiltonian: Hb = (Vc − 3.24)
Performing the lattice summations for the NN. as shown in ﬁgure 5. NNN and NNN conﬁgurations one ﬁnds Vph = 1.38(E p /ω0 ).23E p )
l
[ A† A l + Bl† Bl + C l† C l + Dl† Dl ] l + Bl† C l + C l† Dl + Dl† A l + H. The polaron shift is given by the lattice sum (4.
(5. after summation over polarizations. 0).60E p /ω0 ).148) which.23E p . one should retain only these lowest-energy conﬁgurations and discard all the processes that involve conﬁgurations with higher energies. we obtain V a net inter-polaron interaction as vNN = Vc − 1. |m − n |3 |m − n|3
(5.26) l−x
.5.23E p . The polaron–polaron attraction is
2 Vph (n − n ) = 4ω0 κx m
h 2 + (m − n ) · (m − n) . For reference. C and D. m x .15κx ω0
(5. As a result. m y .38E p /ω0 ) and 2 NNN hopping t = TNNN exp(−0.178
Competing interactions in unconventional superconductors
neighbour (NNN) hopping across copper TNNN and the NNN hopping between the pyramids TNNN. 0. vNNN =
2 0. Since in general t > t .] (5.) A single-polaron spectrum is. In the ﬁrst order in the renormalized hopping integral.]
−t
l
[ A † Bl l
−t
n
† † [ A† Bl + Bl+ y C l + C l†+x Dl + Dl− y A l + H. The polarons tunnel in the square lattice with NN hopping t = TNN exp(−0.59(E p /ω0 ). gNNN = 0. respectively.60(E p /ω0 ) and (gNNN)2 = 0. h) and 2 2 n x . At Vc > 1.80E p . therefore. the mass is mostly determined by the NN hopping amplitude t. no bipolarons are formed and the system is a polaronic Fermi liquid.c. V √c 2 2 2 − 0. The bipolarons tunnel in the plane via four resonating (degenerate) conﬁgurations A. Let us now discuss different regimes of the model.23E p . n y . B. yields
2 E p = 2κx ω0 m
h2 1 + 4 |m − n| |m − n|6
2 = 31. If Vc < 1. m y are integers.25)
The polaron mass is m ∗ = 1/(t + 2t ). n y + 1 .

Four degenerate in-plane bipolaron conﬁgurations A. The bipolaron binding energy is ≈ 1. The bipolaron already moves in the ﬁrst order of polaron hopping.Bipolaron bands in high-Tc perovskites
179
Figure 5.26E p − 4t 2 + t 2 . 0) and y = (0.71Vc − 3. x = (1. We recall that in the Holstein model m ∗∗ ∝ exp(2E p /ω) (section 4. C and D.60E p /ω). This remarkable property is entirely due to the strong on-site repulsion and long-range electron–phonon interaction that leads to a non-trivial connectivity of the lattice. In the ﬁrst order. The ﬁrst-order polaron hopping processes mix the states resulting in √ ground-state linear combination with the energy a E 3 = 2. 1). The dominance of the potential energy over kinetic in the transformed Hamiltonian enables us to readily investigate these many-polaron cases.23E p − Vc . It is essential that between the squares such triads could move only in higher orders of polaron hopping.5. Some single-polaron hoppings are indicated by arrows.
where l numbers octahedra rather than individual sites. With an even stronger e–ph interaction. This fact combines with a weak renormalization of t yielding a superlight bipolaron with the mass m ∗∗ ∝ exp(0. (5. Vc < 1.and four-particle clusters. B.27)
There are four bipolaronic sub-bands combined in the band of width 8t . The effective mass of the lowest band is m ∗∗ = 2/t . Thus the mass of the Fr¨ hlich o bipolaron in the perovskite layer scales approximately as a cubic root of that of the Holstein one.16E p . NNN bipolarons become stable. A Fourier transformation and diagonalization of a 4×4 matrix yields the bipolaron spectrum: E 2 (k) = Vc − 3.1).80E p . More importantly.6.
.23E p ) + Vc / 2 − 0. holes can now form three.23E p ± 2t [cos(k x /2) ± cos(k y /2)]. Three holes placed within one√ oxygen square have four degenerate states with the energy 2(Vc − 1.

stripes or strings is strictly prohibited (section 8.3) are perfectly applied in the normal and superconducting states. triplet n t and polaron. Its energy is E 4 = 4(Vc − 1.80E p ) = 5. have enough space to move.4) and the formation of inﬁnite clusters. 5. The system is effectively a charged Bose gas.7).7.41Vc − 6. the system quickly becomes a charge segregated insulator. so only the Coulomb repulsion is relevant.23E p ) + 2(Vc / 2 − 0.5 Bipolaron model of cuprates
The considerations set out in sections 5.4 leads us to a simple model of cuprates [179]. as well as all bigger ones. where y is found using the thermal equilibrium of singlet and triplet bipolarons and polarons: 2n s + 2n t + n p = x.180
Competing interactions in unconventional superconductors
they are immobile. The hard-core interaction does not play any role in this dilute limit.16E p .23E p . (5. When the carrier 1 (as in cuprates)). This indicates that the superconducting state in cuprates requires a rather ﬁne balance between electronic and ionic interactions.3 and 5. This cluster. The fact that within the window. are also immobile in ﬁrst-order polaron hopping.1). so that the canonical Boltzmann kinetics (section 1. A cluster of four holes has only one state √ within a square of oxygen atoms. Above Tc a material such as YBa2 Cu3 O6+x contains a non-degenerate gas of hole bipolarons in singlet and triplet states.16E p < Vc < 1. The population of singlet.7. respectively.52E p . We would like to stress that at distances much larger than the lattice constant the polaron–polaron interaction is always repulsive (section 4.1) and the Bogoliubov excitations of the charged Bose gas (section 4. As the undoped plane has a half-ﬁlled Cu3d9 band. that we have found.
5. Hole pairs. n s . The main assumption is that all electr ons are bound into small singlet and triplet inter-site bipolarons stabilized by e–ph and spin–ﬂuctuation interactions. 1. Triplets are separated from singlets by a spin-gap J and have a lower mass due to a lower binding energy (ﬁgure 5. and in their band-width renormalization as described in chapter 4.2. and they are responsible for low-energy kinetics. there are no three or more polaron bound states means that bipolarons repel each other. bipolaronic operators are almost density is small (n b bosonic (section 4. We conclude that at Vc < 1. there is no space for bipolarons to move if they are inter-site. is quite narrow. which appear with doping. This repulsion is signiﬁcantly reduced due to a large static dielectric constant in oxides ( 0 1). The superconductivity window. Hence. The main part of the electron–electron correlation energy (Hubbard U and inter-site Coulomb repulsion) and the electron–phonon interaction are taken into account in the binding energy of bipolarons .6). Their Brillouin zone is half the original electron one and is completely ﬁlled with hard-core bosons. n p bands is determined by the chemical potential µ ≡ T ln y. carriers are almost-free charged bosons and thermally excited nondegenerate fermions.6). which is a superconductor (section 4.28)
.

w is the half-bandwidth of the polaron band.Bipolaron model of cuprates
181
Figure 5. Here x is the total number of holes per unit area.29). we obtain for 0 y < 1: πx T (5. which naturally includes the addition of an extra excitation band. Bipolaron picture of high-temperature superconductors. A indicates degenerate singlet and triplet inter-site bipolarons. The crosses are copper sites and the circles are oxygen sites. because doping inevitably introduces some disorder. and y = 1 in the superconducting state. t is the half-bandwidth of the bipolaronic bands. /2 is the bipolaron binding energy per polaron and J is the exchange energy per bipolaron.29) in the normal state. We should also take into account localization of holes by a random potential. If the polaron energy spectrum is (quasi-)onedimensional (as in section 6.6.
Applying the effective-mass approximation for quasi-two-dimensional energy spectra of all particles.5). an additional T −1/2 appears in front of the logarithm in the third term on the left-hand side of equation (5. B shows non-degenerate singlet and triplet inter-site bipolaron. The Coulomb repulsion −m ∗∗ ln(1 − y) − 3m ∗∗ ln(1 − ye−J /T ) + m ∗ ln(1 + y 1/2 e− s t
/(2T )
)=
.

182

Competing interactions in unconventional superconductors

restricts the number of charged bosons in each localized state, so that the distribution function will show a mobility edge E c [194]. The number of bosons in a single potential well is determined by the competition between their long-range Coulomb repulsion ( 4e2/ξ ) and the binding energy, E c − . If the localization length diverges with the critical exponent ν < 1 (ξ ∼ (E c − )−ν ), we can apply a ‘single-well–single-particle’ approximation assuming that there is only one boson in each potential well (see also section 8.1). Within this approximation localized charged bosons obey the Fermi–Dirac statistics, so that their density is given by Ec NL (E) dE (5.30) n L (T ) = y −1 exp(E/T ) + 1 −∞ where NL (E) is the density of localized states. Near the mobility edge, it remains constant NL (E) n L /γ , where γ is of the order of the binding energy in a single random potential well, and n L is the number of localized states per unit area. The number of empty localized states turns out to be linear as a function of temperature in a wide temperature range T º γ from equation (5.30). Then the conservation of the total number of carriers yields for the chemical potential: π(x − 2n L ) = − m ∗∗ ln(1 − y) − 3m ∗∗ ln(1 − ye−J /T ) s t T 2πn L ln(1 + y −1 ). + m ∗ ln(1 + y 1/2e− /(2T ) ) − γ If the number of localized states is about the same as the (n L ≈ x/2), a solution of this equation does not depend on wide temperature range above Tc . With y to be a constant (y range of parameters in equation (5.31)), the number of singlet Bloch states is linear in temperature: n s (T ) = (x/2 − n L ) + T nL ln(1 + y −1 ). γ

(5.31)

number of pairs temperature in a ≈ 0.6 in a wide bipolarons in the

(5.32)

The numbers of triplets and polarons are exponentially small at low temperatures, T J, . The model suggests a phase diagram of the cuprates as shown in ﬁgure 5.7. This phase diagram is based on the assumption that to account for the high values of Tc in cuprates, one has to consider electron–phonon interactions larger than those used in the intermediate-coupling theory of superconductivity (chapter 3). Regardless of the adiabatic ratio, the Migdal–Eliashberg theory of superconductivity and the Fermi-liquid theory break at λ ≈ 1 (section 4.2). A many-electron system collapses into a small (bi)polaron regime at λ ≥ 1 with well-separated vibration and charge-carrier degrees of freedom. Even though it seems that these carriers should have a mass too large to be mobile, the inclusion of the on-site Coulomb repulsion and a poor screening of the long-range electron–phonon interaction leads to mobile inter-site bipolarons (section 4.6).

Bipolaron model of cuprates

183

Figure 5.7. Phase diagram of superconducting cuprates in the bipolaron theory (courtesy of J Hofer).

Above Tc , the Bose gas of these bipolarons is non-degenerate and below Tc , their phase coherence sets in and hence superﬂuidity of the double-charged 2e bosons occurs. Of course, there are also thermally excited single polarons in the model (ﬁgure 5.6). There is much evidence for the crossover regime at T ∗ and normal-state charge and spin gaps in cuprates (chapter 6). These energy gaps could be understood as half of the binding energy and the singlet–triplet gap of preformed bipolarons, respectively [185] and T ∗ is a temperature, where the polaron density compares with the bipolaron one. Further evidence for bipolarons comes from a parameter-free estimate of the renormalized Fermi energy F , which yields a very small value as discussed in section 5.3. There might be a crossover from the Bose–Einstein condensation to a BCS-like polaronic superconductivity (section 4.5) across the phase diagram [179]. If the Fermi liquid does exist at overdoping, then it is likely that the heavy doping causes an ‘overcrowding effect’ where polarons ﬁnd it difﬁcult to form bipolarons due to a larger number of competing holes. Many experimental observations can be explained using the bipolaron model (chapters 6, 7 and 8).

Chapter 6 Normal state of cuprates

**6.1 In-plane resistivity and Hall ratio
**

Thermally excited phonons and (bi)polarons are well decoupled in the strongcoupling regime of the electron–phonon interaction (chapter 4), so that the standard Boltzmann equation of section 1.1 for renormalized carriers is applied. We make use of the τ approximation [193] in an electric ﬁeld E = ∇φ, a temperature gradient ∇T and in a magnetic ﬁeld B ⊥ E, ∇T . Bipolaron and single-polaron non-equilibrium distributions are found to be f (k) = f 0 (E) + τ where F = (E − µ)∇T /T + ∇(µ − 2eφ) f 0 (E) = [y −1 exp(E/T ) − 1]−1 for bipolarons with the energy E = k 2 /(2m ∗∗ ) and with the Hall angle ∗∗ and b = 2eBτb /m F = (E + /2 − µ/2)∇T /T + ∇(µ/2 − eφ) f 0 (E) = {y −1/2 exp[(E + /2)/T ] + 1}−1 E = k 2 /(2m ∗ ) and = p = eBτ p /m ∗ for thermally excited polarons. Here n = B/B is a unit vector in the direction of the magnetic ﬁeld. The Hall angles are assumed to be small ( 1), because the polarons and bipolarons are heavy. Equation (6.1) is used to calculate the electrical and thermal currents induced by the applied thermal and potential gradients: jα = aαβ ∇ β (µ − 2eφ) + bαβ ∇ β T wα = cαβ ∇β (µ − 2eφ) + dαβ ∇β T. 184 (6.2) (6.3) = ∂ f0 v · {F + ∂E n × F} (6.1)

**In-plane resistivity and Hall ratio
**

Equation (6.2) deﬁnes the current with the polaronic conductivity σ p e2 τ p n p /m ∗ , where kinetic coeffﬁcents are given by ax x = a yy = a yx bx x b yx 1 σ p (1 + 4 A) 2e 1 = − ax y = σ p ( p + 4 A b ) 2e σp −µ [ p+ + 2 A( b − µ/T )] = b yy = e 2T σp −µ = − bx y = + 2 A b( p p+ e 2T

185 =

(6.4)

b

− µ/T ) .

**Equation (6.3) deﬁnes the heat ﬂow with coefﬁcients given by cx x = c yy = c yx dx x σp [T p + /2 + eφ + 2 A(T b + 2eφ)] 2e2 σp = − cx y = 2 [ p (T p + /2 + eφ) + 2 A b(T b + 2eφ)] 2e σp −µ = d yy = 2 T γ p + p ( − µ/2 + eφ) + ( /2 + eφ) 2T e + A[T γb + d yx = − dx y = +A Here = and γ = σp e2 +
**

b (2eφ

(6.5)

− µ) − 2eφµ/T ]

p(

p

T γp +

− µ/2 + eφ) + ( /2 + eφ)

−µ 2T

b [T γb

b (2eφ

− µ) − 2eφµ/T )] .

∞ 2 0 dE E ∂ f 0 /∂ E ∞ T 0 dE E∂ f 0 /∂ E ∞ 3 0 dE E ∂ f 0 /∂ E ∞ T 2 0 dE E∂ f 0 /∂ E

=

2 (z, 2, 1) (z, 1, 1) 6 (z, 3, 1) (z, 1, 1)

(6.6)

=

**are numerical coefﬁcients, expressed in terms of the Lerch transcendent
**

∞

(z, s, a) =

k=0

zk (a + k)s

p,

with z = y in

b , γb

and z = −y 1/2 exp[− /(2T )] in A= m ∗ τb n b m ∗∗ τ p n p

γ p, (6.7)

186

Normal state of cuprates

is the ratio of bipolaron and polaron contributions to the transport, which strongly depends on temperature. Here, for simplicity, we neglect the spin gap, which is small in optimally doped cuprates [195]. Then the bipolaron singlet and triplet states are nearly degenerate, so that bipolaron and polaron densities are expressed as 2m ∗∗ T | ln[1 − y]| π m∗T np = ln 1 + y 1/2 exp − π 2T nb = (6.8) . (6.9)

Using the kinetic coefﬁcients (equations (6.4) and (6.5)), we obtain ρ= and RH = 1 σ p (1 + 4 A) 1 + 4 A b/ p en p (1 + 4 A)2 (6.10)

(6.11)

for the in-plane resistivity and the Hall ratio, respectively. According to [194], the main scattering channel above Tc is due to particle– particle collisions with the relaxation time τb,p 1/T 2 . In case of the Bose– Einstein statistics, umklapp scattering can be neglected, so the scattering between bosons in extended states does not contribute to the resistivity in a clean sample. However, the inelastic scattering of an extended boson by bosons localized in the random potential and the scattering of extended bosons by each other make a contribution because the momentum is not conserved in two-particle collisions in the presence of the impurity potential. In the framework of the ‘single-well– single-particle approximation’, the repulsion between localized bosons plays the same role as the Pauli exclusion principle in fermionic systems. Then the relaxation rate is proportional to the temperature squared because only bosons within the energy shell of the order of T near the mobility edge contribute to the scattering and because the number of ﬁnal states is proportional to T in a non-degenerate gas. The chemical potential is pinned near the mobility edge, so that y ≈ 0.6 in a wide temperature range, if the number of localized states in the random potential is about the same as the number of bipolarons (section 5.5). This could be the case in YBa2 Cu3 O6+x , where every excess oxygen ion x can localize the bipolaron. Hence, neglecting the polaron contribution to the in-plane transport which is exponentially small below T ∗ , we ﬁnd that ρ and RH T 1 T (6.12)

(6.13)

In-plane resistivity and Hall ratio

187

Figure 6.1. In-plane resistivity and the Hall ratio of bipolarons and thermally excited polarons for /2 = 600 K and b / p = 2m ∗ τb /(m ∗∗ τ p ) = 0.44 (inset), and the Hall Lorenz number as a function of temperature compared with the experimental data by Zhang et al in the optimally doped YBa2 Cu3 O6.95 [196]. The second inset shows the ratio of the Hall Lorenz number to the Lorenz number.

as observed in many cuprates. Taking into account the polaron contribution in equations (6.10) and (6.11) does not change this result, if the binding energy is sufﬁciently large, as shown in ﬁgure 6.1. One can also take into account the transport relaxation rate due to a two-dimensional boson–acoustic phonon scattering, 1 T (6.14) τb−ac and the elastic relaxation rate due to the scattering by unoccupied impurity wells with their density proportional to temperature, 1 τb−im T. (6.15)

Hence, neglecting the polaron contribution, we ﬁnally obtain [194] RH ≈ 1 2e(x/2 − n L + bn L T ) (6.16)

These observations have been explained [202] as a result of the resonance scattering of bipolarons by lattice defects. Remarkably. one would require a measurement of the normal-state conductivity down to very low temperatures. are only a tiny part of the non-Fermi-liquid characteristics of cuprates. with the anticipated divergent resistivity and temperature-independent anisotropy at low temperatures.6). This then allows one to scrutinize the nature of the electronic structure of cuprates without the ‘complication’ of superconductivity at elevated temperatures. of course. On the basis of this simple—but potent—deﬁnition. However. These expressions contain important information about the number of bosons. they lead to a smaller C in equation (6. Hence. The answer to this vexing question clearly depends on how precisely one can deﬁne a metal.2 Normal-state resistivity below T c
The scaling of the in-plane resistivity and of the Hall ratio with doping tells us that cuprates are doped semiconductors. many high-Tc cuprates studied to date by this technique appear to be doped three-dimensional anisotropic insulators. This important experiment has been carried out by Boebinger et al [201] by applying an ultra-high magnetic ﬁeld (up to 60 T) to destroy the superconducting state. C is a temperatureindependent constant. 197]. since the conductivity at T = 0 K is inﬁnite. which might be metallic Fermi liquids. discussed earlier.4). they predict the Hall ratio and resistivity scaling with doping as 1/x because n L x. any superconductor must clearly be a metal. while the electrical conductivity is zero in non-metals as T → 0 K. Importantly.17) and to a smaller temperature gradient of the in-plane resistivity at temperatures above J. This change was explained by Mott [200]: the spin gap is the singlet–triplet separation energy in the bipolaron model (ﬁgure 5. They ﬁt well the experimental data in a number of cuprates [194.5.
6. To arrive at one single deﬁnition. Thus. one can suppress the superconducting state by applying a large external magnetic ﬁeld to measure a ‘true’ normal-state conductivity down to a very low temperature well below a zero ﬁeld Tc . one can ask a fundamental question: ‘What are high-Tc cuprates in their normal state—metals or something else?’.
. Because triplets are lighter than singlets.17)
where σb is the relative boson–boson scattering cross sections. where γ was introduced in section 5. This scaling was experimentally observed in underdoped and optimally doped cuprates [198]. where the spin gap is observed in NMR (section 6. But the linear (in temperature) resistivity and the temperature-dependent Hall ratio. [199]) below the temperature. for example. There is a characteristic change in the linear slope of ρ(T ) (see.188
Normal state of cuprates
ρ≈C T + σb T 2 x/2−n L + bn L T (6. localized states and the relative strength of the different scattering channels. metals show a non-zero low-temperature conductivity. where their density becomes comparable with the singlet density. proportional to the carrier mass squared and b is about 1/γ .

This simple expression ﬁts the Hall ratio in La2−x Srx CuO4 at optimum doping (x = 0. The number of extended bosons n b (T ) above the mobility edge is determined in the ‘single-well–single-particle’ approximation (equation (5. The Hall ratio. so that 1 RH (T ) = (6.15) (ﬁgure 6.20)) over U
. Let us now extend the theory towards low temperatures. U is the well depth and Umin = π 2 /8m ∗∗ a 2 with the well size a. RH . For slow particles.20)
π2 U Umin −1 16 Umin
(6. There is a wide distribution of potential wells with respect to both U and a in cuprates.18)
where 1 x is the total number of pairs and n L (T ) n L − NL (0)T is the number 2 of bosons localized by the random potential with NL (0) the density of localized states near the mobility edge. If the total number of carriers 1 x is 2 larger than the number of the potential wells n L . one has to integrate the Wigner cross section (equation (6.Normal-state resistivity below Tc
189
As we discussed in the previous section. there is no need to abandon the Boltzmann kinetics to explain the linear in-plane resistivity and the temperaturedependent Hall effect above Tc in cuprates in the framework of the bipolaron theory. The characteristic kinetic energy of non-degenerate bipolarons in the normal state is of the order of the temperature rather than the Fermi energy in metals and in conventional semiconductors. Therefore.30)) as n b (T ) = 1 x − n L (T ) 2 (6.19) RH (0) 1 + T /TL where TL = (x − 2n L )/2NL (0). while the boson–boson inelastic scattering rate is proportional to T 2 . A fraction of bipolarons is localized by disorder. The most effective scattering at low temperatures is then caused by attractive shallow potential wells. This allows us to explain that both the in-plane resistivity and the inverse Hall ratio are proportional to T . so that the number of delocalized carriers is proportional to T at sufﬁciently high temperatures. where the transport relaxation time of bipolarons is determined by elastic boson–impurity scattering and single polarons are frozen out.21)
is the energy of a shallow virtual (U < Umin ) or real (U > Umin ) localized level. measures the inverse carrier density.2) with TL = 234 K and the constant RH (0) = 2 × 10−3 cm3 C−1 . this is described by the familiar Wigner resonance cross section as σ (E) = where =− 1 2π ∗∗ E + | | m
2
(6. The transport relaxation rate is the sum of the scattering cross sections from √ different potential wells within the unit volume multiplied by the velocity v = 2E/m ∗∗. the carrier density is practically temperature independent at low temperatures.

single impurities are screened. we can extend the integration over a up to inﬁnity.15 CuO4 (triangles [198]) described by equation (6.23)
for E E 0 . The result is: π2 √ 8a 2m ∗∗ E 0 (6. because the U -averaged cross section (equation (6.22)) behaves like 1/a 2 at a large a.85 Sr0.19). NL (0) n L /γ .190
Normal state of cuprates
Figure 6. However. Here A is the width of the size distribution of the random potential 2 wells. one has to take into account the fact that the Wigner formula (equation (6. which are not screened. we take into account only shallow wells with U < Umin because the deeper wells are occupied by localized carriers and cannot yield a resonant scattering.22) where the width of the U -distribution. As a result. However. ˚ We expect a very large value of A of the order of 10 A or more due to the twin boundaries and impurity clusters. γ . E 0 = π 4 /128m ∗∗amin and amin is the minimum size. is supposed to be large compared with Umin .
and over a. we obtain the inverse mean free path: σ (E) ≡ γ −1
Umin
σ (E) dU =
4π tan−1 √ m ∗∗ γ a 2m ∗∗ E
l −1 (E) =
nL A
∞ amin
σ (E) da
π 2 NL (0) E0 √ ln ∗∗ A 2m ∗∗ E E m
(6. The Hall ratio in La1. Performing the integration over U . Integrating over the size a.2. A simple estimate of the screening radius by the use
.20)) is applied only to slow particles with a ≤ √ / 2m ∗∗ E.

.25) ρ(T ) = ρ0 ln T
−1 where ρ0 = 2(x − 2n L )e2 τ0 /m ∗∗ .18)).24)
where τ0 = m ∗∗2 A/π 2 NL (0) is a constant. which corresponds to quite a large E 0 = 1500 K if m ∗∗ = 10m e .87 Sr0. In-plane resistivity of La1.Normal-state resistivity below Tc
191
Figure 6. at low temperatures.1) scattering rates and taking into account the temperature dependence of the extended boson density n b (T ) (equation (6.2 × 10−5 cm. The value of E 0 agrees well with the estimate. Hence. Because the bipolaron energy spectrum is three dimensional at low temperatures.3. The full curve in ﬁgure 6. The low-temperature resistivity is now derived by the use of the Boltzmann theory: E0 (6. E 0 = 1900 K and Tb = 62 K which appears to be remarkably good.
of the classical expression.3 is a ﬁt to the experimental data with ρ0 = 7. Combining both elastic (equation (6.25)) and inelastic (section 6. there is no temperature dependence of the anisotropy ρc /ρab at low T as observed.13 CuO4 (diamonds [201]) described by equation (6. we ﬁnd that
ln(E 0 /T ) + (T /Tb )2 ρ(T ) = ρ0 1 + T /TL
(6. we arrive at the logarithmic transport relaxation rate as 1 1 E0 ≡ vl −1 (E) = ln τ τ0 E (6.26). rD = T 0 /(16πn b e2 ) yields a value of amin of ˚ the order of the interatomic spacing (∼2 A).26)
with a constant Tb .

However. and the phonon term.192
Normal state of cuprates
6. The contribution from the carriers. the value of L H just above Tc was found to be about the same as that predicted by the bipolaron model (L H ≈ 0.4) and (6. 207]. Takenaka et al [204] found that κe is constant or weakly T -dependent in the normal state of YBa2 Cu3 O6+x . As a result. the experimental L H showed a strong temperature dependence which violates the Wiedemann–Franz law. the bipolaron model explains the violation of the Wiedemann–Franz law in cuprates and the Hall Lorenz number as seen in the experiment. κe . Remarkably. Unfortunately. 205]. Here κe .15L e —due to the double charge of carriers and also due to their nearly classical distribution function above Tc . and the elementary charge. implies the violation of the Wiedemann–Franz law (since resistivity is found to be a nonlinear function of temperature) in the underdoped region.3 Lorenz number: evidence for bipolarons
Kinetic evidence for 2e bipolarons in cuprates (sections 6.5).2) is strong but direct evidence that these materials contain a charged 2e Bose liquid in their normal state is highly desirable. respectively. The breakdown of the Wiedemann–Franz law has also been seen in other cuprates [206. σ and e are the electronic thermal conductivity. Here we demonstrate that the Wiedemann–Franz law breaks down because of the interference of polaron and bipolaron contributions to the heat transport [209].15L e). Reference [203] predicted a very low Lorenz number L b of bipolarons—L b = 6L e /(4π 2 ) ≈ 0. This experimental observation is hard to explain in the framework of any Fermiliquid model. and the thermal Hall
. Then the thermal conductivity.95 because the transverse thermal κx y and the electrical σx y conductivities involve only electrons. the electrical conductivity. This constraint allows us to express the electric and chemical potential gradients ∇(µ − 2eφ) via the temperature gradient ∇T using equations (6. When thermally excited polarons are included. As a result. Mott and the author [203] discussed the thermal conductivity κ. κph . are comparable to each other in cuprates. There is no electric current ( j = 0) in the thermal conductivity measurements. κ.1 and 6. therefore. the extraction of the electron thermal conductivity has proven difﬁcult since both the electron term. depends strongly on the elementary charge as ∼(e∗ )−2 and should be signiﬁcantly suppressed in the case of e∗ = 2e compared with the Fermiliquid contribution. the ‘Hall’ Lorenz number (L H ≡ L x y = (e/kB )2 κ yx /(T σ yx )) has been directly measured in YBa2 Cu3 O6. In particular. A new way to determine the Lorenz number has been realized by Zhang et al [196] based on the thermal Hall conductivity. given by the Wiedemann– Franz ratio. the Lorenz number L (= (e/kB )2 κe /(T σ )) differs signiﬁcantly from the Sommerfeld value L e (= π 2 /3) of standard Fermi-liquid theory. if the carriers are bipolarons. Some experiments have attemped to get around this problem in a variety of ways [204. This approximately T -independent κe . The thermal Hall effect allowed for an efﬁcient way to separate the phonon heat current even when it is dominant.

which describe an interference between the polaron and bipolaron contributions in the heat ﬂow. However. The bipolaron model. As discussed in section 6. are found to be κ = dx x − κ yx = d yx cx x bx x dx x c yx bx x cx x (ax x b yx − a yx bx x ) − + 2 ax x ax x
193
(6. In the limit of a purely polaronic system (i. this contribution is exponentially small because the number of unpaired polarons is small. These terms are mainly responsible for the breakdown of the Wiedemann–Franz law in the bipolaronic system. The model gives
. equations (6.e.6 in a wide temperature range. which is reminiscent of the contribution of the electron-hole pairs to the heat ﬂow in semiconductors [193]. In the opposite limit of a purely bipolaronic system (i.1. we obtain a reduced Lorenz number [203]: L = L H = L b .29)) are L = L H = L p . (1 + 4 A)2 (1 + 4 A b/ p ) (6. The contribution appears as a result of the recombination of a pair of polarons into a bipolaronic bound state at the cold end of the sample.29)
2 Here L p = (γ p − 2 ) and L b = (γb − b )/4 are the polaron and bipolaron p Lorenz numbers. A = ∞). To reduce the number of ﬁtting parameters.p 1/T 2 . also ﬁts the Hall Lorenz number measured by Zhang et al [196]. in general. As a result.27)
and the Lorenz and Hall Lorenz numbers to be A[2 p − b + /T ]2 L p + 4 AL b + 1 + 4A (1 + 4 A)2 L p + 4 AL b b / p LH = 1 + 4 A b/ p L= + A(4 A + b / p )[2 p − b + /T ]2 . so that y ≈ 0.28) and (6.28)
(6.29) yield temperature-dependent Lorenz numbers that differ signiﬁcantly from both limits.29)) which is the ratio of the bipolaron and polaron Hall angles b / p . which ﬁts the in-plane resistivity and the Hall ratio. In the low-temperature regime (T ). there is only one ﬁtting parameter in L H (equation (6. it is enhanced by the factor ( /T )2 and becomes important in the intermediate-temperature range Tc < T < T ∗ . The main difference originates in the second terms in the righthand side of equations (6. we take the charge pseudogap /2 = 600 K. However. the main scattering channel above Tc is due to particle–particle collisions with a relaxation time τb. The chemical potential is pinned near the mobility edge.Lorenz number: evidence for bipolarons
conductivity. κx y .29). A = 0).28) and (6. if the number of localized states in the random potential is about the same as the number of bipolarons.28) and (6. as found by Mihailovic et al [195] for nearly optimally doped YBa2 Cu3 O6+x in their systematic analysis of charge and spin spectroscopies.e. the Lorenz numbers (equations (6.

