Foundations of statistical mechanics. Volume I: Equilibrium theory (Q1210892)

[For Volume II see the following review.] This two-volume set should be in every physics and mathematics library. Just as Sigmund Freud asked in concern, ``What does woman want?'' so do some mathematicians ask, ``What do physicists (female and male) want?'' In statistical mechanics these volumes will answer a lot of the latter questions. While these books are intellectually rigorous, they are not mathematically rigorous. And it is noted that there are actually two very different books here. One is a high-quality short monograph on the physical formulations of the maximum entropy approach to statistical mechanics, as pioneered by E. T. Jaynes. The second is a two-volume textbook on equilibrium and nonequilibrium statistical mechanics. The introductory three chapters could be taught to advanced undergraduates in mathematics and physics. Most of the material in both volumes could be taught to first-year graduate students. The maximum entropy principle is a way to deal with ignorance and/or missing information. For this reason it is useful for generalized inverses, estimation and prediction theory, as well as statistical mechanics. In statistical mechanics at each instant of time t one has \(N=10^{21}\) to \(10^{24}\) particles with positions \(\vec r_ i(t)\) and momenta \(\vec p_ i(t)\) and the observables studies depend upon them all. There are only a few observables, such as pressure, entropy, volume and temperature for equilibrium systems or hydrodynamic velocity, sound attenuation and electrical conductivity for nonequilibrium systems. These leave the experimentalist ignorant of most of the 6N positions and momenta and many constraints as well. The maximum entropy method for a positive, additive functional F(\(\vec x_ i,\vec p_ i,...)\) is that the best choice of configuration of (\(\vec x_ i,\vec p_ i)'s\) is the one which maximizes the Shannon-Jaynes entropy. In 1957 Jaynes showed that this approach, when combined with Bayes statistics, extended and clarified the Gibbs ensemble approach. The contents of Volume I, after the preface, are the following: Chapter 1. Introduction, Chapter 2. Theory of probability, Chapter 3. Equilibrium thermodynamics, Chapter 4. Quantum statistical mechanics, Chapter 5. Noninteracting particles, Chapter 6. External fields, Chapter 7. Interacting particles I. Classical and quantum clustering, Chapter 8. Interacting particles II. Fock space, Chapter 9. The phases of matter, Appendix A. High points in the history of statistical mechanics, Appendix B. The law of succession, Appendix C. The method of Jacobians, Appendix D. Convex functions and inequalities, Appendix E. Euler-Maclaurin summation formula, Appendix F. The first four Ursell functions and their inverses, Appendix G. Thermodynamic form of Wick's theorem. The index seems reasonably complete. This volume is well written and could be seriously considered as a course text, were it not for price. All of the topics which are required for an educated graduate student (or faculty member) are here. [Reviewer's remark: Without wanting to criticize the author for the book which he did not choose to write, a few comments are appropriate. Phase transitions seemed to get short shrift in Chapter 9, and the complexity ideas being developed by L. P. Kadanoff and many others could have been an additional section in Chapter 7 in addition to Volume II. The history in Appendix A seems questionable in omitting Planck's justly famous black-body formula and Onsager's exact solution to the two-dimensional Ising model. In spite of these minor criticisms, the reviewer enjoyed this book a great deal. One (of many) example of the author's scholarship appears at the bottom of page 71, where he says, ``In a very real sense, this is the real ergodic problem of probability theory, and its resolution is far from being understood.'' The mathematician who wonders what physicists ``want'' would do well to ponder the discussion which leads up to this statement.]