STATISTICAL MECHANICS An Advanced Course with Problems and Solutions RYOGO KUBO Tokyo HIROSHI ICHIMURA —TSUNEMARU USUI NATSUKI HASHITSUME Tokyo Institute of Technology Kyoto University Ochanomicu University e CY eo S fevoREGIS TERED ON VERSION ADDS NO WA TERMARK &, & NORTH-HOLLAND, AMSTERDAM - OXFORD - NEW YOM : re >© isvier Sience Publishers BLV, 1965, All rights reserved. No par of this publication may be reproduced. stored ina retrieval system, oF transmitted, in any form or by any means, clecttonic, mechanical, photocopying, recocing oF ‘otherwise, without the prior permission of the publisher, Ese Holland Physics Publishing Disision), RO. ox 105, 1000 AC Amsterdam sin the USAT publication has been registered with the Copyright Information can be obtained from the CCC ‘Special regulations fr rea ‘Clearance Center tne. (CCC), my, Massachusetts et Sience Publishers BLY. (North he Netherlands bout conditions under which photocopies of pats of this publication may be made in the USA. Al other copsright questions, including photocopying outside of the USA, should be referred to the publisher. ISBN: 0 444 87103 9 Paperback First edition (hardbound) 1965 Second edition (hardbound) 1967 Third edition (hardbound) 1971 Fourth edition (hardbound) 1978 Fifth edition (hardbound) 1978 Sith edition (hardbound) 1981 Seventh edition (paperback) 1988 Published by Nont-tolland Physics Publishing a division of Elsevier Science Publishers LV. PO, Box 108 1000 AC Amsterdam The Netherlands Sole distributors fr the USA and Canada: Eleever Science Publishing Company, Ine 52 Vanderbilt Avenue New York, NY 10017 USA ‘North-Holland Personal Library: paperback’ Library of Congress Catalog Card Number 67-20003, Printed in The Netherlands 45 y REGISTERED ©) VERSION ADDS NO WA TERMARK &, Op, <5 Tint-drwePREFACE TO ENGLISH EDITION The original text ofthis volume is part of the book “Problems and Solutions in Thermodynamics and Statistical Mechanics", itself one of the “University Series” published by the Shokabo Publishing Company. At the request of the present publisher, the English edition is being published in two volumes, cone on thermodynamics and the other on statistical mechanics. Considering the more urgent interest of university students in statistical mechanics, this Volume has been translated and published first. The volume on thermo- dynamics is expected to be published within a year. ‘The translation was made from the Japanese text by the original authors, together with a few collaborators. AS the editor of the original Japanese edition and of the English edition, I wish to express my deep appreciation to Drs. Masaji Kubo, Toshihiko Tsuneto and Satoru Miyake who did the trans- lation work with the authors, and particularly to Professor Donald C. Worth of International Christian University, Tokyo, who kindly took the trouble of helping us with linguistic difficulties. ‘The authors are also indebted to Miss N. Tokuda for the preparation of the manuscript. 1964 Ryoao Kuso VERSION ADDS NO WA TERMARK &, < eint-driNgPREFAC! E TO JAPANESE EDITION ‘Thermodynamics and statistical mechanics are indispensable tools in studying the physics of the properties of matter. Statistical mechanies, together with quantum mechanics, provides a foundation for modern physics which aims at the thorough understanding of physical phenomena from the microscopic viewpoint of atomic physics. Fundamental knowledge and training in statis tical mechanics are therefore of vital importance not only for students studying the physical properties of matter but also for those who study nuclear physics or even astrophysics. Outside the realm of physics, its im- portance is rapidly penetrating into chemistry, biology and into those vast areas of technology which owe their growth to the advances in modern physics. ‘Thermodynamics belongs completely to classical physics and is some- times regarded as unimportant by students of physics who are over-occupied. in learning modern physics. Even for students in chemistry, the present is different from the time some decades ago when physical chemistry was al- ‘most nothing but chemical thermodynamics. However, it must be stressed hhere that the usefulness and unique significance of thermodynamics as a fundamental science remain as basic today as they were in the latter half of the last century. Thermodynamics teaches us the value of @ phenomenological approach. It avoids explicit use of physical images oF models such as atoms and molecules. Instead it deals with relations between somewhat abstract quantities such as energy, entropy, free energy and so forth. Admittedly it does not give intuitive pictures as atomic theories do, which is one of the reasons why students find it difficult to gain sufficient understanding and familiarity to use thermodynamics in real problems. But the simplicity of the logic of thermodynamics sometimes makes us see more clearly into the nature of the basic physics of a given problem from very general principles. This is the great advantage of a phenomenological approach, Obviously, however, it is impossible to explore more deeply the ung attention to thermodynamics. Such progress is made quantum mechanics and statistical mechanics. Statistical me§ > Ye REGISTERED ON VERSION ADDS NO WATERMARK & Ss < Tint-drweus with a means to link the physical laws of the microscopic world to those of the macroscopic world, Without close cooperation with statistical mecha- nies, quantum mechanics itself would not be able to represent the physics of the real world. In this sense, statistical mechanics is indispensable as one of the keystones of modern physics. Like any other science, statistical mechanics cannot be mastered easily just by learning its principles once. One has to think by oneself a great deal before ‘one grasps the way in which to use the statistical approach in one’s thin- king, and to apply statistical mechanics to real physical problems, In statis- tical mechanics and in thermodynamics, there are certain aspects which are quite different from other fields of physics. We often meet students who find difficulty in mastering thermodynamics or statistical mechanies, lacking confidence in applying it to real problems, although they know the prin- ciples. Such difficulties are due to insufficient and inadequate training. The purpose of the present book isto provide a guide for students studying and acquiring facility in thermodynamics and statistical mechanics. Thus it contains fundamental topics, examples and a fairly large number of problems with complete solutions. The fundamental topics are rather condensed, but still they cover all of the points which are basic. This bookis meant to be read- able without reference to other textbooks. By reading through these topics only, one would be able to obtain fundamental knowledge of thermo- dynamics and statistical mechanics. The examples are partly to supplement the fundamental topics, but they are primarily meant to show the reader how the principles are applied to physical problems. The problems are classified into three groups, A, B and C, in order of in- creasingdifficulty. [fa readerhasenoughtime he may gothroughall problemsin each chapter. But, if not, it is recommended that he studies first the problems in group A throughout the whole book and then later comes back to try Band C. By just finishing group A problems, he will ind himself to have obtained @ much better understanding of physics. The number of group A problems is fairly large, so that he may even select about half of these and come back later to the other half. The subjects in the fundamental topics an . which are marked by * are not needed in solving problems it + The reader is reminded that this text is a translation o Japanese edition, in which thermodynamics and statistical fone volume, VERSION ADDS NO WATERMARK & < Tint-drwewhich apply to non-equilibrium problems. We had, however, to content our- selves in treating such topics in a limited way only in the last chapter (Chapter 6 of the present English edition). This is because the whole volume had become much larger than the original plan and also because such non- equilibrium problems are certainly somewhat advanced. ‘As mentioned previously in this preface, quantum mechanics is the funda- ‘mental dynamies of the microscopic world. In this sense, statistical mechanics ought to be essentially quantum-statistics. However, since the present book devotes itself to clear understanding of the nature of statistical considera- tions, only an elementary knowledge of quantum mechanics is required in studying problems in groups A and B. Therefore, even those students who are not specializing in physics but have only an elementary background in quantum mechanics will not find any serious difficulty in starting to study this book. ‘What is most important in studying a physical problem is to grasp it as a problem in physics. Mathematical manipulations may sometimes be tedious and sometimes may require specialized techniques. Training in mathematical methods should not be ignored, but it would be a serious mistake if one was to be dazzled by the mathematics and to