Because the thermal Hall conductivity directly measures the Lorenz number. lower inset).30)
ˆ where Ai is an operator acting on the nuclear spin and j is its nearest-neighbour site.c. Let us see how the bipolaron model describes the temperature dependence of 1/T1 [185].
6. l = 0. When we subtract the electronic contribution determined by using the Lorenz number in ﬁgure 6.32)
where f (E) = [exp(E + J − µ)/T −]−1 is the distribution function and 2t is the bandwidth of triplet bipolarons. It also describes well the (quasi-)linear in-plane resistivity and the inverse Hall ratio.8. we obtain BT sinh(t/T ) 1 = 2 T1 t [cosh[(t + J )/T − ln y] − cosh(t/T )] where B is a temperature-independent hyperﬁne coupling constant. it can be used to measure the lattice contribution to the heat ﬂow as well. B 1 =− 2 T1 t
2t
dE
0
∂ f (E) ∂E
(6.l
j. j
(6. As a result.1. the lattice contribution to the diagonal heat ﬂow appears to be much higher than anticipated in the framework of any Fermi-liquid model. The conventional contact hyperﬁne coupling of nuclear spin on a site i with electron spins is described by the following Hamiltonian in the site representation: ˆ Hi = A i
j
c†↑ c j ↓ + H.31) j. (6.l=l
Here l.194
Normal state of cuprates
quite a good ﬁt (ﬁgure 6.4 Spin pseudogap in NMR
Pairing of holes into singlets well above Tc should be seen as a drop of magnetic susceptibility. we obtain the effective spin–ﬂip interaction of triplet bipolarons with the nuclear spin: Hi b† b j.1). Indeed. ±1 are the z-components of spin S = 1. a rapid decrease in the uniform magnetic susceptibility and of the nuclear magnetic relaxation rate 1/T1 with temperature lowering is a common feature of the normal state of novel superconductors. The NMR width due to the spin-ﬂip scattering of triplet bipolarons on nuclei is obtained using the Fermi– Dirac golden rule as in section 2.c. as observed in cuprates (inset in ﬁgure 6. (6.33)
. we take their DOS as a constant (= 1/(2t)).44. For simplicity.l + H. Performing transformations to polarons and bipolarons as described in chapter 4.1) with a reasonable value of b / p = 0.1. The Hall Lorenz number appears to be slightly larger then the Lorenz number (ﬁgure 6.

also drops well above Tc in many copper oxides. The value of the singlet–triplet bipolaron exchange energy J . It ﬁts the experimental data with reasonable values of the parameters.
6. 213]. 212.c-axis transport and charge pseudogap
195
Figure 6. Temperature dependence of the nuclear spin relaxation rate 1/T1 [209] compared with the theory for J = 150 K and t = 250 K showing the absence of the Hebel–Slichter peak of Cu NMR in YBa2 Cu4 O8 .4. There is only a tiny change in the slope in 1/T1 (T ) near Tc due to a kink in the temperature dependence of the chemical potential µ(T ) caused by the Bose–Einstein condensation of singlets (appendix B). 197.
Equation (6. determined from the ﬁt to the experimental NMR width. as shown in the inset. and polarized [211] neutron scattering.4). NMR is almost unaffected by the superconducting transition in the bipolaron model.33) describes all the essential features of the nuclear spin relaxation rate in copper-based oxides: the absence of the Hebel–Slichter coherent peak below Tc . because there are no coherence factors (section 2. A similar unusual behaviour of the NMR was found in underdoped YBa2 Cu3 O6+x . is close to the ‘spin’ gap observed above and below Tc in YBa2 Cu3 O6+x with unpolarized [210]. The Knight shift. which measures the spin susceptibility of carriers. in agreement with the bipolaron model. and in many other cuprates.5
c-axis transport and charge pseudogap
The bipolaron theory quantitatively explains the c-axis transport and anisotropy of cuprates [186. the temperature-dependent Korringa ratio (1/T T1 ) above Tc and a large value of 1/T1 due to the small bandwidth 2t. The latter can propagate across the
.8) for triplets. The crucial point is that polarons dominate in the c-axis transport at intermediate and high temperatures because they are much lighter than bipolarons in the c-direction. t and J (ﬁgure 6.

we have a mixture of non-degenerate quasi-two-dimensional bosons and thermally excited fermions. and µB is the Bohr magneton. respectively. Polarons mainly contribute to c-axis transport. which are capable of propagating along the c-axis. Polarons are not degenerate at any temperature neither are bipolarons degenerate above Tc . x) = − 2 ∂E 0 ∞ ∂ fp χs (T. σpc (E) (6. so that f b ≈ y exp − and E T (6.38) T If the scattering mechanism is the same for polarons and bipolarons.4).40) N p (E) 4πda 2E
. x) = − 4 ∂ fb ∂E 0 ∞ ∂ fp dE σpc (E) σc (T. (6. we expect a highly anisotropic polaron energy spectrum with a quasi-one-dimensional polaron density of states as also observed in high-resolution ARPES experiments (section 8. x) = − 2µ2 dE N p (E) B ∂E 0 dE σb (E)
∞
(6. Quite generally the in-plane bipolaron and c-axis polaron. if the temperature is not very low. conductivities (σab and σc respectively) and the uniform spin susceptibility (χs ) are expressed as follows: σab (T. let us consider a system containing only singlet pairs and thermally excited polarons.39)
There is a large difference in the values of the ppσ and ppπ hopping integrals between different oxygen sites (section 5. Therefore. which leads to a fundamental relation between anisotropy and magnetic susceptibility [186].34) (6.35) (6. polarons and inter-site bipolarons propagate with about the same effective mass. the polaron contribution to the in-plane transport is small at low temperatures due to their low density compared with the bipolaron density. respectively) is independent of the energy and doping f p ≈ y 1/2 exp − σb (E) ≡ A = constant. To illustrate the point. Along the planes.37)
E + /2 . the ratio of the differential bipolaron and polaron conductivities (σb (E) and σpc (E). which is much less probable than a single-polaron tunnelling (chapter 4).36)
where f b = [y −1 exp(E/T ) − 1]−1 and f p = [y −1/2 exp(E/T + /2T ) + 1]−1 are the bipolaron and polaron distribution functions.4): m ∗ 1/2 1 (6. Hence. As a result.196
Normal state of cuprates
planes only due to a simultaneous two-particle tunnelling.

The bipolaron density (n b (T. suceptibility are expressed as
Then the anisotropy and the spin
ρc (T. x)/(T m ∗∗ ).41)
exp −
2T
. (a) Experimental anisotropy [214] compared with equation (6.c-axis transport and charge pseudogap
197
Figure 6. x) = µ2 B 2da ym ∗ 2π T
1/2
(6. x) 2T and χs (T. x) = 2 Ay 1/2 exp ρab (T.
where a is the in-plane lattice constant. x) per cm2 ) is linear in temperature and doping
.41) and (b) using experimental magnetic susceptibilities of La2−x Srx CuO4 [215]. is calculated taking ab into account the localization of bipolarons in a random potential (section 5. y = 2πn b (T.42)
The chemical potential.5.
(6.5).

where lowenergy charge and spin degrees of freedom are due to holes doped into a parent insulating matrix with no free carriers and no free spins. They also argued that a similar spectral shape and systematic trends in both the photoconductivity of optically doped dielectric samples and the infrared conductivity of chemically doped high-Tc oxides indicate that carriers in a
. YBa2 Cu3 O6 .15). in particular in Bi2 Sr2 CaCu2 O8+δ [197].44)
exp −
2T
We expect some dependence of the binding energy on doping because of screening. different experiments are consistent with 0 (6. x) x T 1/2 χ x T
s (T. Importantly. x) ρab (T.5.5(b)). the uniform magnetic susceptibility above Tc increases with doping (ﬁgure 6. x) ≈ x T /γ a 2) in a wide temperature range above Tc which ﬁts the temperature and doping dependence of the Hall ratio RH 1/(2en b) (section 6. which well screen the electron–phonon interaction with low-frequency phonons. (6.2). as shown in ﬁgure 6. In a wide range of doping near the optimum value (x = 0. Bipolarons are heavy non-degenerate particles.3 in ﬁgure 6.43). It proves once again that cuprates are doped insulators.6 Infrared conductivity
Studies of photoinduced carriers in dielectric ‘parent’ compounds like La2 CuO4 .5 is explained by the contribution of polarons to the in-plane transport. If the binding energy is below 100 K. One can describe all qualitative features of c-axis resistivity and magnetic susceptibility of La2−x Srx CuO4 using equations (6.
6.5.44) and (6.45) = x where 0 is doping independent.45) without any ﬁtting parameters.198
Normal state of cuprates
(n b (T. x) χs (T. Anisotropy is quantitatively described by equation (6. YBa2 Cu3 O6+x [213] and HgBa2 CuO4+δ [212]. and others demonstrated the formation of self-localized small polarons or bipolarons.43) . Mihailovic et al [170] described the spectral shape of the photoconductivity with the small polaron transport theory. The temperature-independent anisotropy of a heavily doped sample.43) using the experimental values of χs (T. A rather low magnetic susceptibility of parent insulators in their paramagnetic phase is presumably due to a singlet pairing of copper spins as discussed in section 5. x = 0. As a result. x) 1/2
(6. An exponential temperature dependence for the c-axis resistivity and anisotropy was also interpreted within the framework of the bipolaron model in other cuprates. x). (6. the polaron contribution ﬂattens the temperature dependence of the anisotropy. In particular. we ﬁnd the temperature-independent y x and ρc (T.

Its characteristic maximum is due to the incoherent spectral weight of the polaron Green function (section 4.6(b).7. One of the qualitative observations. the peak frequency shifts toward lower values as the Tc of the material increases. The critical temperature of the superconducting transition is proportional to 1/m ∗ in bipolaron theory (section 4. therefore. to a higher Tc . (a) Photoinduced infrared conductivity in the insulator precursors for Tl2 Ba2 CaCu2 O8 (top). YBa2 Cu3 O7 (middle) and La2−x Srx CuO4 (bottom) compared with the polaron infrared conductivity (broken curve).
. to the (bi)polaron mass. therefore.6(a).5). The infrared conductivity of high-Tc superconductors (broken curves) is compared with the photoinduced infrared conductivity (full curves) in respective insulator precursors in ﬁgure 6. (b) the infrared conductivity for superconducting samples (broken curves) compared with (a). The measured photoinduced infrared conductivity σ (ν) (full curves) in insulating parent compounds is shown in ﬁgure 6.6. A lower peak frequency corresponds to a lower mass of bipolaronic carriers and. The peak position in σ (ν) is related to the polaron level shift and.6(a).
concentrated (metallic) regime retain much of the character of carriers in a dilute (photoexcited) regime. which follow from a comparison of the infrared conductivities of insulating and superconducting materials is that in all perovskites exhibiting superconductivity.Infrared conductivity
199
Figure 6.5) as follows from a comparison with a theoretical small-polaron infrared conductivity (broken curves) in ﬁgure 6.

It monitors the time evolution of the muon spin polarization. Tc is fairly approximated by McMillan’s formula (3.1 Parameter-free description of T c
An ultimate goal of the theory of superconductivity is to provide an expression for Tc as a function of some well-deﬁned parameters characterizing the material. because the experimental Tc was to be found about three or more times lower than the BEC temperature. Since bare electron bands are narrow. He concluded that cuprates are neither BCS nor BEC superﬂuids but that they exist in the crossover region from one to the other.Chapter 7 Superconducting transition
7. Hence. Uemura [218] established a correlation of Tc with the in-plane magnetic ﬁeld penetration depth measured by the µs R technique in many cuprates as Tc 1/λ2 . 200
. However.96). Just before the discovery [1]. But applying a theory of this kind to high-Tc cuprates is problematic. one cannot increase λ without accounting for a polaron collapse of the band (chapter 4). strong correlations result in the Mott insulating state of undoped parent compounds. In the framework of BCS theory. Also. the Eliashberg equation (3. we predicted Tc would be as high as ∼100 K using an estimate of the bipolaron effective mass [105]. an estimate of Tc in cuprates within BCS theory appears to be an exercise in calculating µ∗ rather than Tc itself. µ∗ and λ are ill deﬁned in doped cuprates and polaronic effects are important as in many doped semiconductors [13]. The technique is based on the ab implantation of spin-polarized muons. which ﬁts the experimentally measured Tc in many cuprates for any level of doping.74) for the gap function properly takes into account a realistic phonon spectrum and retardation of the electron–phonon interaction. This appears at λ 1 for uncorrelated electrons (holes) and even at a smaller value of bare electron–phonon coupling in strongly correlated models [216]. which works well for simple metals and their alloys. As a result. the bipolaron theory provides a parameter-free expression for Tc [217]. Tc is calculated using the density sum rule as the Bose–Einstein condensation (BEC) temperature of 2e charged bosons on a lattice.

y ∗∗ 2m x
+
2K 2 y. t⊥ /(kB Tc ) and m ∗∗ = 2 /(2|t⊥ |d 2 ). we can express the band-structure parameters via the in-plane magnetic ﬁeld penetration depth λab = m ∗∗ m ∗∗ x y 8πn B e2 (m ∗∗ + m ∗∗ ) x y
1/2
and out-of-plane penetration depth.i=(x.Parameter-free description of Tc
201
Here we calculate the Tc of a bipolaronic superconductor taking properly into account the microscopic band structure of bipolarons in layered cuprates as derived in section 5. Fortunately. but also the out-of-plane. λc .x ∗∗ 2m y
+ 2t⊥ [1 − cos(K z d)]
(7. It describes the experimental data for a few dozen different samples clearly indicating that many cuprates are in the BEC rather than in the crossover regime. [exp(E iK /Tc ) − 1]−1 = n b
K . equation (5.31 2(n B /2)2/3 (m ∗∗ m ∗∗ m ∗∗ )1/3 x y c (7. magnetic ﬁeld penetration depth and a normal state Hall ratio RH just above the transition. One can apply the effective mass approximation at T Tc . We arrive at a parameter-free expression for Tc .y)
(7.4)
.4. The energy spectrum of bipolarons is twofold degenerate in cuprates (section 5. λab . c This expression is rather ambiguous because the effective mass tensor as well as the bipolaron density n b are not well known.22). The bipolaron density is expressed through the in-plane Hall ratio (above the transition) as RH = 4m ∗∗ m ∗∗ 1 x y × ∗∗ 2en b (m x + m ∗∗ )2 y (7.4). which in contrast to [218] involves not only the in-plane.1)
where d is the interplane distance and t⊥ is the inter-plane bipolaron hopping integral. Also a three-dimensional correction to the spectrum is important for BEC (appendix B) which is well described by the tight-binding approximation: EK =
x.2)
one readily obtains Tc (in ordinary units): kB Tc = f t⊥ kB Tc × 3.y 2K 2 x. Substituting the spectrum (equation (7.1 as a function of the anisotropy. because Tc should be less than the bipolaron bandwidth. λc = m ∗∗ c 16πn b e2
1/2
(we use c = 1).3)
where the coefﬁcient f (x) ≈ 1 is shown in ﬁgure 7.1)) into the density sum rule.

202
Superconducting transition
Figure 7.2.64 e RH λ4 λ2 ab c
1/3
.1 and ﬁgure 7. The Hall ratio has a strong temperature dependence above Tc (section 6.64 f t⊥ kB Tc e RH λ4 λ2 ab c
1/3
.01. For almost all samples. then the theoretical Tc would be almost the same as the experimental values for these samples as well. In a few cases (mercury compounds).1.5)
Here Tc is measured in kelvin. If we assume that the degeneracy of the bipolaron spectrum is removed by the random potential of Zn. The coefﬁcient f is about unity in a very wide range of t⊥ /(kB Tc ) ≥ 0. There are a few Zn-doped YBCO samples (ﬁgure 7. where RH (Tc + 0) is unknown. Correction coefﬁcient to the three-dimensional Bose–Einstein condensation temperature as a function of anisotropy. Hence.
(7. the theoretical Tc ﬁts experimental values within an experimental error bar for the penetration depth (about ±10%). the bipolaron theory yields a parameter-free expression. we take the inverse chemical density of carriers (divided by e) as RH .1).6)
We compare the last two expressions with the experimental Tc of more than 30 different cuprates.
which leads to Tc = 1. we use the experimental Hall ratio just above the transition.1.
. ﬁgure 7. Therefore. which unambiguously tells us how near the cuprates are to the BEC regime: Tc ≈ Tc (3D) = 1.
(7.2) whose critical temperature is higher than the theoretical one. for which both λab and λc have been measured along with RH (Tc + 0) in table 7. e RH in cm3 and λ in cm.

2 1.75 1.097) Hg(0.055) Hg(0.7 4.21) Y(0. Should this be the case.2) La(0. for La2−x Srx CuO4 (La).0 56.8 60.5 87.1.3 94.08) Y(0.16) Y(0.3 2. respectively.5 4.45 1. The KT critical temperature can be expressed through the in-plane
.9 63.5 95.75 1.85 4.1 93. and calculated values of Tc .Parameter-free description of Tc
203
Table 7.26) Y(0.2
Tc (3D) 38 35 32 33 25 17 111 45 35 26 31 14 87 75 83 47 46 48 43 35 28 23 43 28 56 79 61 66 70 88 69 58
Tc 41 36 32 39 31 19 114 46 36 26 51 28 88 82 89 59 58 60 54 50 42 60 92 69 105 120 99 105 107 127 106 90
TKT 93 95 88 65 36 47 172 50 38 30 77 40 98 97 108 73 73 83 77 82 65 115 206 134 229 254 197 216 220 277 220 171
36.23) Y(0.7 73. YBaCuO (x%Zn) (Zn).2 8. YBa2 Cu3 O7−x (Y) and HgBa2 CuO4+x (Hg) compounds.48 4.0 46.0 24.3) Y(0.0 91. due to a large anisotropy.2 1.24) La(0.108) Tc
exp
λab 200 198 205 240 320 280 140 260 300 370 210 290 186 186 177 216 215 202 210 204 229 216 161 200 153 145 165 158 156 139 156 177
λc 2 540 2 620 2 590 3 220 4 160 3 640 1 260 1 420 1 550 1 640 4 530 7 170 1 240 1 565 1 557 2 559 2 630 2 740 2 880 3 890 4 320 16 200 10 300 12 600 7 040 3 920 4 390 4 220 3 980 3 480 3 920 3 980
RH 0.62 0.9 89.2 78.1
One can argue that.0 92.9 2.4 67.066) Hg(0.2 1.52 1.0 58.0 56.7 1.5 9. cuprates may belong to a twodimensional ‘XY ’ universality class with the Kosterlitz–Thouless (KT) critical temperature TKT of preformed bosons [219.055) Hg(0. Compound La(0.4 66.4 92.1) La(0.12) Y(0.101) Hg(0.22) La(0.8 0.55 1. Experimental data on Tc (K).3) Y(0.101) Hg(0.9 83.2 1.0 37. ab and c penetration depth (nm).2 55.0 1.5 68.25) Zn(0) Zn(2) Zn(3) Zn(5) Y(0.0 0. 220] or Cooper pairs [164].096) Hg(0.5 20.35) Y(0.049) Hg(0.0 78.1) Hg(0.2 6.1 2. Hall coefﬁcient (10−3 (cm3 C−1 )).2 8.5 90.2 27.4) Hg(0.4 4.6 1.0 70. one could hardly discriminate the Cooper pairs with respect to bipolarons.7 4.43) Y(0.0 30.6 95.105) Hg(0.66 4.8 1.3 4.5 88.15) La(0.

Rev.204
Superconducting transition
Figure 7. 61 1093 and Panagopoulos C Private communication in underdoped and overdoped La2−x Srx CuO4 . Mod. Phys.1). from Grebennik V G et al 1990 Hyperﬁne Interact. respectively.4) favour 3D Bose–Einstein condensation of charged bosons as the mechanism for high Tc rather than any low-dimensional phase-ﬂuctuation scenario. for Zn-substituted YBa2 Cu1−x Znx O7 (circles). Theoretical critical temperature compared with the experiment (the theory is exact for samples on the straight line) for LaSrCuO compounds (squares).3 and 7. our parameter-free ﬁt of the experimental critical temperature and the critical behaviour (sections 7.2. it does not mean that
. for YBa2 Cu3 O7−δ (triangles). which proves that the phase transition is not the KT transition. There are also quite a few samples with about the same λab and the same d but with very different values of Tc . Rev. The Hall coefﬁcient above Tc is taken from Carrington A et al 1993 Phys. However. B 48 13 051 and Cooper J R Private communication (YBa2 Cu3 O7−δ and YBa2 Cu1−x Znx O7 ) and from Hwang H Y et al 1994 Phys.9d
2
16πe2λ2 ab
. and from Hofer J et al 1998 Physica C 297 103 in HgBa2 CuO4+δ . In contrast. B 12 1007 and Janossy B et al 1991 Physica C 181 51 in YBa2 Cu3 O7−δ and YBa2 Cu1−x Znx O7 . Lett.
(7. and for HgBa2 CuO4+δ (diamonds).7)
It appears signiﬁcantly higher than the experimental values in many cases (see table 7. The ﬂuctuation theory [221] further conﬁrms the three-dimensional character of the phase transition in cuprates. 72 2636 (La2−x Srx CuO4 ).
penetration depth alone [164] kB TKT ≈ 0. J. Experimental data for the London penetration depth are taken from Xiang T et al 1998 Int.

0) Y(0. path integral [226]) and numerical [227] studies have addressed the intermediate-coupling regime beyond a variational approximation [223]. in particular overdoped samples.15) La(0. Moreover. an assumption [141] that the number of carriers is determined by Luttinger’s theorem (i.6 and 5.2. Mass enhancement in cuprates. we have to realize that this model. We note.5. They well argee with the inter-site bipolaron mass (sections 4.5 (or less) (section 2.Isotope effect on Tc and on supercarrier mass
Table 7.e. however. cannot be applied to a strong electron– phonon interaction and polaron–bipolaron crossover (section 4. However. signiﬁcantly reducing the crossover region.1) Y(0. n = 1 + x) would lead to much heavier carriers with m ∗ about 100m e. 224].2) La(0.2.0 11. As discussed in section 4. It is interesting to estimate the effective mass tensor using the penetration depth and the Hall ratio. that an absolute value of the effective mass in terms of the free electron mass does not describe the actual band mass renormalization if the bare (band) mass is unknown.2 Isotope effect on T c and on supercarrier mass
The advances in the fabrication of the isotope substituted samples made it possible to measure a sizeable isotope effect. Compound La(0.12) Y(0.6 mc 3558 2698 1909 584 600 1994
205
all cuprates are in the BEC regime with charged bosons as carriers. α = −d ln Tc /d ln M in many high-Tc oxides.2 8. which makes the BCS–BEC crossover problem relevant.
7. Several compounds showed
. including two-dimensional systems [227–229]. An essential effect of the polaron band-narrowing is missing in the Hubbard model.3 10. Some of them. which predicts a smooth BCS-BEC crossover.1 15. It strongly affects the BCS–BEC crossover. These estimates for in-plane and out-of-plane boson masses are presented in table 7.5). the polaron collapse of the bandwidth is responsible for high Tc . in using the negative Hubbard U model.4).5). Both analytical (diagrammatic [225].3) m ab 22. Starting from the pioneering works by Eagles [72] and Legget [222]. the isotope effect in cuprates was found to be quite different from the BCS prediction: α = 0. Nevertheless. this problem received particular attention in the framework of a negative Hubbard U model [223.3 7. This led to a general conclusion that phonons are relevant for high Tc .7. might be in the BCS or intermediate regime.

according to equations (4. and a small negative value of α was found in Bi-2223 [231].3.222)) which is linked with the isotope effect on Tc .226) for the isotope exponents of (bi)polaronic superconductors [82].225) and (4. Another prediction of the bipolaron theory is an isotope effect on the carrier mass (equation (4. could have α > 0. in particular large values in low-Tc cuprates. Moreover.
. the isotope exponent in Tc can distinguish the BCS-like polaronic superconductivity with α < 0 and the BEC of small bipolarons with α > 0. The isotope effect on the magnetic ﬁeld penetration depth in two samples of La2−x Srx CuO4 [173] (courtesy of J Hofer). Essential features of the isotope effect.226)) as observed.5 [230].226). the bipolaron mass increases and the BEC temperature Tc decreases in the bipolaronic superconductor. In contrast. underdoped cuprates. an overall trend to lower value as Tc increases [232] and a small or even negative α in some high-Tc cuprates were understood using equations (4. which are deﬁnitely in the BEC regime (section 5. an increase in the ion mass leads to band narrowing and an enhancement of the polaron density of states and to an increase in Tc in polaronic superconductors.3). Hence. With increasing ion mass.206
Superconducting transition
Figure 7.
α > 0.5 (equation (4.224) and (4.

1 1. 4 He) show the characteristic λ-point singularity in their speciﬁc heat but superﬂuid Fermi liquids like BCS superconductors exhibit a sharp second-order phase transition accompanied by a ﬁnite jump in the speciﬁc heat (section 2.6 2. cannot be adequately treated by mean-ﬁeld BCS theory even if the Gaussian ﬂuctuations are included. As was stressed by Salamon et al [238].53)) of the order of one. consequently.025 YBa2 Cu3 O7−0.3.Speciﬁc heat anomaly
207
Remarkably. the heat capacity is logarithmic near the transition and. where n is the hole density.27).5 Cu2 O7 Tl1. Using equation (5.895Sr0. so that the isotope effect on λab directly measures the isotope’s effect on the carrier mass.2 Sr2 Tl0. estimates using the Gaussian ﬂuctuations yield an unusually small coherence volume (table 7.
˚ Table 7.6).05 YBa2 Cu3 O7−0. The in-plane magnetic ﬁeld penetration depth.2 Cu3 O10 400 309 250 143 84
2 ˚ ξab (A2 )
˚ ξc (A) 3.
7.3).1 Ca0. the polaron mass 5 in this material. λab 320 nm. It was established beyond doubt [233–237] that the anomaly in high-Tc cuprates differs qualitatively from the BSC prediction.3 Speciﬁc heat anomaly
Bose liquids (or. according to equation (4.222). more precisely. In particular. The carrier density is unchanged with the isotope substitution of O16 by O18 . comparable with the unit cell volume [234] and Gi number (equation (1.8 Ba2 Ca2.105 CuO4 . enhancement should be m ∗∗ /m we obtain an in-plane bipolaron mass as large as m ∗∗ ≈ 10m e with the bare hopping integral T (NNN) = 0.2 1. we obtain Tc = 31 K in astonishing agreement with the experimental value Tc = 30 K in this compound. Compound YBa2 Cu3 O7 YBa2 Cu3 O7−0. The effect was observed in the London penetration depth of isotope-substituted cuprates (ﬁgure 7.2 <0. In particular. λc = 4160 nm and RH = 4 × 10−3 cm3 C−1 (just above Tc ). This agrees well with the experimental one.2 2. Using the measured values of λab = 320 nm.2 eV.3). Then.9
125 119 119 119 70 40
. Coherence volume in A3 . the in-plane ξab and out-of-plane ξc coherence lengths derived from a Ginzburg–Landau analysis of the speciﬁc heat [234]. this prediction was experimentally conﬁrmed by Zhao et al [173] providing compelling evidence for polaronic carriers in doped cuprates. the carrier mass isotope exponent αm = d ln m ∗∗ /d ln M was found to be as large as αm = 0. calculated with this mass is λab = [m ∗∗ /8πne2 ]1/2 ≈ 316 nm.8 in La1.5 Pb0.8 Y0.

