§ Aléa & SaclayIntroduction to the Replica Theory of Disordered Statistical Systems This book provides a detailed introduction to the statistical mechanics of classical spin systems with quenched disorder. The first part of the book describes the physics of spin-glass states using results obtained within the framework of the mean field theory of spin glasses. The technique of replica symmetry breaking is explained in detail, along with a discussion of the underlying physics. The second part is devoted to the theory of critical phenomena in the presence of weak quenched disorder. Here one can find a systematic derivation of the traditional renormalization group theory, which is then used to obtain a new ‘random’ critical regime in disordered vector ferromagnets and in the two-dimensional Ising model. The third part of the book describes other types of disordered systems, with relation to new results at the frontiers of modern research. This book is suitable for graduate students and researchers in the field of statistical mechanics of disordered systems. It will also be useful as a supplementary text for graduate level courses in statistical mechanics. Viktor Dorsenko is a professor at the Laboratoire de Physique Théoretique des Liquides, Université Paris VI, and has been a researcher at the Landau Institute for Theoretical Physics in Moscow since 1983. He has lectured at universities all over the world and from 1990 to 1992 was a visiting professor at the universities of Rome, La Sapienza and Tor Vergata, and then at the Ecole Normal Supérieure, Paris. He has published widely in the area of statistical mechanics and disordered spin systems.Collection Aléa—Saclay: Monographs and texts in statistical physics C. Godréche (ed.): Solids Far from Equilibrium P. Peretto: An Introduction to the Modeling of Neural Networks C. Godréche and P. Manneville (eds.): Hydrodynamics and Nonlinear Instabilities A. Pimpinelli and J. Villain: Physics of Crystal Growth D. H. Rothman and S. Zaleski: Lattice-Gas Cellular Automata B. Chopard and M. Droz: Cellular Automata Modeling of Physical Systems J. Marro and R. Dickman: Nonequilibrium Phase Transitions in Lattice Models J.-F. Sadoc and R. Mosseri: Geometrical Frustration V. Dotsenko: Introduction to the Replica Theory of Disordered Statistical SystemsIntroduction to the Replica Theory of Disordered Statistical Systems Viktor Dotsenko Université Paris VI ‘and Landau Institute for Theoretical Physics, Moscow CAMBRIDGE INIVERSITY PRESSPUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE ‘The Pitt Building, Trumpington Street, Cambridge, United Kingdom CAMBRIDGE UNIVERSITY PRESS ‘The Edinburgh Building, Cambridge, CB2 2RU, UK 40 West 20th Street, New York, NY 10011-4211, USA 10 Stamford Road, Oakleigh, VIC 3166, Australia Ruiz de Alarcon 13, 28014 Madrid, Spain Dock House, The Waterfront, Cape Town 8001, South Africa http://www.cambridge.org © Viktor Dotsenko 2001 This book is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2001 Printed in the United Kingdom at the University Press, Cambridge Typeface Times 11/13pt System W4TeX [UPH] A catalogue record of this book is available from the British Library ISBN 0 521 77340 7 hardbackContents Preface page ix 1 Introduction 1 1.1 General principles of the statistical mechanics 1 1.2 The mean-field approximation 4 1.3 Quenched disorder, selfaveraging and the replica method 8 Part one: Spin-glass systems 13 2 Physics of the spin-glass state 15 2.1 Frustrations 15 2.2 Ergodicity breaking 7 2.3. Continuous sequence of phase transitions 19 24 Order parameter 21 2.5 Ultrametricity 22 3 Mean-field theory of spin glasses 26 3.1. Infinite range interaction model 26 3.2. Replica symmetric solution 27 3.3 Replica symmetry breaking 31 3.4 Parisi RSB algebra 37 3.5 RSB solution near T, 39 3.6 de Almeida-Thouless line 41 4 Physics of replica symmetry breaking 46 4.1 The pure states 46 4.2 Physical order parameter P(q) and the replica solution 47 5 Ultrametricity 54 5.1 Ultrametric structure of pure states 54 5.2 Tree of states 57 5.3 Scaling in the space of spin-glass states 60 5.4 Phenomenological dynamics 62 vvi ML Contents Experiments 6.1 Aging 6.2 Temperature cycles and the hierarchy of states 6.3 Temperature dependence of the energy barriers Part two: Critical phenomena and quenched disorder Scaling theory of the critical phenomena 7.1 The Ginzburg-Landau theory 7.2. Critical exponents 73 Scaling 74 Renormalization-group approach and e-expansion 75 Specific heat singularity in four dimensions Critical phenomena in systems with disorder 8.1 Harris criterion 8.2 Critical exponents in the ¢‘-theory with disorder 8.3 Critical behavior of the specific heat in four dimensions Spin-glass effects in the critical phenomena 9.1 Non-perturbative degrees of freedom 9.2 Replica symmetry breaking in the RG theory 9.3 Scaling properties and the replica symmetry breaking 9.4 Discussion Two-dimensional Ising model with disorder 10.1 Two-dimensional Ising systems 10.2 The fermion solution 10.3 Critical behavior in the disordered model 10.4 Numerical simulations 10.5 General structure of the phase diagram Part three: Other types of disordered system Ising systems with quenched random fields 11.1 The model 11.2 General arguments 11.3. Griffiths phenomena in the low-temperature phase 11.4 The phase transition One-dimensional directed polymers in random potentials 12.1 General scaling arguments 12.2. Replica solution for 5-correlated potentials 12.3, Random potentials with finite correlation radius Vector breaking of replica symmetry 13.1 Zero-dimensional systems 13.2 Directed polymers in a random potential 22 66 69 75 17 a 82 84 87 94 97 97 105 107 107 112 118 125 128 128 129 135 141 145 151 153 153 154 155 164 164 167 171 178 179 193Contents vii 13.3. Replica instantons and Griffiths singularities in the disordered Ising model 195 14 Conclusions 209 Bibliography 213 Index 219Preface This book is devoted to the special area of statistical mechanics that deals with the classical spin systems with quenched disorder. It is assumed to be of a pedagogical character, and it aims to help the reader to get into the subject starting from fundamentals. The book is supposed to be selfcontained (the reader is not required to go through all the references to understand something), being understandable for any student having basic knowledge of theoretical physics and statistical mechanics. Nevertheless, because this is only an introduction to the wide scope of statistical mechanics of disordered systems, in some cases to get to know more details about a particular topic the reader is advised to refer to the existing literature. Although throughout the book I have tried to present all the unavoidable calculations such that they would look as transparent as possible and have given everywhere (where it is at all possible) physical interpretations of what is going on, in many cases certain personal efforts and/or use of imagination are still required. The first part of the book is devoted to the physics of spin-glass systems, where the quenched disorder is the dominant factor. The emphasis is made on a general qualitative description of the physical phenomena, being mostly based on the results obtained in the framework of the mean-field theory of spin-glasses with long-range interactions [1]. First, the general problems of the spin-glass state are discussed at the qualitative level [2]. In Chapters 3-5 the ‘magic’ of the replica symmetry breaking (RSB) scheme is explained in detail, and the physics behind it is discussed. This part also contains the detailed derivation of the ultrametric structure of the space of spin-glass states as well as its scaling properties (the reader interested in more details about ultrametricity may refer to the review [3]). Chapter 6 is devoted to the review of particular experimental works on real spin-glass materials, which on a qualitative level confirm the basic theoretical predictions. It should be stressed that in this book I have no intentions of giving ixx Preface any kind of review of the very many different problems that exist in the scope of spin glasses. For that the reader should refer to the many excellent books and reviews that exist [1, 4, 5]. My intentions are much more modest. Here I would mostly like to present in a systematic way the methods and physical ideas of the replica theory, and in particular the powerful technique of the replica symmetry breaking. On the other hand (keeping in mind the pedagogical character of this book), I would also like to demonstrate how this general method works when applied to various (which often look quite different) particular problems of statistical mechanics, such as critical phe- nomena, directed polymers, Griffiths singularities, etc., which are considered in subsequent parts of the book. For those reasons I do not discuss here the exciting current debates about the validity of the considered mean-field approach for finite-dimensional spin glasses [5, 6], as well as the alternative point of view on the spin-glass state, which is called the droplet model [7]. For the same reasons I do not include consideration of the very important and presently actively developing field of dynamical behavior of spin glasses [8, 9]. At the moment these very interesting problems are still too far from being resolved. The second part of the book is mainly devoted to the theory of the critical phenomena at the phase transitions of the second order in the presence of weak quenched disorder. The theory of the critical phenomena deals with macroscopic statistical systems in the close vicinity of the phase-transition point where spontaneous symmetry breaking takes place, and the situation is characterized by large-scale fluctuations. According to the traditional scaling theory the large-scale fluctuations are characterized by a certain dominant scale, the correlation length, R,, which grows as the critical point is approached, where it becomes infinite. The large-scale fluctuations lead to singularities in the macroscopic characteristics of the system as a whole. These singularities are the main subject of the theory. Chapter 7 is devoted to the systematic consideration of the traditional renormalization group (RG) theory of the critical phenomena, including e-expansion. The reader familar with this technique (which is well explained in many books, see e.g. [10]) may easily skip this chapter. Originally, many years ago, it was generally believed that quenched dis- order could either completely destroy the long-range fluctuations, such that the singularities of the thermodynamical functions become smoothed out, or it could produce only a shift of the critical point but could not affect the critical behavior itself. Later it was realized that an intermediate situation is also possible, in which a new critical behavior, with new universal critical exponents, is established sufficiently close to the phase-transition point. InPreface xi terms of the RG approach the standard procedure for obtaining a new universal ‘random’ critical regime for the vector ferromagnetic spin systems is considered in Chapter 8. However, according to recent developments in this field, the effects of the quenched disorder on the critical behavior could appear to be more compli- cated, and in certain cases a completely new type of critical phenomenon of the spin-glass nature could be established in the close vicinity of the critical point. In Chapter 9 the RG theory for the vector ferromagnetic spin systems is generalized to take into account these non-perturbative spin-glass-type phenomena. It is demonstrated that whenever the disorder is relevant for the critical behavior there exist no stable fixed points, and the RG flows lead to the so-called strong coupling regime at a finite spatial scale, The physical consequences of the obtained RG solutions are discussed. In Chapter 10 we consider the critical properties of the two-dimensional disordered Ising model. It is well known that in the critical region the two- dimensional Ising model can be reduced to the free-fermion theory [11]. Here this reduction will be demonstrated in very simple terms by means of the Grassman variables technique (for detailed treatment of this new mathematics see [12-14]). The resulting continuum theory, to which the exact lattice-disordered model is equivalent in the critical region, appears to be simple enough, and its specific heat critical behavior can be found exactly. Here we also consider the results of the recent numerical simulations of the two-dimensional disordered Ising model, and describe the general structure of its phase diagram. In the third part of the book some other types of disordered system are described. Here the reader is taken to the frontier of modern research, because this part is mostly devoted to problems which are not yet solved. In Chapter 11 the Ising spin systems with quenched random fields are considered. This type of disorder is essentially different from that with random interactions, because external magnetic fields break the symmetry with respect to the change of the signs of spins, Despite extensive theoretical and experimental efforts during the past 20 years or so (for reviews see e.g. [S, 15]) very little is understood about even the basic thermodynamic properties of the random field Ising model (in particular the nature of the phase transition in this model is still a mystery). Here it is demonstrated why the statistical systems of this type exhibit many qualitatively different properties compared with those considered before. In a wide variety of physical systems one is interested in the behavior of a fluctuating linear object (with finite line tension) interacting with a quenched random potential. This class of problem is traditionally discussedxii Preface in terms of a directed polymer in random media, and it has been much studied during recent years [16]. Quite naturally the best understanding has been achieved for the simplest one-dimensional case when the displacements of a directed polymer can occur only in one direction. Chapter 12 is devoted to the derivation of the scaling properties of the one-dimensional directed polymers in random potentials with finite correlation radius. Here it is demonstrated that in the low-temperature limit the solution of the problem can be described in terms of the effective one-step replica symmetry breaking ansatz. In Chapter 13, I present a new method to study statistical systems with quenched disorder in the low-temperature limit, the so-called vector replica symmetry breaking ansatz. At first sight in the low-temperature limit the situation must be simplified because the partition function could be analyzed in terms of the saddle-point approximation. However, it is easy to see that generically, even at zero temperature, there still exist sample-to-sample fluctuations. It appears that in a large class of strongly disordered systems, it is necessary to include saddle points of the Hamiltonian that break the replica symmetry in a vector sector, as opposed to the usual matrix sector breaking of