forget the physics. Teachers often ‘meet students’ papers in which the student seems to be in no doubt about the numerical answers although they are in error by two or three orders of magnitudes or are dimensionally incorrect. Professor H. Nagaoka (a pioneer physicist in Japan) was carrying out calculations on a blackboard in his class. He changed the sign of his answer saying “It is plus rather than minus. Isn't it?” Mathematical calculations may very often be in error. A physical mind is very important, for this can give you the right sign even when your calculation betrays you. An answer obtained by calculation is in many cases easily understood, at least qualitatively. It may not be guessed before making, calculations, but one should not forget to think it over again in order to see if one can see some physical meaning contained in it. Such remarks are not given in each solution of the problems, so that we should like to empha- size here the importance of such reasoning. Here and there between the pages some comments t are ingart rites te po ten 0 chat om te aa ge coffe or ust smoking [= REGISTERED VERSION ADDS NO WATERMARK & + These are revised in this English edition. < eint-driNgThe fundamental topics were mostly written by R. K. Examples and prob- lems were selected after repeated discussion by all the authors. The final check of the solutions was made by R. K. and the whole design of the book was made by N. H. The authors would appreciate it if readers would kindly point out any mistakes which may have escaped our notice. Five years have passed since this book was originally planned, and two years since we started actually to write it. The undertaking proved to be ‘much more difficult than we anticipated. The authors are particularly grate- ful to Mr. K. Endo, editor of Shokabo Publishing Company, for his continual encouragement and help. January, 1961 Ryoco Kuso erereD us 3 és REGISTERED ©) VERSION ADDS NOCONTENTS CHAPTER 1, PRINCIPLES OF STATISTICAL MECHANICS Fundamental Topics § 1.1. Microscopic States 1 § 1.2. Statistical Treatment . Geo O § 1.3. The Principle of Equal Weight and the Microcanonical Ensemble 3 § 1.4. The Thermodynamic Weight of a Macroscopic State and Entropy ‘i . a7 § 1.5. Number of States and the Density of States 8 § 1.6. Normal Systems in Statistical Thermodynamics 10 § 1.7. Contact between Two Systems . . ed § 1.8. Quasi-Statie Adiabatic Process 4 § 1.9. Equilibrium between Two Systems in Contact Is § 1.10. Fundamental Laws of Thermodynamics 18 § LIL, The Most Probable State and Fluctuations. 19 § 1.12. Canonical Distributions . 2 4 § 1.13. Generalized Canonical Distributions 2B § 1.14. Partition Functions and Thermodynamic Functions. 25 § 1.15. Fermi-, Bose-, and Boltzmann- Statistics 28 § 1.16. Generalized Entropy. : - i Examples . 2.2 Problems 56 Solutions 64 CHAPTER 2. APPLICATIONS OF THE CANONICAL DISTRIBUTION Fundamental Topics § 2.1. General Properties of the Partition Function Z() + 102 § 2.2% Asymptotic Evaluations for Large Systems Thermodynamic Functions §2.4%, Grand Partition Function (2). és REGISTERED VERSION ADDS NO WATERMARK & < efint-arwe§25% §26. Partition Functions for Generalized Canonical Distributions . 109 Classical Configurational Partition Functions. §2.7. Density Matrices Examples. Problems Solutions CHAPTER 3, Fundamental Topics 109 110 112 130 139 STATISTICAL THERMODYNAMICS OF GASES §3.1. Partition Functions of Ideal Gases. 183 $3.2. Internal Degrees of Freedom and Internal Partition Functions, 184 $3.3. Mixtures of Ideal Gases 188 §34. Molecular Interactions 189 §35. Cluster Expansion 190 Examples 193 Problems 199 Solutions 20s CHAPTER 4, APPLICATIONS OF FERMI- AND BOSE- STATISTICS Fundamental Topies § 4.1. Fundamental Formulae of Fermi-Statisties 228 § 42. Fermi Distribution Function 29 § 43. Electronic Energy Bands in Crystals 233 $44. Holes 24 §4.5. Semiconductors 235 $4.6. Bose-Statistcs. Liquid Helium 238 Examples: 240 Problems 251 Solutions 259 CHAPTER 5, STRONGLY INT Fundamental Topics gS § 5.2. $5.3. 55.4. $5.5. $5.6. Molecular Field Approximation Bragg-Williams Approximation Co-operative Phenomena : Average Potential in Charged Particle Systquas Debye-Hiickel Theory Distribution Functions in a Particle Syst RACTING SYSTEMS ADDS NO WATERMARK & o ‘2 tint-dtweExamples Problems... Solutions 313 +. 322 329 CHAPTER 6. FLUCTUATIONS AND KINETIC THEORIES Fundamental Topics § 6.1. Fluctuations § 6.2. Collision Frequency §6.3, Boltzmann Transport Equation. Examples Problems Solutions Index . 361 362 363 + 366 315 - 381 = 418 ERED |e S eS &j S REGIS 6 JS REGISTERED ©) VERSION ADDS NO WA TERMARK &, & : re >CHapTeR 1 PRINCIPLES OF STATISTICAL MECHANICS ‘Thermodynamics is a phenomenological theory based upon a few funda- mental Jaws derived from empirical facts. In contrast to this, statistical ‘mechaniesaimsto providea deductive method which leads us from the micro- scopic physical world to the macroscopic world starting from the atomic or ‘molecular structure of matter and the fundamental dynamical principles of the atomic world and combining with these the logic of probability theory. It answers the questions what are the physical laws of the microscopic world behind the thermodynamic laws, how the thermodynamics can be “explained” from such laws and why a specific physical system exhibits such thermo- dynamic characteristics. The fundamental principles of statistical mechanics involve, in fact, very profound and difficult questions if one meditates upon them, but it would not be very wise for the beginners to be too much con- cerned with such questions. The most important thing is to learn how one thinks in statistical mechanics and how one applies statistical considerations to physical problems. Fundamental Topics 5.1... MICROSCOPIC STATES ‘Microscopie and macroscopic states: A physical system which one observes usually consists of a great number of atoms or molecules and so has an enormously large number of degrees of dynamical freedom. But in the usual case, only a few physical quantities, say the temperature, the pressure and the density, ae measured, by means of which the “state” of the system is specie fied. A state defined in this crude manner is called a macrascopic state (example: a thermodynamic state). On the other hand, from a dynamical point of view, each state of a system can be defined, at least in principle, as precisely as possible by specifying all ofthe dynamical variables ofthe sys Such a state is called a microscopic state. Classical statistical mechanics and quantum statistical VERSION ADDS NO WATERMARK & < efint-arwe2 [PRINCIPLES OF STATISTICAL MECHANICS (cast and so classical statistical mechanics may be said to be useful only as a certain approximation to quantum statistical mechanics. But the classical theory has even today a great value from theoretical and educational points of view because it makes us understand more clearly the basic ways of thinking in statistical mechanics. Classical phase space: Let (41, das.» - 4;) be the generalized coordinates of a system with f degrees of freedom and (p,, 3... Py) their conjugate ‘momenta. A microscopic state ofthe system is defined by specifying the values Of (Gus das ++ Gyo PrsPrs +» Py)s The 2f-dimensional space constructed from these 2 f variables as the coordinates is the phase space of the system. Each point in the phase space (phase point) corresponds to a microscopic stat. ‘Therefore the microscopic states in classical statistical mechanics make a continuous set of points in phase space. If the Hamiltonian of the system is denoted by 28 (a, p), the motion of the system is determined by the canonical equation of motion ae an Bem ae f) an aay ee, apne Fig 1 This determines the motion of the phase point, P,, defining the state of the system at time 1. This motion of P, will be called the natural motion in the phase space. The trajectory of the phase point occurring during natural motion is called a phase orbit. For a conservative system, the energy is constant, i. (gq, p) = E. specified simultaneously (the uncertainty principl classical phase space loses its rigorous meaning. mechanics, a microscopic state is a state defined, ADDS NO WATERMARK & < efint-arwech .$2), STATISTICAL TREATMENT 3 sense, In particular, a stationary dynamical state of a system must be one of the quantum states determined by the equation. 