6 .4 .8
T/Tc
2. N( .4) has a striking resemblance with the Geneva group’s experiments on DyBa2 Cu3 07 and on YBa2 Cu3 O7 [239]. B) exp[( − µ)/T ] − 1
calculated with this DOS and with µ determined from n b = d N( . ﬁgure 7. the lowest one traces the resistive transition. It barely shifts in the magnetic ﬁeld.266)).5
1.0 1. 241]. Within the bipolaron model.7.4 ω /T = 0.208
4.5
Superconducting transition
H* a 4.4 .2 0. tracing precisely the resistive transition (section 7. This shifts with the magnetic ﬁeld. (b) shows two anomalies.4(b). We can use the DOS in a magnetic ﬁeld with an impurity scattering as in section 4. when the magnetic ﬁeld is applied.03 0.2
Hc2(T)
b
C/T
3. which is absent in the ideal gas. Temperature dependence of the speciﬁc heat divided by temperature (arbitrary units) of the charged Bose gas scattered off impurities for several ﬁelds (ωH = 2eB/m ∗∗ ).2). there is a second anomaly at lower temperatures.T)-C(0.2 0 -.01 H c 0.4.3 .1 0. B) = T T dT d N( . while the highest anomaly is the normal-state feature.5
3.01 0. Calculations of the speciﬁc heat of charged bosons in a magnetic ﬁeld require an analytical DOS. where both anomalies were observed.4
1. it hardly changes the
.6 0.4.8 0. which allows for an analytical result (equation (4.2 -.2 1. as clearly seen from the difference between the speciﬁc heat in the ﬁeld and zero-ﬁeld curve. B) f ( ) is shown in ﬁgure 7.
The magnetic ﬁeld dependence of the anomaly [239] is also unusual but it can be described by the bipolaron model [240. The speciﬁc heat coefﬁcient d C(T. However.3 -1.0
C(H. scattered by other particles and/or by a random potential of impurities.2 1. The speciﬁc heat (ﬁgure 7.6 .4).4.8.0
ωH/Tc= 0 0.8
T/Tc
-.0 1. The broad maximum at T Tc is practically the same as in an ideal Bose gas without scattering [240] (appendix B.T)
.1 0.4
Figure 7. B) of a particle.03 -.0 .

the energy spectrum of low-energy bosons is strongly perturbed even by a weak magnetic ﬁeld. 245–247]. the higher one is just a ‘memory’ about the zero-ﬁeld transition. while the broad higher anomaly is the normalstate feature of the bosonic system in an external magnetic ﬁeld.3 led to the conclusion that the resistive Hc2 and the thermodynamically determined Hc2 are very different in bosonic superconductors. In contrast. We believe that many unconventional superconductors are in the ‘bosonic’ limit of preformed real-space bipolarons.4 Universal upper critical ﬁeld of unconventional superconductors
The upper critical ﬁeld (Hc2(T ) = 0 /2πξ(T )2 ) is very different in a BCS superconductor (section 1. This is because their characteristic energy (of the order of Tc ) remains huge compared with the magnetic energy of the order of 2eB/m ∗∗ . While the magnetic ﬁeld destroys the condensate of ideal bosons. these two anomalies are well separated in the bosonic superconductor at any ﬁeld but zero. A comprehensive scaling of resistive Hc2 measurements in unconventional
. spin-ladders [248] and organic superconductors [249]. A nonlinear temperature dependence in the vicinity of Tc was unambigously observed in a few samples [241. 246. As a result the chemical potential ‘touches’ the band edge at lower temperatures.7. Importantly.84Tc (in tesla) in the BCS theory but the limit can be exceeded by many times in CBG. it hardly shifts the speciﬁc heat anomaly as observed. 249]. While the lower anomaly corresponds to the true longrange order due to BEC. Also at zero temperature. Hc2(0) is normally below the Pauli pair-breaking limit given by H p 1. This microscopic consideration shows that a genuine phase transition into a superconducting state is related to a resistive transition (section 7. it has a positive curvature (Hc2(T ) (Tc − T )3/2 ) in a CBG. a thermodynamically determined Hc2 turned out to be much higher than the resistive Hc2 [250] due to the contrasting magnetic ﬁeld dependences of the speciﬁc heat anomaly and of resistive transition.4) and in a charged Bose gas (section 4. the Pauli limit was exceeded several times. 248. Calculations carried out for the heat capacity of a CBG in section 7. While Hc2(T ) is linear in temperature near Tc in the Landau theory of second-order phase transitions.6.
7. In cuprates [241–247]. When measurements were performed on low-Tc unconventional superconductors [243.Universal upper critical ﬁeld of unconventional superconductors
209
temperature dependence of the chemical potential near the zero-ﬁeld Tc because the energy spectrum of thermally excited bosons is practically unchanged. 244. Differing from the BCS superconductor.4) and to the lower speciﬁc heat anomaly. high-magnetic ﬁeld studies revealed a non-BCS upward curvature of resistive Hc2(T ).8). while having almost the same ‘kink’-like temperature dependence around Tc as in a zero ﬁeld. so their resistive Hc2 is actually a critical ﬁeld of the BEC of charged bosons [121].

8). As shown in section 4. the scaling constant H0 depends on the mean-free path l.02 (chain). Resistive upper critical ﬁeld (determined at 50% of the transition) of cuprates.
superconductors is shown in ﬁgure 7. 2 (l/n b (Tc ))1/4 .1
Tl-2201 Bi-2201 Sr2Ca12Cu24O41 (TMTSF)2PF6 LSCO YBCO+Zn L2-xCexCuO4-y Bi-2212 YBCO-123
Hc2(T)/H0
. The parameter b is 1 (dots). spin-ladders and organic superconductors scaled according to equation (7. The generalized equation (4.8) but the low-temperature behaviour depends on a particular scattering mechanism and a detailed structure for the density of localized states.261).6 . H0 = 0 /2πξ0 .8 1.0012 (full curve).01
.001
.2 .0
T/Tc
Figure 7.
.4 . accounting for a temperature dependence of the number of delocalized bosons n b (T ). The inset shows a universal scaling of the same data near Tc on the logarithmic scale. can be written as follows: Hc2(T ) = H0 n b (T ) − t 1/2 tn b (Tc )
3/2
(7. and 0 (dashes).8)
where Tc is the zero-ﬁeld critical temperature and t = T /Tc . 0.5. with the characteristic (coherence) length ξ0 In the vicinity of Tc .5 [241] in the framework of the microscopic model of charged bosons scattered off impurities (section 4.7. 0.210
Superconducting transition
1 2 .8.01
1
.7.1
1
1-T/Tc
0 0 . one obtains a parameter-free Hc2(T ) ∝ (1 − t)3/2 using equation (7.9).

9).1). All measurements reveal a universal (1 − t)3/2 behaviour in a wide temperature region (inset).
.
(7.6). where the Fermi–Dirac golden rule in the scaling fails. when they are ﬁtted by equation (7. Exceeding the Pauli pair-breaking limit readily follows from the fact that the singlet-pair binding energy is related to the normal-state pseudogap temperature T ∗ . and organic superconductors. spin-ladders. It naturally becomes rather small in overdoped cuprates where randomness is more essential. so that Hc2(T ) can be described by a single-parameter expression: Hc2(T ) = H0 b(1 − t) + 1 − t 1/2 t
3/2
. T ∗ is higher than Tc in bosonic superconductors and cuprates (ﬁgure 5. n b (T ) can be parametrized as n b (T ) = n b (0) + constant × T . b.5. so almost all bosons are localized (at least in one dimension) at zero temperature. We expect that this expression is applied in the whole temperature region except at ultra-low temperatures.9)
The parameter b is proportional to the number of delocalized bosons at zero temperature. The universal scaling of Hc2 near Tc is conﬁrmed by resistive measurements of the upper critical ﬁeld of many cuprates. rather than to Tc . as shown in ﬁgure 7.Universal upper critical ﬁeld of unconventional superconductors
211
As suggested by the normal-state Hall measurements in cuprates (section 6. The low-temperature behaviour of Hc2(T )/H0 is not universal but well described using the same equation with the single ﬁtting parameter. The parameter is close to one in high-quality cuprates with a very narrow resistive transition [247].

superconductor–insulator–normal metal (SIN) and superconductor– insulator–superconductor (SIS) tunnelling studies (sections 8. 254] was found to be linear in a few cuprates as expected for a d-wave BCS superconductor. However. Indeed their superconducting state is as anomalous as the normal one. the caxis Josephson tunnelling [255] and some high-precision magnetic measurements [256] show a more usual s-like symmetry or even reveal an upturn in the temperature dependence of the penetration depth below some characteristic temperature [257]. the symmetry of the order parameter and superconducting stripes [263] in the framework of the bipolaron theory. In this chapter we discuss the low-temperature London penetration depth [258]. In particular. the low-temperature magnetic penetration depth [253.3).Chapter 8 Superconducting state of cuprates
Independent observations of normal-state pseudogaps in a number of magnetic and kinetic measurements (chapter 6) and unusual critical phenomena (chapter 7) tell us that many cuprates may not be BCS superconductors. A number of phasesensitive experiments [252] provide unambiguous evidence in this direction. then the number in the condensate n c is nc = 1 x − nL − nb 2 212 (8.
8. Strong deviations from the Fermi/BCS-liquid behaviour are suggestive of a new electronic state in cuprates. furthermore. tunnelling [259–261] and ARPES [262].2 and 8.1)
.1 Low-temperature penetration depth
If the total number of bipolarons in one unit cell is x/2 of which n L are in localized states and n b are in delocalized states. which is a charged Bose liquid of bipolarons.5) or in its intermediatecoupling generalization (chapter 3). Also both angle-resolved photoemission spectroscopy (ARPES) [264] and scanning tunnelling microscopy (STM) [265] have shown that the maximum energy gap is many times larger and the 2 /Tc ratio is well above that expected in weak-coupling BCS theory (∼3. there is strong evidence for a d-like order parameter (changing sign when the CuO2 plane is rotated by π/2) in cuprates [251].

the total energy of p bosons in a localized state of energy E is taken to be w(E) = p E + p( p − 1)κ(−E)ν where κ > 0.Low-temperature penetration depth
213
√ and the London penetration depth λ ∝ 1/ n c . The partition function Z for such a system is then the product of the partition functions z(E) for each of the wells.7) (8.3) (8. λ has the same temperature dependence as n L . for non-interacting bipolarons moving in d dimensions. (8. this would give λ ∝ T d/2 . z(E) = eαp0
2
∞ p=0
e−α( p− p0 )
2
(8. we can neglect n b and make the approximation n c ≈ x/2 − n L . 0ξ (8. In our picture. For convenience we choose E c = 0. In this limit. When two or more charged bosons are in a single localized state of energy E.4)
where ν > 0. Thus in the low-temperature limit. For small amounts of disorder delocalized bipolarons may contribute to the low-temperature dependence of λ(T ) as well. We take the total energy of charged bosons in localized states to be the sum of the energies of the bosons in individual potential wells. Taking the delocalized bipolarons to be a three-dimensional gas.2)
i. We can thus deﬁne an energy scale E 1 : E 1 = κ 1/(1−ν).8)
. there may be signiﬁcant Coulomb energy and we take this into account as follows. interacting bosons ﬁll up all the localized single-particle states in a random potential and Bose-condense into the ﬁrst extended state at the mobility edge. The localization length ξ is thought to depend on E via ξ∝ 1 (−E)ν (8. E c .6)
From here on we choose our units of energy such that E 1 = 1. λ(T ) − λ(0) is small and so λ(T ) − λ(0) ∝ n L (T ) − n L (0) (8.5)
Thus.e. The Coulomb potential energy of p charged bosons conﬁned within a radius ξ can be expected to be potential energy 4 p( p − 1)e2 . we have n b ∝ T 3/2.

We now focus on the temperature dependence of n L at low temperature (θ 1) for the case where the (dimensionless) width of the impurity tail γ is large (γ > 1). ≈1+ NL 2(2 − ν)γ (2 − ν)γ (8. we consider the case ν > 1 ﬁrst and then ν < 1. we can approximate n L as 2θ ν−1 nL + ln 2. 2 2(2 − ν)γ γ (8.12)
where the mean occupancy p of a single localized state is taken to be p =
∞ −α( p− p0 )2 p=0 pe ∞ −α( p− p0 )2 p=0 e
(8. in this case. E1 The average number of bosons in localized states n L is θ= nL =
0 −∞
(8.13)
and NL (E) is the one-particle density of localized states per unit cell below the mobility edge (section 5.1(b) compares this analytical formula with the numerical calculation for the case ν = 0.5).9) (8. If ν > 1. We note that such a value of ν is typical for amorphous semiconductors [266]. If ν < 1 we obtain. n L (θ ) will still be linear with the same slope provided θ γ . E 1 . Figure 8.1(a) compares this analytical formula with an accurate numerical calculation for the case ν = 1.
. Figure 8. n L decreases linearly with increasing temperature (in the low-temperature limit). γ = 20. We note that the parameter γ is the ratio of the width of the tail to the characteristic Coulomb repulsion. γ = 20.5.214 where
Superconducting state of cuprates
p0 = 1 {1 + (−E)1−ν } 2 (−E)ν α= θ and T .10)
(8.65.11)
dE p NL (E)
(8. keeping only the lowest power of θ (valid provided θ 1/ν θ) nL 1 = + NL 2 (2 − ν)γ 1−ν 1−ν θ + − ln 2.14)
So we expect n L to be close to the total number of wells NL and to increase linearly with temperature.15)
Hence. In the following. A large value of γ is also expected in disordered cuprates with their large static dielectric constant. We also note that even when γ < 1.

The excitation spectrum of the charged Bose liquid determines.
Figure 8. Both ν < 1 and ν ≥ 1 are observed in doped semiconductors [266]. We believe that ν < 1 is more probable for a rapidly varying random potential while ν ≥ 1 is more likely for a slowly varying one.Low-temperature penetration depth
215
Figure 8.15) while the broken lines are derived from an accurate numerical calculation. Hence. Fit to the London penetration depth measured by Walter et al [257] for a YBCO ﬁlm.
Figure 8.65. The full lines correspond to the low-temperature predictions from equations (8. It is more usual to see λ increase linearly with temperature [253. the more usual linear increase of λ with temperature is observed in samples which either have no disorder or a shallow and smooth random potential proﬁle.254] and this would correspond to ν > 1 (or to the predominance of the effect of delocalized bipolarons moving in two dimensions). λ(T ) is related to the localization of carriers at low temperatures rather than to any energy scale characteristic of the condensate or to its symmetry. it is not surprising that a drastically different lowtemperature dependence of the London penetration depth is observed in different samples of doped cuprates.14) and (8. (b) ν = 0.1. the temperature dependence of λ(T ) at higher temperatures. however. The parameter values from the ﬁt were E 1 = 74 K. In the framework of this approach.2 shows that the low-temperature experimental data [257] on the London penetration depth λ of YBCO ﬁlms can be ﬁtted very well by this theory with ν < 1. The key parameter of the temperature dependence of the London penetration depth is the exponent ν of the localization length of bosons. Here we
.5. Experimentally. Dependence of the density of localized bosons n L on temperature θ for γ = 20: (a) ν = 1.2. γ = 20 and ν = 0.67.

The gap in tunnelling and photoemission is almost temperature independent below Tc [265. The ﬁrst band in cuprates to be doped is the oxygen band inside the Hubbard gap (section 5.16)
−U
† † dx ψ↑ (x)ψ↓ (x)ψ↓ (x)ψ↑ (x)
where s = ↑. It is then that real-space bipolarons are formed.216
Superconducting state of cuprates
expect ν ≥ 1 and a substantial contribution from delocalized bipolarons moving principally in two dimensions and thus. as observed [254].4. ↓ is the spin. This would explain the upturn in the temperature dependence of λ in the disordered ﬁlms [257]. 269–271].13). However. the clearest being uniform susceptibility [186. reﬂection experiments. heavy-ion bombardment [257] introduces rather deep and narrow potential wells for which we might expect ν < 1. that λ increases linearly with temperature. We assume that this radius is less than the inter-carrier distance in cuprates. This band is quasi-onedimensional as discussed in section 5. the ground state of the system is
. revealed a much smaller gap edge than the bias at the tunnelling conductance maxima in a few underdoped cuprates [273]. 275] also showed distinctly different superconducting. V (x − x ) = −U δ(x − x ). as H=
s † dx ψs (x) −
1 d2 − µ ψs (x) 2m ∗ dx 2 (8.2 SIN tunnelling and Andreev reﬂection
There is compelling experimental evidence that the pairing of carriers takes place well above Tc in cuprates (chapter 6).16)) is a realistic starting point. 268] and exists above Tc [265.2). In the framework of bipolaron theory. so that a one-dimensional approximation (equation (8. Other tunnelling measurements [274. in which an incoming electron from the normal side of a normal/superconducting contact is reﬂected as a hole along the same trajectory (section 2. Kinetic [198] and thermodynamic [272] data suggest that the gap opens in both charge and spin channels at any relevant temperature in a wide range of doping.
8. The assumption is that the attraction potential in cuprates is large compared with the (renormalized) Fermi energy of polarons. The model is a generic one-dimensional Hamiltonian including the kinetic energy of carriers in the effective mass (m ∗ ) approximation and a local attraction potential.and normal-state gaps. At the same time. Solving a two-particle problem with the δ-function potential. we can consider a simpliﬁed model. 267] and tunnelling [265]. which describes the temperature dependence of the gap and tunnelling spectra in cuprates and accounts for two different energy scales in the electron-hole reﬂection [260]. one obtains a bound state with the binding energy 2 p = m ∗ U 2 /4 and with the radius of the bound state r = 2/(m ∗ U ). If three-dimensional corrections to the energy spectrum of pairs are taken into account.

explained in [259]. The single-particle gap. Ai(x) and Bi(x) the Airy functions and 0 is the scattering rate. Taking into account the scattering of single-particle excitations by a random potential. dI /dV is proportional to the density of states.21) π2 [Ai2 (−2ξ ) + Bi2 (−2ξ )]2 A is an asymmetry coefﬁcient. A normal metal–superconductor (SIN) tunnelling conductance via a dielectric contact.3). so that the normal state single-particle gap is p . of the spectrum (equation (8. E(k). The single-particle excitation energy spectrum.2) we reduce the Hamiltonian (8.19). p c It varies with temperature from (0)/2 = [ 2 + 2 (0)]1/2 at zero temperature p c down to the temperature-independent p above Tc . is found using the Bogoliubov transformation as E(k) = [(k 2 /2m ∗ +
p) 2
+
2 1/2 c]
(8.19) /2 = [ 2 + 2 (T )]1/2.SIN tunnelling and Andreev reﬂection
217
the Bose–Einstein condensate (BEC). The theoretical temperature dependence.18)). and n = 4.4. is given by (8. ρ(E). describes well the pioneering experimental observation of the anomalous gap in YBa2 Cu3 O7−δ in the electron-energy-loss spectra by Demuth et al [277] (ﬁgure 8. This spectrum is quite different from the BCS quasi-particles because the chemical potential is negative with respect to the bottom of the single-particle band. We note that n = 4 is an exponent which is expected in the two-ﬂuid model of any superﬂuid [276].20)
4 Ai(−2ξ )Ai (−2ξ ) + Bi(−2ξ )Bi (−2ξ ) × (8.17) where the coherent pairing potential. c = −U ψ↓ (x)ψ↑ (x) . as well as thermal lattice and spin ﬂuctuations (see section 8. is proportional 1/2 to the square root of the condensate density. Applying the Bogoliubov approximation (section 2. ρ(ξ ) =
. deﬁned as the minimum of E(k). equation (8. c = constant × n c (T ).16) to a quadratic form: H=
s † dx ψs (x) −
1 d2 − µ ψs (x) + 2m ∗ dx 2
dx [
† † c ψ↑ (x)ψ↓ (x) + H.c. Now we take into account that in the superconducting state (T < Tc ) single-particle excitations interact with the condensate via the same potential U . µ = − p . The chemical potential is pinned below the band edge by about p both in the superconducting and normal states.3) with 2 (T ) = 2 (0) × [1 − (T /Tc )n ] c c below Tc and zero above Tc .]
(8. at T = 0 dI /dV = constant × ρ with 2eV −
0
+ Aρ
−2eV −
0
(8. one ﬁnds. /2.18)
if one assumes that the condensate density does not depend on position.

20)) with one of the best STM spectra measured in Ni-substituted Bi2 Sr2 CaCu2 O8+x single crystals by Hancottee et al [268] in ﬁgure 8. We note that the scattering rate.20)) also ﬁts well the conductance curve obtained on ‘pure’ Bi-2212 single crystals. while a simple d-wave BCS density of states cannot describe the excess spectral weight in the peaks and the shape of the conductance outside the gap region.3. as shown in ﬁgure 8.4(b). with the only difference being that the chemical potential with respect to the bottom of the band is a function of the
.19) (line)) compared with the experiment [277] (dots) for p = 0. A simple theory of tunnelling into a bosonic (bipolaronic) superconductor in a metallic (no-barrier) regime follows from this model.4(a). We approximate this buffer layer by a layer with a constant chemical potential µb (− p < µb < µ) and with the same strength of the pairing potential c as in a bulk superconductor. 0 . This experiment shows anomalously large /Tc > 12 with the temperature dependence of the gap similar to that in ﬁgure 8. where the chemical potential with respect to the bottom of the conduction band changes gradually from a positive large value µ in metal to a negative value − p in a bosonic superconductor.7 (0). injection/emission asymmetry.20)) describes well the anomalous gap/Tc ratio.4(a)) is s-like. Temperature dependence of the gap (equation (8. There is also a buffer layer of the thickness l at the normal metal–superconductor boundary (x = 0). allowing for a reﬂected electron and (Andreev) hole in the normal metal (x < 0) (section 2. As in the canonical BCS approach applied to normal metal–superconductor tunnelling by Blonder et al [47]. the incoming electron produces only outgoing particles in the superconductor (x > l). The Bogoliubov equations may be written as usual (section 2.
We compare the conductance (equation (8. is apparently smaller in the ‘pure’ sample than in the Ni-substituted sample.3 below Tc .218
Superconducting state of cuprates
Figure 8. There is no doubt that the gap (ﬁgure 8.13). zerobias conductance at zero temperature and the spectral shape inside and outside the gap region.13). as expected. The conductance (equation (8. The theoretical conductance (equation (8.

A = 1.23)
.SIN tunnelling and Andreev reﬂection
219
Figure 8.22)
Thus.05.4. 0 = 40 meV and (b) in ‘pure’ Bi-2212 [268] with = 43 meV.20) (line)) compared with the experimental STM conductance (dots) in (a) Ni-substituted Bi2 Sr2 CaCu2 O8+x [268] with = 90 meV. (8.
coordinate x: Eψ(x) = −(1/2m) d2/dx 2 − µ(x)
c c
(1/2m) d2/dx 2 + µ(x)
ψ(x). the two-component wavefunction in normal metal is given by ψn (x < 0) = 1 0 eiq
+x
+b
1 0
e−iq
+x
+a
0 1
e−iq
−x
(8. A = 1.2 and 0 = 18 meV. Theoretical tunnelling conductance (equation (8.

which are the continuity of ψ(x) and its derivatives at x = 0 and x = l. b. δ are determined from the boundary conditions. l = 4 (thin line). c. The coefﬁcients a.220
Superconducting state of cuprates
Figure 8. α. l = 1 (bold dashes).25)
E+ξ
E−ξ
where the momenta associated with the energy E are k ± = [2m(− p ± ξ )]1/2 .24)
E+ξ
E−ξ
where the momenta associated with the energy E are q ± = [2m(µ ± E)]1/2 and p± = [2m(µb ± ξ )]1/2 with ξ = (E 2 − 2 )1/2 . it has the form ψb (0 < x < l) = α +γ 1
c
ei p 1
c
+x
+β +δ
1
c
e−i p 1
c
−x
E+ξ
E−ξ
e
−i p + x
ei p
−x
(8.5. γ . Transmission versus voltage (measured in units of p /e) for c = 0. A well-behaved solution in the c superconductor with a negative chemical potential is given by ψs (x > l) = c 1
c
eik
+x
+d
1
c
eik
−x
(8.
.2 p . µ = 10 p . µb = 2 p and l = 0 (bold line). Applying the boundary conditions and carrying out an algebraic reduction. β. d. and l = 8 (thin dashes) (in units of 1/(2m p )1/2 ).
while in the buffer layer.

5 for different values of l. p < c . 1 + |a|2 − |b|2 is shown in ﬁgure 8. µ = 10 c . l = 4 (thin line). when the coherent gap c is smaller than the pair-breaking gap p and in ﬁgure 8.
we ﬁnd a=2 ( p + f − g + − p − f + g − )/D (8.26) 2q + [(E + ξ ) f + (q − f − − p− g − ) − (E − ξ ) f − (q − f + − p + g + )] b = −1+ D (8. The transmission coefﬁcient in the electrical current.6.2 c .28)
(8.SIN tunnelling and Andreev reﬂection
221
Figure 8. l = 1 (bold dashes). In the ﬁrst
± ± ± ± ±
(8. Transmission versus voltage (measured in units of c /e) for p = 0.27)
cq +
with D = (E + ξ )(q + f + + p + g + )(q − f − − p− g − ) − (E − ξ )(q + f − + p− g − )(q − f + − p+ g + ) and f ± = p± cos( p±l) − ik ± sin( p±l) g = k cos( p l) − i p sin( p l). and l = 8 (thin dashes) (in units of 1/(2m c )1/2 ).29)
.6 in the opposite case. µb = 2 c and l = 0 (bold line).

ω) = Z
l 2 r=1 |γ (q r )|
(2N)l l!(ω −
l r=1
ωq r − (k +
l r=1
(k) = [ξ 2 (k) +
2 1/2 . there is only one gap. between two parts of a superconductor separated by an insulating barrier is proportional to a convolution of the Fourier component of the single-hole retarded Green’s function. which can be seen in the second case.222
Superconducting state of cuprates
case. G R (k.6).
l=0 q 1 . where the Migdal–Eliashberg theory cannot be applied (section 4.5) while p ≤ c in optimally doped cuprates (ﬁgure 8. However. p ∞ −∞
dω Im G R (k. with itself as I (V ) ∝
k.5). I (V ). is also found in the theoretical conductance at ﬁnite l (ﬁgure 8.6).6) are due to a coherent tunnelling into the condensate and into a single-particle band of the bosonic superconductor. λ > 1.q l
G R (k.6. ω).3 SIS tunnelling
Within the standard approximation [76].2.30)
where V is the voltage in the junction. observed in underdoped YBa2 Cu3 O7−δ [273]..32)
. l (2m p )−1/2 (ﬁgure 8. The problem of a hole on a lattice coupled with the bosonic ﬁeld of lattice vibrations has a solution in terms of the coherent (Glauber) states in the strongcoupling limit. the model accounts for two different gaps experimentally observed in Giaver tunnelling and electron-hole reﬂection in underdoped cuprates and for a single gap in optimally doped samples [273].31) The hole energy spectrum. which might explain a signiﬁcant featureless background in the subgap region [273].3.
8. the tunnelling current. c]
(8. we ﬁnd two distinct energy scales.. q r ) + iδ) (8.5) or l are some oscillations of the transmission due to the bound states inside the buffer layer. We note that the transmission has no subgap structure if the buffer layer is absent (l = 0) in both cases. ω) Im G R ( p.. In the extreme case of a wide (2m c )−1/2 (ﬁgure 8.4). We expect that p c in underdoped cuprates (ﬁgure 8. There is an incoherent transmission into localized single-particle impurity states and into incoherent (‘supracondensate’) bound pair states as well.5.
∞
. which is a singleparticle band edge.. one is c in the subgap region due to the electron-hole reﬂection and the other one is /2. there buffer layer. Thus. ﬁgure 8.5 and 8. An oscillating structure. The transmissions (ﬁgures 8. ﬁgure 8. (k). c . e|V | − ω)
(8. is renormalized due to the polaron narrowing of the band and (in the superconducting state) also due to interaction with the BEC of bipolarons as discussed in section 8.

Its particular energy/momentum dependence is essential in the subgap region of tunnelling.31) and (8.
(8. which is seen in the Green fucntion as phonon sidebands with l ≥ 1.SIS tunnelling
223
Here ξ(k) = Z E(k) − µ is the renormalized polaron band dispersion with the chemical potential µ and E(k) = m T (m) exp(−ik · m) is the bare band dispersion in a rigid lattice. Quite differently from the BCS superconductor. To perform the remaining integrations and summations.
.l =0 q.31) by a ﬁnite i /2. However.3). their average number g 2 in the polaronic cloud and their dispersion δω.34)
where L[x. compatible with the ARPES data [262].4)) by replacing the q-sums in equation (8. Substiting equation (8. δω] A(ω)
(8. there is no damping (‘dephasing’) of low-energy polaronic excitations in equation (8.33) into equations (8. ω0 .q
l r=1
l 2 r =1 |γ (q r )| |γ (q r (2N)l+l l!l ! l l
)|2
× L e|V | −
−
r=1
ωq r −
r =1
ω(q r ). of phonons strongly coupled with holes. elastic scattering by impurities yields a ﬁnite life-time for the Bloch polaronic states. so that the edge of the single-hole band is found above the chemical potential at −µ = p .30). where it determines the value of zero-bias conductance. (8. it hardly plays any role in the peak region and higher voltages. However. For the sake of analytical transparency. 2 kx + /2.31) due to the electron–phonon coupling alone. This coupling leads to a coherent dressing of low-energy carriers by phonons. ] = /(x 2 + 2 ).4. we obtain for the tunnelling conductance.33) k 2m ∗ Differently from the canonical Migdal–Eliashberg Green function (chapter 3). σ (V ) = dI /dV . replacing iδ in equation (8. the ‘elastic’ self-energy can be found explicitly as a function of energy and momentum (see section 8. In this way we introduce the characteristic frequency.
∞
σ (V ) ∝
l. In fact.34) by 1 2N |γ (q)|2 A(ωq ) =
q
g2 π
dω L[ω − ω0 . here we model this scattering as a constant imaginary self-energy. we introduce a model analogue of the Eliashberg spectral function (α 2 F(ω) (section 3. the chemical potential µ is negative in the bipolaronic system. and performing the intergration with respect to frequency and both momenta. Near the edge the parabolic one-dimensional approximation for the oxygen hole is applied.35)
for any arbitrary function of the phonon frequency A(ωq ).

they agree with the experimental SIS tunnelling spectra in cuprates [265. and there is an unusual shape for the quasi-particle peaks. (8. phonon densities of states.7 for three different values of coupling. as shown in ﬁgure 8. there is a substantial incoherent spectral weight beyond the quasi-particle peak for the strong coupling. the gap /2 about 30–50 meV and the impurity scattering rate of the order of 10 meV.
+ δω(l + l )]. The coherent (l = l = 0) contribution to the current with no elastic scattering ( = 0) is given by I0 ∝
∞
d ( − /2)1/2
∞
d ( − /2)1/2
δ( +
− e|V |)
(8.268.36) by the Fourier integral.5625.7 for g 2 = 0.9. All these features as well as the temperature dependence of the gap are beyond BSC theory no matter what the symmetry of the gap is. the theory quantitatively describes one of the best tunnelling spectra obtained on Bi2 Sr2 CaCu2 O8+δ single crystals by the break-junction technique [268] (ﬁgure 8. the phonon frequency dispersion about a few tens meV. both s-wave and d-wave. the characteristic phonon frequency about 30–100 meV.275]. we can extend the integration over phonon frequencies from −∞ to ∞ and obtain
∞
σ (V ) ∝
l. However. g 2 ≥ 1.38)
. the dip/hump features (due to phonon sidebands) are clearly seen already in the ﬁrst derivative of the current. one estimates the coupling strength g 2 to be of the order of one.269.l =0
g 2(l+l ) L[e|V | − l!l !
− (l + l )ω0 .8). The dynamic conductance of Bi-2212 mesas [274] at low temperatures is almost identical to the theoretical one in ﬁgure 8.37)
× cos[2g 2e−δωt sin(ω0 t) − (e|V | −
From the isotope effect on the carrier mass. e|V | > . we perform the summation over l and l with the ﬁnal result for the conductance as follows: σ (V ) ∝
0 ∞
dt exp[2g 2e−δωt cos(ω0 t) − t] )t]. The conductance shape is remarkably different from the BCS density of states.36)
By replacing the Lorentzian in equation (8.37)) calculated with parameters in this range is shown in ﬁgure 8. experimental values of the normal-state pseudogap and the residual resistivity (chapters 6 and 7). The exact (energy-dependent) self-energy provides an agreement in this subgap region as well.
(8. There is no Ohm’s law in the normal region. The SIS conductance (equation (8. In particular. The unusual shape of the main peaks (ﬁgure 8. Some excess zerobias conductance compared with the experiment is due to our approximation of the ‘elastic’ self-energy.274.224
Superconducting state of cuprates
As soon as δω is less than ω0 .8) is a simple consequence of the (quasi-)one-dimensional hole density of states near the edge of the oxygen band.

of the total gap (equation (8. and /2 = 29 meV.5 meV. SIS tunnelling conductance in a bipolaronic superconductor for different values of the electron–phonon coupling. = 8. ω0 = 55 meV.SIS tunnelling
225
Figure 8. Below Tc . the width of the main peaks in the SIS tunnelling (ﬁgure 8. The disappearance of the quasi-particle sharp peaks above Tc in Bi-cuprates can also be explained in the framework of bipolaron theory. The temperature-dependent part.39)
Hence.7.5). c (T ). (8.19)) is responsible for the temperature shift of the
. This sudden increase in in the normal state washes out sharp peaks from tunnelling and ARPES spectra. δω = 20 meV. g 2 . while above Tc the scattering rate might increase by many times (section 8. bipolaronic condensate provides an effective screening of the long-range (Coulomb) potential of impurities.
so that the conductance is a δ function: σ0 (V ) ∝ δ(e|V | − ).8) measures the elastic scattering rate directly.