spin-glass mean-field theory. First, I demonstrate how this new method works on the examples of some elementary problems which can be solved directly. As for its further application to more difficult problems, the zero-temperature fluctuations of a particle in a random potential are studied, and the scaling exponents of the directed polymers in random media with long-range correlations are rederived using this new method. In the last section of this chapter I consider the problem of the existence of non-analytic (Griffiths-like) contributions to the free energy of a weakly disordered Ising ferromagnet (in a zero external magnetic field) from the point of view of the replica theory. Here it is demonstrated that in the paramagnetic phase (away from the critical point) such contributions appear as a result of non-linear localized (instanton-like) solutions of the mean-field saddle-point equations that are characterized by the vector type of the replica symmetry breaking.1 Introduction 1.1 General principles of the statistical mechanics In the most simple terms the basic statements of the statistical mechanics can be introduced in the following way. Let the microscopic state of a macroscopic system having many degrees of freedom be described by the configurations of N variables {s;}, (i = 1,2,...,N). The basic quantity characterizing the microscopic states is called the energy, H, and it is defined as a function of all the microscopic variables {s;}: H = H(s),82,...,5n) = H[s} The microscopic dynamic behavior of the system is defined by some dynamic differential equations such that, in general, the energy of the system tends to a minimum. Besides, it is assumed that no observable system can be perfectly isolated from the surrounding world, and the effect of the interaction with the surroundings (the thermal bath) is believed to produce the so-called thermal noise in the exact dynamical equations. The thermal (white) noise acts as random and uncorrelated fluctuations, which produce the randomization and the mixing of the exact dynamical trajectories of the system. Let A[s] be some observable quantity. The quantities, which are of interest in statistical mechanics, are the averaged values of the observables. In other words, instead of studying the exact evolution in time of the value A[s(¢)], one introduces the averaged quantity: (4) = fim + [ae atte (uu) which could be formally obtained after the observation during an infinite time period. The fundamental hypothesis of the equilibrium statistical mechanics lies 1es Introduction in the following. It is believed that, owing to the mixing of the dynamic trajectories, after an infinitely long observation time, the system in general ‘visits’ its different microscopic states many times, and therefore the averaged quantity in Eq. (1.1) could be obtained by averaging over the ensemble of the states instead of that over the time: (A) = [ asidsy...dswAlS]P 61 52. --05%) (1.2) Here P{s] is the probability distribution function of the microscopic states of the system. In other words, it is believed that because of the mixing of the dynamic trajectories, instead of solving the exact dynamics, the system could be statistically described in terms of the probabilities of its microscopic states given by the function P[s]. The probability distribution function, whatever it is, must be normalized: ff dsidsy... ds PCs1 50-558) =1 (13) The fundamental quantity of the statistical mechanics that characterizes the probability distribution itself is called the entropy. It is defined as the average of the logarithm of the distribution function: = —(log(P [s])) = = f dsid50...dswP Is) log(P {s]) (1.4) In general, the value of the entropy could tell to what extent the state of the system is ‘ordered’. Consider a simple illustrative example. Let the (discrete) microscopic states of the system be labeled by an index «, and let us assume that the probability distribution is such that only L (among all) states have non-zero and equal probability. Then, owing to the normalization (1.3), the probability of any of these L states must be equal to 1/L. According to the definition of the entropy, one gets: L S=—-3- Pz log P, = log L @ Therefore, the broader the distribution (ic. the larger L) is, the larger the value of the entropy is. On the other hand, the more concentrated the distribution function is, the smaller the value of the entropy is. In the extreme case, when there is only one microscopic state occupied by the system, the entropy is equal to zero. In general, the value of exp(S) could be interpreted as the averaged number of the states occupied by the system with finite probability. Now let us consider what the general form of the probability distribution1.1 General principles of the statistical mechanics 3 function must be. According to the basic hypothesis, the average value of the energy of the system is: E=(H)= 3" PxHz (1.5) The interaction of the system with the thermal bath produces the following fundamental effects. First, the averaged value of its energy in the thermal equilibrium is conserved. Second, for some reason Nature is constructed in such a way that irrespective of the internal structure of the system, the value of the entropy in the equilibrium state tries to attain a maximum (bounded by the condition that the average energy is constant). In a sense, it is natural: random noise makes the system as disordered as possible. Let us now consider the form of the probability distribution function, which would maximize the entropy. To take into account the two constraints— the conservation of the average energy, Eq. (1.5), and the normalization Yq Px = 1~one can use the method of the Lagrange multipliers. Therefore, the following expression must be maximized with respect to all possible distributions P,: Spy[P] = — 2 Palog(P.) — B (= PyHy — e) -Y (= Py 1) (1.6) where f and y are the Lagrange multipliers. Variation with respect to Py gives: Pra 7 ex(—PHe) (17) where Z = exp(—BH2) = exply + 1) (18) is called the partition function, and the parameter f, which is called the inverse temperature, is defined by the condition: a Yo Ae exp(—BHa) = E (19) In practice, however, it is the temperature that is usually taken as an inde- pendent parameter, whereas the average energy is obtained as the function of the temperature by Eq. (1.9). The other fundamental quantity of the statistical mechanics is the free energy defined as follows: F=E-TS (1.10)4 Introduction where T = 1/8 is the temperature. Using Eq.(1.7), one can easily derive the following basic relations among the free energy, the partition function, the entropy and the average energy: F =—T log(Z) (1.11) ae 2 OF S= B55 (1.12) é OF E = —ppglosiZ) =F + Bap (1.13) Note, that according to the definition given by Eq. (1.10), the principle of the maximum of entropy is equivalent to that of the minimum of the free energy. One can easily confirm that taking the free energy (instead of the entropy) as the fundamental quantity which must be minimal with respect to all possible distribution functions, the same form of the probability distribution as given by Eq. (1.7) is obtained. 1.2 The mean-field approximation In magnetic materials the microscopic state of the system is supposed to be defined by the values of the local spin magnetizations. In many magnetic materials the electrons responsible for the magnetic behavior are localized near the atoms of the crystal lattice, and the force that tends to orient the spins is the (short range) exchange interaction. The most popular classical models, which describe this situation qualita- tively, are called the Ising models. The microscopic variables in these systems are the Ising spins oj, which by definition can take only two values: +1 or —1. The traditional form for the microscopic energy (which from now on will be called the Hamiltonian) as the function of all the Ising spins is in the following: H=— XO Jyoia;—h oo; (1.14) i Here the notation indicates the summation over all the lattice sites of the nearest neighbors, Jj; are the values of the spin-spin interactions, and h is the external magnetic field. If all the Js are equal to a positive constant, then one gets the ferromagnetic Ising model, and if all the Jys are equal to a negative constant, then one gets the antiferromagnetic Ising model. In spite of the apparent simplicity of the Ising model, an exact solution (which means the calculation of the partition function and the correlation1.2 The mean-field approximation 5 functions) has been found only for the one- and the two-dimensional systems in the zero external magnetic field. In all other cases one needs to use approximate methods. One of the simplest methods is called the mean-field approximation. In many cases this method gives results that are not too far from the correct ones, and very often it is possible to get some qualitative understanding of what is going on in the system under consideration. The starting point of the mean-field approximation is the assumption about the structure of the probability distribution function. It is assumed that the distribution function in the equilibrium state can be factorized as the product of the independent distribution functions in the lattice sites: Plo] = zexPt- Ho) = T] Fe) (1.15) The normalized site distribution functions take the form: Pi(oi) = 1+ 15¢6,—1) +1 1 F541 (1.16) where ¢; are the parameters that have to be specified. The factorization of the distribution function, Eq. (1.15), means that the average of any product of any functions at different sites is also factorizing on the product of the independent averages: (f(oi)e(o,)) = (F(a) {e(e,)) (1.17) where, according to the ansatz (1.15): goa) = $4 yay + Shey (1.18) In particular, for the average site magnetizations, one easily gets: (oi) = $3 (1.19) Therefore, the physical meaning of the parameters {¢;} in the trial distri- bution function is that they describe the average site spin magnetizations. According to the general principles of the statistical mechanics, these pa- rameters must be such that they would minimize the free energy of the system. Using Eqs. (1.15) and (1.16) for the entropy and for the average energy, one gets: 5 = —(log(P[o])) ~ ~ Liles (Pi(oi))) ~ [toe (4) 42 =H 08 (4 *)| (1.20)6 Introduction E => ibid hd (1.21) 2 Sp For the free a Eq. (1.10), one obtains: F=-5 py Jijdids bY ; a3 [Ee +! Sti ce (* 5*)| (1.22) To be more specific, consider the ferromagnetic system on the D-dimen- sional cubic lattice. In this case all the spin-spin couplings are equal to some positive constant: Jij = (1/2D)J > 0, (the factor 1/2D is inserted just for convenience) and each site has 2D nearest neighbors. Since the system is homogeneous, it is natural to expect that all the ¢js must be equal to some constant ¢. Then, for the free energy (1.22) one gets: 1 1 1 =f) =—j1e ho +T | + tog ( 148) LF og (# =) (1.23) where V is the total number of sites (which is proportional to the volume of the system) and f is the density of the free energy. The necessary condition for the minimum of f is afd) _ roa —J¢—h+T arctanh(¢) = 0 (1.24) The resulting equation, which defines the order parameter ¢ is: o = tanh[B(J¢ + h)) (1.25) Note, that the minimum of the free energy is conditioned by df /d¢? > 0. Using Eq. (1.24), this condition can be reduced to 1 Tog? (1.26) Consider first the case of a zero external magnetic field (h = 0). One can easily see that if T > T. = J, the only solution of the Eq. (1.25) @ = tanh(BJ¢) is ¢ = 0, and this solution satisfies the condition (1.26). Therefore, at all temperatures higher than T, the minimum of the free energy is achieved in the state in which all the site spin magnetizations are zero.1.2 The mean-field approximation 1 i f T>T. > oe ale T T, (t > 0) and at T < T, (t < 0) are shown in Fig. 1.1. Note, that since the total free energy F is proportional to the volume of the system, the value of the free energy barrier separating the states with ¢ = +(T) at T < T, is also proportional to the volume of the system. Therefore, in the thermodynamic limit V — oo (which corresponds to the consideration of the macroscopic systems) the barrier separating the two states is becoming infinite. The simple considerations described above demonstrate on a qualitative level the fundamental phenomenon called spontaneous symmetry breaking. At the temperature T = T;, a phase transition of the second order occurs,8 Introduction such that in the low-temperature region T < T, the symmetry with respect to the global change of the signs of the spins is broken, and two (instead of one) ground states appear. These two states differ by the sign of the average spin magnetization, and they are separated by the macroscopic barrier of the free energy. 1.3 Quenched disorder, selfaveraging and the replica method In this book we will consider the thermodynamical properties of various spin systems that are characterized by the presence of some kind of a quenched disorder. In realistic magnetic materials such disorder can exist, for example, owing to the oscillating nature of the exchange spin-spin interactions combined with the randomness in the positions of the interacting spins (such as in metallic spin-glass alloys AgMn), or owing to defects in the lattice structure, or because of the presence of impurities, etc. Because we will mostly be interested in the qualitative effects produced by the quenched disorder, the details of the realistic structure of such magnetic systems will be left aside. Here we will concentrate on extremely simplified model description of the disordered spin systems. In what follows we will consider two essentially different types of dis- ordered magnet. First, we will study the thermodynamic properties of spin systems in which the disorder is strong. The term ‘strong disorder’ refers to the situation when the disorder appears to be the dominant factor for the ground-state properties of the system, so that it dramatically changes the low-temperature properties of the magnetic system as compared with the usual ferromagnetic phase. These types of system, usually called the spin glasses, will be considered in the first part of the course. In the second part of the course we will consider the properties of weakly disordered magnets. This is the case when the disorder does not produce notable effects for the ground-state properties. It will be shown, however, that in certain cases even a small amount of disorder can produce dramatic effects for the critical properties of the system in close vicinity of the phase transition point. The main problem in dealing with disordered systems is that the disorder here is quenched. Formally, all the results for the observable quantities for a given concrete system must depend on the concrete interaction matrix Jjj, ie. these results are defined by a macroscopic number of random para- meters. Apparently, results of this type are impossible to calculate and, moreover, they are useless. Intuitively it is clear, however, that the quantities1.3 Quenched disorder, selfaveraging and the replica method 9 which are called the observables should depend on