9, =F, (I= 1,2.) (13) Here fis the Hamiltonian of the system, £, the energy of the quantum state Tand 9, is the wave function representing the quantum state / ‘The set of microscopic states in quantum statistical mechanics is thus a discrete denumerable set of quantum states denoted by the quantum number 1. (Un statistical mechanics, one usually considers a system confined in a limited space, so that the quantum number /is usually discrete. A system with infinite extension is considered as the limit of one of finite extension.) § 12. STATISTICAL TREATMENT ‘Whenever a system is kept in equilibrium and remains constant according to ‘macroscopic observations, it never stays constant from the microscopic point of view, and so one can never say precisely in which microscopic state the system is found. One can only define the probability for the set ofall possible microscopic states of the system. Fundamental assumption for observed values of physical quantities: Suppose ‘a physical quantity A is observed for the system under consideration. A is a dynamical quantity from the microscopic point of view and is a function of microscopic states. The microscopic value of A is represented by A(q,p) = A(P) in classical mechanics (P is a phase point) and by the expec- tation valuet Ai= Joi Agide = C1 AID a4) in the quantum state /in quantum mechanics. The observed value 4p, in the macroscopic sense must be a certain average of microscopic A: i. Aone = 4. ) Realization probability of a microscopic state: Let 3R be the set of all possi- ble microscopic states which can be realized by the system undergs cenfit + The integration in the following expression is carried outggadm C2 . Oe Jaa pent the ve funciona gv Hoc too SCS TERED ), Space of ewe variables Note that a quantum stat coreespongMlagpase oP incassteat esac ands. conepones ashe eres then MST DE. EPICA ADDS NO4 PRINCIPLES OF STATISTICAL MECHANICS [en 1,92 probability that one of the microscopic states in the volume element AT of hase space is realized: PrAD) =f f(P)AP, (ATER) (1.62)+ or the probability that the quantum state /is realized PD=f(), (le (1.66) that is, by giving the probability density f(P) = f(g, p) ot the probability ‘S(O.F(P) and f () are sometimes called simply the distribution functions. tt ‘When the distribution functions are given, the average value (1.5) is explicitly written as Ao =A APS (PIAL, (1.72) A=LDASO (1.76) Statistical ensembles: In order to make the probabilistic idea as clear as possible, let us consider an hypothetical ensemble consisting of a great ‘number of systems each of which has the same structure as the system under observation, and assume that the probability that a system arbitrarily chosen from this ensemble is found to be in a particular microscopic state is siven by (1.6a) or (1.66). For this hypothetical ensemble, (1.5) may be written as Asay = ensemble average of A= 4. (1.8) A statistical ensemble is defined by the distribution function which character izes it. The most fundamental ensemble is the micro-canonical ensemble to be discussed later, but many other ensemblescan be considered corresponding, to various physical conditions (see (1.12) and (1.13). Ideal gas ~ P space and u-space: So far the whole system in question is considered as the object of statistical treatment. This is the general stand- point of statistical mechanics established, in particular, by Gibbs. If the to take each molecule as.a statistical unit and regard the gas agg regia ee lS REGISTERED VERSION ADDS NO WATERMARK & 41 Inmathematial probability theory, a dstbution function § ponee ~ Fon in the one-dimensional ease, for example. The te mechanics is usually used in'a loose way, < efint-arwech.1,93) PRINCIPLE OF EQUAL WHIOHTT 5 consisting of such units. Tis point of view was taken in the kinetic theory of gases which became the prototype of statistical mechanics, From this stand- Point, the important thing is, in classical statistical mechanics, the istibue tion function of the position x and the momentum p of a molecule, i, the probability that a molecule chosen from the ensemble of gas molecules is found to have the coordinate and momentum values between x and x + dx, pand p + dp is equal to Se pas dp as) Most of the properties of dilute gases can be derived from a knowledge of this distribution function. Ths isa distribution ina six-dimensional space, which is often called the sespace, The phase space of the N molecules ofthe gas is called the I*space. Maxwell distribution: In a thermal equilibrium state at high temperatures, the distribution function ffor a dilute gas is given by 1 Gamer where Tis the absolute temperature, m the mass of a molecule, and k the Boltzmann constant. This Maxwell distribution can be derived by various ‘methods. The most general derivation will be described later§ 1.15, e4.(I.100). S(P) 1 2 2 2) §13. THE PRINCIPLE OF EQUAL WEIGHT AND THE MICROCANONICAL ENSEMBLE When a system consisting of a great number of particles (more generally a system having a great number of degrees of freedom) is isolated for a long time from its environment, it will finally reach a thermal equilibrium state. In this case, the energy of the system is constant, so that it is presumed to be fixed at the value £ with a certain allowance SE. This is the prescribed macroscopic condition. The set IR(E, 5E) of the microscopic states to be considered under such conditions ti classically: the shellike subspace of the phase space between the tye constant-energy surfaces for 2 = Eand.#” = E + 6E; and quantum-mechanically: the set of quantum states haya eigenvalues in the interval E < E, < E + dE. S uy gota eS The principle of equal weight: In a thermal equilibriut Si TERED © SCORERS ZN VERSION ADDS NO WATERMARK & + Under certain circumstances, other constants of motion s tum or the total angular momentum may be preserited. In 4 restricted, < efint-arwe6 PRINCIPLES OF STATISHICAL MECHANICS [on g3 system, each of the microscopic states belonging to the set IR(E, 5E) is realized with equal probability, namely (classically) f(P) = constant [scrtesee S| + PeaWEaE) (L118) (quantum-mechanically) Sil) = constant [e2eeuul] , TARRE,SE). (1.118) The mierocanonical ensemble: A statistical ensemble defined by the principle of equal weight, or more precisely by the probability distribution given by (11a) of (1.11b) is called a microcanonical ensemble, and the distribution is called a microcanontcal distribution. A merocanonical ensemble thus rep- resents an isolated system which has reached thermal equilibrium. ‘Classical limit (SE -» 0): Using classical considerations, one may goto the Limit DE ~ 0 and take the set ¢(£) on the surface of constant energy in- stead of IRE, 6E), Then one has, instead of (1.68) and (1.118) Prido) = { f(P)do, — Peo(E) (1.12a) . lo {de SPIO = | a LT ieee (1.126) were dei surface element on the consan nergy surface, and 2 vaeynt an woo -[3{(5) Ga) 7) * Vea, Equation (11a) becomes ata.poe | Terad% " a ye RoroN ADDS NO WATERMARK & o ‘2 tint-dtwecus) TE THERMODYNAMIC WEIGHT AND ENTROPY 1 it appears possible to justify the principle of equal weight as Aggy = 8 long time average of A, = the phase average (1.13) of 4(P), (4) ‘The second equality of the above equation, vine avenge = Apiase evra: als) is called the ergodic theorem. This theorem has been studied as a mathe- ‘matical problem. But there exist various arguments about its physical signifi- cance 4s to whether it really provides convincing grounds for the principle of equal weight. The finite allowance of the energy BE: In quantum mechanics, the energy of a system has an uncertainty E)yy ~ hit (16) determined by the length of the time of observation, ¢. Here h is the Planck constant, Therefore one has to choose dE > (Eq GE may be small, but in this range there exist a great number of quantum states if the system has a macroscopic size so that statistical considerations become possible. (If the system stays certainly in one quantum state, there would be no need of a statistical treatment.) Note: The existence of the allowance dE is necessary. But its ambiguity may cause concern for some readers. It can be seen, however, that the magnitude of dE does not affect thermodynamic properties in macroscopic systems (see § 1.6, (1.27) $1.4, THE THERMODYNAMIC WEIGHT OF A MACROSCOPIC STATE AND ENTROPY Variables defining a macroscopic state: As the variables defining a macro- scopie state of a system under consideration, one may choose the energy E (with an allowance 5£), the numbers Ng, Ng --. of particles of various kinds existing in the system, the volume V of the box which contains the syste and other parameters x, ... tospecify external forces, such asthe el W(E, bE, N, V, x) of the possible quantum states values of the variables (SE), Ny Na -- Vix, ADDS NO WATERMARK & o ‘2 tint-dtwe5 PRINCIPLES OF STATISTICAL MECHANICS (cass (quantum-mechanically) WEE,SE,N, V,x) yt (1.17) ‘The thermodynamical weight in classical statistical mechanics should be defined as the limit of quantum statistical mechanics. It is given by (classically): ar, [SOR NON ANGI W(E,SE.N, Vx) = (1.176) ‘The denominator dividing the volume element dof phase space is derived from the correspondence of the classical phase space and quantum states (ee § 1.5). ‘Statistical definition of entropy: The entcopy defined by the Boltzmann relation SCE, Ng Novos Vox) = blog W(E,SEN, Vs) (1.18) is called the statistical entropy. Here is the Boltzmann constant. One has to show the well-known thermodynamic equation ASUE,N gy Nayons Vox) = Z(GE + pd = EX dx = Sp AND (1.19) in order to convince oneself that the statistical entropy is identical with the ‘thermodynamic entropy. This will be seen later. Note: The allowance dE of the energy does not affect the numerical values of the entropy 5 (§ 1.6). § 15, NUMBER OF STATES AND THE DENSITY OF STATES. Number of states: The zero of the energy E is now so chosen that the ‘energy levels of the system