Some authors [278] refer to the peak as an extended van Hove singularity (evHs) arising from a plane (CuO2 ) strongly correlated band.3. and Y123 [279]. where the ﬁrst band to be doped is the oxygen band lying within the Hubbard gap. cuprates and many other transition metal compounds are charge transfer Mott–Hubbard insulators. as shown in ﬁgure 5. They also implicate the resulting (quasi-)one-dimensional DOS singularity as a possible origin for the high transition temperature. YBa2 Cu4 O8 (Y124) and several other materials. 275]. while the incoherent temperature-independent part p survives well above Tc as observed [274.
8. Theoretical conductance of ﬁgure 8.5625 (full line) compared with the tunnelling spectrum obtained on Bi2 Sr2 CaCu2 O8+δ single crystals by the break-junction technique [268] (dots).
.2.226
Superconducting state of cuprates
Figure 8. polarized ARPES studies of untwinned Y123 crystals of exceptional quality [279] unambiguously show that the peak is a narrow resonance arising primarily from the quasi-one-dimensional CuO3 chains in the buffer layers rather than from the planes.
main peaks in the superconducting state of an SIS junction.8. a very similar narrow peak was observed by Park et al [280] in high-resolution ARPES near the gap edge of the cubic semiconductor FeSi with no Fermi surface at all.4 ARPES
Let us discuss ARPES in doped charge-transfer Mott insulators in the framework of bipolaron theory [262] to describe some unusual ARPES features of high-Tc YBa2 Cu3 O7−δ (Y123). However. Perhaps the most intriguing feature of ARPES in cuprates is an extremely narrow and intense peak lying below the Fermi energy. Interestingly. Its angular dependence and spectral shape as well as the origin of the featureless (but dispersive) background remain unclear. As discussed in section 5.7 for g 2 = 0. which is most clearly seen near the Y and X points in Y124 [278].

k k
(8.ARPES
227
A single photoexited oxygen hole is described by the polaron spectral function of section 4.42)
and v0 is the unit cell volume (G is a reciprocal-lattice vector). we consider
. f is the energy of the initial and ﬁnal states and .1 Photocurrent The interaction of the crystal with the electromagnetic ﬁeld of frequency ν is described by the following Hamiltonian (in the dipole approximation): Hint = (8π I )1/2 sin(νt)
k. the sum in equation (8. E i.43) k−G
where E is the binding energy.43) is n(E)A(k − G. This allows us to explain the shape of the narrow peaks in ARPES using the spectral density A(k. In the following. −E).e. Due to the translational symmetry of the Bloch states. E) π (8. µ. E) = − 1 Im G R (k − G. respectively.4. we suppress k k
(8. 8.G with d(k) = ie(N/v0 )1/2 ∇k
v0
where I is the intensity of the radiation with polarization e. Its low-energy part is affected by low-frequency thermal lattice. For simplicity. The latter can be described as a ‘Gaussian white noise’ potential. Also p-hole polarons in oxides are almost one-dimensional due to a large difference in the ppσ and ppπ hopping integrals. spin and random ﬂuctuations.40)
the band index in h † . The Fermi–Dirac golden rule gives the photocurrent to be I (k. By deﬁnition. where the spectral function A(k − G. k is the (quasi-)momentum of the hole remaining in the sample after the emission and c† and h † are their creation operators.4. E) = 4π 2 I |e· d(k)|2
i. E) of a one-dimensional particle in a Gaussian white noise potential [281].k
(e · d kk )c† h † + H. of the photoelectron registered by the detector).44)
is proportional to the imaginary part of the retarded GF (appendix D) and n(E) = [exp(E/T )+1]−1 is the Fermi distribution.41)
e−iG·r u k−G (r) dr
(8. k |k ≡ u −k (r) exp(−ik · r) (appendix A).c. d kk = d(k)δk+k . k is the momentum of the ﬁnal state (i. f
e
+µNi −E i
| f |h † |i f |2 δ(E + E f − E i ) (8.3. Ni are the thermodynamic and chemical potentials and number of holes. respectively. there is a momentum conservation in the dipole matrix element.

E) Z .
(8. Within this approximation the self-energy is k-independent for a shortrange scattering potential like a deformation or a screened impurity potential. E)]2 (8. so that the coherent part of the spectral function is given by Ac (k.y
G R (k. The present experimental resolution [264] allows the intrinsic damping of the coherent quasi-particle excitations to be probed. spectral and polarization dependence of the ﬁrst coherent term in equation (8. E) ∼ |d(k)|2 n(E)Z δ(E + ξk ) + Iincoh (k.45)
where Iincoh(k.
= 2t cos(k x. E) = − Im (k. are found at the Brillouin zone boundary in X (π. so that n(E) = 1. 0) and Y (0. (k.2). so that G R (k.x a) − µ. which spreads from about −ω0 down to −2E p . π [E + Re (k.2). the theory of narrow ARPES peaks is reduced to determining the selfenergy of a hole.46)
Hence.79). The most important interaction in oxides is the Fr¨ hlich electron–phonon interaction with c-axis polarized high-frequency o phonons (section 5. There might be some multi-phonon structure in Iincoh (k.
. small polarons exist in the Bloch states at temperatures below the optical phonon frequency T < ω0 /2 (section 4. Here we concentrate on the angular.2 Self-energy of one-dimensional hole in a non-crossing approximation Due to energy conservation. The spectral function depends on essential interactions of a single hole with the rest of the system.y a) − 2t cos(k y. π). E) (8. E). and zero otherwise.3). E) − ξk )]2 + [Im (k. E) (8.48)
We assume that the minima of two polaron bands (equation (8. which leads to the polaron spectral function (4. E). we obtain I (k. a low-frequency deformation potential and spin ﬂuctuations. if E is negative. A ﬁnite polaron self-energy appears due to (quasi-)elastic scattering off impurities. The damping appears due to a random ﬁeld and low-frequency lattice and spin ﬂuctuations described by the polaron self-energy. and we put G = 0. First we apply the simplest non-crossing (ladder) approximation (chapter 3) to deﬁne an analytical (k.47) (E)
k
where E) = [E − ξk − The hole energy spectrum is parametrized in a tight-binding model as ξk
x.48)). 8.3.4. E) is an incoherent part.
(E)]−1 . As a result.45).228
Superconducting state of cuprates
temperatures well below the experimental energy resolution. E) as observed in tunnelling (section 8.

k = 0. δ > 0.y one-dimensional parabolic bands near the X and Y points. −E) exp −
(k − k )2 .04π/a (almost Y or X points of the Brillouin zone) and k = 0.9 for two momenta. where V0 is the amplitude of the deformation potential. π).e.49) ( ) = −2−3/2 [ ( ) − ]−1/2 for each doublet component.28π/a and the damping 0 = 19 meV. δ2
(8.9449.50)
While the energy resolution in the present ARPES studies is almost perfect [264].1π/a.52)
2Z Im{ ( )[w(z 1 ) + w(z 2 )]} δ
2
where z 1. The solution is ( )= 3 − 1+ i31/2 2 − 1 − i31/2 2
3 1 + − 16 27 3 1 + 256 216
1 + + 16 27
3
1 + 256 216
3
1/2
1/3
1/2
1/3 . respectively. The imaginary part of the self-energy (equation (8.48) to two x. µ = −20 meV.51)
The integral is expressed in terms of I (k. w(z) = e−z erfc(−iz) and = (−E + µ)/ 0 .ARPES
229
As previously mentioned. n im is the impurity density and α is the coefﬁcient of the δ-function impurity potential (i. E)
∞ −∞
dk Ac (k .2 = [±k − i/2 ( )]/δ. (8. 2 comprising thermal and random ﬂuctuations as D = 2(V0 T /M + n im α 2 ). The chemical potential is placed in the charge transfer gap below the bottom of the hole band. This photocurrent is plotted as broken lines in ﬁgure 8. reducing equation (8. E) −
( ) and the error function w(z) as follows: (8. the momentum resolution remains ﬁnite in most experiments. The constant D is the second moment of the Gaussian white noise potential. we have to integrate the spectral function (equation (8.46)) with a Gaussian momentum resolution to obtain the experimental photocurrent: I (k.47) for the self-energy in the non-crossing approximation takes the following form: (8. Hence. M is the elastic modulus. Hence. Here we introduce a dimensionless energy (and self-energy). 0) and (0. This allows us to apply a one-dimensional approximation.50)) disappears below = −3/25/3 −0. ξk = k 2 /2m ∗ − µ ∗ = 1/2ta 2 and k taking relative to (π. the oxygen hole is (quasi-)one-dimensional due to a large difference between the oxygen hopping integrals for the orbitals elongated parallel to and perpendicular to the oxygen–oxygen hopping t t. with m equation (8. the strength of the scattering potential). ≡ (E + µ)/ 0 using 0 = (D 2 m ∗ )1/3 as the energy unit.3π/a. Then. this
. the momentum resolution is taken as δ = 0.

28π/a and the polaron mass m = 9.
(8.4. ) = 4
∞ −∞
p0 (−z) Re p1(z) dz.1(z) obeys two differential equations: d2 d + (z 2 + 2 ) p0 = 0 dz 2 dz and d2 d + (z 2 + 2 ) − z − ik p1 + p0 = 0 2 dz dz (8. We have to go beyond a simple ladder to describe the single-electron tunnelling inside the gap and the ARPES spectra at small binding energies.54)
.9. The bipolaron binding energy 2|µ| = 40 meV. 8.04π/a (upper curves) and k = 0. the non-crossing approximation fails to describe the localized states inside the gap (i. k = 0. the momentum resolution δ = 0.30π/a with the damping ∗ 0 = 19 meV. integrated with the momentum resolution function for two angles.
approximation gives a well-deﬁned gap rather than a pseudogap.53)
Here p0.55) (8.230
Superconducting state of cuprates
Figure 8. Actually.3 Exact spectral function of a one-dimensional hole The exact spectral function for a one-dimensional particle in a random Gaussian white noise potential was calculated by Halperin [281] and the density of states by Frisch and Lloyd [282]. Halperin derived two pairs of differential equations from whose solutions the spectral function may be calculated: Ac (k. a tail of the density of states).e.9m e . The broken curves are the spectral density integrated with the momentum resolution in the non-crossing approximation. The polaron spectral function.

must be solved numerically. hence. provides a quantitative ﬁt to the ARPES spectra in Y124 along the Y– direction.60) DOS ρ( ) = dK 0 ( )/d ﬁts well the voltage–current tunnelling characteristics of cuprates.59)
is expressed analytically [282] in terms of the tabulated Airy functions Ai(x) and Bi(x) as K 0 ( ) = π −2 [Ai2 (−2 ) + Bi2 (−2 )]−1 .4 ARPES in Y124 and Y123 With the polaronic doublet (equation (8. as discussed in section 8.9 for two values of the momentum (full lines). the exact spectral function (averaged with the momentum resolution function) has a tail due to the states localized by disorder within the normal-state gap. The angular dispersion is described with the polaron mass m ∗ = 9. The exact one-dimensional polaron spectral function. In contrast to the non-crossing approximation. it is practically exact.ARPES
with the boundary conditions
z→∞
231
lim z 2−n pn (z) = lim z 2−n pn (z)
z→−∞
(8. ∞ π 1/2 0 u −1/2 exp(−u 3 /12 − 2u ) du
(8. 2π(−2 )1/2 exp − (−2 )3/2 cosh2 3 (−8 )1/2 (8. −E) integrated with the Gaussian momentum resolution is shown in ﬁgure 8. the non-crossing approximation gives very good agreement and for a binding energy greater than about 30 meV.2.56)
where k is measured in units of k0 = (D 1/2 m ∗ )2/3 .57)
The equation for p1 (z) has no known analytic solution and. However. ) in the tail where | | 1: Ac (k.48)) placed above the chemical potential. There is.9m e . integrated with the experimental momentum resolution (shown in ﬁgure 8. )
(8. however. ) πk 4 . (8. The ﬁrst equation may be integrated to give p0 (z) = exp(−z 3 /3 − 2z )
z 3 −∞ exp(u /3 + 2u ) du .4. besides this tail.10).58)
The result for Ac (k. The spectral shape is reproduced well with 0 = 19 meV. we can quantitatively describe high-resolution ARPES in Y123 [279] and Y124 [278]. an asymptotic expression for Ac (k. The cumulative DOS (appendix B) K 0 ( ) = (2π)−1
∞ −∞
d
−∞
dk A p (k. There is also quantitative agreement
. 8.

Figure 8. Theoretical ARPES spectra for the Y– direction (b). Parameters are those of ﬁgure 8.11.
. The theoretical ﬁt agrees well with experiment (a) [278] in a restricted range of k x near the Y-point and.10.9. Theoretical ARPES spectra in Y124 for the Y–S direction (b). outside this range. The theory provides a quantitative ﬁt to experiment (a) [278] in this scanning direction.232
Superconducting state of cuprates
Figure 8. the theoretical peaks are somewhat higher than in experiment.

where almost no dispersion is observed around Y (ﬁgure 8. Therefore. including different marginal Fermiliquid models.3(b)).
∞ −∞
dE I (k. Also the experimental spectral shape of the intensity at k = kF is incompatible with any theoretical scenario.
between theory and experiment in the perpendicular direction Y–S. there is a signiﬁcant loss of the energy-integrated intensity along both directions (ﬁgure 8. as shown in ﬁgure 8. However. Energy-integrated ARPES intensity in Y124 in the Y– (a) and Y–S (b) directions. The energy-integrated ARPES spectra obey the sum rule. If the loss of spectral weight were due to a Fermi-surface crossing. the peaks in the Y–S direction are all 15 meV or more below the Fermi level—at a temperature of 1 meV. Indeed. one would expect the peaks to approach much closer to the Fermi level. The rapid loss of the integrated intensity in the Y–S direction was interpreted by some authors [283] as a Fermi-surface crossing.ARPES
233
Figure 8. In contrast.62)
because the imaginary part of the self-energy behaves as |E|β with 0 ≤ β ≤ 2 in a Fermi or in a marginal Fermi liquid. the experimental intensity
. E) ∼ |E|β E 2 + constant × E 2β (8. E)
|d(k)|2
(8. in a restricted region of small k x .12) which the theoretical spectral function alone cannot account for. While the Fermi-surface crossing might be compatible with our scenario (see ﬁgure 5.13. The spectral function on the Fermi surface should be close to a simple Lorentzian: Ac (kF .11). it is hard to believe that it has been really observed in Y124.61)
if the chemical potential is pinned inside the charge-transfer gap.12. we have to conclude that the dipole matrix element depends on k. Momenta are measured relative to the Y-point of the Brillouin zone.

13. Then ARPES show intense peaks near large k imitating a large Fermisurface. charge e) scattered by a static Coulomb potential V (r ) screened by a charged Bose gas. We assume particle–hole symmetry to obtain the spectral function for negative binding energy. 275]. let us calculate the scattering cross section of a charged particle (mass m. To illustrate the point. 285]. While. Here we show that a large increase in the quasi-particle lifetime below Tc is due to the screening of scatterers by the Bose–Einstein condensate of charged bipolarons [203. If there is indeed no Fermi-surface crossing.234
Superconducting state of cuprates
Figure 8. A very sharp increase in the quasi-particle lifetime in the superconducting state has been observed in the thermal Hall conductivity measurements [286]. in principle.13). this allows scattering cross sections to be calculated for an arbitary potential. which is drastically incompatible with their kinetic and thermodynamic properties? It might be due to the fact that the oxygen hole band has its minima at large k inside or even on the boundary of the Brillouin zone. The experimental ARPES signal (full line) on the alleged Fermi surface does not correspond to a Fermi-liquid spectral function (broken line).
8. 287]. The abrupt appearence of the quasi-particle state below Tc is also implied in the tunnelling I –V characteristics [274. in practice the equations for the radial part
.5 Sharp increase of the quasi-particle lifetime below T c
It has been observed at certain points in the Brillouin zone that the ARPES peak in bismuth cuprates is relatively sharp at low temperatures in the superconduncting state but that it almost disappears into the background above the transition [284.
shows a pronounced minimum at the alleged Fermi surface (ﬁgure 8. why then do some determinations point to a large Fermi surface in cuprates. The general theory of potential scattering in terms of phase shifts was developed in the earliest days of quantum mechanics [288].

The ‘variable phase’ approach [289] solves this problem by taking the phaseshift functions of the radial coordinate.71)
. satisﬁes the phase equation δl (r ) = −k −1 V (r )[cos δl (r ) jl (kr ) − sin δl (r )nl (kr )]2 with δl (r ) −→ −
r→0
(8. In the l = 0 case. (8.69)
and jl (x) and n l (x) are the Riccati–Bessel functions. the Schr¨ dinger equation for the radial part of the angular momentum component (l) of the wavefunction of a particle with wavevector k undergoing potential scattering is u l (r ) + [k 2 − l(l + 1)/r 2 − V (r )]u l (r ) = 0. the phase equation reduces to δ0 (r ) = −k −1 V (r ) sin2 [kr + δ0 (r )]. we must satisfy the condition that V (r ) −→ V0r −n with n < 2.
l=0
(8.63)
The scattering phase shift.65)
In the variable phase method [289]. δl .66)
(8. δl (r ).64) u l (r ) −→ sin(kr − lπ/2 + δl ) and the scattering cross section is then σ = 4π k2
∞
sin2 δl . in the standard formulation.70)
In the slow-particle limit. The Schr¨ dinger equation for each radial o component of the wavefunction then reduces to a ﬁrst-order differential equation for the corresponding phase shift.68)
(kr )2l+3 V0r −n 2 (2l + 3 − n)[(2l + 1)!!]2 k
(8.67)
where the phase function. o In dimensionless units ( = 2m = 1). k2 (8. is obtained by comparison with the asymptotic relation r→∞ (8. The angular momentum phase shift is then δl = lim δl (r )
r→∞ r→0
(8. so that σ = 4π sin2 δ0 .Sharp increase of the quasi-particle lifetime below Tc
235
of the wavefunction may only be solved analytically for a few potentials and. (8. are not in a suitable form for numerical computation. we may neglect higher-order contributions to the scattering cross section.

If the unscreened p scattering potential is the Coulomb potential (V (r ) = V0 /r ). for T → ∞.76)
lim V (r ) =
1/2
V0 exp[−K n r ] ≡ Vn (r ) r ζ(3/2) 1 + 3/2 2 Tc T
3/2 1/2
(8. for T < Tc [290]. The second is the well-known
. 0) = 1 + 4π(e∗ )2 q2 0 f k − f k−q 2 /2m ∗∗ − k · q/m ∗∗ q (8.74)
with ω2 = 4π(e∗ )2 n b /( 0 m ∗∗ ) and n b the boson density. 0) = 1 + ζ (3/2) 1 m ∗∗ ω2 p 1 + 3/2 q2 T 2 Tc T
3/2
+ · · · + O(q 0 )
(8.
r→∞
lim V (r ) =
V0 exp[−K s r ] cos[K s r ] ≡ Vs (r ) r T Tc
3/2 1/4
(8.78)
The T < Tc result is exact for all r at T = 0. Its static dielectric function is: (q.
r→∞
(8.75)
with K s = (m ∗∗ ω p )1/2 1 − and for T → ∞.
(8. (q. the ﬁrst (Levinson’s theorem [289]) states that for ‘regular’ potentials (which include all those with which we shall be concerned). then performing the inverse Fourier transforms.72)
k
in which e∗ = 2e the boson charge. There are two further important analytical results. 0) = 1 + (8. the zero-energy phase shift is equal to π multiplied by the number of bound states of the potential. one ﬁnds that.236
Superconducting state of cuprates
The effective potential about a point charge in the charged Bose gas (CBG) was calculated by Hore and Frankel [290].
Eliminating the chemical potential.73) 4 q Tc q3 and. fk = exp
is the volume of the system and 1
k 2 /2m ∗∗ −µ T
−1
. for small q the dielectric function for T < Tc is 4(m ∗∗ ω p )2 T 3/2 1 1− +O (q.77)
with Kn =
m ∗∗ ω2 p T
+ ···
.

In each case the units are those used to derive the phase equation. m E +| | (8.
Wigner resonance scattering formula. (b) Plot of σn /σs for a range of K = K n = K s in which both potentials have no bound states. which states that for slow-particle scattering of a particle with energy E off a potential with a shallow bound state of the binding energy º E.79)
The zero-energy scattering cross sections for the potentials Vn (r ) =
.Sharp increase of the quasi-particle lifetime below Tc
237
Figure 8. the total scattering cross section is σ = 2π 1 .14. (a) Plots of zero-energy scattering cross sections (i) σn and (ii) σs against screening wavevector K for the potentials (i) Vn (r ) = −(1/r )e−K n r and (ii) Vs (r ) = −(1/r )e−K s r cos(K s r ).

14(a). about three orders of magnitude). It can also be seen that as K decreases. This is the origin of the peaks in ﬁgure 8. the screening wavevector is K 0 = (m ∗∗ ω p )1/2. a Bose condensate of bipolarons might be d-wave. are of the same order of magnitude. As described in section 4.79)) as K decreases.14(a). with these parameters it would then follow that the quasi-particle lifetime is much greater in the superconducting state. the ﬁrst few bound states appear at higher K in the ordinary Yukawa potential.14(b) it may be seen that this is the case when K is large enough. be deeper in a non-oscillatory potential. p Substituting ω p and Tc 3.81)
With a realistic boson mass m ∗∗ = 10m e and dielectic constant 0 = 100.14(b). so substituting for n b = 1021 cm−3 . as there will then be a minimum in the binding energy of the shallowest bound state.238
Superconducting state of cuprates
−(V0 /r )e−K r and Vs (r ) = −(V0 /r )e−K r cos(K r ) are shown in ﬁgure 8. in each case. the equivalent graph for arbitary V0 may be found by rescaling σ and K . the symmetry of the BEC on a lattice should be distinguished from the ‘internal’ symmetry of a single-bipolaron wavefunction and from the symmetry of a single-particle excitation gap. The d-wave condensate
. in general. if bipolaron bands have their minima at ﬁnite K in the centre-of-mass Brillouin zone. e and m e . and at a temperature T = Tc + 0 just above the transition. K Tc = (m ∗∗ ω2 /Tc )1/2 .
(8. and the enhancement of the thermal conductivity in cuprates below Tc [203] due to a many-fold increase in the quasi-particle lifetime. we obtain
2/3
K Tc = K0
√ 2.
8. Another intuitive expectation which is also borne out is that for a given V0 and K .0 K0 me 0
1/4
.80)
From this.9. when a new bound state appears there should be a peak in the cross section.3n b /m ∗∗ . we obtain K αTc m ∗∗ = 3.6 Symmetry of the order parameter and stripes
In bipolaron theory. This effect could then explain the appearance of sharp quasi-particle peaks in ARPES and tunnelling. we see that the ratio is only marginally dependent on the boson density.
(8. K Tc and K 0 . At zero temperature. which may be checked using Levinson’s theorem. while different. These graphs are plotted for V0 = 1.1e m ∗∗
1/2 1/6 0 nb
1/2
. If the screening wavevectors are such that neither the normal state nor condensate impurity potentials have bound states. so that both potentials have no bound states (the difference in cross sections is then. in fact. the non-oscillatory potential should be the stronger scatterer. this agrees with the intuitive conclusion that the bound states should. ﬁgure 8. According to the Wigner formula (equation (8.7. In ﬁgure 8.

the single-particle excitation spectrum might be an anysotropic s-wave providing an explanation of conﬂicting experimental observations.6. this interaction is signiﬁcantly reduced to vi j = e2 . In this case the problem is reduced to narrow-band fermions with a repulsive interaction (equation (8. Because the net long-range repulsion is relatively weak. taking into account the Coulomb repulsion and lattice deformation for a single-polaron wavefunction. the relevant dimensionless parameter rs (= m ∗ e2 / 0 (4πn/3)1/3) is not very large in doped cuprates and the Wigner crystallization does not appear at any physically interesting density. At the same time.2 but.19)) in the real (Wannier) space (ﬁgure 4. However.3 and 5. polarons could be bound into small bipolarons and/or into small clusters as discussed in sections 4. the d-wave bipolaron condensate is ‘striped’.4. o and allows much weaker short-range electron–phonon and antiferromagnetic interactions to bound carriers into small bipolarons. Here we show that the Fr¨ hlich electron–phonon o interaction combined with the direct Coulomb repulsion does not lead to charge segregation like strings or stripes in doped insulators and the antiferromagnetic exchange interaction is not sufﬁcient to produce long stripes either [294].Symmetry of the order parameter and stripes
239
reveals itself as a checkerboard modulation of the hole density and of the global gap (equation (8.82)
Optical phonons reduce the bare Coulomb repulsion at large distances in ionic solids if 0 1. the long-range repulsion would prevent any clustering in inﬁnitely charged domains. taking into account both the long-range attraction of polarons due to the lattice deformations and the direct Coulomb repulsion. Originally antiferromagnetic interactions were thought to give rise to spin and charge segregation (stripes) [291]. The two-dimensional pattern (ﬁgure 4.293] in cuprates and the symmetry of the order parameter [123]. so their deformation ﬁelds overlap at a ﬁnite density. one can apply a discrete version [80] of the continuous nonlinear Pekar equation [61].e. the Fr¨ hlich interaction signiﬁcantly reduces the Coulomb repulsion. i. o However. the role of long-range Coulomb and Fr¨ hlich interactions has o not been properly addressed. which is the case in oxides. However. there is a fundamental connection between stripes detected by different techniques [292.14) is oriented along the diagonals. At distances larger than the lattice constant (|m − n| ≥ a ≡ 1).82)) at large distances and a short-range attraction at atomic distancies. 0 |m − n| (8.4. ψn (the
. Then the d-wave Bose condensate of bipolarons naturally explains superstripes in cuprates. the extention of the deformation surrounding (Fr¨ hlich) polarons is large. the net long-range interaction is repulsive. In contrast.14). Let us ﬁrst consider a non-adibatic and intermediate regime when the characteristic phonon energy is comparable with the kinetic energy of holes. As discussed in section 4. in any case. Hence. In the opposite adiabatic limit.

p ] −
2πe2 κ
|ψn.84) Both potentials satisfy the discrete Poisson equation:
M
κ and
∞
l φn. p. Here the integral I N is given by IN = π (2π)3
π −π π π
dx
−π
dy
−π
dz
sin(N x/2)2 (3 − cos x − cos y − cos z)−1 . p − ψn+m.240
Superconducting state of cuprates
amplitude of the Wannier state |n ): −
m=0
t (m)[ψn − ψn+m ] − eφn ψn = Eψn . p=k
|ψn.k .85)
M c φn. p.k = φn. and U =− e2 e2 2 M IN + M(M − 1)I N κ ∞ (8.q
−1
|ψm. φn. φn. p t (m)[ψn. where T = −2t (N − 1) sin(π M/N)/[N sin(π/N)] is the kinetic energy.k l c φn.m=0 ∗ ψn. (8.31 + ln N . p |2
(8. N 2 sin(x/2)2 1: 1.
The single-particle function of a hole trapped in a string of the length N is ψn = N −1/2 exp(ikn) with periodic boundary conditions. at the site n is created by the l polarization of the lattice.
(8. k.k
= 4πe
p=1
|ψn.87)
corresponds to the polarization and Coulomb energies.88)
It has the following asymptotics at N IN =
.q |2
+
2πe2
∞ n. proportional to t. p |2
(8.83)
The potential φn.q = p
|ψn. N (8. so that the functional J is expressed as J = T + U . p |2
−1
|ψm.k .k + φn.m.m. describing the total energy in this self-consistent Hartree approximation. p.86)
with φn = m (φn − φn+m ).k
= −4πe
p=1. Then the functional J [61]. and by the Coulomb repulsion with the other c : M − 1 fermions. p |2
n. is given by J= −
n.q |2 .k acting on a fermion.