some general averaged characteristics of the random interactions. This brings us to the concept of selfaveraging. The traditional speculation about why the selfaveraging phenomenon should be expected to take place, is as follows, The free energy of the system is known to be proportional to the volume V of the system. Therefore, in the thermodynamic limit V — oo the main contribution to the free energy comes from the volume, and not from the boundary, which usually produces effects of the next orders in the small parameter 1/V. Any macroscopic system could be divided into a macroscopic number of macroscopic subsystems. Then the total free energy of the system would consist of the sum of the free energies of the subsystems, plus the contribution that comes from the interactions of the subsystems, at their boundaries. If all the interactions in the system are short range (which takes place in any realistic system), then the contributions from the mutual interactions of the subsystems are just the boundary effects, which vanish in the thermodynamic limit. Therefore, the total free energy could be represented as a sum of the macroscopic number of terms. Each of these terms is a random quenched quantity because it contains, as parameters, the elements of the random spin-spin interaction matrix. In accordance with the law of large numbers, the sum of many random quantities can be represented as their average value, obtained from their statistical distribution, multiplied by their number (all this is true, of course, only under certain requirements on the characteristics of the statistical distribution). Therefore, the total free energy of a macroscopic system must be selfaveraging over the realizations of the random interactions in accordance with their statistical distribution. The free energy is known to be given by the logarithm of the partition function. Thus, in order to calculate the observable thermodynamics, one has to average the logarithm of the partition function over the given distribution of random Jijs after the calculation of the partition function itself. To perform such a program the following technical trick, which is called the replica method, is used. Formally, the replicas are introduced as follows. In order to obtain the physical (selfaveraging) free energy of the quenched random system we have to average the logarithm of the partition function: FeFj= 5 in(Zs) (1.29) where (...) denotes the averaging over random interactions {Jj;} with a given10 Introduction distribution function P [J]: Ws ( Il fax) PUM.) (1.30) and the partition function is Zs =D exp(-BHU,o)) (1.31) To perform this averaging procedure, the following trick is invented. Let us consider the integer power n of the partition function (1.31). This quantity is the partition function of the n non-interacting identical replicas of the original system (ic. having identical fixed spin-spin couplings Jij): Zj= (i x) exp (- > HU, aa) (1.32) wai “a ost Here the subscript a labels the replicas. Let us introduce the quantity: 1 F,= —pylntZn) (1.33) where Z, =Zj (1.34) Now, if a formal limit n + 0 is taken in the expression (1.33), then the original expression for the physical free energy (1.29) will be recovered: lim Fy = ~ tim Fe IntZ) . tin Ftn [exp@nZ)] = Jingi (1.35) Thus, the scheme of the replica method can be described in the following steps. First, the quantity F, for the integer n must be calculated. Second, the analytic continuation of the obtained function of the parameter n should be made for an arbitrary non-integer n. Finally, the limit n + 0 has to be taken. Although this procedure may look rather doubtful at first, actually it is not so crazy! First, if the free energy appears to be an analytic function of the temperature and the other parameters (so that it can be represented as the series in powers of B), then the replica method can be easily proved to be correct in a strict sense. Second, in all cases, when the calculations can be performed by some other method, the results of the replica method are confirmed. One could also introduce replicas another way [2, 29, 30]. Let us consider a general spin system described by a Hamiltonian H[J;c], which depends1.3 Quenched disorder, selfaveraging and the replica method ul on the spin variables {o;} and the spin-spin interactions Jj; (the concrete form of the Hamiltonian is irrelevant). If the interactions Jj; are quenched, the free energy of the system depends on the concrete realization of the Jijs: i B Now, let us assume that the spin-spin interactions are partially annealed (ie. not perfectly quenched), so that they can also change their values, but the characteristic time scale of their changes is much larger than the time scale at which the spin degrees of freedom reach the thermal equilibrium. In this case the free energy given by (1.36) would still make sense, and it would become the energy function (the Hamiltonian) for the degrees of freedom of Jiys. “Besides, the space in which the interactions Jj; take their values should be specified separately. The interactions Ji; could be discrete variables taking values +Jo, or they could be continuous variables taking values in some restricted interval, or they could be something else. In the quenched case this space of Ji; values is defined by a statistical distribution function P [J]. In the case of partial annealing, this function P[J] has a meaning of the internal potential for the interactions Jjj, which restricts the space of their values. Let us now assume that the spin and the interaction degrees of freedom are not thermally equilibrated, so that the degrees of freedom of the interactions have their own temperature T’, which is different from the temperature T of the spin degrees of freedom. In this case for the total partition function of the system one gets: g= i DsPUlexp(—B'F (UI) FU) = log(Z,) (1.36) - [vrPtnexo§ 1092 = [orien =e (137) where n = T/T’. Correspondingly, the total free energy of the system would be: F =—T' log [z ') (1.38) In this way we arrive at the replica formalism again, in which the ‘number of replicas’ n = T/T’ appears to be the finite parameter. To obtain the physical (selfaveraging) free energy in the case of the quenched random Jjjs one takes the limit n — 0. From the point of view12 Introduction of partial annealing, this situation corresponds to the limit of the infinite temperature T’ in the system of Jjjs. This is natural in a sense that in this case the thermodynamics of the spin degrees of freedom produces no effect on the distribution of the spin-spin interactions. In the case when the spin and the interaction degrees of freedom are thermally equilibrated, T’ = T (n = 1), we arrive at the trivial case of the purely annealed disorder, irrespective of the difference between the characteristic time scales of the interactions and the spins. This is also natural because the thermodynamic description formally corresponds to the infinite times, and the characteristic time scales of the dynamics of the internal degrees of freedom become irrelevant. If n # 0 and n ¥ 1, one gets the situation that could be called partial annealing, and which is the intermediate case between quenched disorder and annealed disorder.Part one Spin-glass systems2 Physics of the spin-glass state Before starting doing detailed calculations, first it would be useful to get a qualitative understanding of the general physical phenomena taking place in statistical mechanics of spin systems with strong quenched disorder. There- fore, in this chapter we will discuss the problem of spin-glass state only in simple qualitative terms. 2.1 Frustrations There are quite a few statistical models of spin glasses. Here we will concen- trate on one of the simplest models, which can be formulated in terms of the classical Ising spins, described by the following Hamiltonian: a. Lo VL dyoio; (2.1) This system consists of N Ising spins {g;} (i = 1,2,...,N), taking values +1 which are placed in the vertices of some lattice. The spin-spin interactions Ji are random in their values and signs. The properties of such a system are defined by the statistical distribution function P(Jij] of the spin-spin interactions. For the moment, however, the concrete form of this distribution will not be important. The motivation for the Hamiltonian (2.1) from the point of view of realistic spin-glass systems is well described [4]. The crucial phenomena revealed by a strong quenched disorder, which make such systems so hard to study, are as follows. Consider the system of three interacting spins (Fig. 2.1). Let us assume for simplicity that the interactions among them can be different only in their signs, being equal in the absolute value. Then for the ground state of such a system we can find two essentially different situations. If all three interactions Jj2,J23 and J)3 are positive, or two of them are 1516 Physics of the spin-glass state Fig. 2.1. The frustrations in the system of three spins. (a) No frustration: the product of the interactions along the triangle is positive. (b) The frustated triangle: the product of the interactions along the triangle is negative. negative while the third one is positive, then the ground state of this three- spin system is unique (except for the global change of signs of all the spins, Fig. 2.1(a)). This is the case when the product of the interactions along the triangle is positive. However, if the product of the interactions along the triangle is negative (one of the interactions is negative, or all three interactions are negative, Fig. 2.1(b)), then the ground state of such a system is degenerate. One can easily check, going from spin to spin along the triangle, that in this case the orientation (‘plus’ or ‘minus’) of one of the spins remains ‘unsatisfied’ with respect to the interactions with its neighbors. One can also easily check that a similar phenomenon takes place in any closed spin chain of arbitrary length, provided that the product of the spin— spin interactions along the chain is negative. This phenomenon is called Srustrationt (17). One can easily see that not any disorder induces frustrations. On the other hand, it is the frustrations that describe the relevant part of the disorder, and which essentially affect the ground-state properties of the system. In other words, if the disorder does not produce frustrations, it can be considered as being irrelevant. In some cases an irrelevant disorder can just be removed by a proper redefinition of the spin variables of the system. A simple example of this situation is illustrated by the so-called Mattice magnet. This also is a formally disordered spin system, which is described by the Hamiltonian (2.1), where the spin-spin interactions are defined as Ji; = fj, and the quenched és take values +1 with equal probability. In such a system the + This term is quite adequate in its literal meaning, as the triangle discussed above might as well be interpreted as the famous love triangle. Besides, the existence of frustrations in spin glasses removes any hope for finding a simple solution of the problem.2.2 Ergodicity breaking 17 interactions Ji are also random in signs, although one can easily check that no frustrations appear here. Moreover, after simple redefinition of the spin variables, oj > o;¢;, an ordinary ferromagnetic Ising model will be recovered. Thus, this type of disorder (called the Mattice disorder) is actually fictitious for the thermodynamic properties of the system. It is crucial that the ‘true’ disorder with frustrations can not be removed by any transformation of the spin variables. Because in a macroscopic spin system, in general, one can draw a lot of different frustrated closed spin chains, the total number of frustrations must also be macroscopically large. This, in turn, would result either in a tremendous degeneracy of the ground state, or, in general, it could produce a lot of low-lying states with the energies very close to the ground state. 2.2. Ergodicity breaking Formally, according to the general selfaveraging arguments (Section 1.3), to derive the observable thermodynamics of a disordered spin system one has to find a way for averaging the logarithm of the partition function over random parameters Ji; simultaneously with the calculation of the partition function itself. It is clear that this problem is not easy, but nevertheless it looks just like a technical problem (well, presumably a very hard one), and not more than that. Actually, for spin-glass type systems this is not just a technical problem. To realize this, let us consider again a few general points of statistical mechanics. Everything would be rather simple if the free energy in the thermodynamic limit was an analytic function of the temperature and the other parameters. Actually, for most of non-trivial systems that are of interest in statistical mechanics, this is not so. Very often, owing to spontaneous breaking of some kind of a symmetry in the thermodynamic limit, there exists a phase transition, and this makes the free energy a non-analytic function of the parameters involved. Let us consider again the ordinary ferromagnetic Ising model (Chapter 1), which in very simple terms illustrates the physical consequences of this phenomenon. Because the Hamiltonian of this system is invariant with respect to the global change of the signs of all the spins, any thermodynamic quantity that is odd in spins must be identically equal to zero. In particular, this must be true for the quantity that describes the global magnetization of the system. If the volume N of the system is finite, these arguments are indeed perfectly correct. However, in the thermodynamic limit N — 00 we are facing a rather non-trivial situation. According to simple calculations performed in Section 1.2, the free energy as a function of the global magnetization18 Physics of the spin-glass state acquires the double-well shape (Fig. 1.1) at low temperatures, The value of the energy barrier separating the two ground states is proportional to the volume of the system, and it is getting infinite in the limit N > oo, In other words, at temperatures below T, the space of all microscopic states of the system is divided into two equal valleys separated by the infinite barrier. On the other hand, according to the fundamental ergodic hypothesis of statistical mechanics (Section 1.1), it is assumed that in the limit of infinite observation time the system (following its internal dynamics) visits all its microscopic states many times, and it is this assumption that makes it possible to apply the statistical mechanical approach: for the calculation of the averaged quantities we use averaging over the ensemble of states with the corresponding probability distribution instead of that over time. In the situation under consideration, when the thermodynamic limit N — oo is taken before the observation time goes to infinity (it is this order of limits that corresponds to the adequate statistical mechanical description of a macroscopic system), the above ergodic assumption simply does not work. Whatever the (reasonable) internal dynamics of the