under consideration are non-negative: OSE, SE:SEy ‘The number of quantum states with energy values between O.and E{aienfi™ QENV I= Td Classically this is redefined by 2(E,N, V2) * VERSION ADDS NO WA TERMARK & 6 ‘2 tint-dtweca85) [NUMBER OF STATES AND DENSITY OF STATES ° State density: a (E,W Vax) J OWEN, V2) 21 is called the state density of the system. If 6E is small enough, one may write QUE,N, Vx) 5E = W(E,SE,N,V,x) (1.22) Classical number of states (Phase integral) (Correspondence of classical and quantum mechanics): Quantum mechanics reduces toclassical mechanics in the limit of h - 0, so that (1.20) must be derived from (1.20a) by ths lim iting process. A general proof of ths is somewhat too advanced to be given here (see problem 33, Chapter 2). Therefore only some explanation of (1.206) will be given in the following. There are two elements in the cor- respondence of classical and quantum mechanics, which give rise to the two factors appearing in (1.206). (1) Quantization of phase space: For a system with f degrees of freedom, the set of microscopic states held in a volume element A/"corresponds, in the limit h + 0 (not considering another factor to be stated in (2)), to a set of arin (1.23) ‘quantum states (as long as h remains finite, this correspondence is an approxi- mation). This can be understood by means of the uncertainty principle Apag~ k, since the classical states in a cell of h per degree of freedom, or M per f degrees of freedom, merge into a single quantum state which cannot be further distinguished (see problems 7 and 8). Q) Indistinguishability of identical particles: In quantum mechanics, ‘identical particle are indistinguishable in principle. For instance, the state in Which two identical particles have coordinates(p’,x’)and (p", x") respectively is distinguished classically from another state in which the second particle has coordinates (p',x') and the first particle has coordinates (p", ver, these are simply the same single state from a quantum-mechs in equation (1.206). For example, the expression Lf, [ Seti nel M i Ss REGISTERED SD VERSION ADDS NO WATERMARK & ase < efint-arwe10 PRINCIPLES OF STATISTICAL MECHANICS (ch. 1,56 gives an approximation to the number of states below E for a system con- sisting of N identical particles. The indistinguishability of identical particles introduces the denominator N! in the above expression because the N! classical states t arising from a given phase point p,, ty, ... pss ty must be identified with each other by this principle (see the Note to Chapter 2, problem 33 for a more rigorous discussion). Nore: The denominator V! was very difficult to understand before the principle of the indistinguishability of identical particles was introduced into quantum mechanics. In spite of thi, the necessity for this denominator term hhad long been recognized in order to make the entropy defined by (1.18) an extensive quantity as it should be. 516. NORMAL S' STEMS IN STATISTICAL THERMODYNAMICS, Asymptotic forms of the number of states and state density of a macroscopic system: A system consisting of a great number of particles, or of a system with an indefinite number of particles but with a volume of macroscopic extension usually has a number of states 2q(E) which shows the following properties (in which case the system will be called normal in the statistical- thermodynamic sense): (1) When the number V of particles (or the volume ¥)is large, the number of states £29(E) approaches asymptotically to a-wofra(s}} a onfr(f)}- cam avseofxa(EOt oe enfew(®8)h caw If EJN (or E]V)is looked upon as a quantity of the order of O(1) +t, 4 is also O(1) (the same holds for y), and g>0, g>0, go" <0. (4.25) (2) Therefore Q = dQa/dE = $ exp(Nd) >0, (yt t oe ges PSXBRED >. Ale number of & REGISTERED © “* VERSION ADDS NO WATERMARK & 4+ When some of ps, (ps, x8). (py, x3) coincide with e classical states produced by the permutation of particle stagagies chance for such coincidence is negligible in thelist of h > Tf One writes y = (8) and s = o(x) if lim yx = finite o ‘2 tint-dtwe©m.1,871 [NORMAL SYSTEMS IN STATISTICAL THERMODYNAMICS. " When N (or V) is large, 25 oF @ increases very rapidly with energy E. No general proof of these properties will be attempted here. Ifa system existed which did not have these properties, it would show a rather strange macro- scopic behavior, very different from ordinary thermodynamic systems (see example 4, Chapter 1). Entropy of a normal system: For the statistical entropy defined by (1.18), one finds the following from (1.24)-(1.26): o S = klog{Q(E}5E} ~ klog Qy(E) = KN. (1.27) The error involved here is o(N) (or o(V/)), and so is negligible for a macro- scopic system (for which N, V, or E is very large). 2) The statistical temperature T(E) is introduced by means of the defini- tion, 1 ea (1.28) T(E) =, >0. (1.29) By (1.24) and (1.25) it will be showa later that this temperature in fact agrees with the thermodynamic temperature (see § 1.9) The allowance of the energy and the definition of entropy: By (I.24)-(1.26), the function 2 (£) is positive and increases monotonically with E. Therefore one has Q(E)SE < 24(8) < EVE, thus $= klogQ(E)SE < klogQo(E) < klogQXE)E. ‘Also by (1.24) and (1.25) and using the fact that £ = O(N), one finds: k (log Q(E)E — log0,(E)} = klog £4’ = O(logN) = o(N) (or o(¥)) and K flog Q(E)E ~ log Q(E)SE} = klog E/é Therefore (1.27) is seen to be vali =0(N)t (or o(¥). §17. CONTACT BETWEEN TWO SYSTEMS the uncertainty prinsple(I16)the time ofthe observation iy It's — O(), this 1 astronomically long for a macroscorh fof ordinary length, 3E cannot be so small and thus one migi#hdee (wamely x = of). fib ADDS NO WATERMARK & o ‘2 tint-dtwe2 PRINCIPLES OF STATISTICAL MECHANICS towns? In phenomenologial thermodynamics, these intacion are idealized as thermrynamic contacts, Le, mechani, hemal or materi-ransterring contact Instat mecha, eorespndingly on comer the follow: ing pes of eontact (1) Mechta contact with a work source: I the ouside word (environ men) isconected simply asa sure which exerts afore on the ytem unde consideration, hs mechanical or electromagnet) eft canbe represented by the Hamiltonian (g, p, x), in which x is the coordinate (for instance, the Position of a piston in a box containing a gas), to describe the interaction between the system and the environment and is regarded as a variable of the tem, Then wane) te 4.30) represents the force which the system exerts on the outside world, Q) Thermal contact between two systems: When two systems with Hamil- tonians and 2 are in contact and are interacting with the interaction Hamiltonian”, then the (otal Hamiltonian for the composite system 1 + is written as ign SIE to A a3t) The two systems are said to be in thermal contact if the interaction 2” satisfies the following two conditions 4a) 27 is so small (the interaction is so weak) that each of the microscopic states of the composite system 1 + 1, say J is specified by giving the micro- scopic states ' and !" ofthe subsystems 1 and 1, and the energy E, istoa good approximation the sum of the energies of the subsystems, namely IAQ), Bp Bb + eR (132) ) On the other hand, the existence of #” allows a sufficiently frequent exchange of energy between the systems 1 and 11. As a result, after waiting ong enough one can certainly expect that the composite system + ttreaches a final state, whatever the initial state was. In this final state eve scopic state (I, !") of the composite system is realized with in accordance with the principle of equal weight. This final s statistical equilibrium of the two systems and cor ADDS NO WATERMARK & < efint-arwecn.871 CONTACT BETWEEN TWO SYSTEMS, 1B still the microscopic states of the composite system can be represented by (WDEWTLNT), EN) = ELOY) FEM) 3t to a good approximation (namely the interaction is weak enough), then the interaction is called a material-transferring contact. ‘DIvERTISSEMENT 1 ‘Maxwell's demon, Entropy never decreases spontaneously in an isolated system. Who in the world has ever seen water in a kettle boiling by itself taking heat from a block of ice on which the kettle has been placed? Who thas ever seen two boxes filed with gas at the same temperature and pres sure spontaneously producing an unbalance of temperature, one be- coming automatically heated up and the other cooled down, when they fare connected by an open window on the wall between the boxes? Of course, nobody. Wouldn't it be possible, however, to find an extremely clever creature who stands by the window watching gas molecules coming. byandwho opens the window only when hot molecules, namely those mole- cules having kinetic energy greater than the average, approach the window from one side or when cold molecules come from the opposite side? Maxwell imagined such a clever demon. ‘The second law of thermodynamics denies the existence of Maxwell's demon. You may be able to find a demon who starts to do an extremely fine job of selecting molecules passing through the window. But he will never be able to continue the work indefinitely. Soon he will become dazzled and get sick and lose his control. Then the whole system, the