the bipolaron condensate might be striped owing to the bipolaron energy band dispersion. one can expect only short strings (if any) with realistic values of E att (about 0. We conclude that there are no strings in ionic narrow-band doped insulators with only the Fr¨ hlich interaction.
(8.90)
Minimizing this energy yields the length of the string as N = exp
0 E att δω e2 ω
− 2. the hole density. then it becomes clear that the net interaction between polarons remains repulsive in the adiabatic regime as well because κ > ∞ .8 A. with the ˚ 2 in equation (8.3 eV. a string of ﬁnite length can appear (sections 4. Hence. However. one can pose the key question of how one can see stripes at all. yield a short-range polaron–polaron attraction: vatt (n − m) = −E att (δω/ω)δ|n−m|. And this shows the absence of strings or stripes.91) typical values of 0 = 30. For example.89)
where E att = γ 2 ω/2. If (bi)polaronic carriers in many cuprates are in a liquid state. The energy of M well-separated polarons is lower than the energy of polarons trapped in a string. Short-range electron–phonon and/or o antiferromagnetic interactions could give rise to a bipolaronic liquid and/or short strings only because the long-range Fr¨ hlich interaction signiﬁcantly reduces the o Coulomb repulsion in highly polarizable ionic insulators.6.87) into two parts by replacing M 2 for M + M(M − 1). is striped. In this scenario. Taking into account the long-range repulsion as well.31 . the potential energy of the string with M = N polarons is now U= e2
0
N 2 IN −
N E att δω . which is about twice that of the condensate density at low temperatures. ω
(8. dispersive phonons. at least not in all high-Tc cuprates.91).3 and 5.1 (8. Due to the numerical coefﬁcient in the exponent in equation (8. one obtains N even with E att as high as 0. this expression provides a fair estimate of the string length for any kind of attraction (not only generated by phonon dispersion) but also by antiferromagnetic exchange and/or by Jahn–Teller type of e–ph interactions.Symmetry of the order parameter and stripes
241
If we split the ﬁrst (attractive) term in equation (8.87) for any type of short-range interaction. with the
. the short-range attractive forces are not strong enough to segregate charges into strings of any length. as previously discussed. When a short-range e–ph (or exchange) interaction is taken into account.4. In fact. a = 3. We can readily estimate its length by the use of equation (8.1 eV) and with a static dielectric constant 0 ≤ 100.2).91)
Actually. ωq = ω + δω(cos qx + cos q y + cos qz ) with a q-independent matrix element γ (q) = γ .

as observed. Importantly. where a characteristic distance needs to be observed in the normal state as well.
. where the incommensurate peaks were observed only in the superconducting state [154]. a checkerboard modulation (ﬁgure 4. Such an interpretation of stripes is consistent with the inelastic neutron scattering in YBa2 Cu3 O7−δ . The vanishing at Tc of the incommensurate peaks is inconsistent with any other stripe picture.14) has been observed in STM experiments with [295] and without [296] an applied magnetic ﬁeld in Bi cuprates. the incommensurate neutron peaks should disappear above Tc .242
Superconducting state of cuprates
characteristic period of stripes determined by inverse wavevectors corresponding to bipolaron band minima. with the d-wave striped Bose–Einstein condensate. In contrast.

which are real-space electron (hole) pairs dressed by phonons. of the superconducting critical temperature and the upper crtitical ﬁeld. These and some other experimental observations have been satisfactorily explained using this particular approach. A direct measurement of a double elementary charge 2e on carriers in the normal state could be decisive.e. While the BCS theory gives a qualitatively correct description of some novel superconductors like magnesium diborade and doped fullerenes. which provides evidence for a novel state of electronic matter in layered cuprates. tunnelling and the Andreev reﬂection. if the phonon dressing of carriers (i. We have discussed the multi-polaron approach to the problem based on our recent extension of BCS theory to the strong-coupling regime. High-temperature superconductors present a challenge to the conventional theory. Cooper and Schrieffer taken further by Eliashberg to intermediate coupling solved the major scientiﬁc problem of Condensed Matter Physics in the ﬁrst half of the 20th century. We have discussed a few applications of the bipolaron theory to cuprates.
243
. the isotope effect. polaron formation) is properly taken into account. normal and superconducting gaps.Chapter 9 Conclusion
The main purpose of this book was to lead the reader from the basic principles through detailed derivations to a description of the many facinating phenomena in conventional and unconventional (high Tc ) superconductors. They are itinerant quasi-particles existing in the Bloch states at temperatures below the characteristic phonon frequency. Here strong antiferromagnetic and charge ﬂuctuations and the Fr¨ hlich and Jahn–Teller o electron–phonon interactions have been identiﬁed as an essential piece of physics. The seminal work by Bardeen. the magnetic-ﬁeld penetration depth. in particular the bipolaron theory of the normal state. stripes and the symmetry of the order parameter. This is the charged Bose liquid of bipolarons. The low-energy physics in this regime is that of small bipolarons. cuprates remain a real problem. angle-resolved photoemission.

and take the Coulomb interaction between carriers in the mean-ﬁeld (Hartree–Fock) approximation (similar to the central-ﬁeld approximation for atoms) into account. let us deﬁne the translation operator Tl . The eigenvalues of the translation operators are related as follows: c(l + l ) = c(l)c(l ) 244 (A.4)
where c(l) is a number depending on l. the eigenstates of H can be chosen to be simultaneously eigenstates of all the Tl : Tl ψ(r) = c(l)ψ(r) (A. which shifts the argument by the lattice vector l while acting upon any function F(r): Tl F(r) = F(r + l). we arrive at a one-electron Hamiltonian with a periodic (crystal ﬁeld) potential V (r). The wavefunction obeys the Schr¨ dinger equation o − ∇2 + V (r) ψ(r) = Eψ(r).2) To prove the theorem. (A. H (r + l) = H (r).Appendix A Bloch states
A. Hence. it commutes with Tl .1)
The Bloch theorem states that one-particle eigenstates in the periodic potential are sorted by the wavevector k in the ﬁrst Brillouin zone and by the band index n and have the form ψnk (r) = u nk (r)eik·r .1 Bloch theorem
If we neglect the interaction of carriers with impurities and ion vibrations. 2m e (A.5)
. (A.3)
Since the Hamiltonian has the translation symmetry.

8)
Here.3
Ni /2 and ei are unit vectors parallel to the primitive lattice where −Ni /2 < n i vectors. for example to the region −π < ka j π (A. 2. a linear combination of the bi with integer coefﬁcients). n i are integers and bi are primitive vectors for a reciprocal lattice. This requires that exp(iN j k · a j ) = 1. of the reciprocal k-space per allowed value of k is just the volume of a little parallelepiped with edges bi /Ni : 1 k = b1 · b2 × b3 . N1 N2 N3 (A. The number of allowed wavevectors in the ﬁrst Brillouin zone is equal to the number of cells. which satisfy bi a j = 2πδi j (A. if i = j. Replacing k by k + G does not change any of the c(l) if G is a reciprocal lattice vector (i. N. and zero otherwise). (A. 3.13) N
.7)
where j = 1.12) k= a Ni
i=1. in the crystal. Therefore. Hence.6)
with any k. such that N = N1 N2 N3 is the total number of primitive cells in the crystal (N 1023 cm−3 ).11)
which is the ﬁrst Brillouin zone.10) where δi j is the Kronecker delta symbol (δi j = 1. In particular for a simple cubic lattice with period a. The most convenient condition is the Born–von Karman periodic boundary condition ψ(r + N j a j ) = ψ(r) (A. k. Tl+l = Tl Tl . the a j are three primitive vectors of the crystal lattice and N j are large integers. Imposing an appropriate boundary condition on the wavefunction makes the allowed wavevectors k real and conﬁned to the ﬁrst Brillouin zone. the general form for allowed wavevectors is k= n2 n3 n1 b1 + b2 + b3 .2. This relation is satisﬁed if c(l) = exp(ik · l) (A.Bloch theorem
245
since shifting the argument by l + l leads to the same function as two successive translations.e. The volume. 2π ni ei (A.9) (A. all eigenstates of the periodic Hamiltonian can be sorted by the wavevectors conﬁned to the primitive cell of the reciprocal lattice.

the eigenstates of an electron in a crystal are described by almost continuous wavevectors k in the ﬁrst Brillouin zone. the momentum operator p = −i∇ yields −i∇ψk (r) = kψk (r) − ieik·r ∇u k (r) (A. Substituting ψk (r) into the Schr¨ dinger equation yields the following equation for the periodic part of the wavefunction: − 1 (∇ 2 + 2ik · ∇) + V (r) u k (r) = 2m e Ek − k2 2m e u k (r). If we write ψ(r) in the form ψk (r) = u k (r)eik·r (A. F(k) =
k
V (2π)3
F(k) k =
k
V (2π)3
F(k) dk
BZ
(A. similar to that of the free electron momentum p. k. k can be written in the alternative form: k= (2π)3 .
. . The wavevector.246
Bloch states
Since the volume of the reciprocal lattice primitive cell is (2π)3 N/V . 1. When acting on ψk (r). . However. . (A.16)
then u k (r) should be periodic (u k (r + l) = u k (r)) as follows from o equation (A.18)
This equation should be solved within the primitive unit cell and extended periodically for the rest of the crystal. where V is the volume of the crystal. Hence. Hence.14)
This allows us to replace the sum over all allowed k by an integral over the ﬁrst Brillouin zone.19)
which is not a constant multiplied by ψk (r) because of the second term. ψk (r) is not a momentum eigenstate.16). Different eigenstates of the periodic Hamiltonian are also the eigenstates of the translation operator. and by a discrete band index n. they transform under translations as ψk (r + l) = eik·l ψk (r).15)
for any analytical function F(k). although the free-electron wavevector and the momentum are the same (if = 1). . the Bloch k is not proportional to the electron momentum on a lattice. Just like for particles in a box.17) (A. V (A. plays a fundamental role in the electron energy band structure and dynamics. 2. the eigenstates of the equation are each identiﬁed by a discrete quantum number n = 0. Hence.

Effective mass approximation
247
A. 2m e (A.18). At the point k = 0. states in the vicinity of special points of the Brillouin zone matter only at low temperatures and doping. Two of them correspond to the conduction (c) and light hole (lh) valence band: 2 Eg Eg k 2 p2 k2 ± + E c.22)
to obtain a secular equation: Eg + det
k2 2m e kx p me ky p me kz p me
− Ek
k2 2m e
kx p me
ky p me
kz p me
− Ek 0 0
k2 2m e
0 − Ek 0
k2 2m e
0 0 − Ek
=0
(A. the band structure should be calculated numerically. According to the Bloch theorem. we consider a simple cubic semiconductor with the gap E g . (A. Different bras and kets correspond to the four different Bloch functions u n0 (r).x.20) ψnk (r) = eik·r u nk (r) where u nk (r) is periodic in r.23)
where p ≡ s| − i∇x |x = s| − i∇ y |y = s| − i∇z |z . However. of the conduction (empty) band has a minimum and the valence band dispersion has a maximum. It satisﬁes equation (A.lhk = + (A. the energy dispersion nearby can be calculated analytically by applying the so-called ‘k · p’ perturbation theory. u n0 (r) has the symmetry of the crystal space group.y. To illustrate the theory. There is no spin–orbit interaction in this example. so that spin is irrelevant. u nk (r) =
n =s. E nk . in doped semiconductors. y and z under the rotation of the crystal space group. this equation for u n0 (r) is − 1 2 ∇ + V (r) u n0 (r) = E n0 u n0 (r). In general. While their positions and band edges are determined experimentally or numerically. These are the points where the energy dispersion. one can expand u nk in the series. For a small k.z
an k u n 0 (r)
(A.21)
Hence.2 Effective mass approximation
Electron wavefunctions in semiconductors and narrow-band metals differ signiﬁcantly from plane waves because the periodic part of the Bloch function is strongly modulated with a characteristic length of the order of the lattice constant. which has a non-degenerate conduction band of s or d-like symmetry at k = 0 ( point) and three p-like valence bands degenerate at k = 0 and transforming at this point like x.24) 2m e 2 4 m2 e
. There are four solutions to this equation.

it does not necessarily follow that the electron–phonon interaction is small and the renormalization of the
.1. Therefore.25)
One can split two degenerate heavy-hole bands and change the sign of their effective mass if one includes the states at the -point split-off from those under consideration by an energy larger than E g as shown by a thin line in ﬁgure A.27)
In many semiconductors.lh 1 . when experiment does not show heavy carriers. The effective mass approximation follows from equation (A.
and the other two.1.248
Bloch states
(F
N (KK (OK
Figure A.26) E ck − E g 2m c and E lhk where the effective mass is m c. which are degenerate. ‘k · p’ energy bands near the -point of the Brillouin zone in the cubic lattice. the interband dipole moment p is large and the bandgap is small.24) for small k m e E g / p. correspond to heavy holes (hh): E hhk = k2 2m e (A.28) − k2 2m lh (A. = 2 me p /(m e E g ) ± 1 (A. p2 m e E g . (A.29) me me As a result. k2 (A. the electron and hole masses can be signiﬁcantly smaller than the free-electron mass: m lh mc 1.

30)
where m = j m j a j is the lattice vector with the integer m j and we drop the band quantum number n.3 Tight-binding approximation
For narrow-band semiconductors and metals.31)
because the Bloch functions are orthogonal. Indeed calculating the integral in equation (A. Here each of the multipliers is the sum of a geometric progression:
N j /2 Nj
exp(2iπn j l j /N j ) = eiπli
n j =−N j /2+1 n j =1
exp(2iπn j l j /N j ) (e2iπn j − 1) exp(2iπl j /N j ) . If the external or internal ﬁelds are strong and contain high-frequency and (or) short-wave Fourier components. deﬁned as 1 wm (r) = √ N e−ik·m ψk (r)
k
(A. one obtains dr w∗ (r)wn (r) = m The sum is
N1 /2
1 N
eik·(m−n) . e2iπl j /N j − 1
= eiπli
(A. The Wannier wavefunctions are orthogonal. it is convenient to replace the Bloch states by the site (Wannier) wavefunctions.34)
. In this case.31). they involve large momenta of the order of the reciprocal lattice constant making all states of the electron band relevant.32)
e
k
ik·(m−n)
=
n 1 =−N1 /2+1 N3 /2
e
2πin 1 l1 /N1
N2 /2 n 2 =−N2 /2+1
e2πin2 l2 /N2
×
n 3 =−N3 /2+1
e2πin3 l3 /N3
(A.Tight-binding approximation
249
band mass is absent.n m (A.
A.
k
(A. a tight binding approximation is more appropriate. dr w∗ (r)wn (r) = δm.33)
where l j ≤ N j − 1 are integers. The effective mass approximation is applied if the external ﬁeld varies slowly in time and smoothly in space.

m
(A. Many electronic structures. is negative.250
Bloch states
The denominator of this expression is not zero for any allowed l j . except l j = 0. in particular perovskite ones. However. However. o then. after integrating with respect to the electron coordinate r. the sum equation (A. (A. atomic-like Wannier orbitals (equation (A. E k .30) by exp(ik · m) and performing the sum with respect to m. In this case the sum is trivially equal to N which proves the orthogonality of the Wannier functions. True atomic orbitals are not orthogonal for different sites and they do not provide a quantitative description of bands in solids. is expressed via the hopping integrals T (m − n) as
Ek = where T (m − n) =
1 N
T (m)eik·m
m
(A. p. the bottom of the band is found at k = 0. one can prove that 1 N exp[i(k − k ) · m] = δkk .32) is non-zero only for m = n. In the same way.and d-like orbitals. the approximation yields Ek = 1 N T (m)eik·m = 2T (a)[cos(k x a) + cos(k y a) + cos(k z a)] (A. the energy band dispersion. The hopping integrals could not be calculated by using tabulated atomic wavefunctions and potentials estimated for various solids. For a non-degenerate band in a cubic lattice. If the nearest-neighbour hopping integral. the numerator is zero because n j is an integer.39)
|m|=a
if the middle of the band is taken as zero (T (0) = 0). can be ﬁtted with only the nearestneighbour matrix elements between s-. 2m e
(A.30)) can provide a very good description already in a tight-binding nearest-neighbour approximation.40)
. we express the Bloch state as a linear combination of the Wannier states: 1 ψk (r) = √ eik·m wm (r). (A.37)
∗ dr wn (r) −
∇2 + V (r) wm (r). Near the bottom the dispersion is parabolic with the effective band mass m= 1 2a 2|T (a)| .38)
The idea behind the tight-binding approximation is to ﬁt the band dispersion calculated numerically with a ﬁnite number of hopping integrals. T (a).35)
Then multiplying equation (A.36) N m
∗ If we substitute this sum into the Schr¨ dinger equation and multiply it by wn (r). Hence.

the DOS has a square-root singularity near the bottom of the band:
N(E) =
1 2π|T (a)|
|T (a)| . 4π|T (a)| (A. Replacing the sum by the integral. Many physical properties of solids depend on the density of states (DOS) per unit cell. we obtain E 1 (A.Tight-binding approximation
251
Half the bandwidth is D = z|T (a)|. E
(A. E k ∝ (k x + k 2 ).44)
.43)
2 Finally.42) N(E) = 2 |T (a)| |T (a)| 4π if the dispersion is parabolic. for a one-dimensional parabolic spectrum (E k ∝ k x ). the DOS is energy independent: y N(E) = 1 . 1 N(E) = δ(E − E k ) (A. where z = 6 is the number of nearest neighbours. If the dispersion is parabolic but two2 dimensional.41) N
k
rather than on a particular dispersion E k . E k ∝ k 2 .

N) = ln P(U1 . The particles interact with the walls of their container (i. If the two parts do not interact with each other. N) = P(U. N) in thermal equilibrium. N1 ) and P2 (U2 . The statistical probability P of ﬁnding the system in a particular quantum state Q with the total number of particles N is proportional to this time. Then taking the logarithm of both parts of equation (B. the life history of a single particle is not so important. (B.Appendix B Quantum statistics and Boltzmann kinetics
B.2 and the number of particles N1. N) = P2 (U. respectively. Due to the interaction with the thermostat. N1 ) that system 1 should be in the state U1 . N2 ).e. we ﬁrst assume that we know the exact eigenstates |Q and energy levels U Q of the manyparticle system. N1 )P2 (U2 . The probability for two independent events to occur is equal to the product of their separate probabilities.2 . N) = P1 (U1 .1) The thermal equilibrium is described by a uni ver sal probability.1). the probability P1 (U1 . N is the product of P1 (U1 . N1 ) + ln P(U2 . It depends on U Q and N: P = P(U Q . with a thermostat) so as to establish a thermal equilibrium but not to such an extent as to affect the whole set of quantum numbers Q and energy levels. P1 (U. N). What really determines the physical properties of a macroscopic system is the average distribution of particles in real and momentum space. A statistical approach combined with quantum mechanics allow for a full description of the macroscopic system. which depends only on U and N but not on the particular system. Within this approach. Let us consider a system containing two independent parts 1 and 2 with the energies U1. N). This means that the probability P(U. N2 ). we obtain ln P(U. N2 ). It is not hard to ﬁnd the probability P(U Q . N2 ). 252 (B. N) of ﬁnding the whole system in the state U .2)
. the energy and the total number of particles (N) ﬂuctuate.1
Grand partition function
If the number of particles is large. The system takes some time staying in every allowed quantum state Q. P(U. N1 is independent of the probability P2 (U2 .

The quantum state of an ideal Fermi or Bose gas is fully described by the numbers n ν of identical particles in every singleparticle quantum state |ν .2
Fermi–Dirac and Bose–Einstein distribution functions
The quantum statistics of non-interacting identical particles is readily derived using the grand partition function. . the equilibrium density matrix is given by (B. The constant Z (the grand partition function) depends on β and µ because the probability is normalized by the condition
∞
P(U Q . 1. β and µ are independent of U and N. . It describes the interaction with the thermostat and determines the average energy of the macroscopic system. N) = 1. 2. .8)
The quantity β = 1/T is the inverse temperature. It determines the average equilibrium number of particles in an open system. .7)
Hence. The energy levels are given by UQ =
ν
Eν nν
(B.
B. N) = Z −1 e−β(U Q −µN) . Q = {n ν }.3) (B. ∞ for bosons.Fermi–Dirac and Bose–Einstein distribution functions
253
The energy of non-interacting systems and the number of particles are additive: U = U1 + U2 N = N1 + N2 and the only possibility to satisfy equation (B. 1 for fermions and n ν = 0.6) P(U Q .
(B. the constant µ is the chemical potential.4)
where Z .10)
. N) = − ln Z − β(U − µN) (B.
(B.
N=0 Q
(B.2) is given by ln P(U. Hence. where n ν = 0. This means that proper quantum numbers of the whole gas are different sets of the occupation numbers of single-particle states.5) (B. Finally.9)
and the total number of particles is N=
ν
nν . the grand partition function is obtained as
∞
Z=
N=0 Q
e−β(U Q −µN) .

(B. The average energy and the average number of particles of a gas depend on the average values n ν of the occupation ¯ numbers of each single-particle state. (B. N Q )
(B.11)
can be expressed as nν = − ¯ 1 Zβ
∞ N=0 Q
∂ exp − β ∂ Eν
n ν (E ν − µ) = −
ν
∂ ln Z . The distribution function nν = ¯
Q
n ν P(U Q .1 Fermi energy The density of states (DOS) (per unit of volume) of a single particle in a macroscopic box is proportional to its kinetic energy E = k 2 /(2m) in a power depending on the dimensionality d of the box (appendix A.
(B.12)
The partition function is readily obtained by using Z=
{n ν } ν
e−βnν (Eν −µ) =
ν nν
e−βnν (Eν −µ) . and Z= [1 + e−β(Eν −µ) ]. there are only two terms in the sum with n ν = 0 and n ν = 1.3): N(E) E (d−2)/2. ∂ Eν
(B. (B.254
Quantum statistics and Boltzmann kinetics
Here E ν are the single-particle energy levels.3
Ideal Fermi gas
B.3. eβ(Eν −µ) − 1
B.15)
The partition function of ideal bosons is determined by the sum of the geometric progression as ∞ 1 e−βnν (Eν −µ) = (B.13)
In the case of fermions.14)
ν
Hence the (Fermi–Dirac) distribution of ideal fermions is nF = ¯ν 1 eβ(Eν −µ) +1 .17)
.16) −β(E ν −µ) 1−e
n ν =0
and the (Bose–Einstein distribution) of ideal bosons is nB = ¯ν 1 .

respectively.23)
becomes
β 4
∞ −∞
dE
cosh [β(E − µ)/2]
2
. the distribution is a step function eβ(E−µ) 1 ≈ eβ(E−µ) + 1 (E F − E) (B.
(B. For µ > 0 and β → ∞. otherwise the Fermi–Dirac distribution function would be zero for any positive E. we can expand the total energy in powers of T /TF to calculate the electron contribution to the speciﬁc heat of a metal. Then integrating by parts. 2 their total energy U and particle density n (= NF /L d ) are given by U =2 and n=2 dE N(E) dE N(E) E eβ(E−µ) + 1 1 (B.2 Speciﬁc heat At almost all temperatures of interest.1 into a box of volume L d . Let us introduce a function of energy deﬁned as K r (E) =
E −∞
dE N(E )E r . which is the number of states with an energy less than E.and one-dimensional box. (B.18)
. T remains much smaller than the Fermi temperature of ordinary metals.Ideal Fermi gas
255
If we place NF non-interacting fermions with the spin. and (x) = 1 for x > 0 and zero otherwise.20)
where E F ≡ µ(0) is the Fermi energy. 2 takes into account the degeneracy of every orbital state due to the spin. Following Sommerfeld. Calculating the integrals with a step-like distribution yields EF = (3π 2 n)2/3 2m πn EF = m π 2n2 EF = 8m (B. The chemical potential should be positive at T = 0. B. the integral
∞ −∞
dE N(E)
Er eβ(E−µ) + 1 K r (E) . Let us calculate these integrals at zero temperature.22)
In particular K 0 (E) represents the cumulative DOS. and the total energy U n (d+2)/d . (B.21)
for a three-. TF = E F /kB ² 1000 K.3. two.19) +1 Here.

only a narrow region of the order of a few T in the vicinity of the Fermi energy contributes to the integral.24)
Here the dimensionless coefﬁcients a p are given by ap = 2 p−1 p!
∞ −∞
dx
xp cosh2 (x)
.30)
The chemical potential differs from the Fermi energy by small terms of the order of T 2 . In particular. we can keep only this.256
Quantum statistics and Boltzmann kinetics
Since cosh−2 (x/2) falls exponentially for |x| > 1.
(B.26) t z−1 et + 1 (B. we obtain U (T ) ≈ 2
0 µ
dE N(E)E +
π2 2 T [N(µ) + µN (µ)] 3 π2 2 T N (µ). p = 0. we can expand it in a Taylor series around E = µ to obtain
∞ −∞
dE N(E)
Er eβ(E−µ) + 1
∞
=
p=0
T pa p
∂ p K r (µ) .28)
At low temperatures. so we write U (T ) ≈ U (0) + 2E F N(E F )(µ − E F ) + n ≈ n(0) + 2N(E F )(µ − E F ) + π2 2 T [N(E F ) + E F N (E F )] 3 (B. Provided that K r (E) is non-singular in the neighbourhood of E = µ.
(B. for the total energy (r = 1) in any dimensions.27)
(1 − 21−z ) (z) 0 ∞ 1 t z−1 = dt t (z) 0 e −1
dt
and (z) is the gamma function.29)
where N (µ) ≡ dN(µ)/dµ and for the particle density (r = 0) n≈2
0 µ
dE N(E) +
(B. and the quadratic terms with p = 2 and a2 = π 2 /6. dµp
(B. The zero-order coefﬁcient of the Taylor expansion is obtained by a straightforward integration: a0 =
0 ∞
dx
1 cosh2 (x)
= 1. 3
(B. The coefﬁcients with p 2 are usually written in terms of Riemann’s zeta function as a p = [2 − 22− p ]ζ( p) where ζ (z) = 1
∞
(B.25)
Only terms with even p contribute because a p = 0 for odd p.31)
π2 2 T N (E F ) 3
.

one can separate these two contributions plotting C(T )/T against T 2 . The energy levels of electrons are quantized (see section 1. we take n to be temperature independent (n = n(0)) and ﬁnd that µ = EF − Then the total energy becomes U (T ) ≈ U (0) + π2 2 T N(E F ) 3 (B. In practical terms. B.33) π2 2 T N (E F )/N(E F ).34)
The lattice contribution to the speciﬁc heat falls off as the cube of the temperature and at very low temperatures. it drops below the electronic contribution. Landau diamagnetism and de Haas–van Alphen quantum oscillations Let us consider a two-dimensional ideal electron gas with the parabolic energy 2 dispersion E k = (k x +k 2 )/2m on a lattice in an external magnetic ﬁeld H directed y along z. Here ω = eH /m is the Larmour frequency and µB = e/(2m e ) is the Bohr magneton. therefore.35)
where each level is g-fold degenerate with g = L 2 eH /2π and the last term takes into account the splitting of the levels due to the spin with σ = ±1. Thus one can ﬁnd γ ≡ 2π 2 N(E F )/3 and the DOS at the Fermi level by extrapolating the C(T )/T curve linearly down to T 2 = 0 and noting where it intercepts the C(T )/T axis.6. This sharp oscillatory structure of the DOS imposed by the quantization of levels in a magnetic ﬁeld combined with a step-like Fermi– Dirac distribution at low temperatures leads to the famous de Haas and van Alphen oscillations of magnetization of two. The DOS becomes a set of narrow peaks:
∞
N(E) = d
n=0
δ[E − ω(n + 1/2) ∓ µB H ]
(B.4) as E n = ω(n + 1/2) + µB σ H (B.3 Pauli paramagnetism.36)
rather than a constant.and three-dimensional metals with
.Ideal Fermi gas
257
where U (0) and n(0) are the total energy and the particle density at zero temperature. The band mass m might be different from the free-electron mass m e . which is linear in T . 6 (B. To calculate the speciﬁc heat at constant density. Ce (T ) = 2π 2 T N(E F ).3.32)
and the speciﬁc heat of a degenerate electron gas is. 3 (B.

H ) = − ∂H For non-interacting fermions. The Poisson formula is derived by Fourier transforming the sum of δ-functions
∞
(x) =
n=−∞
δ(x − n).39)
where Fr is the transform of F(x).41)
by F(x) and integrating it with respect to x from −δ to +∞ we obtain the Poisson formula (B. (x).42)
. (B. It is described by the thermodynamic potential = F − µN = −T ln Z .258
Quantum statistics and Boltzmann kinetics
the magnetic ﬁeld.40)
and δ → +0.37) M(T. we can expand it into the Fourier series
∞
(x) =
r=−∞
re
2πir x
where the Fourier coefﬁcients are
r
=
1/2 −1/2
dx e−2πir x (x) = 1. The magnetization of the gas is ∂ . Fr =
∞ −δ
dx F(x)e2πir x
(B.14) and
∞
= −dT
n=0 σ =±1
ln 1 + exp
µ − ω(n + 1/2) − µB σ H T
.38)
We calculate this sum by applying the Poisson formula for an arbitrary real function F(n):
∞
F(n) =
n=0
∞
∞
−δ
dx F(x) + 2
r=1
Re Fr
(B.39).
∞
Multiplying the Fourier expansion of
∞
δ(x − n) =
n=−∞ r=−∞
e2πir x
(B. Applying the formula for F(n) = −gT σ =±1 ln{1 + exp[(µ − ω(n + 1/2) − µB σ H )/T ]} yields =
P
+
L
(B. the partition function Z is the product equation (B.
This sum is periodic as a function of x with the period 1. Hence.
(B. We consider an open system (grand canonical ensemble) with a ﬁxed chemical potential independent of the applied ﬁeld.