system, it could never make it possible to jump over the infinite energy barrier separating the two valleys of the space of states. Thus, in the observable thermodynamics, only half of the states contribute (these are the states that are on one side from the barrier), and that is why in the observable thermodynamics the global magnetization of the system appears to be non-zero. In the terminology of statistical mechanics, this phenomenon is called. ergodicity breaking, and it manifests itself as spontaneous symmetry breaking: below Tz the observable thermodynamics becomes non-symmetric with re- spect to the global change of signs of all the spins. As a consequence, in the calculations of the partition function below 7, one has to take into account not all, but only one half of all the microscopic states of the system (the states that belong to one valley). The above example of the ferromagnetic system is very simple because here one can easily guess right away what kind of symmetry could be broken at low temperatures. In spin glasses, spontaneous symmetry breaking also takes place. However, unlike the ferromagnetic system, here it is much more difficult to guess which one. The main problem is that the symmetry which might be broken in a given sample can depend on the quenched disorder parameters involved. In this situation the calculation of the observable free energy becomes an extremely difficult problem, because now one must take into account only the states belonging to one of the many valleys, while the structure of these valleys depends on a concrete realization of the random disorder parameters.23 Continuous sequence of phase transitions 19 2.3 Continuous sequence of phase transitions Of course, the existence of many local minimum states in the frustrated spin system does not automatically means that at low temperatures these states create their valleys separated by the infinite barriers of the free energy. Because of thermal fluctuations (which are usually rather strong in the low- dimensional systems) the energy barriers could effectively ‘melt’, and in this case the ground state of the free energy could appear to be unique. Then there will be no spontaneous symmetry breaking, and at any finite temper- ature the system will be in the ‘symmetric’ paramagnetic state. Of course, from the point of view of the anomalously slow dynamic relaxation prop- erties this state can be essentially different from the usual high-temperature paramagnetic state, but this problem would lead us well beyond the scope of pure statistical mechanics. It could also happen that because of some symmetry properties the global minimum of the free energy of a given sample is achieved at low tempera- tures at some unique non-trivial spin configuration (of course, in this case the ‘counterpart’ spin state that differs by the global change of the signs of the spins must also be the ground state). It would mean that at low enough temperatures (below a certain phase-transition temperature T,), the system must ‘freeze’ in this unique random spin state, which will be characterized by the non-zero values of the thermally averaged local spin magnetizations at each site (o:). As this ground state is random, the values of the local magnetizations (o;) will fluctuate in their values and signs from site to site, so that the usual ferromagnetic order parameter, which describes the global magnetization of the system, m = 1/N >),(o;), must be zero (in the infinite volume limit). However, this state can be characterized by the other ther- modynamic order parameter (usually called the Edwards-Anderson order parameter [18]): 7 Lev #0 (2.2) The properties of systems of this type are studied in detail in the papers by Fisher and Huse [7], and we will not consider them here. In the subsequent chapters we will concentrate on a qualitatively different situation, which arises when there exist macroscopically large numbers of states in which the system could get ‘frozen’ at low temperatures. Moreover, unlike ‘ordinary’ statistical mechanical systems, according to the mean-field theory of spin glasses the spontaneous symmetry breaking in the spin-glass state takes place not just at a certain T., but it occurs at any temperature below T;. In other words, below T, a continuous sequence of the phase20 Physics of the spin-glass state Space of states Fig. 2.2. The qualitative structure of the spin glass free-energy landscape at different temperatures. transitions takes place and, correspondingly, the free energy appears to be non-analytic at any temperature below T.. In general qualitative terms this phenomenon can be described as follows. Just below a certain critical temperature Te, the space of spin states is divided into many valleys (their number diverges in the thermodynamic limit), separated by infinite barriers of the free energy. At the temperature T = T, —6T each valley is characterized by the non-zero values of the average local spin magnetizations (g;)(a) (which, of course, fluctuate in sign and magnitude from site to site). Here (...), denotes the thermal average inside a particular valley «. The order parameter, which describes the degree of freezing of the system inside the valleys, could be defined as follows: 4(T) = ¥ Lovie (23) According to the mean-field theory of spin glasses, the value of q depends only on the temperature, and it appears to be the same for all the valleys. At T > T, q(T) +0. At a further decrease of the temperature, new phase transitions of er- godicity breaking take place, so that each valley splits into many new ones separated by infinite barriers of free energy (Fig. 2.2). The state of the system in all new valleys can again be characterized by the order parameter (2.3), and its value grows as the temperature decreases. As the temperature goes down to zero, this process of fragmentation of the space of states into smaller and smaller valleys goes on continuously. In a2.4 Order parameter 21 sense, it means that at any temperature below T, the system is in the critical state. To what extent this situation is realistic from the experimental point of view remains open, although the series of recent experiments (which will be discussed in Chapter 9) gives a strong indication in favor of it. In any case, this new type of physics is very interesting in itself, and it is worthy of study. 2.4 Order parameter It is clear that the order parameter (2.3) defined for one valley only does not contain any information about the other valleys, and it does not tell us anything about the structure of space of the ground states as a whole. Let us try to construct the other physical order parameter, which would describe this structure as fully as possible. Consider the following series of imaginary experiments. Let us fix an arbi- trary spin state, and then at a given temperature T below T, let the system relax to thermal equilibrium. For each experiment a new starting random spin state should be taken. Then each experiment will be characterized by some equilibrium values of the average local spin magnetizations (oi)(a), where « denotes the number of the experiment. Since there exists a macro- scopically large number of valleys in the phase space in which the system could get ‘trapped’, these site magnetizations, in general, could be different for different experiments. Let us assume that we have performed an infinite number of such exper- iments. Then, we can introduce the quantity that would describe to what extent the states which have been obtained in different experiments are close to each other: N op = 4 Lode (aig (24) It is clear that |q,g| < 1, and the maximum value of qag is achieved when the two states in the experiments « and f coincide (in this case, the overlap (2.4) coincides with that of (2.3), which has been introduced for one valley only). It is also clear that the less correlated the two different states are, the smaller the value of the overlap (2.4) they have. If the two states are not correlated at all, then their overlap (in the thermodynamic limit) is equal to zero. In a sense the overlap qug defines a kind of a metric in the space of States (the quantity Gp could be conditionally called the ‘distance’ in the space of states). To describe the statistical properties of these overlaps one can introduce22 Physics of the spin-glass state the following probability distribution function: P(g) = > 5(4ap — 9) (2.5) ob It appears that it is in terms of this distribution function P(g) that the spin-glass state looks essentially different from any other ‘ordinary’ thermodynamic state. Possible types of the function P(q) are shown in Fig. 2.3. The paramagnetic phase is characterized by the only global minimum of the free energy, in which all the site magnetizations are equal to zero. Therefore, the distribution function P(q) in this phase is the 6-function at q = 0 (Fig. 2.3(a)). In the ferromagnetic phase there exist two minima of the free energy with the site magnetizations +m. Thus, the distribution function P(q) in this phase must contain two 5-peaks at q = +m? (Fig 2.3(b)). It is clear that in the case of the ‘fake’ spin-glass phase in which there exist only two disordered global minima (the states that differ by the global reversal of the local spin magnetizations) the distribution function P(q) must look the same as in the ferromagnetic state. According to the mean-field theory of spin glasses, which will be considered in the subsequent chapters, the distribution function P(q) in the ‘true’ spin- glass phase looks essentially different (Fig. 2.3(c)). Here, between the two 6-peaks at q = +4max(T) there is a continuous curve. The value of gmax(T) is equal to the maximum possible overlap of the two ground states, which is the ‘selfoverlap’ (2.3). Because the number of valleys in the system is macroscopically large and their selfoverlaps are all equal, the function P(q) has two 6-peaks at g = +qmax(T). The existence of the continuous curve in the interval (0, tgmax(T)) is the direct consequence of the ‘origin’ of the spin states involved: because they appear as the result of a continuous process of fragmentation of the valleys into the smaller and smaller ones, the states that form such a type of hierarchy are necessarily correlated. Thus, it is the distribution function P(q) that can be considered as the proper physical order parameter, adequately describing the peculiarities of the spin-glass phase. Although the procedure of its definition described above looks somewhat artificial, it will be shown later that the distribution function P(q) can be defined as the thermodynamical quantity and, moreover, in terms of the mean-field theory of spin glasses it can be calculated explicitly. 2.5 Ultrametricity According to the qualitative picture described above, the spin-glass states are organized in a kind of a hierarchical structure (Fig. 2.2). It can be proved2.5 Ultrametricity 2B PU} (a) o q #14) (b) -m2 m2 9 a) (co) “Fox Fmax, Fig. 2.3. The probability distribution function P(q): (a) in the paramagnetic phase; (b) in the ferromagnetic phase; (c) in the spin-glass phase. that this rather sophisticated space of states could be described in terms of well-defined thermodynamical quantities. In the previous section we introduced the distribution function P(q), which gives the probability of finding two spin-glass states having an overlap equal to q. Now let us introduce a somewhat more complicated distribution function P(q1.q2, 43), which gives the probability for three arbitrary spin-glass states to have their overlaps equal to qi,q2 and q3: P(q1,42,43) = (dap — 41)5(4xy — 42)6(4py — 93) (2.6) By In terms of the mean-field theory this function can also be calculated explicitly (Chapter 5). It can be shown that the function P(q1,q2,q3) is not equal to zero only if at least two of the three overlaps are equal to each other and their value is not larger than the third one. In other words, the function P(q1,42,43) is non-zero only in the following three cases: q, = q2 < 43;24 Physics of the spin-glass state T>T, Ty+ AT an ali NLL A La ~o SLeth To Fig. 2.4. The hierarchical tree of the spin-glass states. qi = 43 S 92; 43 = 42 < 1. In all other cases the function P(qi,q2,q3) is identically equal to zero. It means that in the space of spin-glass states there exist no triangles with all three sides being different. The spaces having the above metric property are called ultrametric. Ultrametricity from the point of view of physics (in mathematics, ultrametric structures have been known since the end of the nineteenth century) is described in detail in the review by Rammal et al. [3]. The simplest illustration of the ultrametric structure can be made in terms of the hierarchical tree (Fig. 2.4). Here the space of the spin-glass states is identified with the set of the endpoints of the tree. The metric in this space is defined in such a way that the overlap (the distance) between any two states depends only on the number of generations to their closest ‘ancestor’ on the tree (as the number of the generations increases, the value of the overlap decreases). One can easily check that the space with such metrics is ultrametric. In the mean-field theory of spin glasses, such an illustrative tree of states actually describes the hierarchical fragmentation of the space of the spin- glass states into the valleys, as has been described above (Fig. 2.2). If for the vertical axis in Fig. 2.4 we assign the (discrete) value of the overlaps q, then the set of spin-glass states at any given temperature T < T, can2.5 Ultrametricity 25 be obtained at the cross-section of the tree at the level q = qmax(T). After decreasing the temperature to a new value T’ < T, each of the states at the level qmax(T) gives birth to numerous ‘descendants’, which are the endpoints of the tree at the new level gmax(T’) > gmax(T). Correspondingly, after increasing the temperature to a higher value T” > T, all the states having their common ancestors at the level qmax(T”) < qmax(T) merge together into one state. As T > Te, qmax(T) — 0, which is the level of the (paramagnetic) “‘grandancestor’ of all the spin-glass states. Because the function gmax(T) is determined by the temperature, it means that it is the temperature that defines the level of the tree at which the ‘horizontal’ cross-section should be made, and this, in turn, reveals all the spin-glass states at this temperature. All the states that are below this level are ‘indistinguishable’, while all the states that are above this level form the ‘evolution history’ of the spin-glass states at a given temperature. In this sense the temperature defines the elementary (‘ultraviolet’) scale in the space of the spin-glass states. This creates a kind of scaling in the spin-glass phase: by changing the temperature one just changes the scale in the space of the spin-glass states.3 Mean-field theory of spin glasses 3.1 Infinite range interaction model The Sherrington and Kirkpatric (SK) model of spin glasses [19] is defined by the usual Ising spins Hamiltonian: N H =O Jyoi0; GB. i Soyo syofes — 54) G5) (here and in what follows irrelevant pre-exponential factors are omitted). Integration over the Jijs gives: 2~ So] fe ¥ (Sherer) ] (3.6) i T, = 1 the only solution of this equation is g = 0. On the other hand, at T < T, there exists the non-trivial solution q(T) # 0. In the vicinity of the critical temperature, at (1— T) =t < 1, this solution can be found explicitly: g(t) = t + O(r”). It is also easy to check that in the low-temperature limit T — 0, q(T) > 1. According to Eqs. (3.16) and (3.9), the obtained solution for q(T) gives us the physical order parameter: 140 A(T) = % Lila) (3.23) As q(T) is not equal to zero in the low-temperature region, T < T,, the3.3 Replica symmetry breaking 31 spins of the system are getting frozen in some random state. Besides, as only one solution exists for g(T), such a disordered ground state must be unique. Using the result obtained for the free energy, one can easily perform further straightforward calculations to find all the other observable thermodynamical quantities, such as specific heat, susceptibility, entropy etc. Thus, in terms of the considered replica symmetric ansatz, a complete solution of the problem can easily be obtained. All that would be very nice, if it were correct. Unfortunately, it is not. One of the simplest ways to see that there is something fundamentally wrong in the obtained solution is to calculate the entropy. One can easily check that at sufficiently low temperatures the entropy is becoming negative! (At T = 0 the entropy S = —1/2n ~ —0.17.) Moreover, the calculations of the Hessian 5°f/5Q? for the obtained RS solution (see Section 3.6) demonstrate that this solution becomes unstable (det(5”f/5Q) < 0) in the entire low-temperature region T < T, [21]. It means that the true solution must be somewhere beyond the replica symmetric subspace. 