gas ‘molecules and the demon himself, will again approach a final equilibrium, where the temperature difference the demon once succeeded in building up disappears and the demon will run a fever at a temperature equal to that of the gas. A living organism may look like a Maxwell's demon, but itis ‘not. A living organism is an open system, through which material, energy, ‘and entropy are flowing. But life itself cannot violate thermodynamic aws. (4) Pressure-transmitting contact: When two systems are separated by a movable wall, then an exchange of volume is possible between the two systems. If the wall allows only volume exchange but not energy or particle exchange, this is an example of a purely mechanical contact, but it may Wi D=(VIVT), BV) = BV) + EE is possible. respectively "VERSION ADDS NO WATERMARK & < eint-driNg4 PRINCIPLES OF StATIStICAL MECHANICS [on 1.98 § 18, QUASLSTATIC ADIABATIC PROCESS Quasi-static adiabatic process in statistical mechanics +: This is a process involving a very slow change of a parameter x which determines a purely mechanical interaction of a system with an external work source, namely: (4, p.x) + (qs P.x + Ax), dxjdt +0 (1.35) Adiabatic theorem in dynamics: Dynamical quantities which are kept in- variant in a (quasi-static) adiabatic process are called adiabatic invariants. It can be shown, in particular, that the number of states 24(E) is an adiabatic invariant. Namely, QfE,x) = O(E + dE,x + dx), (dxjdt+0), (1.36) where dE is the increment of energy of the system in the process involving an adiabatic increase dx of x. For this, it holds that GE = CX ame wnerne dx = PAX 5 37) X here means the phase average of X, or the average of X over a micro- canonical ensemble Adiabatic theorem in statistical mechanics: It isa rather debatable question whether the adiabatic theorem in pure dynamics retains its original signifi- cance for systems with enormously large number of degrees of freedom such as treated in statistical mechanics, but (1.36) and (1.37) will be assumed valid here without examining any detailed arguments about this question (sce problem 34). Therefore one has, from (1.36) and (1.37), = 0, (dxiét +0) 138) for a quasi-static adiabatic process. In other words, as =e + , dE = Rdx (1.39) for an adiabatic reversible process. This equation gives the formuly r-(% Ss REGISTERED 2 VERSION ADDS NO WA TERMARK &, + In mechanics, this is often called simply an adiabatic pga < efint-arwe89) EQUILIBRIUM BETWEEN TWO SYSTEMS IN CONTACT 1s for the statistical mechanical force or the average of X. Here Tis the statistical temperature (1.28). Also one has as Ode Fen) (1.82) § 19. EQUILIBRIUM BETWEEN TWO SYSTEMS IN CONTACT Partition of energy between two systems in thermal contact: The principle of equal weight (see § 1.3) can be applied to the composite system + 11 rep= resented by (I-32), whereby one obtains the probability that the subsystems and have the energies and Ey,(E, + Ey = E) respectively. Let the state densities of 1 and 11 be denoted by 2, and 2, respectively and that of t + 11 by @ then one has QEWE= ff OE NA (En) AEE, (1.43) E J (EYE ~ E) dE, ‘Therefore the probability that the system 1 has energy in the range between E,and E, + dE; is given by QUE NQ(E — E)AE SE NESE (SENSE =. (44) J(E) GE, = < %5 y FSi REGISTERED © rea Z VERSION ADDS NO WA TERMARK & The most probable partition of energy: (1.44) gives partition when 1 +11 are in statistical equilibriugy 6 oy tint-dtwe16 PRINCIPLES OF STATISTICAL MECHANICS [cn.1,$9 ‘The function 2,(E,) increases very rapidly with E, (see (1.26)) and 2,,(E —E,) decreases very rapidly with £;, so that this probability has a very steep maximum ata certain partition (E, Eq = E ~ E}) (see Fig. 1.2). This is the partition to be expected almost certainly in the equilibrium state of 1 + 1 This is determined by the condition 2(E)Q(E — E) AEE = max. 4) which is, by (127), equivalent © SE) + SE — E) = max. (1.46a) orto eC (1.460) This may be also expreted at TKEn wEn, EY+ En (L471) (ee (1.28). Equilibrium of two systems with a material-transferring contact: When pat- ticle exchange is made possible between two systems 1 and 1 through materiabtransferring contact, the probability of partition (Ny, Ny) of a certain kind of particles (with the total number N) is derived in the same way as (1:44) and is given by “f(6, Nab = MEPNOMAE ~ EN ~ NDSEBE ~~ QE NE ‘hich represents the probability thatthe subsystem 1has an energy E, and N; particles, In this expression QUE, NE = 5E Y [A(EpN)Q(E — EyN — NYSE (1.49) (1.48) is the thermodynamic weight for the composite system 1 + 1, The quantities Q(B N,) dE, and 2y (Eq, Nu) Ey refer, respectively, to system 1 having Ey and N; and to system 1 having Ey and Nj, the energy allowance being and dy respectively. The most probable partition is obtained inf way as (1.45), (1.46) and (1.47) from the conditions S(EwN) + Su(Ew: EytEn=E, Ni+Nu=¥ )= max, 2. REGISTERED VERSION ADDS NO WATERMARK & or by the conditions that the temperatures of the subs (1.47), and that < efint-arwech 1,89), [2QUILIDRIUM BETWEEN TWO SYSTEMS IN CONTACT 7 Ma 7 bnN, Na+ Ni (st) (1.52) Equilibrium of two systems exerting pressure upon each other: When two systems are in contact through a movable wall, so that an exchange of volume is made possible (§ 1.7,(4)), the condition to determine the most probable partition of volumes is given by PEN MD: PUNY), Wt Kie¥ (153) where p is defined by é P WEN =F (1.54) This is derived in the same way as before. (If heat exchange is possible, then ‘one has also the condition (1.47),) Additivity of entropy for systems in equilibrium: If the subsystems | and 11 are normal in a statistical-thermodynamic sense, the additivity of entropy holds for the composite system 1 + 1, namely, SpenlE, N,V) = S(E3, Nin Vi) + Su(Biu Nin Vad (4.55) where Ei, etc., represent the most probable partitions (all quantities which are not allowed to be exchanged by virtue of the nature of the contact are fixed). Equation (1.55) means thatthe entropy of the composite system is the sum of the entropies of the subsystems (example 5). Increase of entropy by establishment of new equilibrium: Two separate systems 1 and 1 ae given with the partition (E?, N?, V9; E%, N&, V8) which is different from (Ej, Nj, Vj; Ein, Ni, Va). Then the two are brought ingg the entropy of the total system becomes Sion(.N.¥) (Ei, Nis Vid+ SulEnw Niv Vad 2 S(ET, NP, VE)+ Su(Ety NB, Via) S <> REGISTERED ein eR SION ADDS NO WATERMARK & If the systems 1 and 1 are separated again, they will o ‘2 tint-dtwe18 PRINCIPLES OF STATISTICAL MECHANICS [cn $10 partition (Ej, Nj, Vii Bit, Ni. Vg). Therefore one has Srna Vi) + Su(Eie Ni Vi) 2 Siac (1.56) In other words, the entropy has been increased as a result of bringing t and into contact, and remains unchanged only when 1 and were originally in equilibrium with each other (namely the initial partition is equal to the most. probable partition after contact). When the two systems are separated after contact, the entropy of the two systems is almost certainly found to be greater than before contact, and they are almost certainly in equilibrium, § 110, FUNDAMENTAL LAWS OF THERMODYNAMICS. ‘The first law of thermodynamics is, so to say, a self-evident consequence of the dynamical laws which satisfy the conservation of energy. The internal energy U is nothing but the energy £ of the system, ifthe mechanical energy is not included, and is therefore a function of the state. Entropy: By the definitions (1.27), (1.28), (1.52) and (1.54), the entropy S satisfies the relation 1 dS = He + paV ~ Rex — ud) asy which gives the difference of entropy in the two neighboring equilibrium states (E, V, x, N) and (E + dE, V + d¥,x + dx, N + dN). When the system suffers a change between these two states in a quasi-static process, (1.57) can be written as dE = — pdV + Rdx + pan +4'Q, (1.58) with dQ=Tas (1.59) which represents the heat introduced into the system from the outside in the course of the process (because the first two terms represent the mechanical work given to the system from the outside, the third term is the energy it inn S REGISTEREDchem) [FUNDAMENTAL LAWS OF THERMODYNAMICS 9 first is a characteristic general property for the temperature and the second agrees with the definition of absolute temperature as introduced by the second law of thermodynamics. The second law of thermodynamics: This law is interpreted by a probabilistic concept in statistical mechanics. When a certain inhibition (which, for in- stance, prohibits mutual contact of two systems) is removed and two systems are brought into contact, the entropy of the whole system will almost certainly increase (see (1.56°)) Therefore a spontaneous decrease of entropy against the second law of thermodynamics is extremely improbable. For any process in which the system under consideration comes into contact with an external environment, it very certainly holds that dS +dS,20 (1.60) If the external environment isso large that its own change can be considered as quasi-static and its temperature is equal to the temperature of the system, then one may put dS, = ~ 4’Q/T. Therefore the inequality Tas zaQ (Le) will be