47)
This result is actually valid for any energy band spectrum of a metal. B (B.Ideal Fermi gas
where
P
259
=−
m L2 T 2π
∞ σ =±1 0
dE ln 1 + exp
µ − E − µB σ H T
(B. It is proportional to the DOS at the Fermi level: χP = 2µ2 N(E F ). we ﬁnd that (B.44) is the ‘quantum’ part due to the quantization of the orbital motion.45) MP = µB (N↓ − N↑ )
L ∞
= −2 dT Re
∞
dx e2πir x ln 1 + exp
where N↑↓ =
L 2m 2π
∞
dE
0
1 exp
E±µB H −µ T
+1
is the number of electrons with spin parallel (↑) or antiparallel (↓) to the magnetic ﬁeld.46) MP = 2µ2 H N(E F ) B where N(E F ) = L 2 m/(2π) is the two-dimensional zero-ﬁeld DOS. Now let us calculate the quantum part L .43)
is the ‘classical’ part containing no orbital effect of the magnetic ﬁeld but only spin splitting and µ − ω(x + 1/2) − µB σ H T r=1 σ =±1 0 (B.
(B. we obtain (B. Taking the derivative of P . we obtain the Pauli paramagnetic contribution to the magnetization (B. The spin susceptibility χP = ∂ MP /∂ H is positive and essentially independent of temperature.48) L = LD + dHvA where
LD
=
dω Re 2π 2
∞ r=1
1 r2
1
σ =±1
1 + exp
ω/2+µB σ H −µ T
(B. Integrating equation (B.50)
.49)
and
dHvA
=
dω2 Re 8π 2 T
∞ r=1
1 r2
∞
dx
σ =±1 0
e2πir x cosh2
ω(x+1/2)+µBσ H −µ 2T
. Performing the integration for T TF with the Fermi step-function.44) twice by parts.

However. The remaining part dHvA is responsible for the de Haas–van Alphen (dHvA) effect. so that (B. µ
LD
=
L 2 (eH )2 ζ(2).52)
which might be larger or smaller than the paramagnetic Pauli term depending on the band mass. Differentiating this expression twice with respect to H yields r=1 r the diamagnetic (Landau) orbital susceptibility χD = − ∂ 2 LD 1 me =− 2 3 m ∂H
2
χP
(B. Then using
∞ −∞
dt
eiαt cosh (t/2)
2 ∞
=
4πα sinh(πα) rf + πr H
we obtain
dHvA
=
r=1
Ar cos
(B.
(B.50) yields the region of x µ/ω 1.56) (1/H ) = S 2 It directly measures the cross section of the Fermi surface S = πkF . where 1/2 is the Fermi momentum. The main contribution to the integral in equation (B.57)
. the oscillating part of magnetization at zero temperature is larger than the monotonic part as µ/ω. With increasing temperature. The oscillating part dHvA is small compared with the (ω/µ)2 at any temperature.54)
and f = 2πmµ/e is the frequency of oscillations with respect to the inverse magnetic ﬁeld 1/H .55)
The period of oscillations with the inverse magnetic ﬁeld of the ﬁrst harmonic (r = 1). the kF = (2mµ) oscillating part of magnetization falls exponentially as Ar e−2π
2 r T /ω
. (B. so that we can replace the lower limit in the integral by −∞. MdHvA = − ∂
dHvA
∂H
≈−
f H2
∞
r Ar sin
r=1
rf + πr . 2mπ 3
Here we applied the series representation of the Riemann’s function ζ(2) = ∞ −2 = π 2 /6.51)
At all temperatures and ﬁelds of interest. due ‘classical’ part P as dHvA/ P to the high frequency of oscillations. 2πe .260
Quantum statistics and Boltzmann kinetics
EF T and µ ω. H
(B.53)
where the amplitudes of the Fourier harmonics are Ar = L 2 eH T πmr cos 2r T /ω) me πr sinh(2π (B.

the condensate density). However.59) is written as n = n s (T ) +
0 ∞
dE
√ where N(E) = m 3/2 E 1/2 /(π 2 2). If the value of µ is of the order of level spacing 1/(m L 2 ) → 0. Let us take the number of bosons in the ground state per unit of volume as n s (T ). Hence. However. n 3 } is the wavevector. NB → ∞ but n is ﬁnite (the so-called thermodynamic limit). the one-particle energy spectrum becomes continuous. The ratio equation (B.61) n s (T ) = n[1 − (T /Tc )3/2 ]
N(E) exp[β(E − µ)] − 1
(B.1 Bose–Einstein condensation temperature Let us consider NB non-interacting Bose particles with zero spin in a cubic box of the volume L 3 under periodic boundary conditions. we obtain the density of bosons in the ground state (i. Calculating the integral. If µ is negative. a relative contribution of the term with k = 0 and E k = 0 could be ﬁnite. The one-particle energy spectrum is given by Ek = 4π 2 2m L 2
3
n2 j
j =1
(B. we can replace the sum by the integral using the DOS.58)
where n j are positive integers including zero and k = (2π/L){n 1 . every term of the sum is ﬁnite. n) as a function of temperature and boson density n = NB /L 3 is determined using the ‘density sum rule’: n= 1 L3 1 exp[β(E k − µ)] − 1 (B.52) of the Landau diamagnetism and the Pauli paramagnetism does not depend on a particular distribution function and remains the same even for non-degenerate electrons.4
Ideal Bose gas
B. n 2 . one can put µ = 0 in this equation.60)
. Taking into account this singular contribution we replace the sum by the integral for all but the ground state.4. if the chemical potential approaches zero from below. In this limit the contribution of every individual term to the sum is negligible (as 1/N) compared with n. Then equation (B. (B. This equation has a negative solution for µ at any temperature.
B. The chemical potential of a gas µ(T. ωτ º 1. the dHvA effect can be observed only at low temperatures (T ω) in clean metals with a well-deﬁned Fermi surface.Ideal Bose gas
261
The scattering of electrons by impurities also washes away dHvA oscillations as soon as the scattering rate is large enough compared with the Larmour frequency. When L.59)
j
where β = 1/T .e. The observation of dHvA oscillations is considered to be the ultimate evidence for a Fermi liquid.

68)
for 0 < T − Tc Tc .313 2n 2/3 . Tc 3 µ = − T ln . (B.64)
if T Tc .65)
When the temperature approaches Tc from above. the condensate density n s (T ) should be zero at temperatures T > Tc . Here ζ(3/2) = 2. and ﬁnally zero at and below Tc . we have 3.262 where
Quantum statistics and Boltzmann kinetics
n ζ(3/2)m 3/2
2/3
Tc = 2π
(B. To obtain its normal state value in the vicinity of Tc .69)
.612. the chemical potential becomes small. the gas is classical. exp[β(E − µ)] − 1
(B.4. 1 ≈ exp[β(µ − E)] exp[β(E − µ)] − 1 (B. we rewrite the density sum rule as 1 1 − . The total energy of the gas per unit of volume is U=
0 ∞
dE N(E)
E . the population of every individual level remains microscopic. E(E − µ)
(B. 2 T (B. At temperatures far above Tc .2 Third-order phase transition To satisfy equation (B.63) kB Tc = m B.62)
is the Bose–Einstein condensation temperature. Now we can calculate the free energy and the speciﬁc heat near the transition. where the chemical potential is negative. so that the integral becomes n[1 − (T /Tc )3/2 ] = dE N(E) n[1 − (T /Tc )3/2 ] = T µ
0 ∞ ∞
dE
N(E) . In ordinary units. exp[β(E − µ)] − 1 exp[β E] − 1 0 (B. In this normal state.66) The function under the integral can be expanded for small E and µ. The magnitude of the chemical potential is logarithmically large compared with the temperature in this limit. Here the Bose–Einstein distribution is almost the same as the Maxwell–Boltzmann distribution.67)
The remaining integral yields µ=− 9π 2 n 2 2m 3 Tc2 T − Tc Tc
2
(B.60).

71).77) 9π 2 n 3 4m 3 Tc2 T − Tc Tc
2
(B.4). Tc 2m 3 Tc4 (B. The free energy density F(T ) = −T
0 T
dT
U (T ) T2
(B.74) ∂µ 2 and 3 27π 2 n 3 Un (T ) = Us (T ) + nµ = Us (T ) − 2 4m 3 Tc2 T − Tc Tc
2
. 3 (B. 2ζ (3/2)
263
(B.128m 3/2 T 5/2. We can take µ = 0 on the right-hand side of this equation because the chemical potential is small just above Tc .75)
Using equation (B. ∂Cs ∂Cn − ∂T ∂T = 27π 2 n 3 n ≈ 3. we ﬁnd that 3 ∂Un = n (B.Ideal Bose gas
In the condensed state below Tc .71)
is found to be Fs (T ) = − 2 Us (T ).70)
Remarkably the energy density of the condensed phase does not depend on the particle density.
(B.76)
While the speciﬁc heat of the ideal Bose gas is a continuous function (see ﬁgure 7. the chemical potential is zero and U = Us (T ) = 3ζ (5/2) nT (T /Tc )3/2 ≈ 0.73) dE N(E)E ∂µ ∂ E exp[β(E − µ)] − 1 0 in the normal state. its derivative has a jump at T = Tc .72) The speciﬁc heat Cs (T ) = 5Us (T )/(2T ) is proportional to T 3/2 in the condensed phase.69) with respect to the chemical potential. we arrive at the free-energy density in the normal state close to the transition as Fn (T ) = Fs (T ) + and the normal state speciﬁc heat Cn (T ) = Cs (T ) − 27π 2n 3 2m 3 Tc3 T − Tc . Then integrating by parts. Tc (B.78)
. Differentiating equation (B.66 . one obtains ∞ 1 ∂ ∂Un =− (B. This is because the macroscopic part of particles in the condensate does not contribute to the energy.

y. t) · ∇ k f (r.4). ∂t (B. t) − f (x + dx. the change in the number of particles in the elementary volume of the phase space is δ N = δ N r + δ Nk . The single-particle distribution function f (r. k z . y. (B. k y + dk y . which ﬁnally yields ∂ f (r. we obtain equation (1. t)]k x dt dk y dk z dr ˙ + [ f (r. k x . k. Their number in the elementary volume dr dk changes due to their motion. The equation can be derived by considering a change in the number of particles in an elementary volume dr dk of the phase space. k x . z. k y . k x . k y . z. k. where the collision integral describes the effect of the internal forces. t) − f (r.79)
This change can also be expressed via a time derivative of the distribution function as δ N = (∂ f /∂t) dt dr dk. k. k y . t) − f (x. t)]k z dt dk x dk y dr ˙ where k = F(r.80) into the external and internal contributions. z. t)]vz dt dx dy dk. k.80)
The force acting upon a particle is the sum of the external force and the internal force of other particles of the system.
. y. y + dy. z. k. k z . The change during an interval of time dt due to the motion in real space is given by δ N r = [ f (x. k. z. It satisﬁes the celebrated Boltzmann kinetic equation. k. k. k z . z + dz. k x . k z + dk z . t) − f (r. k y . Particles with the momentum k move with the group velocity v = dE k /dk. t)]v y dt dx dz dk + [ f (x. t) dr dk. Not only do the coordinates but also the momentum k change under the inﬂuence of a force F(r. which is given at the time t by f (r. k y . Splitting the last term in equation (B. As a result. t) − f (r. k z . t).5
Boltzmann equation
An external perturbation can drive a macroscopic system out of the thermal equilibrium. t)]k y dt dk x dk z dr ˙ + [ f (r. k. t) − f (x. y. k. t) + v · ∇ r f (r. t) = 0. t)]vx dt dy dz dk + [ f (x. t) + F(r.264
Quantum statistics and Boltzmann kinetics
B. t). k z . y. k x . t) in real r and momentum k space is no longer a Fermi–Dirac or a Bose–Einstein distribution. k x + dk x . The change in the number of particles during the interval dt due to the ‘motion’ in the momentum space is ˙ δ Nk = [ f (r. k.

which satisﬁes the stationary Schr¨ dinger equation: o
N
−
i=1
1 2m
i
+ V (r i ) +
1 2
N
V (r i − r j )
i=1 j =i
(q1 . . q2 . or of n. .1)
= E (q1 . q2 . q N ).
Q (q1 . σi ) is a set of position r i and spin σi coordinates. which is the sum of one-particle energies. a spin s and the interaction potential between any two of them V (r i − r j ). The spin coordinate (z-component of spin sz ) describes the ‘internal’ state of the particle. . . qi . There is also an external ﬁeld V (r). . .4) 265
. if the particles are free.2)
o where u k (q) is a solution of a one-particle Schr¨ dinger equation − 1 2m + V (r) u k (r. Direct substitution of equation (C. which might be a composite particle like an atom. σ ) =
k u k (r. We can readily solve this equation for non-interacting particles when V (r i − r j ) = 0. manyparticle eigenstates are products of one-particle wavefunctions. q N ) (C. . .
N
EQ =
i=1
ki . σ )
(C. . for example the wavevector k and sz . . .1) yields the total energy of the system. .Appendix C Second quantization
C. 2. In this case. . Here r i are three position coordinates {x i .3)
k is a set of one-particle quantum numbers. if V (r) is periodic (appendix A). q j . j = 1. . . . q N )
= u k1 (q1 )u k2 (q2 ) × · · · × u k N (q N )
(C. q2 . yi . . .2) into equation (C.1 Slater determinant
Let us consider a quantum mechanical system of N identical particles with a mass m. q2 . i. The system is described by a many-particle wavefunction (q1 . . k and sz . . . z i }. . . .
(C. . 3. N. q N )
where qi ≡ (r i .

... . k j .266
Second quantization
The quantum number Q of the whole system is the set of single-particle quantum numbers. q j . . . . qi . . . . Q = {k1 . .. . qi . . . . . . . 3/2. . . . q2 . . . . Bosons have integer spins. 5/2. Any state of bosons is symmetric and any state of fermions is antisymmetric. . . . . q2 . . . 1. . so that ˆ Pi j (q1 . q N ) will also be a solution of the Schr¨ dinger equation with the same total energy. .. . The eigenstates with the eigenvalue P = 1 are called symmetric and the eigenstates with P = −1 antisymmetric. . u k N (q N ) This function is totally antisymmetric because the permutation of the coordinates of any two particles (let us say q1 and q2 ) corresponds to the permutation of two (C. There is an extreme case of a totally symmetric (or totally antisymmetric) ˆ state which does not change (or changes its sign) under any permutation Pi j . . . . Relativistic quantum mechanics connects the symmetry of the many-particle wavefunction with the spin.7)
. (C. only one of them is acceptable for identical particles. . o q j . Importantly.. q N ). . . qi . q2 . qi . A = √ . . s = 1/2. H ] = 0. . . . q2 . and fermions have half odd integer spins. .. . . which swaps the coordinates qi and q j . . qi . . ˆ Hence. . . s = 0. . While in classical mechanics the existence of sharp trajectories makes it possible to distinguish identical particles by their paths. . 2. the eigenfunctions of H are also the eigenfunctions of Pi j . (C. .5) If we swap the coordinates of any two particles. q N ) = (q1 . u k N (q2 ) Det k1 2 (C. . (C. .9) . . . . the new state Q (q1 . Any linear combination of such wavefunctions solves the problem as well. N! u k1 (q N ) u k2 (q N ) . q j . qi . . . . . . . . k2 . k N }. . . Let us introduce an operator Pi j . q N ). . . q2 . q N ) (equation (C. 3. q2 . .6)
The Hamiltonian (C. . Only these extreme cases are realized for identical particles.2)) with coefﬁcients determined by the use of the Slater determinant: u k1 (q1 ) u k2 (q1 ) . . .1) is symmetric with respect to the permutation of any pair ˆ of the coordinates. q j . . A totally antisymmetric wavefunction A of non-interacting identical fermions is readily constructed as linear superpositions of Q (q1 . . . . ˆ Pi j (q1 ... we have P 2 = 1 and P = ±1. q j . there is no way of keeping track of individual particles in quantum mechanics. q N ) = P (q1 . . ki . u k N (q1 ) 1 u (q ) u k2 (q2 ) . that is Pi j and H commute: ˆ [ Pi j . This quantum ‘indistinguishability’ of identical particles puts severe constraints on our choice ˆ of a many-particle wavefunction. .8) Since two successive permutations of qi and q j bring back the original conﬁguration. q j . .

. . .
C.11)
while two bosons are described by
S (q1 . . so that N1 + N2 + · · · = N. dq1 dq2 |
S. q2 )
1 = √ {u k (q1 )u p (q2 ) + u k (q2 )u p (q1 )} 2
S (q1 . q N )
(C.10). These functions are normalized. The totally symmetric S of bosons is obtained as
S
=
N1 !N2 ! .12)
if k = p. Only one fermion can occupy a given one-particle quantum state. p (q) are normalized and orthogonal. .10) for bosons. q j . q2 )| 2
=1
if u k. . . k (C.10)
where the sum is taken over all permutations of different single-particle quantum numbers ki in Q. . and by
= u k (q1 )u k (q2 )
(C. It is solved in the form of the Slater determinant for fermions (equation (C.14)
It is apparent that if two or more single-particle quantum numbers ki are the same. For example the system of two identical non-interacting fermions. A = 0. is described by
A (q1 . . qi . N2 . the interaction term makes the problem hard.Annihilation and creation operators
267
rows in the matrix N × N which changes the sign of the whole determinant. which can be only in two one-particle states u k (q) and u p (q). . .A (q1 . are the numbers of identical ki in Q.9) or (C. .13)
if k = p. However. .2 Annihilation and creation operators
The collection of identical particles which do not mutually interact is a trivial problem. . q2 )
1 = √ {u k (q1 )u p (q2 ) − u k (q2 )u p (q1 )} 2
(C. √ The factor 1/ N! is introduced to make A properly normalized. dq u ∗ (q)u p (q) = δkp . Then a set of algebraic equations for the expansion coefﬁcients may be treated
. Exact many-particle wavefunctions can be expanded in a series of ‘non-interacting’ functions equation (C. N!
Q (q1 . q2 . q2 )
(C. the totally antisymmetric wavefunction vanishes. N1 . This statement expresses the Pauli exclusion principle formulated in 1925.9)) and of the sum (C.

. . . .
=
n ki + 1δm k1 . . . .268
Second quantization
perturbatively. . . m ki . . . . q N ) A {n i } (q1.nk1 δm k2 . where the initial coordinates {q1. . if all matrix elements of the many-particle Hamiltonian are known. .
(C. . . . . . . . . . n ki . |bki |n k1 . . . n k2 . q N ) are written as ˆ m k1 . . . n k2 . which are functions of the position and spin coordinates of particles qi . . | A|n k1 . 1.17) There are two major types of many-particle symmetric Hermitian operators: ˆ A(1) ≡ ˆ A(2) ≡
N i=1 N i. The wavefunction of non-interacting particles (equations (C. .10) are fully identiﬁed by a set {n k1 . . . q j ). q2 . . whose matrices are deﬁned in the second quantization as m k1 . m k2 . . . |
(C.nk j × · · ·
.nk2 × · · · × δm ki . . .16)
where n ki = 0. . . A representation. . n ki . n k2 . n ki . 1 for fermions.
∗ {n i }
(C. . m k2 . .nki −1 × · · · × δm k j . n k2 . . q N ). . q N } are replaced by new ‘coordinates’ {n k1 .nk2 × · · · × δm ki . q2 . dq N
∗ ˆ {m i } (q1. . . n k2 .15)
and its complex conjugate as a bra. |bki |n k1 . . . n ki . n ki . Differing from the original wavefunctions. . .nk j × · · · (C. . .9) and (C. .10)) is written in the second quantization using the Dirac notations as a ket. m ki . . . ≡ dq1 dq2 .
{n i }
= |n k1 . q2 . n k2 . . . .18)
ˆ While A(1) is the sum of identical one-particle operators (like the kinetic energy). m ki . . q2 . the expansion coefﬁcients are functions of the occupation numbers {n k1 . . . .19)
† m k1 . . . . . . respectively. . . A Their matrix elements can be readily calculated by the use of annihilation † † bk (ak ) and creation bk (ak ) operators. . . ∞ for bosons and n ki = 0. ˆ (2) is the sum of identical two-particle operators (like the potential energy). . . n ki . . These coefﬁcients and matrices perfectly replace the original wavefunctions and operators. = n k1 . . . 2. . m k2 . n ki . j =1
ˆ h(qi ) ˆ V (qi .} of the occupation numbers n ki of each single-particle quantum state. n k2 . √ = n ki δm k1 . . 3. n ki . is known as the second quantization. .
(C. . n k2 . . . .}. .9) and (C. . The non-interacting eigenstates (C. .nk1 δm k2 .nki +1 × · · · × δm k j . . . The matrix ˆ elements of any operator A(q1 .}.

nk1 δm k2 . n k2 . m k2 . Then multiplying a † p † and ak . the matrix elements of fermionic annihilation and creation operators are:
† 0k |ak |1k = 1k |ak |0k = (−1) N(1. . |aki |n k1 . which are n k .nki −1 × · · · × δm k j .k−1)
(C. n k2 .nk j × · · · for fermions. . |aki |n k1 . m ki . . . of the number of one-particle occupied states. b p ] = 0 p {ak . . We see that bk (ak ) decreases the number of particles in the state k † † † † by one and bk (ak ) increases this number by one. 0k |a † |0 p . . bk ] = 1
(C. respectively.nk1 δm k2 . . . .21)
and
† † † ak ak + ak ak ≡ {ak ak } = 1. . These are the occupation number † † operators. .24)
where N(1. . we obtain a non-zero matrix element 1 p . a † } p = {ak . The operators bk bk and ak ak are also diagonal and their eigenvalues are 1 + n k and 1 − n k . (C.Annihilation and creation operators
for bosons and m k1 .nki +1 × · · · × δm k j .22)
The bosonic annihilation and (or) creation operators for different single-particle states commute and the fermionic annihilation and (or) creation operators anticommute: [bk . m k2 . . respectively. 0k |a † ak |0 p .
(C. It follows from † † this deﬁnition that bk is the Hermitian conjugate of bk and ak is the Hermitian conjugate of ak . √ = ± n ki δm k1 . Here + or − depend on the evenness or oddness. For example. m ki . . which precede the state k in the adopted ordering of one-particle states and the occupation numbers of all other single-particle states are the same in both bra and ket.nk2 × · · · × δm ki . . . . The products bk bk and ak ak have diagonal matrix elements only.20)
† m k1 .nk2 × · · · × δm ki . . . k − 1) is the number of occupied states. b† ] = [bk . 1k p p = (−1) N(1.25)
. . 0k |ak |0 p . . . the commutation rules for these operators are:
† † † bk bk − bk bk ≡ [bk . .nk j × · · ·
269
(C. p−1) (−1) N(1. a p } = 0 (C. which precede the state ki . 1k = 1 p . n ki .
= ± n ki + 1δm k1 . .k−1) = (−1) N( p+1. As a result. 0k 0 p .23)
if k = p. n ki .k−1) .

k . p
(C. the eigenstates of the whole system are |2m . 2l = u l (q1 )u l (q2 ) 1 |1m . 1l = √ [u l (q1 )u m (q2 ) + u l (q2 )u m (q1 )].30)
. And an operator of the second type is expressed as ˆ A(2) ≡
k.26) Equations (C. p k. 1k p p = (−1) N(1.29)
For example. we obtain ak a † + a † ak = 0 p p for k = p. respectively. q )u p (q )u k (q).26) have opposite signs because the p-state of the intermediate bra and ket in equation (C.28)
ˆ A(2) ≡
ˆ ˆ Here.k−1) (−1) N(1. If bosons do not interact. 0k |ak |1 p .27)
ˆ A(1) ≡
for bosons and fermions. 2 (C. p † h k k bk bk † h k k ak ak k. which can be in the two different one-particle states u m (q) and u l (q).270
Second quantization
Multiplying these operators in the inverse order yields 1 p . h pk and V pk are the matrix elements of h and V in the basis of one-particle wavefunctions:
pk
hk k ≡ V pk ≡
pk
ˆ dq u ∗ (q)h(q)u k (q) k dq ˆ dq u ∗ (q)u ∗ (q )V (q. let us consider a system of two bosons. p pk pk
(C. 1k 1 p . Hence. k p
(C. p. N = 2. while it is empty in the intermediate bra and ket in equation (C. 0k |ak a † |0 p . Any operator of the ﬁrst type is expressed in terms of the annihilation and creation operators as ˆ A(1) ≡
k.25). 1k |a † |0 p . 1k = 1 p . p−1) = (−1) N( p+1.k . (C.26) is occupied.25) and (C. p. p † V pk bk b† b p bk p † V pk ak a † a p ak . 0l = u m (q1 )u m (q2 ) |0m .k−1)+1 .

0l |bk bk |1m . In many cases we do not know the eigenstates ˆ of the one-particle Hamiltonian h(q). 0l | A(1)|1m . 1l | A(1)|1m .3
-operators
The second quantization representation depends on our choice of a complete set of one-particle wavefunctions.19)). we obtain the result of direct calculations (equations (C. The off-diagonal elements are: ˆ 0m . These operators are deﬁned as (q) =
k †
u k (q)bk
† u ∗ (q)bk k k
(C. 1l = 2h lm √ ˆ 2m . 2l |bk bk |0m .28) allow us to replace the ﬁrst (coordinate) representation by the second quantization (occupation numbers) representation. 1l = h mm + h ll . 0l = ˆ ˆ dq1 dq2 u ∗ (q1 )u ∗ (q2 )[h(q1 ) + h(q2 )] m m (C. 0l = 0 √ † 0m . 2l | A(1)|0m .34)
(q) =
. 1l = 2δkm δk l √ † 2m .27). calculated by the use of the matrix elements of the annihilation and creation operators (equation (C. 1l = δkm δk m + δkl δk l † 0m .
(C. 1l = 2δkl δk m . 0l = 0 √ ˆ 0m . 2l = 2h ll ˆ 1m . are ˆ 2m . 1l |bk bk |1m .31) and (C. and a representation in terms of the ﬁeld -operators is more convenient. 1l = 2h ml .33)
Substituting these matrix elements into equation (C.
C. 2l | A(1)|1m . 2l = 2δkl δk l † 1m . 0l = 2δkm δk m † 0m .32)
† However. 2l |bk bk |2m . The expressions (C.31)
× u m (q1 )u m (q2 ) = 2h mm ˆ 0m . 0l | A(1)|2m . calculated directly by the use of the orthogonality of u m (q) and u l (q) for m = l.27) and (C. 0l |bk bk |2m .32)).
(C. 2l | A(1)|2m .-operators
271
ˆ The diagonal matrix elements of A(1). 2l |bk bk |1m . are: † 2m . the matrix elements of bk bk .

[ (q)
†
(q )] =
k.
(C.36)
dq
(q)
†
ˆ (q )V (q. p
u k (q)u ∗ (q )[ak a † ] p p u k (q)u ∗ (q ) = δ(r − r )δσ σ k
k
= [ (q) (q )] = 0 for bosons and anticommutators
(C. The Hamiltonian of identical interacting particles can be expressed in terms of -operators: H= dq
†
ˆ (q)h(q) (q) +
dq
dq
†
(q)
†
ˆ (q )V (q.28).s
ψnk (r)χs (σ )anks
(C.35) into equations (C.36) we obtain ˆ ˆ equations (C.272
Second quantization
for bosons and in the same way for fermions.
Indeed. while the ‘true’ operators are the ﬁeld operators acting on the occupation numbers.40) Now we can use any complete set of one-particle states to express it in terms of the annihilation and creation operators.38)
{ (q) †(q )} = δ(r − r )δσ σ { (q) (q )} = 0
(C.39)
for fermions.37) (C.27) and (C. q ) (q ) (q). c-numbers).k. Both A(1) and A(2) many-particle operators are readily expressed in terms of -operators: ˆ A(1) = ˆ A(2) = dq dq
†
ˆ (q)h(q) (q)
†
(C.41)
. Here the operators h(q) and V (q.e. if we substitute equations (C. -operators do not depend ˆ ˆ on a particular choice of one-particle states. q ) (q ) (q). for fermions on a lattice a convenient set is the set of Bloch functions (appendix A). The commutation rules for the ﬁeld operators are derived using the commutators of the annihilation and creation operators. For example. (q) =
k †
u k (q)ak
† u ∗ (q)ak . which are parameters (i.35)
(q) =
Differing from the annihilation and creation operators. k k
(C.34) or (C. q ) act on the coordinates. (q) =
n.

s † [T (m − n) − µδm.44) n Here T (m − n) is the hopping integral (equation (A.k .m .
. One can equally use the Wannier functions (A.n ]ams ans + m. p pk
(C.38)) and Vnmm are the matrix elements of the interaction potential. the Hamiltonian takes the following form H=
k.s p. (q) = wm (r)χs (σ )ams (C.s pk † ξk aks aks + k.30) as another complete set.
(C. If we drop the band index n in the framework of the single-band approximation.-operators
273
where we also include the spin component χs (σ ).s † nm † Vnm am s an s ans ams . r ) does not depend on the spin. We assume that V (r.s † V pk ak s a † s a ps aks .42)
Here V pk is the matrix element of the particle–particle interaction V (r. p . calculated by the use of the Wannier functions.n .43)
m.n. r ) calculated by the use of the Bloch states.s n.s
to express the same Hamiltonian in the site representation: H=
m.

t − t ) = − i Tr{e(
−H )/T
Tt ψ(r. Nowadays Green’s functions (GFs) serve as a very powerful tool in many-body theory. σ ) representing both the momentum and spin.1) (D. Now the development of the system proceeds according to e−i H t cν |n . a one-particle GF is applied to any interacting system of identical particles. Thus. where cν is the creation operator for the ˆ † state |ν .e. We use the occupation number representation and suppose that the quantum state is described by |n . Such is the idea behind the following deﬁnition of the one-particle GF at any temperature: G(r. their applications became remarkably broad in theoretical physics.2)
≡ − i Tt ψ(r. so the particle in the state ν gets scattered. Later on. the system is described † † immediately after this addition by cν |n . Usually. i. ν is not usually an eigenstate of H . If at time t = 0 we add a particle with the quantum number ν. If we require the probability amplitude for the persistence of the added particle in the quantum state ν. ˆ However.Appendix D Analytical properties of one-particle Green’s functions
Green’s functions were originally introduced to solve differential equations and link them to the corresponding integral equations. t)ψ † (r . t )}
†
(D. r . it is assigned as the quantum number of a free particle. ˆ † and immediately after the addition of the particle it is described by cν e−i Ht |n . t)ψ (r . t ) where ψ(r. when at a later time we measure to see how much of a probability amplitude is left in the state ν. we must take the ˆ † scalar product of e−i H t cν |n with a function describing the quantum state plus ˆ the particle added at the time t. At this time. Regarded as a description of the time development of a system. the measurement provides information about the interaction in the system. ν = (k. The quantum number ν can be anything depending on the problem of interest. t) = eiH t (r)e−iH t
274
. the state is described by e−i Ht |n .