3.3 Replica symmetry breaking Because the RS solution appears to be unsatisfactory, we should try for the matrix @ with some other structure that contains more parameters. Within this new subspace we have to calculate the extremum of the replica free energy f[Q]. After that, to check the stability of the obtained solution, we have to calculate the corresponding Hessian 5f /5Q? Actually, the situation appears to be much more sophisticated because {as we will see later) no ansatz that contains a finite number of parameters can provide a stable solution. Nevertheless, trying with different structures of Q, and calculating the eigenvalues of the Hessian, one at least would be able to judge which ansatz could be better (that is to say, which is less unstable). Such a procedure could point the correct ‘direction’ in the space of the matrices Q towards the true solution. The strategy of finding the true solution for the replica matrix @ in the limit n — 0 is called the Parisi replica symmetry breaking (RSB) scheme [1, 22]. This is the infinite sequence of the ansatzs that approximate the true solution better and better. Eventually, the true solution can be formulated in terms of the continuous function, which is defined as the limit of the infinite sequence. Moreover, in this limit one is able to prove the stability of the obtained solution. Consider now, step by step, the way in which the solution is approximated.32 Mean-field theory of spin glasses Fig. 3.1. The structure of the matrix Qap at the one-step replica symmetry breaking. 3.3.1 One-step RSB At the first step, which is called the one-step RSB, it is ‘natural’ to divide n replicas into n/m groups, each containing m replicas (at this stage it is assumed that both m and n/m are integers). Then, the trial matrix Q is defined as follows: Qus = q1, if the replicas a and b belong to the same group, and Qgp = qo, if the replicas a and b belong to different groups (the diagonal elements are equal to zero). In the compact form such a structure could be represented as follows: w(3-1(2 v(er() where I(x) is the integer valued function, which is equal to the smallest integer larger than or equal to x. The qualitative structure of this matrix is shown in Fig, 3.1. In the framework of the one-step RSB we have three parameters: q1,q2 and m, and these parameters have to be defined from the corresponding saddle-point equations. Using the explicit form of the matrix é for the replica free energy (3.13), one gets: Qar = (3.24) ot Lee * F10) = GB + 958 Sin ~ 7, 8 (z(d1) (3.25) where Z[01 = exp (w yD ones) (3.26) a 0 in Eqs. (3.29) (here the procedure of taking limit n + 0 is similar to that described in Eq. (3.20) and (3.30) for the free energy (Eq. (3.25)) one obtains: Flargom)=—48 [1 +ma3-+(1— ma? 201] -+5 f ex ae 41> 40,m) = a 70. ai G1 mB Vinao ‘P 240 dy ? ; xin if ree (- ery a ) (2cosh p(z + y)) | (3.31)34 Mean-field theory of spin glasses One can easily check that in the extreme cases m = 0 and m = 1 the replica symmetric solution is recovered with g = qo and q = q; correspondingly. It should be noted that in the framework of the RSB formalism one has to look for the maximum and not the minimum of the free energy. The formal reason is that in the limit n + 0 the number of the components of the order parameter @ becomes negative. For example, in the case of the one-step RSB, each line of the matrix Q contains (m— 1) < 0 components that are equal to q), and (n —m) > —m <0 components which are equal to qo. This phenomenon can also be easily demonstrated for the case when the replica free energy (3.25) contains only the trivial term (8/2n) 7,2, @,: lim, [ = 03 = nO = 78 [1 — mat + mad] (32) Apparently, the ‘correct extremum’ of this free energy (in which the Hessian is positive) for 0 < m < 1 is the maximum and not the minimum with respect to qo and q). To derive the saddle-point equations for the parameters qo, qi and m one just has to take the corresponding derivatives of the free energy (3.31). The calculations are straightforward, but as the resulting equations are rather cumbersome, we omit this simple exercise. The results of the numerical solution of these saddle-point equations are as follows. (1) At T < T, = 1 the function f[q1,qo,m] indeed has the maximum at the non-trivial point: 0 < m(T) < 1; 0 < qgo(T) < 1:0 0 has to be taken. Until the parameter n is an integer, according to the36 Mean-field theory of spin glasses Fig. 3.4. The explicit form of the matrix Qu» for the two-step RSB. above definition of the n x n matrix @ the parameters {m,} must satisfy the inequalities 1 < mj, < mj; oo for an arbitrary value of the parameter n < 1 is parametrized by its diagonal element g and the off-diagonal function q(x) (n < x < 1): 0 = (9, q(x). Then for the linear invariants TrQ and 574, Qab one can easily prove: TrO = ng (3.38) and n k 1 Jim 3° Qan= n+ jim [ dtm — mo =ng—n [ dsa(x) 639) Similarly to the above equation one gets: i: i fim 50 Oly = na! —n fo dxal(x) (3.40) ab where the power / can be arbitrary. Now let A and B be two Parisi matrices parametrized respectively by (G,a(x)) and (b, B(x). Then for an arbitrary finite n for the Hadamard product (Q- P),.» = QabPab one easily proves: A+B (ab, a(x)b(x)) (3.41) Let us denote the parametrization of the matrix product of the two matrices as follows: AB — (@,c(x)). Then after somewhat painful algebra one can prove that @ = ab— (ab) (x) = —na(x)b(x) + [& — (a)]b(x) + [5 — (b) a(x) 7 f etacs) — a(y)] [b(x) — d(y)] (3.42) where we have introduced the notation: (a) = [ ” aa(x) (3.43) For the eigenvalues of a Parisi matrix Q and their multiplicities one finds: dg = — (a); with multiplicity 1 1 3x) = B= xalx) — [ dyq(y); with multiplicity - Fax (3.44) : where x € [n,1]. The above algorithms are sufficient to operate quite easily with the Parisi matrices in the continuum RSB representation.3.5 RSB solution near T, 39 3.5 RSB solution near T, Near the critical temperature T, = 1, the solution for the saddle-point function q(x) can be obtained analytically. In the vicinity of the phase transition point the order parameter q(x) should be expected to be small in t = (T.— T)/T. < 1, and consequently one can expand the replica free energy (3.25)-(3.26) in powers of the matrix Qap. This calculation is straightforward, and the result of the expansion up to the fourth order is as follows: 1) fig | pe OP — ETHOP — FOG EO grea (3.45) Here in all the terms but the first one we have substituted T = 1. Detailed study of the stability of the replica symmetric solution shows that it is the term S7qy Q%, that makes the RS solution unstable below T., and it is this term which is responsible for the replica symmetry breaking [20]. This indicates that for the RSB solution near T;, the last two terms of the fourth order in (3.45) must be of higher orders in t than all the other terms. Thus, the easiest way to obtain the solution is by first neglecting these last two terms; and then by using the explicit form of the obtained solution for q(x) one can easily prove a posteriori that these neglected terms are indeed of higher orders in t. Using the rules for the Parisi matrices in the continuum RSB representation described in the previous section one can easily get the explicit expression for the free energy as the functional of q(x). In particular, using Eq. (3.42) for the second term in Eq. (3.45) after simple algebra in the limit n + 0 one gets: 1 - 1 x tim THQ = [lax [sae +3000 [ayer] 8.46) The first and the third terms in Eq, (3.45) can be expressed using Eq. (3.40) (in our case G = 0). For the free energy one finally obtains: 1 x Hava = 5 [a [sae — Fae) — a0 [advair + zatco] G47) Variation of this expression with respect to the function g(x) yields the following saddle-point equation: : eats) — x47) — 2000 [ dvay)— fava + Za =0 (348) The solution of this equation is simple. Taking the derivative of Eq. (3.48)40 Mean-field theory of spin glasses over x one gets: 1 a(x) [> — 2xq(x) —2 i dyq(y) + 29] (3.49) ‘ This equation results in the following: 1 2t — 2xq(x) — af dyq(y) +247(x) =0 (3.50) : or qx) =0 (3.51) The last equation means that q(x) = const, and it corresponds to the replica symmetric solution that has been already studied. Consider Eq. (3.50). Taking the derivative over x again, one gets: 1 q(x) = 3% (3.52) The above simple analysis allows us to build an ansatz for a general form of the solution of the original saddle-point equation (3.48): qo, OSx< x0 1 Ax)= 9 3% XOSx Sx) (3.53) gM u h,(T) the replica symmetric solution is stable, can be obtained for the whole range of temperatures and magnetic fields [21]. This line is usually called the de Almeida-Thouless (AT) line. Let us derive the criteria for the stability of the replica symmetric solution of the SK model in a non-zero external magnetic field. The general expression for the replica partition function of the SK model in the external field h is (see Section 3.2): = [ Ddexp (pn g101) (59) where f[Q] is the _ free energy: s10) = 7 +h FO -> 1 tog ji%8[Doe (ry Ye urcaon + mSe.)| xb (3.60) Here Qqp is the symmetric n x n replica matrix with zero diagonal elements. The replica symmetric solution of this model is defined by the condition: Qc» = q for all a + b, where the value of q(T,h) is given by the equation (see Eq. (3.22): q= ji = hae + Bh) exp (-3#) (3.61)42 Mean-field theory of spin glasses To check whether the RS solution defines the correct extremum of the free energy (3.60) we have to consider small deviations from the RS form of the matrix Quan: Qab = 4 + Nab (3.62) where “ap is also a symmetric nxn replica matrix with zero diagonal elements. The expansion of the free energy i up to the second order in yap is: £149] = frs(q) — ZS Geabyca)Nab Mea (3.63) T (ab)acd) where f;.(q) is the replica symmetric free energy, and Geabyed) = S(abyed) — B° [(oaesaeoa) — (706) oe04)] (3.64) The RS solution will be stable only if the quadratic form in Eq. (3.63) is positively defined. The matrix G has three different types of matrix element: Gabyab) = 1 — BL — {ou95)") = A Gearyacr = —B [(one) — (aon)?] = B, (b #0) Grabyeay = —B? [(o0010-04) — (oa0)"] =C. (ab) # (ed) (3.65) Here (ayo) = and (oaonoeoa) = 1 = [°4 ae ahh ae + Bh) exp (-#) (3.66) Because the matrix G is real symmetric and its order is 4n(n—1), the number of linearly independent eigenvectors must also be equal to 4n(n — 1). The complete set of these eigenvectors can be found for general values of n, and then the limit n — 0 can be taken. Let us consider the corresponding eigenvalue equation: Gn =in (3.67) One can easily check that the complete set of eigenvectors and eigenvalues defined by this equation can be classified in terms of three different types of symmetry. (1) The eigenvectors 9!) that are symmetric under interchange of all indices: ny = (3.68)(2 (3) 3.6 de Almeida-Thouless line 43 for all a # b. Substituting this into the eigenvalue equation (3.67) and taking into account Eq. (3.65), one gets: 1 A+2n—2B + 5(n—2)(n—3)C—aln =O (3.69) This equation yields one non-degenerate eigenvalue: =A+2(n—2)B+ Fn —2)(n—3)C (3.70) The eigenvectors 9) that are symmetric under interchange of all but one of the indices (this fixed index could be taken e.g. number 1): a? = ns foraorb=1 2 = n?; for a,b #1 (3.71) To ensure orthogonality with the eigenvector (3.68) one should impose the condition: Yay 72) = 0, which gives: 9 = (1—4n)y?, There exist (n — 1) such vectors. Substituting (3.71) into the eigenvalue equation (3.67) one obtains: [A+ (n—4)B —(n—3)C — A] n? =0 (3.72) which yields the following eigenvalue: =A+(n—4)B—(n—3)C (3.73) with degeneracy (n — 1). The eigenvectors 7°) that are symmetric under interchange of all but two of the indices (these could be taken to be e.g. number 1 and number 2): (3) 3} ny = ns nf) = =the = =n; fora#1,2 i= =n); for a,b # 1,2 (3.74) Orthogonality with the Been eigenvectors, (3.68) and (3.71), im- poses the conditions: D4, nf in?) = = 0 and Say 12) = 0, which yields: yO) = —nn?); 7" ny) = 5—my (3.75) One can easily see that the number of such vectors is 4n(n — 3).Mean-field theory of spin glasses Substituting (3.74) and (3.75) into the eigenvalue equation (3.67) one obtains the following eigenvalue: =A-2B+C (3.76) which has the degeneracy jn(n — 3). In the limit n > 0 one gets: Ay = 41 =A —4B +3 jy =A-2B+C (3.77) Substituting the values of A, B and C, Eqs. (3.65)-(3.66) into the condition that the first two eigenvalues in (3.77) are positive (which ensures the “any of the RS solution) one finds: te [1—4tanh?(p jaz + Bh) CE v anh*| ae 2 +3tanhi(p Jaz + fh)| exp ( 5? ) 0, which after simple algebra can be represented as follows: fo Feo Bz./q + Bh) exp (-}# ) 1, the corresponding temperature3.6 de Almeida-Thouless line 45 Tar(h) appears to be exponentially small (in this case the value of q is close to 1): = exp (—1i2 T > Tar(h) = ia exp ( 5h ) (3.82)4 Physics of replica symmetry breaking In this chapter the physical interpretation of the formal RSB solution will be proposed, and some new concepts and quantities will be introduced. The crucial concept that is needed to understand physics behind the RSB structures is that of the pure states. 