satisfied almost certainly for the entropy increase dS of the system and the heat d’Q given to the system in the process. The third law of thermodynamics: Entropy is defined by (1.18) in an abso- lute manner, Its always non-negative by definition (WV 2 1), For areal quan- tummechanical system the existence ofthe lowest ground state may be usually- assumed. If the density of the system remains finite, log @(E) will approach, aS E+ 0 (the lowest energy value), a value which is independent of the size, N or V, of the system. This means that the statement 5 +0 (as E = 0) is a consequence of the quantum-mechanical definition (1.18) for real physical systems, It must be noted, however, that this does not necessarily assure that an actual experimental measurement will give S -+ 0.48 -+ 0. It may happen that the lowest state of the system is not attained in the course of the experi T +0. When such freezi value not equal to zero (example: glass). ig occurs, the observed entropy § LIL THE MOST PROBABLE STATE AND FLUCTUAMIISS Probability and entropy: Let 2(24, X25 «».) be thes and V required to specify the macroscopic state of a V, a) be the thermodynamic weight of the state (gy ~dBEGIS TERED 9) 2th WER SION ADDS NO2 PRINCIPLES OF STATISTICAL MECHANICS [usu ability for the realization of the state (E, N, V,a) is, by the principle of equal weight, Pa) x MENV® — geesrane Le = SWE, ¥,a) ~ 7 uo where : S(E.N,V.0) = klog W(E,N, Va) 4.63) is the entropy of the state (E, N, V, 1). ‘Usually the most probable value a” and the average value @ of 2 coincide with each other, because the maximum of the probability P(a) at a = a’ is extremely steep when the system is large. The most probable value a” is determined by the condition that P(a) is maximum or S(E,N,V,a)=max, ama’ (1.64) or ¥, eee _m (1.65) a Determination of mest probable values under constraints: When certain constraints, Poy 2250.)=0, ANZ, (1.66) are imposed on the variables 4, 23, .. gy the most probable values of a's ‘are determined by the condition ae 66) and 542 4,2, (167) 0a, * 5 = where Ay, da, -.« Ap are Lagrange's undetermined multipliers.+ Fluctuations: By (1.62) A) = Ces {SUE,N, Veal +0) — S{EN,V, ») (1.68) ives the probability for a realization of deviations of a’s from their m 5 Forte maximum poi en st ave t= E Sano subject to the conditions: - so F Seon o ‘With + intially undetermined multipliers 2a, 2a... 52 ADDS NO - TERMARK & 6 ‘2 tint-dtweCh. 1,612 ‘CANONICAL DISTRIBUTIONS a probable values. If the a’s are continuous, (1.68) may be written as @s Pe ee rae Pletis as-..) da, dat ~ Cem 5 jersey da a. (1.69) which can be used for deviations which are not too large #112, CANONICAL DISTRIBUTIONS Canonical distribution: When a system with volume V and Nq, Np.» pat= ticles is in equilibrium with a heat bath at temperature 7, the probability of its microscopic states is given by os (1.700) 1 (classically) PAD) = RZ, (quantum-mechanically) Pr(1) 5 (4.708) where B=\JkT, (k= Boltzmannconstant) andof y isthe Hamiltonian of the system, Ey, being the energy of the /-th quan- tum state. This distribution is called a canonical distribution, and an ensemble obviously old Q 2 (es =, On) as+ Ew F (E+E aS) ano ® Now di... dr ae supposed tobe chosen by the conditions as . SF ” oy Then equation) seduced > (as ey) &, ham )m-e. Since m ~ rvariables out ofa, ... on are independent by the constraints (1.66), rss, ‘am can be regarded as independent Variation, so thatthe conditions| as eee = +E ey jerth ‘determined from r ++ m equations (1.66) and (1.67. Generally speaking there exists a weight function (agg depends on the nature of the variables». If they are origi ‘dynamical variables (@,), the explicit form of « can be dork equal weight. Re choice of SREGISTERED J) VERSION ADDS NO WATERMARK & < efint-arwe2 PRINCILES OF SFATISHICAL MECHANICS [ong defined by this distribution is called a canonical ensemble. The system con- sidered here may be a small system with a rather few degrees of freedom, or it may well be a large macroscopic system. Zy in (1.70a, b) is defined by (classically) 2a = pga de ar, (u7tay (quantum-mechanically) y= e*™™" (1b) or more generally by Zy = fe PQUEVGE (76) and is called the partition function Nore: The factor 1/[]Nql#?"* is introduced in (I.7la), as was explained in § 1.5, in order to secure the correct correspondence between classical and quantum mechanics. Derivation of canonical distribution: Let Q(E) be the state density of the heat bath, £, the total energy of the composite system (the system under consideration plus the heat bath), and £, the energy of Lth quantum state of the system (E, = E + E). According to the principle of equal weight, the probability that the quantum state fs realized is proportional to the number of microscopic states allowed, which is equal to Q(E, ~ E)) 6E. Therefore Dc OE, ~ ESE ea FD) © ME, ~ ENE MEE E, ~ E) — StE) pf SEi= BI HE ony Since the heat bath is very large compared to the system considered it may be assumed that £, > E;, and so the exponent may be expanded as StS, | foe t a ae aoe, S(E,— E) ~ S(E) = = where (esjezy* is the temperature of the heat bath and C = cE Because of the assumed large size of the heat bath! Ey < CT & FQ at capacity = TI ° IS REGISTERED ZN VERSION ADDS NO WATERMARK & < efint-arweChg ‘GENERALIZED CANONICAL DISTRIBUTIONS 2B and the second term in the last brackets is so small that it may be ignored. ‘Then (1.72) reduces t0 Sacer, § 1.13. GENERALIZED CANONICAL DISTRIBUTIONS T-u distribution (Grand canonical distribution): When a system enclosed in a volume V is in contact with a heat bath at temperature J and with a particle source characterized by the chemical potentials jig, up, --- for par ticles 4, B, ... , then the number of particles it contains is also indeterminate, The probability that the system has the microscopic state which contains N4, Ng, --- particles is given by eM Dhaee) (Classical) PAN AD) = Fy a — (1.73a) Bb BNana) (quantum mechanical) Pr(N , 1) = — (1.730) ‘where = is called the grand canonical partition function (or may be called the T-u partition function) defined by (classical) “ g 1 ze 5 sD, 2) © to wete TINatH**) onli (quantum-mechanical) Dehetre, (740) or more generally by jo Re “(EE Instead of the chemical potentials, the absolute activities 2y (1.74e) ‘are sometimes used If one wishes only the probability distribution of partich aiven by PrN a Nos VERSION ADDS NO WATERMARK & < efint-arwePy [PRINCIPLES OF STATISTICAL MECHANICS Ich. 613 Derivation of T-u distribution: Let N, be the total number of particles of a iven kind in the system considered and in the particle source, then the probability that the system has a partition of particles and energy (WV, £) becomes PAN, Bc O(N, NE, B) acexpy {S(N,~ NB, —E)~S(M ED}. ‘The exponent may be expanded in the same way as for (1.72) since we may assume that N,> N and E, > E. The higher order terms in the expansion can safely be omitted. Then this reduces to 1.75) with the chemical potentials defined by (1.52) (see example 6) Tp distribution: When a system with given numbers of particles N,, Np, .. isin contact with a bath at the temperature 7 and pressure p through ‘a movable wall, then its volume V is also indeterminate. The probability that the system is found to be in @ microscopic state with a volume V is given by (classically) Pr(aV, ar) = VY a (quantum-mechanically) EWE EY Pr(av, . (1.76b) where 3¢(V) is the Hamiltonian of the system with the volume V specified, E,(V)isits energy quantum energy and Yis the T:p partition function (classically) 1 . vega tam: fav Jerereroren, ay (quantum-mechanically) y= favsermuneen, 7) ‘These may be written as R nen ZO Zy(V)e av The distribution in volume only is given by av Y PHAV) = ZV) < efint-arwe EOUN eS S [> REGISTERED ON VERSION ADDS NO WATERMARK &Ch.1,$14] PARTON FUNCTIONS AND THERMODYRAINC FUNCTIONS 2s ‘This distribution is derived in the same way as the canonical or grand canonical distributions. Nore: The 7-4 distribution and the T:p distribution, like the canonical distribution, can be applied irrespective of the size of the system under consideration. § 1.14. PARTITION FUNCTIONS AND THERMODYNAMIC FUNCTIONS ‘The microcanonical, canonical, 7-u (grand canonical) and T-p distributions are the distributions for given energy (E = constant), given temperature (T= constant), given temperature and chemical potential(T = constant, 1 constant) and given temperature and pressure (T' = constant, p = constant), respectively. If the system is macroscopic, then the thermodynamic function, (potential) for each of these prescribed conditions is derived from each partition function. This is summarized in the following table: Distribution | Patton function Thermodynamic funtion Mircanonzal | m8, ¥,036 | su om) = Ae er mse on BM | chee eN | za,v. | peryany =~ Kron 207, ¥.