τ ) . (D. t) = −ie for t > 0 and G(k. In the following we consider fermions with only one component of their spin and drop the spin index.5)
e−Em /T eiωnm t | n|ck |m |2
(D.Analytical properties of one-particle Green’s functions
is the time-dependent (Heisenberg) ﬁeld operator and ˆ ˆ Tt A(t) B(t) ≡ ˆ ˆ (t − t ) A(t) B(t ) ± ˆ ˆ (t − t) B(t ) A(t)
275
with + for bosons and − for fermions. (D.4)
Let us connect the GF and the Matsubara GF. t) = ie
/T n.8)
. If the system is homogeneous. r . τ ) = exp(H τ ) (r) exp(−H τ ) and the thermodynamic ‘time’ τ is τ 1/T . we can write G(k. By the use of the exact eigenstates |n and eigenvalues E n of the Hamiltonian.3) Here ψ(r.m
e−En /T | n|ck |m |2
0
∞
dt ei(ωnm +ω)t dt ei(ωnm +ω)t . the perturbation theory is easier to formulate for the Matsubara GF deﬁned as (r.7)
e−Em /T | n|ck |m |2
0 −∞
dt eizt = lim =
∞
δ→+0 0
dt eizt −δt
i z + iδ 1 = iP + πδ(z) z
(D. τ ) = − Tτ ck (τ )c† k .m /T n.m /T n.m
e−En /T eiωnm t | n|ck |m |2
(D. (D.6)
for t < 0. However. t) = − i Tt ck (t)c† k (k. The GF allows us to calculate the single-particle excitation spectrum of a many-particle system. ω) = − ie + ie Using the formula
∞ 0 /T n. where ωnm ≡ E n − E m . τ − τ ) = − Tτ ψ(r. τ )ψ † (r . The Fourier transform with respect to time yields for the Fourier component G(k. GFs deﬁned within the interval 0 depend only on the difference (r − r ) and their Fourier transforms depend on a single wavevector k: G(k.

15)
.13)
The spectral function plays an important role in the theoretical description of the photoemission spectra (part 2). 1 π
∞ −∞
dω A(k. ω): A(k. ωnm + ω (D. ω) ≡ π(1 + e−ω/T )e
/T n. |m m| = 1
m
so that we can eliminate the summation over m in equation (D. integrating equation (D.11)
Indeed. ω ) . (D.m /T n
e−En /T n|{ck c† }|n k e−En /T = 1.m
e−En /T | n|ck |m |2 1 + eωnm /T − iπ[1 − eωnm /T ]δ(ωnm + ω) .12). ω) = 1. ω) as an integral. ω). when integrating over ω. It is convenient to introduce the spectral function A(k. 2T
∞ −∞
(D.276 we obtain
Analytical properties of one-particle Green’s functions
G(k. ω) = e =e
/T n.10)
which is real and strictly positive ( A(k.m
e−En /T (D. ω) = e
/T n. It obeys an important sum rule: 1 π
∞ −∞ ∞ −∞
dω A(k. i.9) we can connect the imaginary part of the GF with the spectral function as A(k. ω) = e
/T n. ω) > 0).14)
The real part of the GF can also be expressed via A(k.
(D. k k The set of exact eigenstates is complete.e. Re G(k.m
e−En /T | n|ck |m |2 δ(ωnm + ω)
(D.12)
× ( n|ck |m m|c† |n + n|c† |m m|ck |n ). ω−ω (D. ω) = − coth ω Im G(k.10) yields 1 π dω A(k.9)
× P
Here P means the principal value of the integral. ω) = 1 P π dω A(k. Using equation (D.

τ ) exp(iωk τ ) e−En /T | n|ck |m |2 1 + eωnm /T ωnm + iωk (D. in a similar way. ω) = Re G(k. ω) − i coth Im G(k. ω) and
∞ A(k. t) = − i (t) {ck (t)c† } k G (k. ω) + i coth 2T and ω G A (k. using the deﬁnition (equation (D.23)
It becomes clear that the retarded GF is analytical in the upper half-plane of ω and the advanced GF is analytical in the lower half-plane. we also have
(D. ω) G R (k.18) and (D.4)) of the Matsubara GF.21)
G R (k. τ ) = −e
/T n. we can write.22)
∞ −∞
dω
(D. Finally. ω) = −A(k. ωk ) =
0 1/T
dτ (k. ω) =
(D. (k. G A (k.Analytical properties of one-particle Green’s functions
Let us also introduce retarded and advanced GFs deﬁned as G R (k.18) (D.17)
× P We see that
Im G R (k. π ω−ω −∞ As a result.24)
Transforming this function into a Fourier series yields (k. 2T From equations (D.
(D. ω).16)
{ck (t)c† } k
respectively.19). ω ) ω − ω + iδ A(k. ωnm + ω (D. ω ) . ω) = Re G(k.m
.m
e−En /T | n|ck |m |2 eωnm τ .25)
=e
/T n. ω) = and. By the use of the exact eigenstates and eigenvalues of the manyparticle Hamiltonian. ω − ω − iδ
(D. G R (k. ω ) 1 P dω . we connect the GF with the retarded (or advanced) GF: ω Im G(k.m
e−En /T | n|ck |m |2 (1 + eωnm /T ) 1 − iπδ(ωnm + ω) . t) = i (−t)
A
277
(D.19)
Re G R (k. ω) =
1 π 1 π
∞ −∞
dω
A(k. we obtain.20) (D. for τ > 0. ω) = e
/T n. as before.

−ωk ) =
∗
(D.26). iωk − ω
(D.27)
(k.26) and (D. ±1.278
Analytical properties of one-particle Green’s functions
where ωk = π T (2k + 1).
(D. we obtain. ω) for real frequencies continuing (k. from equation (D. ωk ) = 1 π
∞ −∞
dω
A(k.22). ωk ) and.20).
. for ωk > 0. G R (k. one can obtain G(k. . ±2. ω ) . i ωk ) = (k.28)
These relations allow us to construct G R (k. ω). This Fourier transform can be also expressed via the spectral function: (k. k = 0. ωk ) analytically from a discrete set of points to the upper half-plane. ωk ). . . .26)
By comparison of equations (D. Then using equation (D. (k.

We are looking for a complete orthonormal set of multiparticle eigenstates |n . It displays ions to new equilibrium positions depending on the electron coordinates.2)
The requirement that the states generated by equation (E. Using an appropriate transformation.Appendix E Canonical transformation
Various canonical transformations serve as a powerful tool in the polaron theory [124]. The philosophy of the method is simple.6) 279
. which obey the steady-state Schr¨ dinger equation o H |n = E n |n . ˜ ˜ The orthogonality should conserve ( n | n = n|U †U |n = δnn ) so that U † = U −1 .2) form an eigenbase of the given Hamiltonian H is equivalent to the condition that the transformed Hamiltonian ˜ (E. U = e−S (E.1)
There exists a unitary transformation U such that the eigenstates |n may be generated from another arbitrary complete orthonormal set |n such that ˜ |n = U |n .5) (E. we can approximately diagonalize the Hamiltonian of strongly coupled electrons and phonons and then apply the perturbation theory with respect to residual off-diagonal terms.4)
A frequently imployed transformation in polaron theory is the displacement transformation introduced by Lee et al [125] for a single polaron and by Lang and Firsov [71] for a multi-polaron system. (E. (E. ˜ (E.3) H = U†HU is diagonal with respect to |n : ˜ ˜ ˜ ˜ H |n = E n |n .

ν)dqν )e−S ˆ i q.14) i
.11)
n i u ∗ (q.42): ˆ ˆ X i† X j ≡
q. ν).ν † exp[u ∗ (q. ν)dqν ) i
(E.c. ˆ i i
(E. ˆ i
(E. ν)dqν ) i
q.i † n i [u ∗ (q. dqν ]e−S .ν † (u i (q.28) and (4.7)
is such that S † = S −1 = −S.c. We can simplify equations (E8) scaling all matrix elements by the same amount.12)
and
˜ dqν = dqν − η
i
n i u ∗ (q. are ci = ci exp η ˜
q.10)
† Using commutators [n i . ν)dqν − u i (q. ν)dqν − H. Let us now calculate the statistical average of the multi-phonon operator.ν
(E. ν)dqν − H.13).] ˆ i
(E.ν
ci = e S ci e−S ˜
(E.9)
˜ ∂ dqν = ∂η
† e S n i u ∗ (q. ci ](u ∗ (q. and differentiating the transformed operators with respect to the scaling parameter η as ∂ ci ˜ = ∂η and
† e S [n i .38) and (4.c. S] = 0. which determines the polaron bandwidth in equations (4.] . ν)dqν − H. we obtain equations (4. ν). ci ] = −ci . dqν ] = −1 and [n i . [dqν .280 where
Canonical transformation
S=
q. ν) → ηu i (q. ν)[dqν .] exp[u j (q.29). which respects the ‘boundary’ ˜ ˜ conditions dqν = dqν and ci = ci when η = 0. The electron and phonon operators are transformed as ˜ dqν = e S dqν e−S .12) and (E. we ﬁnd ˆ ˆ
∂ ci ˜ = − ci ˜ ∂η ˜ ∂ dqν = − ∂η
˜† ˜ (u ∗ (q.ν. (E. ν). ν)dqν − u ∗ (q.13)
By taking η = 1 in equations (E.8)
(E. u i (q. respectively. ˆ i
i
The solutions of these differential equations. ν)dqν − u i (q.

B].ν)]/2 j i
e−αdq e−|α|
2 /2
. Applying once again the same identity yields e[u i (q. ν) − u j (q. (E.15) allows us to write e[u i (q.c. starting from the third one.16) eη( A+ B) = eη A eη B eη C where C is a number.20)
Quantum and statistical averages are calculated by expanding the exponents in the trace as eα
∗d†
e−αd
∞
N
= (1 − p)
N=0 n=0
p N (−1)n
|α|2n N(N − 1) × · · · × (N − n + 1) (n!)2 (E. one obtains ˆ ˆ C = − 1 [ A. vanish in the right-hand side of equation (E. Then using equation (E. 2
ˆ ˆ Because [ A.] = e(α
∗ † ∗ d † −αd ) qν qν ∗ ∗
(E.19)
× e[u i (q.16) with respect to the parameter η and multiplying it from the right by e−η( A+ B) = e−η C e−η B e−η A
2
ˆ
ˆ
ˆ
ˆ
one obtains The quantity
ˆ ˆ ˆ eη A Be−η A ˆ
ˆ ˆ ˆ ˆ ˆ ˆ A + B = A + eη A Be−η A + 2ηC. ν).18)
η2 ˆ ˆ ˆ ˆ ˆ ˆ = B + η[ A. we use the parameter differentiation [126] as before. B]] + · · · . B] is a number.q=0
ˆ eη A Be−η A =
(−1)q ηr+q ˆ r ˆ ˆ q A BA r !q! (E.] e[u j (q.ν)dqν −H.] = eα
∗ † ∗d † q
×e
[u i (q.ν)dqν −H. Differentiating both sides of equation (E. B]/2
ˆ ˆ ˆ ˆ ˆ ˆ
281
(E.ν)u j (q.ν)−u ∗ (q.15)
ˆ ˆ is instrumental.ν)u j (q.17).ν)]/2
where α ≡ u i (q. B] + [ A. [ A.ν)u j (q.ν)dqν −H.ν)u ∗ (q.c.ν)−u i (q.17)
is expanded by means of
ˆ ∞ r.
(E.c.ν)dqν −H.Canonical transformation
Here an operator identity e A+ B = e A e B e−[ A. Let us assume that ˆ ˆ ˆ ˆ 2 (E. It is applied when the commutator [ A. 2 The identity (E.21)
. all terms.] e[u j (q.18). B] is a number.c. To prove the identity.

23) eα d e−αd = e−|α| nω where n ω = [exp(ωq /T ) − 1]−1 is the Bose–Einstein distribution function of phonons. after which the series over n turns out equal to ∗ † 2 (E.
N=0
(E.42).21) can be written in the form [71] eα
∗d†
e−αd
N
= (1 − p)
n=0
(−1)n
|α|2n n dn p (n!)2 d p n
∞
pN . Now collecting all multipliers we obtain equations (4.
. Here p = exp(−ωqν /T ). 1− p Equation (E.22)
Taking the sum over N.38) and (4.282
Canonical transformation
where we have dropped the phonon and site quantum numbers for transparency. so that a single-mode phonon partition function (appendix B) is 1 Z ph = .
∞
pN =
N=0
1 1− p
and differentiating it n times yields n! in the numerator.

Fiz. Lett. Cooper L N. Fiz. Rev. Chem.References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]
[12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25]
Bednorz J G and M¨ ller K A 1986 Z. R. Lett. 108 1175 London F 1938 Phys. Lett. Acad. Engl. B 23 1796 Alexandrov A S 1983 Zh. Rev. Rev. 32 1442 Fock V 1928 Z. Rev. 47 446
283
. B 64 189 u Bardeen J. 34 953 Chakraverty B K 1979 J. A149 71 Deaver Jr. 54 947 Bogoliubov N N 1947 Izv. 100 463 Blatt J M and Butler S T 1955 Phys. 7 51 a Pippard A B 1953 Proc. 57 273 Alexandrov A S 1983 Russ. 20 1064 Levanyuk A P 1959 Zh. (USSR) 11 91 Ogg Jr. R. Phys. Teor. J. 27 735 u Meissner W and Ochsenfeld R 1933 Naturw. Phys. A 203 210 Landau L D and Lifshitz E M 1980 Statistical Mechanics 3rd edn (Oxford: Pergamon) Abrikosov A A 1957 Zh. Soc. (USSR) 11 77 Landau L D 1947 J. and Schrieffer J R 1957 Phys. R A 1946 Phys. 7 43 Doll R and N¨ bauer M 1961 Phys. Rev. Fiz. 79 845 o Alexandrov A S and Ranninger J 1981 Phys. Chem. Rev. Fiz. 69 243 Schafroth M R 1955 Phys. Physique Lett. Physique Lett. Int. R. A 216 547 Ginzburg V L and Landau L D 1950 Zh. Phys. Rev. 42 1351 Bednorz J G and M¨ ller K A 1988 Angew. 36 810 Pippard A B 1950 Proc. Rev. Eksp. B S and Fairbank W M 1961 Phys. Khim. Eksp. Phys. J R Schrieffer and M L Chiofalo (Amsterdam: IOS) p 309 Alexandrov A S and Mott N F 1994 High Temperature Superconductors and Other Superﬂuids (London: Taylor and Francis) Alexandrov A S and Mott N F 1995 Polarons and Bipolarons (Singapore: World Scientiﬁc) Anderson P W 1975 Phys. Teor. Rev. Rev. Ed. 21 787 London F and London H 1935 Proc. Soc. Soc. Teor. Eksp. Sci. 57 167 Alexandrov A S 1998 Models and Phenomenology for Conventional and Hightemperature Superconductivity (Course CXXXVI of the Intenational School of Physics ‘Enrico Fermi’) ed G Iadonisi. 100 476 Fr¨ hlich H 1950 Phys. 40 L99 Chakraverty B K 1981 J.

Helfand E and Hohenberg P C 1966 Phys. 64 629 [26] Clogston A M 1962 Phys. F Seitz and D Turnbull (New York: Academic) [52] Ginzburg V L and Kirzhnits D A (ed) 1977 The Problem of High-Temperature Superconductivity (Moscow: Nauka) [53] Tolmachev V V 1959 A new Method in the Theory of Superconductivity ed N N Bogoliubov et al (New York: Consultant Bureau) [54] McMillan W J 1968 Phys.–JETP 7 996 [49] Nambu Y 1960 Phys. Fiz. 113 1504 [39] Geilikman B T 1958 Zh. Fiz. Rev. Lett.–JETP 19 1228 [47] Blonder G E. Rev. 38 178 [43] Gor’kov L P 1958 Zh. B 25 4515 [48] Migdal A B 1958 Zh. Rev. Teor. Eksp.–JETP 9 1364 [46] Andreev A F 1964 Zh. 9 266 Chandrasekhar B S 1962 Appl. Phys. Teor. 34 1438 Migdal A B 1958 Sov. Tinkham M and Klapwijk T M 1982 Phys. Eksp. Phys. Eksp. Eksp. 464 [38] Hebel L C and Slichter C P 1959 Phys. 1 251 [30] Feynman R P. 34 735 Gor’kov L P 1958 Sov. Fiz. Teor. The Commemorative Issue (San Francisco. Phys. Teor. Teor. Phys. 1 7 [27] Werthamer N R. 46 1823 Andreev A F 1964 Sov. Rev. Rev. Rev. Rev. Phys. NJ: Prentice-Hall) [33] de Gennes P G 1989 Superconductivity of Metals and Alloys (New York: AddisonWesley) [34] Bardeen J and Pines D 1955 Phys. Fiz. Phys. Eksp.–JETP 11 696 Eliashberg G M 1960 Sov. Fiz. Phys. CA: Benjamin Cummings) [31] Schrieffer J R 1964 Theory of Superconductivity (Reading. Leighton R B and Sands M 1989 Feynman Lectures on Physics. Rev. Gor’kov L P and Dzaloshinski I E 1963 Methods of Quantum Field Theory in Statistical Physics (Englewood Cliffs. Fiz. 117 648 [50] Scalapino D J 1969 Superconductivity ed R D Parks (New York: Marcel Dekker) [51] Allen P B and Mitrovic B 1982 Solid State Physics vol 37. 99 1140 [35] Cooper L N 1957 Phys. Boyer L L and Papaconstantopoulos D A 1978 Phys. Eksp. Rev. Teor. Fiz. B 167 331 [55] Klein B M. Lett. B 18 6411
. Eksp. Rev. Fiz. Phys. Lett. Teor. 38 966 Eliashberg G M 1960 Zh. 104 1189 [36] Eliashberg G M 1960 Zh. 5 147. 147 295 [28] Schmidt V V 1997 Physics of Superconductors ed P Muller and A V Ustinov (Berlin: Springer) [29] Josephson B D 1962 Phys. Lett.–JETP 7 505 [44] Matsubara T 1955 Prog. 36 1918 Gor’kov L P 1959 Sov. ed H Ehrenreich. Phys. Teor.–JETP 12 1000 [37] Giaever I 1960 Phys. Eksp. MA: Benjamin) [32] Abrikosov A A. 14 351 [45] Gor’kov L P 1959 Zh. Theor.284
References
Landau L D 1930 Z. 34 1042 [40] Lifshits E M and Pitayevski 1978 Statistical Physics Part 2 Theory of Condensed State (Oxford: Butterworth–Heinemann) [41] Svidzinskii A V 1982 Space Inhomogeneous Problems in the Theory of Superconductivity (Moscow: Nauka) [42] Abrikosov A A and Gor’kov L P 1960 Zh. 39 1437 Eliashberg G M 1960 Sov.

Eksp. Rev. B 104 619 Hotta T and Takada Y 1997 Phys. Chem. 21 494 Emin D and Holstein T 1969 Ann. Fiz. 16 335 Fr¨ hlich H 1954 Adv. 8 343 Friedman L and Holstein T 1963 Ann. Phys. Phys. B 52 12 720 Fehske H. Phys. Rev. B 53 8449 Benedetti P and Zeyher R 1998 Phys. Brown D W and Lindenberg K 1998 J. Solids 5 34 Sewell G L 1958 Phil. B 49 9915 Bishop A R and Salkola M 1995 Polarons and Bipolarons in High-Tc Superconductors and Related Materials ed E K H Salje. 2 75 Rashba E I 1985 Excitons ed E I Rashba and D M Struge (Moscow: Nauka) Kabanov V V and Mashtakov O Yu 1993 Phys. 3 446 Fay D and Layzer A 1968 Phys. Rev. B 12 905 Kohn W and Luttinger J M 1965 Phys. Phys.–Usp. Rev. ed F Seitz. Phys. 18 190 Alexandrov A S and Mazur E A 1989 Zh. Teor. Fiz. Fiz. Eksp.–JETP 16 1301 Eagles D M 1963 Phys. Phys. Eksp. Phys. 8 325 Holstein T 1959 Ann. 186 456 Appel J 1968 Solid State Physics vol 21. Phys. Rev. 3 1361 Holstein T 1959 Ann. 96 1773 Rashba E I 1957 Opt. Phys. 43 1843 Lang I G and Firsov Yu A 1963 Sov. Rev. A S Alexandrov and W Y Liang (Cambridge: Cambridge University Press) p 271 Alexandrov A S 1992 Phys. Chem. Rev. Fiz. Spectrosc. 130 1381 Eagles D M 1969 Phys. Rev. Rev. B 56 13 916 Romero A H. Teor. Eksp. 109 6504 La Magna A and Pucci R 1996 Phys. Kabanov V V and Ray D K 1994 Phys. 3 325 o Feynman R P 1955 Phys. Rev. (USSR) 3 664 Pekar S I 1946 Zh. Fiz. Lett. 20 187 Landau L D 1933 J. Lett. Phys. B 46 14 932 Alexandrov A S. Mag. 23 381 Yamashita J and Kurosawa T 1958 J. Teor. B 58 14 320
. 181 1278 Eagles D M 1969 Phys. 115 403 Geilikman B T 1975 Sov. B 47 6060 Aubry S 1995 Polarons and Bipolarons in High-Tc Superconductors and Related Materials ed E K H Salje. Rev. Teor. Rev. 15 524 Friedel H 1954 Adv. A S Alexandrov and W Y Liang (Cambridge: Cambridge University Press) p 353 Marsiglio F 1995 Physica C 244 21 Takada Y and Higuchi T 1995 Phys. 53 439 Lang I G and Firsov Yu A 1962 Zh. Phys. Rev. Phys. Loos J and Wellein G 1997 Z. 97 660 Devreese J T 1996 Encyclopedia of Applied Physics vol 14 (Weinheim: Wiley– VCH) p 383 and references therein Tjablikov S V 1952 Zh. Nauk. Rev.References
[56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72]
285
[73] [74] [75] [76] [77] [78] [79] [80] [81]
[82] [83] [84]
[85] [86] [87] [88] [89] [90] [91]
Allen P B and Dynes R C 1975 Phys. D Turnbull and H Ehrenreich (New York: Academic) Firsov Yu A (ed) 1975 Polarons (Moscow: Nauka) B¨ ttger H and Bryksin V V 1985 Hopping Conduction in Solids (Berlin: Academie) o Mahan G D 1990 Many Particle Physics (New York: Plenum) Geilikman B T 1975 Usp.

Math. 15 601 Alexandrov A S 1999 Phys. 82 807 Proville L and Aubry S 1999 Eur. 90 297 Wilcox R M 1967 J. 93 1007 Alexandrov A S. Rev. Rev. B 11 41 Alexandrov A S 2000 Phys.–JETP 6 560 Vinetskii V L and Pashitskii E A 1983 Fiz. B 60 3252 Bonca J. Lett. B 60 1633 Alexandrov A S and Kornilovich P E 1999 Phys. Solid State 28 631 Bryksin V V and Gol’tsev A V 1988 Fiz. Phys. B 33 4526 Alexandrov A S. 33 730 Vinetskii V L and Giterman M Sh 1958 Sov. B 45 13 109 Alexandrov A S and Ranninger J 1992 Physica C 198 360 Vinetskii V L and Giterman M Sh 1957 Zh. Phys. Matter 14 5337 Alexandrov A S. Teor. Phys. B 54 15 363 Alexandrov A S 1998 Physica C 305 46 Wagner M 1986 Unitary Transformations in Solid State Physics (Amsterdam: North-Holland) Lee T D. Phys. Lett. Phys.286
[92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103]
References
Frank T and Wagner M 1999 Phys. Phys. B 51 5887 Gross E P 1961 Lett. B 64 094507 c Alexandrov A S and Kornilovitch P E 2002 J. Rev. Tela (Leningrad) 25 1744 Vinetskii V L and Pashitskii E A 1983 Sov. Rev. B 48 10 571 Alexandrov A S. Eksp. Soc. Phys. Rev. Phys. Phys. 124 649 Alexandrov A S and Kabanov V V 2002 J. Fiz. T 102 39 and references therein u Alexandrov(Aleksandrov) A S and Kabanov V V 1986 Fiz. Tverd. Japan 54 245 Alexandrov A S 1992 Phys. Fiz. 124 1387 Fetter A L 1971 Ann. Tverd. Rev. Samarchenko D A and Traven S V 1987 Zh.: Condens. Tela 30 1476 Bryksin V V and Gol’tsev A V 1988 Sov. Beere W H and Kabanov V V 1996 Phys. 80 763 Alexandrov A S and Beere W H 1995 Phys. Phys. 33 195 Pines D and Schrieffer J R 1961 Phys. Solid State 30 851 Bonˇ a J and Trugman S A 2001 Phys. Phys. Status Solidi B 109 95 Alexandrov A S 1996 Phys. Samarchenko D A and Traven S V 1987 Sov. Rev. Tela 28 1129 Alexandrov(Aleksandrov) A S and Kabanov V V 1986 Sov. Phys. Trugman S A and Batistic I 1999 Phys. Rev. 8 962
[104] [105]
[106] [107] [108] [109] [110]
[111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126]
. Rev. Rev. Rev. B 61 12 315 Hiramoto H and Toyozawa Y 1985 J. Solid State 25 1005 M¨ ller K A 2002 Phys. B 46 2838 Gogolin A A 1982 Phys. Matter 14 L327 Tonks L and Langmuir I 1929 Phys. Rev. J. Phys. Phys. Scr.: Condens. Ranninger J and Robaszkiewicz S 1986 Phys. 64 1 Bishop R F 1974 J. Eksp. Tverd. Rev. Rev. Nuovo Cimento 20 454 Pitaevskii L P 1961 Sov. Low F and Pines D 1953 Phys. B 60 14 573 Alexandrov A S 1984 Doctoral Thesis MIFI (Moscow) Alexandrov A S 1993 Phys.–JETP 66 567 Peierls R E 1933 Z. 51 188 Alexandrov A S and Ranninger J 1992 Phys.–JETP 13 451 Foldy L L 1961 Phys. Rev. Low Temp. Teor. Rev. Rev. B 53 2863 Alexandrov A S and Sricheewin C 2000 Europhys.

Sci.References
287
[127] Wu M K. Rev. Sidis Y. 58 908 [128] Hebard A F. Rev. 13 735 [155] Abbamonte P.–JETP 28 1267 [137] Schrieffer J R. Engl. 68 530 [153] Forro L. Regnault L P and Sidis Y 2000 J. Somal H S. Koops G E J. Proc. Bossy J. B 61 9166 [147] Hubbard J 1963 Proc. Rosseinsky M J. Gao L. Hor P H. Ren Z F. B 37 5897 [145] Sch¨ tzmann J. B 34 8190 [150] Scherman A and Scrieber M 1995 Phys. Logvenov G and Bozovic I 2002 Science 297 581 Cho A 2002 Science 297 499 and references therein
. Rev. Int. Teor. Kivelson S A and Orgad D 2002 Concepts in high temperature superconductivity Preprint cond-mat/0206217 and references therein [142] Gabovich A M. Ilakovac V. Zettl A. B 52 10 621 [151] Laughlin R B 2002 Preprint cond-mat/0209269 [152] Xiang X D. Rusydi A. Zenitani Y and Akimitsu J 2001 Nature 410 63 [131] Alexandrov A S and Edwards P P 2000 Physica C 331 97 [132] Simon A 1997 Angew: Chem. Murphy D W. Loh J E and Hirsch J E 1986 Phys. Rev. Technol. Regnault L P. B 42 167 [139] Mott N F 1990 Adv. A 134 1416 [135] Fr¨ hlich H 1968 Phys. Lett. Tsvetkov A A. Rev. Lett. Torng C J. Cohen M L. Corkill J L. Palstra T T. Wang Y Q and Chu C W 1987 Phys. Cooper J R. Rev. and Kortan A R 1991 Nature 350 600 [129] Holczer K. 39 55 [140] Anderson P W 1997 The Theory of Superconductivity in the Cuprates (Princeton NJ: Princeton University Press) [141] Carlson E W. Ivanov A and Keimer B 2000 Science 288 1234 Bourges P. Phys. u Koleshnikov N. A 276 238 [148] For a review of the Hubbard and t–J models see Plakida N M 2001 Lectures on the Physics of Highly Correlated Electron Systems V (AIP Conf. Monien H and Pines D 1990 Phys. Keimer B. Rev. Ayache C and Henry J Y 1992 Phys. 14 R1 [143] Mott N F 1990 Metal–Insulator Transitions 2nd edn (London: Taylor and Francis) [144] Wheatley J M. B 46 6626 [154] Bourges P. van der Marel D. Voitenko A I. u Rev. Br¨ ck E and Menovsky A A 1997 Phys. 55 2387 Pashitskii E A 1969 Sov. Eksp. B 55 11 118 [146] Nagaosa N and Lee P A 2000 Phys. Fiz. Kaner R B. 36 1788 [133] Ginzburg V L 1968 Contemp. Wang J H. Phys. Haddon R C. Glarum S H. Nakagawa N. Kijima N and Gronsky R 1992 Phys. Vareka W A. B 39 11 663 [138] Millis A J. Ed. Lett. Phys. Sawatzky G A. 9 355 [134] Little W A 1964 Phys. NY: American Institute of Physics) p 121 [149] Scalapino D J. Fu K J. Ashburn J R. Emery V J. Venema L. Annett J F and Ausloos M 2001 Supercond. Rev. Ramirez A P. A 26 169 o [136] Pashitskii E A 1968 Zh. Klein O. Rev. Meng R L. Hsu T C and Anderson P W 1988 Phys. R. Soc. Whetten R L and Diederich F 1991 Science 252 1154 [130] Nagamatsu J. 580) ed F Mancini (Melville. Supercond. Wen X G and Zhang S C 1989 Phys. Huang S M. Rev. Fong H F. Muranaka T. Huang Z J.