4.1 The pure states Consider again a simple example of the ferromagnetic system. Here, spon- taneous symmetry breaking takes place below the critical temperature T,, and at each site the non-zero spin magnetizations (o;) = -tm appear. As we have already discussed in Section 2.2, in the thermodynamic limit the two ground states with the global magnetizations (6;) = +m and (a;) = —m are separated by an infinite energy barrier. Therefore, once the system has happened to be in one of these states, it will never be able (during any finite time) to jump into the other one. In this sense, the observable state is not the Gibbs one (which is obtained by summing over all the states), but one of these two states with non-zero global magnetizations. To distinguish them from the Gibbs state they could be called the ‘pure states’. More formally, the pure states could also be defined by the property that all the connected correlation functions in these states, such as (ai))¢ = (aja) — (ai) (oj), tend towards zero at large distances. In the previous chapter we obtained a special type of spin-glass ground- state solution. Formally this solution is characterized by the RSB in the corresponding order parameter matrix Q,,, and it automatically means that there actually exist many other solutions of this type. This fact is a direct consequence of the symmetry of the replica free energy (Eqs. (3.25)-(3.26)) with respect to permutations of replicas: if there exists a particular solu- tion for the matrix Q. with the RSB, then any other matrix obtained via 464.2 Physical order parameter P(q) and the replica solution 47 permutations of the replica indices in Q. will also be a solution. On the other hand, as the total mean-field free energy (which is the function of 0) is proportional to the volume of the system, the energy barriers separating the corresponding ground states must be infinite in the thermodynamic limit. Consequently, just like in the example of the ferromagnetic system, all these RSB states could be called the pure states of the low-temperature spin- glass phase. Correspondingly, the Gibbs state (which is formally obtained by summing over all the states of the system) could be considered as being given by the summation over all the pure states with the corresponding thermodynamic weight defined by values of their free energies. For instance, the thermodynamic (Gibbs) average of the site magnetiza- tions can be represented as follows: ) =m = DO want (4.0 (ai, Here m? are the site magnetizations in the pure state 2, and w, denotes its statistical weight, which formally can be represented as follows: wz = exp(—Fa) (4.2) where F, is the free energy corresponding to this pure state. In the same way the two-point correlation function can be represented as the linear combination (0102) = )> walo1o2)2 (4.3) where (o102), is the two-point correlation function in the pure state number a. According to the definition of the pure state (6102)2 = (01)2(02)2 (4.4) Similar expressions can be written for any many-point correlation functions. The representation of the thermodynamic Gibbs state as a linear com- bination of the pure states in which all extensive quantities have vanishing long-distance fluctuations, is actually a central point in the exact definition of the concept of the spontaneous symmetry breaking in statistical mechanics. 4.2. Physical order parameter P(q) and the replica solution To investigate the statistical properties of the spin-glass pure states let us define the overlaps {qzg} among them as follows: 1 N ap = Hy Lomi (4.5) 748 Physics of replica symmetry breaking where m? = (o;), and mf = (oj)g are the site magnetizations in the pure states « and p. Apparently, 0 <| qag |< 1. To describe the statistics of these overlaps it is natural to introduce the following probability distribution function: Ps(q) = > wa p5(4sp — 9) (4.6) op Note that this distribution function is defined for a given sample, and it can depend on a concrete realization of the quenched interactions Jjj. The ‘observable’ distribution function should, of course, be averaged over the disorder parameters: P(q) = Pi(q) (4.7) The distribution function P(q) gives the probability of finding two pure states having their overlap equal to q, with the condition that these states are taken with their statistical thermodynamic weights {wz}. It is the distribution function P(q) that can be considered as the physical order parameter. It should be stressed that P(q) is a much more general concept than ordinary order parameters, which usually describe the phase transitions in ordinary statistical systems. The fact that it is a function is actually a manifestation of the crucial phenomenon that for the description of the spin-glass phase one needs an infinite number of order parameters. The non-trivial structure of this distribution function (it will be calculated explicitly below) demonstrates that the properties of the spin-glass state are essentially different from those of traditional magnets. Consider now which way the order parameter function P(q) can be cal- culated in terms of the replica method. Let us introduce the following set of correlation functions: 1 ay = yD? 7 t Pe 1,0)" ih il os of = de Dou---04)" (48) ing Using the representation of the Gibbs averages in terms of the pure states (4.3){(4.4) for the correlation functions (4.8) one gets: =D (S»o.) (xn) Ge ;4.2 Physical order parameter P(q) and the replica solution 49 1 = Yowaws (= Wlaidaoidp op N“% = Vowawpqap ap Il i dgP3(q)4; = se (Hrlvers) (Cosine i No oem (5 Dee, bins) ( x Tiron] = wews(qep) op = faaPiaas = f daPiiayg (49) For the corresponding correlation functions averaged over the disorder from Eqs. (4.8)-(4.9) one gets: =i = [Pad a =P = Goa) = f daP(aa" (4.10) where i, # in #... Hix. The crucial point in the above consideration is that the function P(q) originally defined to describe the statistics of (somewhat abstract) pure states, can be calculated (at least theoretically) from the multipoint correlation functions in the Gibbs states. Therefore, if we were able to calculate the above multipoint correlation functions in terms of the replica approach, the connection of the formal RSB scheme with the physical order parameter would be established. In terms of the replica approach, the correlator g = (@;)2 can be represented as follows: = Dles)exe Bale] — BHI)50 Physics of replica symmetry breaking = lim, |i SY | (Pos) exp (-eSa00n) =1 0° a=1 Slim Go) #0) (4.11) where a and b are two fixed different replicas (the summation over the remaining (n — 2) replicas in Eq. (4.11) gives the factor Z”~?, which turns into Z~? in the limit n + 0). In a similar way one gets: qQa= Jim (of, otabab (i #3 a#b) MAY 45, oP hn Os (FRA... AiG axd) (4.12) In the calculations of the previous chapter it has been demonstrated that the free energy of the model under consideration is factorizing into the independent site replica free energies. Therefore, the result (4.12) for ¢® can be represented as follows: a 4g! = lim [ (ofa) | = lim Quel" (4.13) where (see Eq. (3.9)) Quy = (ofa?) (4.14) is the replica order parameter matrix introduced in Chapter 3, which is obtained from the saddle-point equation for the replica free energy. Because in the RSB solution the matrix elements of Q,, are not equivalent, one has to sum over all the saddle-point solutions for the matrix Q,» to perform the Gibbs average. Such solutions can be obtained from one of the RSB solutions by doing all possible permutations of rows and columns in Qas. The summation over all these permutations corresponds to the summation over the replica subscripts a and b of the matrix Qa». Thus, the final result for the correlator q) should be represented as follows: 1 Ooi K q ee OatA 2a! (4.15) where n(n — 1) is the normalization factor that is equal to the number of different replica permutations. The results (4.15) and (4.10) demonstrate that by using the formal RSB solution for the matrix Qa, considered in the previous chapter one can calculate the order parameter distribution function P(q), which has been originally introduced on the basis of purely qualitative physical arguments.4.2 Physical order parameter P(q) and the replica solution 51 From these two equations one gets the following explicit expression for the distribution function P(q): : L Pla) = B89 egy 3 Oa ~ (4.16) Using the algorithms of the Parisi algebra (Section 3.4) in the continuous RSB representation this result can be rewritten as follows: A P@= [dda -4) (4.17) Assuming that the function q(x) is monotonous (which is the case for the saddle-point solution obtained in Chapter 3), one can introduce the inverse function x(q), and then from Eq, (4.17) one finally obtains: (4.18) (Note that the same result can be obtained by comparing Eqs. (4.15) and (4.10)). This is a key result, which defines the physical order parameter distribution function P(q) in terms of the formal saddle-point Parisi function q(x). The above result can also be represented in the integral form: x(q)= [Pca (4.19) which gives the answer to the question, as to the physical meaning of the Parisi function q(x). According to Eq. (4.19) the answer is as follows: the function x(q) inverse to q(x) gives the probability of finding a pair of the pure states that would have an overlap not bigger than q. Using the explicit solution for the Parisi function q(x) in the vicinity of the critical point (Eqs. (3.53)-(3.57)), according to Eq, (4.18) for the distribution function P(q), one gets: P(g) = x05(q — go) + (1 — x1)6(q — 41) + PC) (4.20) where p(q) is the smooth function defined in the interval go < q < q1. In the vicinity of the critical point, t < 1, where the solution (3.53) is valid, this function is just constant: p(q) = 2. The result (4.20) shows that the statistics of the overlaps of the pure states demonstrate the following properties: (1) There is a finite probability (1 — x;) ~ (1 — 2) that, taken at random, two pure states would appear to be the same state. The ‘selfoverlaps’ (Eq. (2.3)), of these states is equal to qi; ~ t.52 Physics of replica symmetry breaking Pp HT) GF P % oh q P 1 % 0 qa 0 a? 9 (ce) Fig, 4.1. The qualitative shape of the functions q(x) and P(q): (a) in the zero magnetic field near the critical point (« < 1); (b) in the finite magnetic field h, for 0 that taken at random two pure states would appear to be the most ‘distant’, having the minimum possible overlap 0 ~ hE. (3) There is a finite probability (x; — xo) that, taken at random, two pure states would have the overlap q in the interval go < q < qi. For a given small interval 5g there is a finite probability p(q)dq of finding4.2 Physical order parameter P(q) and the replica solution 53 two pure states with the overlap in the interval (q,q + 6q), where pmsqsn. Although for arbitrary values of the temperature and the magnetic field it is hardly possible to calculate the functions q(x) and P(q) analytically, their qualitative behavior remains similar to the case considered above. The only difference is that the concrete shape of the function P(q) in the interval qo <4 wawgw;0(q — ap)5(q2 — 4x;)5(43 — 9p,) (5.1) apy In terms of the RSB scheme the calculation of this function is quite similar to that for the function P(q). In particular, in terms of the replica matrix Qup instead of Eq. (4.16), in the present case one can easily prove that A 1 P(q1.42,43) = lim a= WN) (de, 10 — 41)8(Qac — 42)5( Qe — 93) (5.2) In terms of the Fourier transform of the function P(q1, 92.43): (V1, ¥2,93) = f dqidqzdqsP (41, 42,93) exp ligiyi + igay2 + igsys] (5.3) 545.1 Ultrametric structure of pure states 55 instead of Eq. (5.2) one gets: (V1 92593) = li So exp [1Qary + iQacy2 + iQbcy3] 1 mao n(n — 1)(n— 2) axbye 7 1 Waray = lim operat Mvnd4on40s)] (64) where oe { SxHiQay)s a#b eG Let us substitute the RSB solution for the matrix Qu, into Eq. (5.4). In the continuum RSB limit the matrix Qa, turns into the function q(x), and according to the Parisi algebra (Section 3.4) the replica matrix Aqp(y) turns into the corresponding function A(x; y): A(x; y) = expfig(x)y} (5.6) Using the algorithms of the Parisi algebra, Eqs. (3.39)-(3.43), after simple calculations one obtains: : 1 a. tim (eat Fonaonvavy)) A x =3 L dx [alas dates velAbesya) + Atcinn f dzA(z; y2)A(z; y3) +Atxiya) [ dedteiyAleiyn) + ACsiys) [deAteiy)ACzsya)] (67) 0 0 Accordingly, for the function P(qi, 42, q3): P (41.92.93) = [andra 8(1, ¥2,¥3) exP[—igiy1 — igay2 — igsy3] (5.8) one gets: 1 1 Plananan) = 5 ff ax [x9tats) — 41)8(408) — 2)8(aCx) — a) +8(q(x) ~ a1) [ deb(ale) — a2)8(al2) — 43) +5(a(%) ~ a2) [ d28(a(2) ~ 8142) ~ a9 +6talx) — an) [ azd(ale)—an)8(402)-a2)] (59) Introducing the integration over q instead of that over x and taking into56 Ultramecricity Fig. 5.1. The ultrametric tree of the spin-glass states. account that dx(q)/dq = P(q) one finally obtains the following result: Plguaa.as) = $Plardx(an)8(as — a2)6(a1 — 9) +$PCQP(a2)0(a — 42)8(q2 — 43) +5P(42)P(43)0(a2 — 93)8(@3 — 41) +5PC@)PCGIOGs—a)8a1-a) (6.10) From this equation one can easily see the following crucial property of the function P(q1,q2.q3). It is non-zero only in the following three cases: UN=2 5933 1=935923 93 =G < 4H. In all other cases the function P(q1, 92,93) is identically equal to zero. In other words, this function is not equal to zero only if at least two of the three overlaps are equal, and their value is not bigger than the third one. It means that in the space of spin-glass states there exist no triangles with all three sides being different. The spaces having the above metric property are called ultrametric. A simple illustration of the ultrametric space can be given in terms of the hierarchical tree (Fig. 5.1). The ultrametric space here is associated with the set of the endpoints of the tree. By definition, the overlaps between any two points of this space depend only on the number of ‘generations’ (in the ‘vertical’ direction) to the level of the tree where these two points have a common ancestor. One can easily check that paired overlaps among three arbitrary points of this set do satisfy the above ultrametric property. The reader can find a detailed description of the ultrametric spaces in the review [3]. Here we are going to concentrate only on the general qualitative properties of the ultrametricity that are crucial for the physics of the spin- glass state.5.2 Tree of states 37 5.2 Tree of states Let us consider how the spin-glass ultrametric structures can be defined in more general terms. Consider the following discrete stochastic process, which is assumed to take place independently at each site i of the lattice. (1) At the first step, with the probability Po(y) one generates mn, random numbers y* (a; = 1,2,...,1), Which belong to the interval [—1, +1]. (2) At the second step, for each y* with the conditional probability P;(y"|y) one generates ny random numbers y%!