™9) ‘Canonical | om Zener ier | wOT¥,N) = klog Z(7, VN) SMTQUE, VN) OE Grand canonical Tn) JT. ¥en) = — pV = F-6 (Ta distribution) | = E emrzcr. | AT og ECT, Ys) Eamzenvmy | att Yo = ktog 507, 09 | | Top disciouion | -Y¢T. 2.) G(T», N) =~ ATI08 YT. B®) jo Jermrzcr nmvay | octp.9) = koe ta. ADDS NO WA cai & 7 cy tint-dtwe26 PRINCIPLES OF STATISTICAL MECHANICS [ou gia This will be shown by the following argument for the canonical distribu- tion (see also § 2.3). Consider the quantum-mechanical case: the generalized force X conjugate to a generalized coordinate x, involved in the Hamiltonian 2 (p, 4, 2)is defined by The quantum-mechanical expectation value of this force in the quantum state Lis given by E(x) (79) where the energy £, is regarded as a function of the parameter x.t The aver~ ages of energy and force in the canonical distribution are given by E= 2 ene [ae -5 Hie. x), (1.808) F-Danetfea DF + The eneray Fi(x) is determined by the eigenvalue equation, at = FypleeZ- (1-806) (9.4, 2)01 > Eps Soom = : at Eo 4 it Tis makin by "om hee and cet oer the vai (9 te wae toon forgenees fora ef a nanfohene ere, nett hee the elations 8 SERED 1 for wee Sal 0 oan ferrede = 1 € Inetear- 8 ecea Sf Sood GS TERED are used. VERSION ADDS NO WA TERMARK & o ‘2 tint-dtwe(Ch.1,§14] PARTITION FUNCTIONS AND THERMODYNAMIC FUNCTIONS 2 In particular, the average pressure is given by ai avpiBZY)- (1.800) If one now writes = kT Nog Z = F(T, V,x,.-) (81) equations (1.80a-c) can be written as (1.82) wT) T This shows that F'is nothing but the free energy of the system if , , and p are identified with the thermodynamic energy, force and pressure. ‘The above can be shown in another way. In the integral AT,V,N) = fe" XE, V,N)GE fe =[oo[-1E-se.n.0f]tese, asp the exponential function varies with E extremely rapidly if £ and S (= ‘klog@ - 6E)are of the order Nin magnitude (isthe total number of particles in the system, which is very large), so that an overwhelming contribution to the integral comes from the neighborhood of £” which is the most probable value of £ in the canonical distribution. £" is determined by EIT — min, thus 1/T=0S/0E (E=6'). exp] ~ — {E' [mi where the relations Ey + } (1.8 esioet = 6T"'/0E Y S Seo. Nese m = REGISTERED ON ‘Y!) VERSION ADDS NO WATERMARK & (1.83) is approximated by C)6E~*exp[— kf Ly Z~ Qk < efint-arwe2% PRINCIPLES OF STATISTICAL MECHANICS [ou g1s the logarithm of which gives F(T,V,N) = E— TS(E,V,N) if the free energy Fis introduced by (1.81). The above may be written as, 1 E — = F(T,V.N) = S(E,V.N) - =, 1.85% ph ViN) = SYN) ~ 7 (1.85a) where £”is written simply as E. E on the right hand side of this equation is regarded as a function of F, T, and N by the-relation aS(E, VN) a (1.850) Equations (1.85a) and (1.85b) show that F/T is the Legendre transform of S by which the independent variables are changed from £ to 1/T. Therefore one finds a (F\_(@_1) eB © ain) #)-G- #5 The second and third equations of (1.82) are easily obtained from (1.85), (1.52) and (1.54). $1.15. FERME, BOSE-, AND BOLTZMANNSTATISTICS One-particle states and the states of a system of particles: Let us consider a system consisting of N particles of a certain kind. If the interaction of the articles is weak enough, each particle has its own motion whichisindependent of all others.t The quantum states allowed for this individual motion one- particle states — are determined by the Schrdinger equation 9 (9.2)03) = 0,9,(8) (1.86) and are represented by the wave functions 9,, the «,'s being the energie these quantum states. ‘one-particle approximation works Surprisingly well when itis not discuss such complicated cases here, but it should be proachis al ADDS NO WA TERMARK & o ‘2 tint-dtweCh 1,915) ERE, DOSE-, AND BOLTZMANNSTATISTICS Fd ‘occupying one-particle states is given precisely. That is, the set of occupation rurnbers fr} = (ay 2p opty) (87) gives the quantum numbers of the whole system. The energy of the system is, then given by 1 (1.88) Fermi-statisties(F.D. and Bose-statistics (B.E.) +: The occupation numbers, or the number of particles in each one-particle state, are strongly restricted by a general principle of quantum mechanics.tt There can be only the following two cases: for F. D. (Fermi-statisties) m= Oorl (1.89) for B. E, (Bose-statistics) n= 0, 1,2, . ‘The difference between the two cases is determined by the nature of the par- ticle. Particles which follow Fermi-statistcs are called Fermi-particles (Fer ‘mions) and those which follow Bose-statistics are called Bose-particles (Boson). Electrons(e), positrons(e*), protons(P), and neutrons(N) are Fermi-par- ticles, whereas photons are Bose-partcles. Generally, @ particle consisting of an odd number of Fermi-particles (example: D = P + N + ¢) isa Fermi-particle, and a particle consisting of an even number of Fermi-particles (example: H = P + e) is a Bose-particle (@ particle consisting of Bose-particles only is a Bose-particle). A Fermi particle has half-integral spin and a Bose-particle has integral spin. Fermi distribution and Bose distribution: When the particle system is in equilibrium (the total number of particles Nis assumed very large, of course), the average occupation number for each one-particle state is shown to be (example 12, problem 3) 4 These are often called Fermi-Diac statistics and Bose-Ei 4 The wave function of system of identical particles or antsymmetrical (Fermi) in permutation of the particle (see, for example, Schiff: Quantum Mechanics). LS REGISTERED VERSION ADDS NO WATERMARK & < efint-arwe30 PRINCIPLES OF STATISTICAL MECHANICS [on 151s where T is the temperature of the system and the chemical potential of the Particle, The energy of the total system is given by Levin (192) while the total number of particles is Neda, (193) Meaning of and T: 1 and T appearing in equation (1.90) and (1.91) may be interpreted in different ways: (1) If the system is isolated (microcanonical ensemble), equation (1.92) and (1,93) determine T and p for prescribed E and N. (2) When the system is in contact with a heat bath at temperature T (canonical ensemble), equation (1.92) gives the average energy and (1.93) determines 1 for prescribed T and N. (3) When the system is in contact with a bath at temperature J and with a particle source characterized by the chemical potential x (grand canonical), (1.92) and (1.93) give the average E and N ‘Thermodynamic functions: The upper sign is for F. D. and the lower sign is for B. E. (see example 12): S=kY[-A,loga, F (1 FA, log (1 FA], (1.94a) F=E-TS, (1.94b) J = pV =PF-Nu=4kTZ log FA) = EATS og(t + 0"), (1.94) The classical limit ( Boltzmann-statistics): When the particle density is so low and the temperature so high that the condition N z (aziz) ft v ca = is satisfied, we have TERED ety: mim “ES & 'S) comesponds tothe clases approximation meno Ses. ARES TERED Ol law called Botamenn satis, ane Sige Boppy ny ADDS NO WA TERMARK &, < efint-arweCh.1, $16) ‘GENERALIZED ENTROPY a distribution, sometimes called the Boltzmann-Maxwell distribution) is given by p= eoreoer an For Boltzmann statistics, the one-particle partition function can be defined by sateesyet (1.98) With this definition the following formulae are obtained (see problem 19) (1.99) pV =NkT The Maxwell velocity distribution: The probability that a molecule (mass m) in an ideal gas at temperature T will be found to have a velocity in a range between (2,,0,, ¥,)and (0, + dv, 0, + do,, 0, + do,) is given by deel SlCx ty 0) de, dv, de, = (gt, Je (lected) do.de, de, (1.100) This is called the Maxwell (Boltzmann) velocity distribution law. $116, GENERALIZED ENTROPY Itis possible to define the entropy for a statistical ensemble by S= —kY flops) = — king? 101) where /(0) is the probability of the realization of the quantum state [It is re- commended that the readers prove for themselves that this definition yig H-theorems which include time explicitly. Divenrissenent 2 sre thd of costing teases amo ees TERED e” JERSION ADDS NO WATERMARK & < efint-arwe2 PRINCIPLES OF STATISTICAL MECHANICS font kinetic method of tracing the exact circumstances ofeach individual mole- cule in all its encounters. It is therefore possible that we may arrive at results which, though they fairly represent the facts as long as we are supposed'to deal with a gas in mass, would cease to be applicable if our faculties and instruments were so sharpened that we could detect and lay hold of each molecule and trace it through all its course.” “For the same reason, a theory of the effects of education deduced from a study of the returns of registrars, in which no names of individuals are given, might be found not to be applicable to the experience of a school- ‘master who is able to trace the progress of each individual pupil.” “The distribution of the molecules according to their velocities is found to be of exactly the same mathematical form as the distribution of observa- tions according to the magnitude of thei errors, as described in the theory of errors of observation. The distribution of bullet-holes in a target ac- ‘cording to their distances from the point simed at is found to be of the same form, provided a great many shots are fired by persons by the same degree of skill.” ~ J. Clerk Maxwell, “Theory of Heat,” 1897 ~ Examples 1. The pressure which a gas exerts on the walls of vessel can be regarded as the time average of the impulses which the gas molecules impart on the wall when colliding with and recoiling from it. From this point of view, calculate the pressure p and