Low Temp. Rev. Kes P H and Li M 2002 Science 295 2239 [158] Labbe J and Bok J 1987 Europhys. Rev. B 42 7989 [171] Calvani P. Kishio K. Supercond. Schmitt-Rink S. 12 9
. 177 543 [162] Radtke R J. Stat. Hussain Z and Shen Z X 2001 Nature 412 510 [175] Egami T 1996 J. 91 113 [172] Timusk T. B 62 14 487 and references therein [157] Molegraaf H J A. Phys. Phys. Sadowski W and Walker E 1994 Solid State Commun. Rice T M and Zhang F C 1987 J. 75 4650 [160] Sachdev S.) 46 128 Alexandrov A S 1987 JETP Lett. Prog. Rev. Foster C M. Capizzi M. Mag. Janssen S. Uchida S. Scharpf O. Maselli P. Kremer R K. Lett. Sch¨ ttler H B and Norman M R 1993 Phys. 20 L229 [168] Emin D 1989 Phys. Technol. Mesot J. Kellar S A. 6 215 Alexandrov A S and Mott N F 1994 J. Cheong S-W. Lett. Supercond. Rev. 63 1996 [164] Emery V J and Kivelson S A Nature 374 434 [165] Leggett A J 1973 Physica Fennica 8 125 Leggett A J 1998 J. van der Marel D. Phys. 84 1990 [177] Shen Z X. 7 599 Alexandrov A S and Mott N F 1994 Rep. Chubukov A V and Ye J 1994 Phys. Lett. Feng D L. Andersen K H. Simon A and Ziebeck K R A 1998 J. Rev. Neumann K U. Pattnaik P C and Tsuei C C 1991 Phys. 62 1544 [169] See different contributions in 1995 Polarons and Bipolarons in High-Tc Superconductors and Related Materials ed E K H Salje. Yoshida T. Rev. A S Alexandrov and W Y Liang (Cambridge: Cambridge University Press) and in 1995 Anharmonic properties of High Tc cuprates ed D Mihailovic et al (Singapore: World Scientiﬁc) [170] Mihailovic D. DiCastro C and Grilli M 1995 Phys. Lanzara A. Keller H and M¨ ller K A 1997 Nature 385 236 u [174] Lanzara A. 57 1197 [180] Allen P B 2001 Nature 412 494 [181] Gor’kov L P 1999 J.288
References
[156] Hirsch J E 2002 Science 295 2226 Hirsch J E 2000 Phys. Beck U. Magnetism Magn. Zhou X J. 46 107 [167] Prelovsek P. Sci. C: Solid State Phys. Presura C. Mater. B 82 1349 [178] Phillips J C 1989 Physics of high Tc Superconductors (New York: Academic) [179] Alexandrov A S and Mott N F 1993 Supercond. Voss K and Heeger A J 1990 Phys. Phys. B 49 11 919 [161] Smith T J. Fiz. Abrahams E and Ruckenstein A E 1989 Phys. Conder K. 93 927 Popov V N 1987 Functional Integrals and Collective Excitations (Cambridge: Cambridge University Press) [166] Alexandrov A S 1987 Pis’ma Zh. Mutka H and M¨ ller K A u 2000 Phys. Rev. B 43 3075 Abrikosov A A 2000 Physica C 341 97 [159] Castellani C. Noda T. Furrer A. Hunt M B. Rev. Paolone A. Bogdanov P V. Teor. Lett. Rev. Capellmann H. (Prilozh. Fujimori A. Ishihara S and Nagaosa N 2002 Phil. Levin K. 3 1225 Newns D M. Roy P. 105 791 [176] Temprano D R. Lupi S. Eisaki H. Lu E D. Homes C C and Reichardt W 1995 Anharmonic Properties of High Tc Cuprates ed D Mihailovic D et al (Singapore: World Scientiﬁc) p 171 [173] Zhao G. Suppl. Littlewood P B. Lett. Shimoyama J I. Eksp. B 48 15 957 u [163] Varma C M.

Rev. Kabanov V V and Mott N F 1996 Phys. Xu Z A. Rev. Rev. Lu J P and Lee W C 1992 Phys. Lett. Fukuzumi Y. B 56 5654 [205] Yu R C. 84 2043 [188] Kim J H. Ando Y. Shimoyama J. Lett. Komiya S and Sun X F 2002 Phys. Lett. A 236 132 [203] Alexandrov A S and Mott N F 1993 Phys. B 52 10 516 a [192] Catlow C R A. Mizuhashi K. Ichikawa N and Uchida S 1996 Phys. Klaser M. Krajewski J J and Peck W F 1994 Phys. Phys. Lett. Lett. Pietronero L and Str¨ ssler S 1995 Phys. Salamon M B. Hwang H Y. Rev. Lett. Okuya M. Boebinger G S. Low Temp. Gagnon R and Taillefer L 2000 Phys. and Demsar J 1999 Phys. Feenstra B J. Phys. Lett. Kishio K. Trans. 77 4796 [187] Alexandrov A S and Bratkovsky A M 2000 Phys. Gerrits A M and Wittlin A 1994 Phys. B 49 13 065 [189] Alexandrov A S 2001 Physica C 363 231 [190] Alexandrov A S and Kabanov V V 1996 Phys. Proust C. 88 077001 [208] Proust C. Kao H L. Phys. Rev. Kwo J. Lett. 68 245 [201] Boebinger G S. Lett. Schweiss P. B 58 14 219 Chen W M. 70 2012 [200] Mott N F 1993 Phil. Schuppler S. Kimura T and Kishio K 1995 Phys. 77 5417 Ando Y. Lett. Zagar K. Krishana K. B 60 6995 and references therein [196] Zhang Y. Rev. Taillefer L. Kabanov V V. 87 721 [186] Alexandrov A S. Lett. Chakarian V. Karpinski J. 75 4662 [202] Alexandrov A S 1997 Phys. Str¨ ssler S and Grimaldi C 1995 Phys. R. Mag. Lett. Bratkovsky A M and Mott N F 1994 Phys. Gui Z. Idzerda Y U. Franck J P and Jung J 2000 Physica C 341 1875 [198] Hwang H Y. Takagi H. Lett. Rev. 72 2636 Batlogg B. 80 5192 Dow J D. Lett. NJ: Prentice-Hall) [194] Alexandrov A S. 43 877 [185] Alexandrov A S 1992 J. B 54 3655 [191] Pietronero L 1992 Europhys. Takagi H. Asaoka H and Takei H 1997 Phys. Yu M. Ruan K Q.References
289
[182] Tempere J. Kaldis E and Wachter P 1993 Phys. Lee W Y. Rev. Rev. 71 1075 [204] Takenaka K. 17 365 Pietronero L and Str¨ ssler S 1992 Europhys. Rev. 101 661 [183] Alexandrov A S 1998 Phil. Kimura T. Islam M S and Zhang X 1998 J. Uchida S. Rev. Hill R W. Boakin E. 69 1431 [206] Hill R W. Rev. Lett. Ando Y. Rev. Somal H S. Kao H L and Kwo J 1994 Physica C 235 130 [199] Bucher B. 75 1158 a Pietronero L. Rev. Rev. Soc. Passner A. Blackstead A 1998 Bull. Rev. Matter 10 L49 [193] Anselm A 1981 Introduction of Semiconductor Theory (Englewood Cliffs. 84 2219 [197] Chen X H. Fournier P and Greene R L 2001 Nature 414 711 [207] Takeya J. Am. 18 627 a Grimaldi C. Tamasaku K. Freeland J. Howard U. Passner A. Soc. Nucker N. A 356 197 [184] Merz M. Cava R J. Chen C T.: Condens. Ong N P. Li S Y. 72 1734 [195] Mihailovic D. van der Marel D. Rev. Rev. Lett. Muller-Vogt G and Wolf Th 1998 Phys. Fomin V M and Devreese J T 1997 Solid State Commun. Taillefer L and Mackenzie A P 2002 Preprint condmat/0202101
. Batlogg B. Steiner P. Cava R J. Lett. Zhang G C and Cao L Z 1998 Phys. Rev.

Levin K and Liu D Z 1996 Phys. 70 3490 [212] Hofer J. 82 2620 Alexandrov A S and Kabanov V V 1999 Phys. Rev. Rev. Rossel C and Keller H 1998 Physica C 297 103 [213] Zverev V N and Shovkun D V 2000 JETP Lett. Rev. Low Temp. Gygax S. 47 5395 [221] Schneider T and Singer J M 2000 Phase Transition Approach To High Temperature Superconductivity (London: Imperial College Press) [222] Legget A J 1980 J. B 44 5190 Gorbar E V. Lett. Harlow R L and Moudden A H 1990 Science 250 1309 [231] Bornemann H J. B 1 158 Maly J. Manabe C. Rev. R¨ der H. Momono N. Pedersen M H and Schneider T 1995 Physica B 230 955 [228] Schmitt Rink S. Mohamed M A K. Rev. Wei J Y T and Ginley D S 1988 Phys. and Sproule G I 1991 Phys. Mitchell D F and Sproule I 1989 Physica C 162 753 Franck J P. Narwankar P and Chandrachood M 1991 Physica C 185–189 1359 [232] Franck J P. Rev. Kopnin E M. Mag. A S Alexandrov and W Y Liang (Cambridge: Cambridge University Press) p 453 [219] Alder B J and Peters D S 1989 Europhys. Lett. B 44 5318 [233] Fisher R A. Jung J. McCarron E M III. B 51 3242 [230] Crawford M K. Bourges P. Willemin M. Kim S. Rev. Sinha A P. Varma C M and Ruckenstein A E 1989 Phys.290
References
[209] Machi T. Lett. Mirza K A and Liu R S 1992 Phil. Yethiraj M. Loktev V M and Sharapov S G 1996 Physica C 257 355 Pistolesi F and Strinati G C 1996 Phys. Pistolesi F and Strinati G C 1998 Eur. B 59 13 628 [218] Uemura Y J 1995 Polarons and Bipolarons in High-Tc Superconductors and Related Materials ed E K H Salje. Oda M. J. B 54 15 657 [226] Alexandrov A S and Rubin S G 1993 Phys. Tomeno I. Rev. Lett. Rev. Miura Y and Ido M 1994 Phys. Jung J. Liu H B. Burlet P and Henry J Y 1992 Phys. Regnault L P. Morris D E. B 65 1405
. Chrzanowski J. Teor. Lett. 63 445 [229] Traven S V 1994 Phys. Rev. Physique 41 C7 [223] Nozieres P and Schmitt-Rink S 1985 J. Salomons G J. Phys. B 53 15 168 Marini M. Scr. Schwer H. Miyatake T. Cooper J R. Karpinski J. Wheatley J M. B 47 8369 [215] Nakano T. Mason T E and Armstrong T 1993 Phys. B 49 16 000 [216] Fehske H. Rev. Morris D E. B 51 16 582 o [217] Alexandrov A S 1999 Phys. Fiz. Irwin J C. Gygax S. Markelz A G. Rev. Rev. 72 73 [214] Nakamura Y and Uchida S 1993 Phys. Aeppli G. 59 195 [224] Micnas R and Kostyrko T 1995 Recent developments in High Temperature Superconductivity ed P W Klamut and M Kazimierski (Berlin: Springer) p 221 and references therein [225] Gyorffy B L. Rev. Meijer G I. Wellein G and Mistriotis A 1995 Phys. Vettier C. B 38 11 942 [234] Loram J W. and Tanaka S 1991 Physica C 173 32 [210] Rossat-Mignod J. Mohamed M A K. 75 312 Singer J M. Phillips N E. 10 1 [220] Pokrovsky V L 1988 Pis’ma Zh. Phys. B 47 5141 [227] Tiverdi N and Randeria M 1995 Phys. Miner W A. Lacy S E. Staunton J B and Stocks G M 1991 Phys. Rev. 45 74 [211] Mook H A. Molinski R. Farneth W E.

Tsydynzhapov G E. Rev. Coffe G. Hughes S D. ed A Narlikar (New York: Nova Science. Goldenfeld N and Legget A J 1996 Physical Properties of High Temperature Superconductors vol 5. 79 1551 [241] Zavaritsky V N. Eisaki H. 76 983 [246] Lawrie D D. B 58 11 153 Roulin M. B 62 716 [251] Annett J. Rev. B 12 1007 [255] Li Q. Liu R S and Sinclair D C 1993 Phys. Braun E. Kamal S. Naumenko I G. Mod. Mod. 83 4160 [256] M¨ ller K A 2002 Phil. 82 279 and references therein u [257] Walter H. Soulen R J. Teor. Bozovic I. Burkhardt H. Rakoto H. Karpinski J. J. Beamish J R. 60 1178 [236] Junod A. Rice J P. Braunisch W. Eckstein J N and Virshup G F 1994 Phys. Gu G D and Koshizuka N 1999 Phys. Beere W H. Prusseit W.References
291
[235] Inderhees S E. Huber H. Ousset J C. Molz E B. Wolf S A. 72 21 [248] Nakanishi T. Askenazy S. Kozeeva L P and Lavrov A N Zh. Assmann W. Junod A and Walker E 1998 Physica C 296 137 [240] Alexandrov A S. Suenaga M. Rev. ed D M Ginsberg (Singapore: World Scientiﬁc) p 375 [252] Tsuei C C and Kirtley J R 1997 Physica C 282 4 and references therein [253] Bonn D A. Inderhees S E. Eckert D. Lee V Y and Muller J 1989 Physica C 159 215 [237] Schnelle W. and Tsydynzhapov G E 2000 JETP Lett. Askenazy S. Rev. Commack) p 1 Revaz B. Kabanov V V and Liang W Y 1997 Phys. Franck J P. 72 3292 [245] Alexandrov A S. Li S L and Zhao Z X 2000 Phys. Pari P. Chen W M and Graf M J 1997 J. Phys. Rev. Chaikin P M and Naughton M J 2000 Phys. Eksp. Baar D J. Rev. Junod A and Erb A 1998 Phys. Lonzarich G G. Fiz. Rev. Mitamura H. Pari P. 107 491 [247] Gantmakher V F. Lett. Eckstein J N and Virshup G F 1993 Phys. Lett. Motoyama N. Rev. Rainer D and Sauls J A 1998 Phys. Rev. Zhang K. Kabanov V V and Alexandrov A S 2002 Europhys. Rakoto H. o Harnischmacher J and Wohlleben D 1990 Physica C 168 465 [238] Salamon M B. Lett. 14 3617 o [249] Lee I J. Lett. Pazol B G and Ginzberg D M 1988 Phys. Uchida S and Mˆ ri N 2000 Int. Emel’chenko G A. Rev. Kinder H. Broto J M. Takahashi H. Ruppel S. Mag. Liang W Y and Nevsky P L 1996 Phys. 71 1238 [244] Osofsky M S. Salamon M B. D¨ mel R. Lett. Semerad R. Broto J M. Klein E and Hardy W N 1994 Phys. Lett. Coffe G. Kaldis E and Wachter P 1990 Physica C 167 324 [243] Mackenzie A P. Carrington A. Lett. Rev. Wolf S A. Tsay Y N. 60 127 [242] Bucher B. Lett. Low Temp. Triscone G. Klemm R A. Bozovic I. Broicher H. B 50 4051 [254] Xiang T. Zavaritsky V N. Lett. Panagopoulos C and Cooper J R 1998 Int. 80 3598 [258] Alexandrov A S and Giles R T 1999 Physica C 325 35 [259] Alexandrov A S 1998 Physica C 305 46
. Phys. 88 148 Gantmakher V F. Lett. Rice J P and Ginsberg D M 1990 Physica A 168 283 [239] Junod A. Liang R X. Takeshita N. Goldenfeld N. 1996 Studies of High Temperature Superconductors vol 19. Phys. B 62 R14 669 [250] Wen H H. J. 71 2315 Osofsky M S. Rev. Julian S R. Soulen R J. Ousset J C.

49 86 Ding H. Kadowaki K and Hoechst H 2000 Phys. Rev. Koelling D and Dabrowski B 1994 Phys. Sato T. Lett. Rev. Lett. Revaz B. Dessau D S.: Condens. B 52 R16 981 Halperin B I 1965 Phys. Randeria M. Delsing P and Claeson T 2000 Phys. Loeser A G. Kadowaki K and Fischer Ø 1998 Phys. Rev. 139 A104 Frisch H L and Lloyd S P 1960 Phys. Lett. 84 1788 Fedorov A V. Rev. 84 5860 Suzuki M and Watanabe T 2000 Phys. Lett. Gu G D and Koshizuka N 1999 Phys. 78 1771 and references therein Loram J W. B 55 R3410 Mourachkine A 2000 Europhys. Ding H. Bonn D A. 64 603 Gofron K. Campuzano J C. Conder K and Keller H 1998 J. Phys. Yokoya T. Mandrus D G. Takahashi T. Mirza K A. Campuzano J C. B 60 15 414 Alexandrov A S 2002 Phil. J R Schrieffer and M L Chiofalo (Amsterdam: IOS) p 115 Kaminski A. Randeria M. Winkler D. Lett.292
[260] [261] [262] [263] [264]
References
Alexandrov A S and Andreev A F 2001 Europhys. Li Q. Fretwell H. 85 4787 Tinkham M 1996 Introduction to Superconductivity (Singapore: McGraw-Hill) p 20 Demuth J E. 73 3302 Schabel M C. Mag. Abrikosov A A. Liang R and Hardy W N 1998 Phys. Bansil A. Takahashi T. Persson B N J. Migliori A. Lett. 82 2179
[265] [266] [267] [268] [269] [270] [271] [272] [273] [274] [275] [276] [277] [278] [279] [280] [281] [282] [283]
[284]
[285]
. Rev. Mochiku T. 120 1175 Randeria M and Campuzano J C 1998 Models and Phenomenology for Conventional and High-temperature Superconductivity (Course CXXXVI of the Intenational School of Physics ‘Enrico Fermi’) ed G Iadonisi. Desgardin G and Monot I 1995 Physica C 250 59 Krasnov V M. Rev. Campuzano J C. Genoud J Y. Davydov D N. Rev. 58 576 Alexandrov A S and Dent C J 1999 Phys. Norman M R. Cooper J R and Liang W Y 1994 Physica C 235 134 Deutscher G 1999 Nature 397 410 Yagil Y. Zhao G M. Matsuura A. Ding H. Valla T. B 81 1397 For a review see Shen Z X 1998 Models and Phenomenology for Conventional and High-temperature Superconductivity (Course CXXXVI of the Intenational School of Physics ‘Enrico Fermi’) ed G Iadonisi. Mesot J. Matter 10 u L291 Hancotte H. Lett. Rev. Rev. Rev. Rev. Lett. Rev. Jansen A G M and Wyder P 1997 Phys. B57 6090 Park C H. Lett. Lindroos M. J R Schrieffer and M L Chiofalo (Amsterdam: IOS) p 141 Renner Ch. Deltour R. and Johnson P D. Rev. 54 373 Alexandrov A S and Sricheewin C 2002 Europhys. Holtzberg F and Chandrasekhar C V 1990 Phys. Hass N. Lett. 80 149 Mott N F and Davis E 1979 Electronic Processes in Non-crystalline Materials 2nd edn (Oxford: Oxford University Press) M¨ ller K A. Mochiku T. Shen Z X. Kadowaki K and Giapintzakis J 1996 Nature 382 51 Shen Z X and Schrieffer J R 1997 Phys. Norman M R. Wade J M. Shen Z X. Park C H. Yurgens A. Rev. Sarrao J and Fisk Z 1995 Phys. Lett.

86 890 [287] Alexandrov A S and Dent C J 2001 J. Anderson P W. Hudson E W. Liang R and Hardy W N 2001 Phys. Uchida S and Davis J C 2002 Science 295 466 [296] Howald C. Madhavan V. B 40 7391 [292] Tranquada J M. Bonn D A. 72 569 [295] Hoffman J E. Axe J D. Phys. Rev. Rev.References
293
[286] Zhang Y. B 12 2619 [291] Zaanen J and Gunnarsson O 1989 Phys. Matter 13 L417 [288] Landau L D and Lifshitz E N 1977 Quantum Mechanics (Non-relativistic Theory) 3rd edn (Oxford: Butterworth–Heinemann) [289] Calogero F 1967 Variable Phase Approach to Potential Scattering (New York: Academic) [290] Hore S R and Frankel N E 1975 Phys. Lang K M. Eisaki H. Lett. Sternlieb B J. Eisaki H. Nakamura Y and Uchida 1996 Nature 375 561 [293] Bianconi A 1999 J.: Condens. Kaneko N and Kapitulnik A 2002 Preprint cond-mat/0201546
. Rev. Physique IV 9 325 and references therein [294] Alexandrov A S and Kabanov V V 2000 JETP Lett. Ong N P.

57 in high-Tc cuprates. 135. 36. 219 anisotropy. 180. 174. 129 theory. 182 model of cuprates.Index
adiabatic ratio. 174 band. 239 ARPES. 76 representation. 180 on-site. 161 u bipolaron apex. 167 master equation. 205 excitation spectrum. 261 boundary conditions. 264 kinetics.119 small. 33 Bednorz and M¨ ller discovery. 245 buble approximation. 181 bipolaronic gas. 76 theorem. 36. 52. 179 coherence
. 198 anomalous averages. 179 superlight. 243 triplet. 177 inter-site. 79. 174 in-plane. 122 repulsion. 177 in high-Tc oxides. 120 Bardeen–Cooper–Schrieffer (BCS) –Bose liquid crossover. 125 self-trapped. 41 ground state. 3 Born–Oppenheimer. 18 Brillouin zone. 174 binding energy. 36 gap. 86 theory. 52. 79 c-axis transport. 135 instability. 179 hopping. 36 -like theories of high-Tc. 21. 140 Boltzmann equation. 180. 181 294
size. 135 Bloch function. 226 attraction non-retarded. 80 polarization. 138 transformation. 119 superconductivity. 98 Bose gas. 4. 136 Hamiltonian. 129 localized. 126. 244 Bogoliubov equations. 58 NS phases. 85 antiferromagnetic interaction. 98 Andreev reﬂection in BCS superconductors. 39 Hamiltonian. 195 clusters. 101 singlet. 37. 119 spectrum. 53.

36
. 80 density of states in cuprates. 45 of ideal Fermi gas. 248 electons in metal. 83 screening. 74 Eliashberg: equations. 5 infrared. 75 condensate wavefunction. 148 energy. 16 collision integral. 81 electron–electron correlations. 83 momentum. 50 pairs. 23. 6 thermal. 203 of polaronic superconductors. 40 permanent. 88 entropy of BCS quasi-particles. 32 phenomenon. 165 excitation spectrum of BCS superconductors. 38 of quasi-particles. 20. 151 ﬁeld thermodynamic. 74 dynamical mean-ﬁeld theory. 254 Fermi–Dirac. 70. 121 crystal ﬁeld. 199 longitudinal. 8. 254 function. 148 high frequency. 75 electron–phonon coupling. 35 interaction. 49 Cooper pairing of repulsive fermions. 102 effective mass. 205 Coulomb pseudopotential.Index
factor. 167 sum rule. 257 dielectric function of CBG. 74. 44 with scattering. 74 current density. 12 volume. 154 dHvA oscillations. 35 coupling constant. 162. 28 ﬁeld lower. 42 of bipolaronic superconductors. 81 distribution Bose–Einstein. 73 static. 5 diamagnetic. 4 commutation rules. 142 conductivity Hall. 45 exchange interaction. 40 paramagnetic. 5 optical. 146 of Bose–Einstein condensation. 45 length. 32 crossover regime. 85 extension of BCS theory. 29 critical temperature of BCS superconductors. 89 repulsion. 12 Drude model. 251 quasi-particle. 17 ﬁeld upper. 262 of high-Tc cuprates. 38. 152 velocity. 91 pairing unconventional. 4 of bare electrons. 169 per unit cell. 72 spectral function. 16. 73 critical current. 6
295
Debye approximation. 37 dirty superconductors. 6 dynamic matrix. 13.

10 free energy London expression. 275 high-Tc superconductors. 107 gap in BCS superconductors. 129 Hamiltonian. 164 hyperﬁne interaction. 206. 255 Fermi liquid breakdown. 40 quantization. 65. 24 Hartree approximation. 13 –Landau theory limitations. 162 Hosltein –Hubbard model. 44 on carrier mass. 129 negative. 45 of high-Tc oxides. 64 formalism. 141 of two-component Bose gas. 38 instability. 205 repulsive. 173. 14 –Landau parameter. 62. 158 3 He. 65 temperature. 23 Fr¨ hlich o –Coulomb Hamiltonian. 5. 98 theories of high Tc . 49 Heisenberg Hamiltonian. 59 one-particle. 95. 206
of charged Bose gas. 277 anomalous. 48. 127 model. 43 Gor’kov equations. 16 Ginzburg –Landau equation. 97 interaction. 85
. 77. 189 harmonic oscillator. 59 matrix. 59 Green’s function advanced. 9 superﬂuid. 10 quantum. 53. 78 formulation of BCS theory. 60. 218 gapless superconductivity. 206 in BCS superconductors. 48. 207 on penetration depth. 86 Matsubara. 112. 195 in BSC superconductors. 208 Hebel–Slichter peak absence in high-Tc oxides. 158 heat capacity of BCS superconductors. 275 normal. 277 Hall ratio. 77 observation. 240 4 He. 33 polaron. 75. 15 –Landau theory. 136 operators. 274 phonon. 99 polaron. 79 retarded. 15 ﬂuctuation theory. 204 ﬂux expulsion. 107 hopping. 194 incoherent background. 15 golden rule. 9 vortex. 55 gauge transformation.296
Index
electron. 17 Gi -number. 145 Fermi energy. 115 isotope effect bipolaronic. 146. 18 Gibbs energy. 41 in high Tc oxides. 164 ﬂuctuations. 99 Hubbard U inﬁnite.

183 overscreening. 148 experiment. 275 Maxwell equation. 52. 281 phonon. 164 marginal. 48. 12 operators annihilation. 234 localization length. 268. 88 –Eliashberg theory breakdown. 164 nuclear spin relaxation rate. 86 operators. 166 Knight shift. 32 kinetic mechanism. 109 Mott insulator. 35. 5 Matsubara frequencies. 50 bipolaronic. 84 tensor. 88 mixture of condensates. 124 nearly Fermi liquid. 157 symmetry. 111 renormalized. 163. 247 Kosterlitz–Thouless temperature. 195 in metals. 48 k · p perturbation theory. 163 multi-phonon correlator. 195. 62. 13 Lang–Firsov transformation. 11 Ogg–Schafroth phenomenology. 75 order parameter BCS. 103 LDA. 164 Luttinger liquid. 142 Monte Carlo algorithm. 75 ﬁeld. 28 diamagnetism. 34 pairing collective. 9 gauge. 9 theory. 195 Kohn–Luttinger. 74 Legendre polynomials summation theorem. 194 Ogg’s pairs. 91 Korringa law breakdown. 71 displacement. 271 identity. 192 mass of small polaron. 6 Lorenz number. 204 Landau criterion. 30 effect. 8 McMillan formula. 268. 269 creation. 183 overdoped cuprates. 236 lifetime of quasi-particles. 7
297
effect in CBG. 213 London equation. 167 RVB. 7 metal–ammonia solution.Index
Josephson current. 98 theorem. 269 current density. 34. 242 non-Fermi-liquid theories interlayer RVB. 257 theory. 167
. 156 overcrowding. 34. 51 Levinson theorem. 113. 9 penetration depth. 90 Meissner–Ochsenfeld effect. 167 N´ el temperature. 165 e neutron scattering. 167. 11 Migdal –Eliashberg theory.

115 Pippard equation. 207 fermions. 241
individual. 230. 110. 257 partition function. 186 resonance. 83 hole. 33 frequency. 48 spin gap. 144 polaron band. 195 pseudospin. 89. 183 phase equation. 161 spin-ﬂip scattering.298
Index
–phonon interaction. 12 theory. 78 phonons. 110
. 254 Peierls substitution. 12 limit. 141 plasmon acoustic. 189 two-phonon. 167 quasi-particle spectrum. 103 phase diagram. 79 second order. 138 Pekar equation. 104 Bloch states. 78 optical. 46 spectral function. 135 quantum phase transition. 25 paramagnetism. 13 non-local relation. 11 Cooper. 119 pseudogap. 195 response function. 187 impurity. 235 phase shift. 235 phonon frequency. 239 perturbation expansion. 101 small. 183 out-of-plane. 142 resistivity below Tc . 102 large. 106 self-energy electron. 194 strings. 228. 12 Pauli exclusion principle. 257 jump. 117. 102 superconductivity. 95 collapse. 103. 117. 109. 182. 144 ionic. 228 phonon. 102 band narrowing factor. 110 radius. 171 pairs Bose. 188 in-plane. 12 plasma excitations. 104 mass. 82 self-trapping. 95 spectral function. 267 matrices. 266 sound attenuation. 111 –polaron interaction. 187 of bosons. 95 Slater determinant. 86 pair-breaking limit. 147 scattering boson–acoustic phonon. 116 polaronic Fermi-liquid. 81 velocity. 105 dynamics. 276 spin bipolaron. 46 damping. 72 damping. 194. 81 speciﬁc heat anomaly. 95 level shift. 77 phonon sidebands.

83 electron–phonon. 95 sum rule. 17 second-order. 17 symmetry. 249 time ordering. 117 vortex charged. 156 susceptibility diamagnetic. 23 of ideal Bose gas. 36 canonical. 192 Wigner crystallization. 29 single-electron. 238 Hall. 237
.Index
stripes. 49 of high Tc cuprates. 156 unconventional. 41 SIS. 29 Wannier orbitals. 221 tunnelling conductance. 23. 6. 43 in high-Tc cuprates. 154. 7 spin. 192 of BCS superconductors. 13 superﬂuid density. 156 of BCS superconductor. 96. 262 transmission. 35 upper critical ﬁeld curvature. 13 surface energy. 16 third-order. 221 Type I. 17. 27 lattice. 163 transformation Bogoliubov. 210 Van Hove scenario. 219 Josephson. 249 representation. 152 core. 118 Wigner cross section. 192 thermally activated hopping. 279 Lang–Firsov. 96 Wiedemann–Franz law. 20 pinning. 164 Tolmachev logarithm. 239 strong coupling theory. 20
299
uncertainty principle. 148. II superconductors. 103 transition continuous. 26 line. 63 t– J model. 111 vertex corrections. 189. 154 of interacting Bose gas. 196 thermal conductivity enhancement. 95 tight-binding approximation. 22 free energy. 167 variational calculations. 152 interaction. 276 supercurrent. 166.