%2 (2 = 1,2,...,m2), belonging to the same interval [—1, +1]. (3) At the third step, for each y*% with the conditional probability P2(y*!2|y) one generates n3 random numbers y*%% (a3 = 1,2,...,13), belonging to the same interval [—1, +1], etc. This process is continued up to the L-th step. Finally, in the interval [-1,+1] one gets mjn2...n, random numbers, which are described by the following set of the probability functions Pray *y%") (= 1,2,...,L) (3.11) This stochastic (Markov) process takes place independently at each site of the lattice. Then, for each set of the obtained random numbers let us define the corresponding site spin states as follows: ott = sign (yZ~*) (5.12) This way one obtains the set of mjnz...nz spin states, which are labelled by the hierarchical ‘address’ 2; ...az. The ‘address’ of a concrete state describes its genealogical ‘history’. Simple probabilistic arguments show that the overlap between any two spin states depends only on the degree of their ‘relativeness’, i.e. it is defined only by the number of generations that separate them from the closest common ancestor. Consider two spin states that have the following ‘addresses’: HD oo RIK 1 Ky? OL and a2... 1B Bi42--- BL where isi # 141. These two ‘addresses’ are becoming different, starting from the generation number |. As the stochastic processes generating the58 Ultrametricity states are independent at each site, for the overlap between these two states : ot te Gace ey Cte (5.13) 7 in the thermodynamic limit N — oo one gets: +1 Wego -[ dy; ...dyPo(v1)Pi(yily2)... Pi-1G—-1ly0) 41 / 2 x fe ayigt dy Pu(yilyiyi)Prea(yrsilyis2)---Pe-i(ye-1lyz) sign (yi) = a (5.14) Therefore, the overlap depends only on the number / of the level of the tree at which the two states were separated in their genealogical history, and does not depend on the concrete ‘addresses’ of these states. One can easily see that it automatically means that the considered set of the states is ultrametric. Note, that this is a general property of the considered stochastic evolution process, and it remains true for any choice of the probability distribution functions (5.11) which describe the concrete tree of states. A general reason for that is very simple. The above stochastic procedure has been defined as the random branching process that takes place in the infinite dimensional space (in the limit N — 00), and it is clear that here the branches, once separated, never come close again. Therefore, it is of no surprise that ultrametricity is observed in Nature very often. Examples are the space of biological species, the hierarchical state structures of disordered human societies, etc. Let us consider the above hierarchical tree of states in some more detail. The equations for the overlaps between two spin states (5.13) and (5.14) can also be represented in terms of the so-called ancestor states m*'~*: Le ayaa? a = yD (mi) (5.15) where the site magnetizations in the ancestor state m*'-* at the level | are defined as follows: mp8 = (o8t BIA) sem ys) (5.16) Here (...)(x,1.0,) denotes the averaging over all the descendant states (branches) of the tree outgoing from the branch %;...% at the level number5.2 Tree of states 59 1. By definition: +1 mi(yi™) = i 1 Att Aye Pi ye yt) x Piai(yisilyi42)---Pr—1(yr—ilyz) sign (yz) (5.17) This equation for the function mj(y) can also be written in the following recurrent form: +1 mi(y) i , a Purly mr) (5.18) where +1 Prob =f dyrgs .-.dyp Puy lYegs Pha (vee i |i142) «+ Pearly’) . (5.19) Therefore, all the concrete properties of the tree of states, and in particular the values of the overlaps {qi}, are fully determined by the set of the probability functions (5.11) or (5.19). For the complete description of a concrete spin-glass system all these functions have to be calculated, or at least the algorithms of their calculations must be derived. In particular, this can be done for the SK model of a spin-glass. Unfortunately, the corresponding calculations for this model are rather cumbersome, and the reader interested in the details may refer to the original papers [23] and [24]. Here only the final results will be presented. The ultrametric tree of states that describes the spin-glass phase of the SK model is defined by the random branching process described above, in which the continuous limit L — oo must be taken. In this limit, instead of the integer numbers /, which define the discrete levels of the hierarchy, it is more convenient to describe the tree in terms of the selfoverlaps {q)} of the ancestor states. In the limit L —> oo the discrete parameters {q/} turn into the continuous variable 0 < q < 1. Instead of the discrete ‘one-step’ functions (5.11) in the continuous limit it is more natural to describe the tree in terms of the functions (5.19) which define the evolution of the tree from the level q to the other level q’. It can be proved (and it is this proof which requires us to go through somewhat painful algebra) that in the continuous limit these functions are defined by the following non-linear diffusion equation: é a -TPH=555 1 é gn = 205 P+ x(a)ma(veP (5.20)60 Ultrametricity with the initial condition: is Pag (vly’) = 5(y — y’) (5.21) Here x(q) is the function inverse to q(x) (which is given by the RSB solution, Chapter 3), and the function m,(y) is the continuous limit of the discrete function (5.18). It can be shown that this function defines the distribution of the site magnetizations in the ancestor states at the level q of the tree. One can easily derive from Egs. (5.18) and (5.20) that the function m,(y) satisfies the following eaten Smt) = 35, Emo) + amd ma() (522) The above equations fully describe the properties of the ultrametric tree of the spin-glass states of the SK model. 5.3 Scaling in the space of spin-glass states Let us summarize all the results obtained for the spin-glass model with the long-range interactions: (1) In terms of the formal replica calculations, the free energy of the system can be represented in terms of the functional F[0], which depends on the n x n replica matrix Q, (Section 3.1), In the ther- modynamic limit the leading contribution to the free energy comes from the matrices o, which correspond to the extrema of this func- tional, and the physical free energy is obtained in the limit n > 0. In this limit the extrema matrices Q* are defined by the infinite set of parameters that can be described in terms of the continuous Parisi function q(x) defined at the interval 0 < x < 1 (Sections 3.3-3.4). In the low-temperature region near the phase-transition point this function can be obtained explicitly (Section 3.5, Fig. 4.1). (2) On the other hand, in terms of qualitative physical arguments, one can define as the order parameter the distribution function P(q), which gives the probability of finding a pair of pure spin-glass states having the overlap equal to q. In terms of the RSB scheme one can show that the distribution function P(q) is defined by the Parisi function q(x): P(q) = dx(q)/dq, where x(q) is the inverse function to q(x) (Section 42). The low-temperature solutions for q(x) and for P(g) show that there exists the continuous spectrum of the overlaps among the pure states.5.3 Scaling in the space of spin-glass states 61 (3) Next, one can introduce the ‘three-point’ distribution function P(q1,92,93) which gives the probability that three arbitrary pure states have their mutual pair overlaps equal to qi,q2 and q3. In terms of the RSB scheme this function can be calculated explicitly, and the obtained result shows that the space of the pure states has the ultrametric topology (Section 5.1). (4 It can be shown that the ultrametric tree-like structures can be de- scribed in terms of the hierarchical evolution tree, which is defined by the random branching process. Based on the above results, the spin-glass phase can be described in the qualitative physical terms as follows (see also Chapter 2). At a given temperature T below Te, the space of spin states is split into numerous pure states (valleys) separated by infinite energy barriers. Although the average site magnetizations mj; are different in different states, the value of the selfoverlap: N 4a(T) = Sone (5.23) appears to be the same in all the states. The value of q is the function of the temperature (q(Te) = 0; (0) = 1), and near Ty it can be calculated explicitly. On the other hand, the overlaps q** of the pure states cover continuously the whole interval 0 < q*? < q(T). (In the presence of an external magnetic field h this interval starts from a non-zero value: go(h, T) < q*° < qi(h, T).) The distribution of the values of the overlaps q*° is described by a probability function P(q), which depends on the temperature (and the magnetic field). The structure of the space of the pure states can be described in terms of the ultrametric hierarchical tree discussed above. Now, if the temperature is slightly decreased T > T’ = T — 5T, each of the pure states is split into numerous new ‘descendant’ pure states. These states are characterized by the new value of the selfoverlap q(T’) > q(T). Correspondingly, the interval of their overlaps becomes bigger: 0 < q%® < aT’). At a further decrease of the temperature, each of the newly born pure states is split again into new descendant pure states, and this branching process continues down to zero temperature (q(T — 0) — 1). The tree of pure states obtained this way has the property of the self-similarity (scaling), and at any given temperature the natural scale in the space of states is given by the value of q(T).62 Ultramerricity Owing to infinite energy barriers separating the valleys, the ‘observable’ physics at the given temperature T is defined by only one of the pure states, which in terms of the hierarchical tree corresponds to one of the ‘ancestor’ states at the level (scale) q(T). All these states are revealed in the horizontal cross-section of the tree at the level q(T). 5.4 Phenomenological dynamics Although the dynamical properties of spin glasses is an extremely hard problem even at the mean-field level, certain (the most simple) general slow relaxation properties of the disordered systems with the hierarchical structure of the free energy landscape can be understood rather easily using a purely phenomenological approach [25]. Assume that the free energy landscape in the spin-glass phase is of the type shown in Fig. 2.2: big wells contain a lot of smaller ones, each of the smaller wells contains a lot of even smaller ones, and so on. This kind of landscape could be characterized by the typical value of the finite energy barrier A(q) separating the wells at the scale q. Assuming that this landscape has the scaling property, the dependence of the typical value of the energy barrier A from the scale q could be described by the following simple scaling law: A(q) = 40(9-4(T))"; (@ > 4(T); ¥ > 0) (5.24) Here q(T) is the value of the selfoverlap of the pure states at the temperature T. The parameter q(T) is the characteristic scale (the typical scale of the valleys) at which the barriers separating the states are becoming infinite. Consider now what kind of relaxation properties could be derived from the above assumptions. The characteristic time needed to overcome the barrier Ais 2(A) ~ toexp (3) (5.25) where to is characteristic microscopic time. Thus, the spectrum of the relax- ation times inside the valley can be represented as follows: 1(q) ~ toexp[BAo(g — 4(T))*) (5.26) Then the long-time relaxation behavior of the order parameter a0 = 7 DleOai(n) (527)5.4 Phenomenological dynamics 63 can be estimated (very roughly) as follows: 4 t q(t) ~ Lon dq qexp ( @ (5.28) Using (5.26), one gets: 1 a(0)~ [aa exp (Inigy— Lexpt-Baota—aiT V1) (528) In the limit of large times t > to the saddle-point estimate of the above integral gives the following result: ify es] (5.30) In(t/to) Therefore at large times the order parameter approaches its equilibrium value q(T) logarithmically slowly. Apparently, the relaxation behavior of other observable quantities should be of the same slow type. Of course, true dynamic properties of spin glasses are much more com- plicated, and they can not be reduced only to the phenomenon of extremely slow relaxation. Actually, the main property of spin glasses is that they can not reach true thermodynamic equilibrium at any finite observation time (the reader can find detailed discussion of this problem in a recent review paper [8]). Because the theoretical achievements in understanding the dy- namical properties of spin glasses are far from being impressive yet, in the next chapter we will consider the results of experimental observations of the relaxation phenomena in real spin-glass magnets. ae) ~ aT) + |6 Experiments In this chapter we will consider classical experiments that have been per- formed on real spin glass materials, aiming to check to what extent the qualitative picture of the spin-glass state described in previous chapters does take place in the real world. The main problem of the experimental ob- servations is that the concepts and quantities that are very convenient in theoretical considerations are rather far from the experimental realities, and it is a matter of the experimental art to invent convincing experimental pro- cedures that would be able to confirm (or reject) the theoretical predictions. A series of such brilliant experiments has been performed by M. Ocio, J. Hammann, F. Lefloch and E. Vincent (Saclay), and M. Lederman and R. Orbach (UCLA) [9]. Most of these experiments have been done on the crystals CdCrj7Ing3S4. The magnetic disorder there is present due to the competition of the ferromagnetic nearest neighbor interactions and the an- tiferromagnetic higher-order neighbor interactions. This magnet has already been systematically studied some time ago [26], and its spin-glass phase transition point T = 16.7K is well established. Some of the measurements have been also performed on the metallic spin glasses AgMn [27] and the results obtained were qualitatively quite similar. It indicates that presumably the qualitative physical phenomena observed do not depend very much on the concrete realization of the spin-glass system. 