show that pain (Bernoulli's formula), where m is the average number of molecules per unit volume and é the mean kinetic evergy per molecule. Souvtion Let the vessel be a cube with edges of length /. Suppose that the wall is perfectly smooth so that a molecule colliding with it is reflected in a com- pletely elastic way. Let us take the axis of an orthogonal coordingss sys I Fe Srsdcle to the walleocer smomentum 2p dered loo SRERINTERED wall (perfect reflection). Hence the time averagggs omen POSS ION ADDS NO WA cai & 7 Z tint-dtwecnt ExAMPLES 3 words, the sum of the momentum imparted to the wall per unit time, is equal t02 | Py | | py |/2ml = p3/ml. Since the sum of these momenta yields a force, the contribution of this molecule to the pressure is given by this sum divided by the area of the wall J? (V = [° is the volume of the vessel). Therefore, adding together the contributions from all the molecules, we obtain the pressure exerted by all the molecules 1 ph ob + oh wo Fig. 13, ALTERNATIVE SOLUTION (Method using the virial theorem). Denoting the position vector of the th molecule by x,, one has fdoxn Saat Date Daat2 de @ where F, is the force acting on the i-th molecule and g its kinetic energy. When the time average is taken of both sides, the left hand side vanishes, TT, fa A = tim f game im! (A+ 9- At) well as the coordinates remain finite). Therefore, one =255, ‘The right hand side is called the virial. Ifthe interm Lr Perce is nealeitle SION ADDS NO WATERMARK & o ‘2 tint-dtwe4 [PRINCIPLES OF STATISICAL MECHANICS [cnt F, is simply the force exerted on the molecule by the walls. Its time average per unit area being the pressure p, the virial due to this force is given by a= — f pweeds () x where y is the outward normal to the walls and the integral is evaluated over the walls. By means of Gauss’ theorem Ja-vds = fdivadv one has Jvixds = feivedy = 3fav =3¥. © Fig, 1.4, Calculation of the virial due to the force exerted by the wall, From (4), (3) and (6) one gets 2NE=3pV, or p= Note: The virial theorem (4) is valid also in the case where one must take into account the intermolecular force explicitly. If Fy is the force the ith molecule exerts on the th, one has Fy, = F,; because of the law of action and reaction. The virial due to the intermolecular force is From (4) one gets ee Pe sne ay When the range of the intermolecular force is muclygg intermolecular distance, @ is almost zero and Bern the general case one has to use (8) (see problem gue’ < =~ BEtalS TERED = VERSION ADDS NOchy xaMPuns 3s 2. An ideal gas consisting of N mass points is contained in a box of volume V. Find the number of states (phase integral) g(£) (1.20b) classically and, using it, derive the equation of state, (Hint: The volume of a unit sphere in a nedimensional space, C,, is equal to x**/7(an + 1).) SoLution ‘The energy of this system is =X pliam a Where we have labelled the momenta conjugate to the Cartesian coordinates, of each mass point with running index number: p;, ... , psy. The number of states is given by (1.20b) where the integral over the coordinates gives V*, I(EN, V) J apy dpay- @ Torsane Since this integral is equal to the volume of a 3N-dimensional sphere of radius (2mE)} itis proportional to (2mE)°*", the proportionality coefficient being the number Cyy given in the hint. Hence, v" Qamey*? = iN TaN + ® 24 By (1.27) the entropy becomes emytet 7 a Cy 2 SEN.) = blog, = wifiogy, + Ho655 + oe where use has been made of Stirling's formula le) = NlogN — (9) log N! ~ N log( From (1.28) and (1.54) we get Li) a? T\GEv.w 2 p_ (as 5 =NKIV, or pV = Ng Tr (=), ra oP sesh hn writen. fe BEBIS TERED « wy" VERSION ADDS NO WATERMARK & Z o ‘2 tint-dtwe36 [PRINCIPLES OF STATISTICAL MECHANICS lout equal to nCyr*"*, Using this relation we evaluate nm Fon Leal aed tot aD) das dy by two different methods. First, because of the well-known formula, J exp(— ax'pdx = (alot, Secondly, by dividing the x, ... x, space into spherical shells we can perform the integration Jemncyr dr nC,(dnya~* = C,P(4n + 1a a inc, Ferny (eve tas = rep. Equating the two expressions we have C= Gn + D), Nore 2: Stirling's formula Pet Daxt~ tte *V2n, (8a) logx! ~ xlogx —x+ (8b) is an asymptotic formula valid for x > 1, which is frequently used in statisti- cal mechanics. . The energy level of an oscillator with frequency v is given by = Phy, dhy, (n+ hy, When a system consisting of N almost independent oscillators has thegstal energy E=4Nhv + Miv @ find the thermodynamic weight Wy, and (ii) dexauuly ween the temperature of this system and E Po\ERED 1 ¢ “o [= REGISTERED Ol Res RESON ADDS NO WA TERMARK & SoLuTION (@) If the quantum number of the ith oscilla RS eint-driNgcn) AMPLE 37 ‘ment that the total energy of the system is equal to $Niv + Mhv implies that tb tye M Therefore, the thermodynamic weight Wy of a macroscopic state with the total energy F is equal to the number of ways of distributing Mf white balls among N labelled boxes. A box may be empty since n; = 0 is possible. As is evident from Fig. 1.5, one can get this number by finding the number of permutations of placing in a row all the white balls together with (V — 1) red balls that designate the dividing walls. If one labels all the balls with the running numbers, 1, 2, ...,.M + N — 1, the number of permutations is (M + N — 1)!. When one erases these numbers there appear indistinguish- able distributions, the number of which is equal to the number of permuta- tions among the (numbered) balls with the same color, M1(N ~ 1)!. There- fore, (M4+N-1) MUN = 1)! Lk UWL a 90 @00000 #00080 Fig. 15, (i) From (1.18) the entropy is S = klog Wy, Substituting (1) and using Stirling's formula under the assumption, > 1, M > 1, one has S = k{(M + N)log(M + N)~ MlogM — log} From (1.28) 1_as_asam T OE OMGE M4N\aM_ kM wg MSEN) OM Ka dg (EN) 45 y [= REGISTERED ©) VERSION ADDS NO WA TERMARK & o ‘2 tint-dtwe38 PRINCIPLES OF STATISHICAL MECHANICS (out Or, inversely, EAANKY nar E-4Nhv Solving this for E, one gets E=N {em tm ® In Fig. 1.6, £/Nhv is plotted against k7/hr. Fig 16 ‘There is a system consisting of N independent particles. Each particle can have only one of the two energy levels, —Z9, éy. Find the thermodynamic weight Wy ofa state with the total energy E = Mey(M = —N,... Nand he as ERED eo" es sin the state SN REGISTERED © VERSION ADDS NO discuss the statistical-thermodynamic properties of the system fo range E <0, deriving especially the relation between temperat fig energy as well as the specific heat SoLvTion If N__ particles are in the state with energy —¢» and MJ with energy é, the energy of the total system is tg = — Nt, + Neto.cat exanPues » Since N= N, +N, one has N. HN-M), Ny=4(N4+M) @ Now, there are V1/(V_1N,!) possible ways of choosing N particles out of N to occupy the state with ~¢, each of which gives a different microscopic state with the energy E. Hence the thermodynamic weight is “T= ME + My ‘According to (I-18) he entropy of the system is Wwe GB) SE) = kiog Wu, = k{N log N ~ 4 — M)log}(N ~ M) —4(N + M)log}(N + M)} = —K{N_log(N_/N) +N lox(N/N)} ® ‘where Stirling's formula has been used. Defining the temperature by (1.28), cone has 1185 jk N=M im 8X 4M 9 ‘As one can see from this equation or from the relation between Sand E (Fig. 1.1), T <0 for M > 0(E > 0), so that this system is not normal in the sense of statistical mechanics, Since, however, it is normal in the range M <0(E <0), one can discuss the property of this system within this range. VERSION ADDS NO«0 PRINCIPLES OF STATISTICAL MECHANICS tout From (5) one gets or Ne © - Therefore, E=—(N-=Ny)eo = ~ Negtanb(eg/kT). oO Equations (6) give the probabilities of finding any one particle in the states to and +60, respectively, and have the form of a canonical distribution (1.70). From (7) the specific heat becomes C=azjar = wa(z2) [oom ® Figs. 1.8 and 1.9 show the graphs of £ versus and C versus 7, respectively. Note: Putting 2e = AE, one rewrites (8) as EY? seary AEIAT)2 cu nn(Ze) ASTI 4 BT) © VERSION ADDS NO WA TERMARK & RS rint-dringcay AMPLE 4“ ‘The specific heat of this form is called the Schottky specific heat. When a body contains a substance having the excitation energy AE, thespecificheat anoma- ly with a peak as shown in Fig. 1.9 is actually observed, (ppp Fig. 19, 5. Suppose that two systems, which are normal inthe sense of statistical thermo- dynamics, have the number of states of the form (I.24a). Prove the following statements concerning the thermal contact of the two system, (i) Iftheir inti temperatures are 7; and (7, > T;),heat will flow from I to2 upon contact and the entropy will increase by the amount d’Q(T3' — 7"). (i) When both systems are in thermal equilibrium with each other, the entropy of the ‘composite system Sy «is equal to the sum of the entropies of the subsystems S, and S,. (ii) The fluctuations in the energy of subsystem 1 or subsystem 2 when in a state of thermal equilibrium, (E, — E,)' or (E, — E3), are given by K77/(Cz! + Cz"), where Ej, E are the mean values (the most probable heats SoLuTion ‘We suppose that the numbers of states 21 (E,), 23 (E:) a} form, ‘> REGISTERED NCVER SION ADDS NO WATERMARK & OYE) ~ exp{NiG(EV/N)}, 3B) < efint-arwe