6.1 Aging The phenomenon of aging in spin glasses has been known for many years [28]. It is not directly connected with the hierarchy of the spin-glass states, but it explicitly demonstrates the absence of true thermodynamic equilibrium in spin glasses. The procedure of the experiments is as follows. The sample is cooled 646.1 Aging 65 TRM/MFC 2 x é 0.35 0.30 0.2 Sa v 10 100 1000 10000 1 (min) Fig. 6.1, The relaxation behavior of the magnetization in the field cooled aging experiments. down into the spin-glass state in the presence of a weak uniform magnetic field H. Then, at a constant temperature T < T, the sample is kept in this magnetic field during some waiting time ty. Finally the magnetic field is switched off, and the measurement of the relaxation of the thermoremainent magnetization (TRM) is performed. The results of these measurements for different values of ty are shown in Fig. 6.1. The first important result of these measurements is that the observed relaxation is extremely slow and non-exponential (note, that the typical values of ty are macroscopic: minutes, hours, days). More important, however, is that the relaxation appears to be non-stationary: the relaxation processes that take place in the system after switching off the field depend on the ‘lifetime’ ty of the system before the measurement was started. The spin glass is getting ‘stiffer’ with the time: the bigger ty is, the slower the relaxation goes on. Therefore, the results of the measurements depend on two time scales: the observation time t, and the time that has passed after the system came into the spin-glass state, the ‘aging’ time ty. It is crucial that at all experimentally accessible time scales no indication that the relaxation curves are reaching saturation at some limiting curve (corresponding to ty = 00) has been observed. Thus, at any experimentally accessible times such a system remains out of the true thermal equilibrium. Note that it is not the presence of the magnetic field that is responsible for the observed phenomenon. The magnetic field here is just the instrument that makes it possible to demonstrate it. One can also perform the ‘mirror’66 Experiments 1.10 SS 0.90 TRM + ZFC 0.70 am(t)/MFC 0.50 035 1 10 100 1000 ¢ (min) Fig. 6.2. The relaxation behavior of the magnetization in the zero field cooled aging experiments, experiment: the system is cooled down into the spin-glass state in the zero magnetic field, then it is kept at a constant temperature T < T, during some waiting time ty, and finally the magnetic field is switched on and the relaxation of the magnetization is measured. Again, the results of the measurements essentially depend on ty. Moreover, for any given value of tw the curves obtained in these two types of the experiment appear to be symmetric: the sum of the values of the magnetizations obtained in these ‘mirror’ experiments appears to be a time-independent constant (Fig. 6.2). 6.2 Temperature cycles and the hierarchy of states Now we consider two types of the experiments which are supposed to reveal the effects connected with the existence of the hierarchical tree of spin-glass states. In the experiments of the first type, the sample, in a weak magnetic field, is cooled down into the spin-glass phase. Then, it is kept at a constant temperature T < T, during some waiting time ty,. After that the temperature is slightly changed down to T’ = (T —AT) (where the value of AT is small), and the sample is kept at this temperature during waiting time ty,. Then the temperature is changed up to the original value T again, and the sample is kept at this constant temperature during waiting time ty,. After that the magnetic field is switched off and the relaxation of the magnetization is measured, The results for different values of AT’ are shown in Fig. 6.3.6.2 Temperature cycles and the hierarchy of states 67 AT (K) ous Si 0.3 ond 0S 1.0 a oat = z 009 ty (min) 1000 0.07 100 30 10 or on + 19 00100010000 (min) Fig. 6.3. The relaxation behavior of the magnetization in the aging experiments with the cooling temperature cycles. The main result of these measurements is the following. It is clear from the plots of Fig, 6.3 that if the value of the temperature step AT is not too small, then all the relaxation curves appear to be identical to those in the usual aging experiments (Section 6.1) with the waiting time ty = tw, + tw). It means that for the processes of equilibration at the temperature T, the system remained effectively completely frozen during all the time period ty, when it was kept at the temperature (T — AT). In the experiments of the second type, again, the sample in the presence of a weak magnetic field is cooled down into the spin-glass phase, and then it is kept at a constant temperature T < T, during waiting time tw,. Next, the sample is slightly heated up to the temperature T’ = (T + AT), (where the value of AT is small) and after a relatively short time interval it is cooled down again to the original temperature T. Then, it is kept at this constant temperature during waiting time ty,, and after that, the magnetic field is switched off and the relaxation of the magnetization is measured. The results for different values of AT are shown in Fig. 6.4. In this case one finds that if the value of the temperature step AT is not too small, then all the relaxation curves appear to be identical to those in the usual aging experiments (Section 6.1) with the waiting time ty = ty,. It means that even slight heating is enough to wipe out all the aging that has been ‘achieved’ at the temperature T during the time period before heating. In other words, after the slight heating jump the equilibration processes start all over again, while all the ‘pre-history’ of the sample appears to be wiped out. (Note that the temperature (T + AT) is still well below T;.)68 Experiments O12 _ 3 z 0.10 2 t= 1000 min 0.08 0.08 o1 0.1 1 10 100 1000 10000, (min) Fig. 6.4. The relaxation behavior of the magnetization in the aging experiments with the heating temperature cycles. Such quite asymmetric response of the system with respect to the con- sidered temperature cycles of cooling and heating can be well explained in terms of the qualitative physical picture of the continuous hierarchy of the phase transitions and the tree-like structure of the spin-glass states. The qualitative interpretation of the results described above is as follows. The process of thermal equilibration, as time goes on, can be imagined as the process of jumping over higher and higher energy barriers in the space of states. After some waiting time ty, the system covers a certain part of the configurational space, which could be characterized by the maximum energy barriers of the order of Amax ~ T log(tw/t) (here t is characteristic microscopic time). It is assumed that any scale in the configurational space is characterized by certain typical values of the energy barriers (see also Section 5.4). Then the results of the experiments with the temperature cycles of cooling can be interpreted as follows. During the time period ty, when the system is kept at the temperature T, it covers a certain finite part of the configurational space inside one of the valleys. After cooling down to the temperature (T — AT) this part of the configurational space is split into several smaller valleys separated by infinite energy barriers. Besides, the finite energy barriers separating the metastable states inside the valleys are getting higher, while some of these metastable states are split into many new ones. Then, during the time t,,,, the system is trying to cover these new states6.3 Temperature dependence of the energy barriers 6 being locked by infinite barriers in a limited part of the configurational space. Therefore, whatever time has passed at the temperature (T — AT) the system can cover only those states, which are the descendants of the states already occupied at the temperature T, and not more. Note that this phenomenon of ergodicity breaking is just the consequence of the phase transition that occurred in the system due to cooling down from the temperature T to the temperature (T — AT). Then, after heating back to the original temperature T all these descendant states are merging together into their ancestors, and the process of thermal equilibration at the temperature T continues again, as if there was no time interval that the system spent at the temperature (T - AT). In the experiments with the temperature cycles of heating, the effects to be expected are different. After heating to the temperature (T + AT) the states occupied by the system during the time ty, at the temperature T merge together into the smaller number of their ancestor states. If the value of AT is not too small, such that q(T + AT) T*) barrier becomes infinite. In this sense the critical temperature T, can be interpreted just as the maximum possible value of T*. In conclusion, the experiments considered above clearly demonstrate the absence of the thermal equilibrium in the spin-glass phase at all experimen- tally accessible time scales. These experiments also demonstrate the existence of the whole spectrum of the free energy barriers up to infinite values, atnR Experiments Fig. 6.6. The dependence of the values of the free energy barriers at the temperature T from their values at the temperature T — AT. a(ayaT, 200 400 500 300 KK T= 10K 00000 T=9.5K 100 K gaooo 7=8K 20 25 30 35 A Fig. 6.7, The dependence of dA/dT from the values of the barriers A.6.3 Temperature dependence of the energy barriers 2B any temperature below Te. The results of the measurements show that the barriers’ heights strongly depend on the temperature, and at any tempera- ture T < Te certain barriers are becoming infinite. This phenomenon clearly indicates the presence of the ergodicity breaking phase transition at any temperature below T,, which results in the continuous process of fragmen- tation of the phase space into smaller and smaller valleys with decrease of the temperature.Part two Critical phenomena and quenched disorder7 Scaling theory of the critical phenomena 7.1 The Ginzburg-Landau theory We begin our study of the critical phenomena at phase transitions of the second order with the mean-field approximation discussed in the Introduction (Section 1.2). The starting point for further consideration is the mean-field expansion of the free energy in the vicinity of the critical point Te, (Eq. (1.28), Fig. 1.1): 1 1 5G) = 50? + Gao" — hd (1) where t = (T — T.)/Tc < 1 is the reduced temperature, h is the external magnetic field. Here the ‘coupling constant’ g is the parameter of the theory, and the order parameter ¢ = (a;) is the average spin magnetization. The value of ¢ is determined from the condition of the minimum of the free energy, df /db = 0: th +ee—h (7.2) and d?f /dd? > 0. In the absence of the external magnetic field (h = 0) at temperatures above T,, (t > 0), the free energy has only one (trivial) minimum at ¢ = 0. Below the critical point, t < 0, the free energy has two minima, and the corresponding solutions of the saddle-point equation (7.2) are: /izl =+ .3) o@) = + : (7.3) As T = T, from below, $(T) > 0. As has already been discussed in the Introduction, this very simple mean- field theory demonstrates on a qualitative level the fundamental phenomenon called the spontaneous symmetry breaking. At the critical temperature T = 71B Scaling theory of the critical phenomena T, the phase transition of the second order occurs, such that in the low- temperature region T < T, the symmetry with respect to the global change of the signs of the spins is broken, and two (instead of one) ground states appear. These two states differ by the sign of the average spin magnetization, and they are separated by the macroscopic barrier of the free energy. Ina small non-zero magnetic field (h < 1) the qualitative shape of the free energy is shown in Fig. 1.1(b). In this case the saddle-point equation (7.2) always has a non-zero solution for the order parameter ¢ at all temperatures. In particular, in the low-temperature region (t < 0) we find: fH tor & helt) g 2 _ (74) h 13 (7) for h > he(t) 8 where hela) = leh? Te) Ve whereas in the high-temperature region (t > 0): t for h < he(t) t (18) h 3 G) for h > hel) Thus, at h # O the phase transition is ‘smoothed out’ in the temperature interval |r| ~ h?/3 (Eq. (7.5)) around Ty. The physical quantity, which describes the reaction of the system on the infinitely small magnetic field, is called susceptibility. It is defined as follows: r= Siew 77) According to Eqs. (7.4}-(7.6) one finds that near the critical point the susceptibility becomes divergent: cathe I: x _. (78) 3ltl atT T, cx 1 (7.11) const= ~— atT Correspondingly, the Hamiltonian in which only small spatial fluctuations of the order parameter are taken into account can be written as follows: a= f ax [0007 + 526°) + eo) —h6Co] 713) The theory that is based on the above Hamiltonian is called the Ginzburg— Landau approach. In fact the Ginzburg-Landau Hamiltonian is nothing but the first few terms of the expansion in powers of @ and (V@). In the vicinity of the (second-order) phase-transition point, where the order parameter is small and the leading contributions come from large-scale fluctuations, such an approach looks quite natural.80 Scaling theory of the critical phenomena Consider the contributions caused by small fluctuations at the background of the homogeneous order parameter ¢o = /|tl/g: G(x) = b0 + 9(x) (7.14) where (x) < ¢o. For simplicity let us consider the case of the zero magnetic field. Then the expansion of the Hamiltonian (7.13) to the second order in yields: Ha Hot f dx[5 ota)? + ito] (7.15) In terms of the Fourier representation Pk (x) = / ‘ago 6) exD(—iex) (7.16) one gets: @k =3/ ome (7 +21cl) | oh) P +Ho (7.17) Therefore, for the correlation function £ Dob) ok)? exp(—H[g]) k) = P) = 7. Golk) = (Noth = EE est Heh (7.18) one obtains the following result: 1 Golk) = atl (7.19) Besides, it is obvious that (p(k) g(k')) = Golk)5(k +k’) (7.20) Therefore, for the spatial correlation function Go(x) = (((0)H(%))) = (G(0)G(%)) — (G(0))(H(%)) @k ' = (o(O)o(2)) = f Sp hotkey exp(ikx) (721) we obtain: jx") for |x| < Re(t) = _ (7.22a) Go(x) ~ 2Ic| exp (—Ix|/Re) for |x| > Re(t); (7.22b) Here the quantity Re(a) ~ |e (7.23) is called the correlation length.7.1 The Ginzburg-Landau theory 81 Thus, the situation near T, (|t| < 1) looks as follows. At scales much larger than the correlation length R.(t) > 1 the fluctuations of the field @(x) around its equilibrium value po (¢o = 0 at T > T., and $9 = V/|t|/g at T < T.) become effectively independent (their correlations decay exponentially, Eq. (7.22b). On the other hand, at scales much smaller than R,(z), in the so called fluctuation region, the fluctuations of the order parameter are strongly correlated, and their correlation functions exibit weak power-law decay, Eq. (7.22a). Therefore, inside the fluctuation region the gradient, or the fluctuation term of the Hamiltonian (7.13) becomes crucial for the theory. At the critical point the fluctuation region becomes infinite. Let us estimate to what extent the above simple considerations are correct. The expansion (7.15) could be used and the result (7.22) is justified only if the characteristic value of the fluctuations g are small in comparison with the equilibrium value of the order parameter ¢o. Because the correlation length R, is the only relevant spatial scale that exists in the system near the phase transition point, the characteristic value of the fluctuations of the order parameter could be estimated as follows: =_ 1 wre The above simple mean-field estimates for the critical behavior are valid only if the value of g? is much smaller than the corresponding value of the equilibrium order parameter $3: i sea MOVPE) ~ REO? (7.24) ROP? a (7.25) Using (7.23) we find that this condition is satisfied if: elt <1 (7.26) Therefore, if the dimension of the system is bigger than 4, near the phase- transition point, t + 0, the condition (7.26) is always satisfied. On the other hand, at dimensions D < 4 this condition is always violated near the critical point. Thus, these simple estimates reveal the following quite important points. (1) If the dimension of the considered system is bigger than 4, then its critical behavior in the vicinity of the second-order phase transition is successfully described by the mean-field theory. (2) If the dimension of the system is less than 4, then, according to Eq. (7.26), the mean-field approach gives correct results only in the