CONGRATULATIONS TO HERBERT KROEMER, 2000 NOBEL LAUREATE FOR PHYSICS
For upper-division courses in thermodynamics or statistical mechanics, Kittel and Kroemer offers a modern approach to thermal physics that is based on the idea that all physical systems can be described in terms of their discrete quantum states, rather than drawing on 19th-century classical mechanics concepts.
Thermodynamics is ridiculous. It isn't 1870 anymore. Use statistical mechanics. This book starts with statistical mechanics immediately. Use it. Then derive thermodynamics from first principles.
Buy the first edition for lower price and better organization and exposition. The second edition is ruined by rewriting (most likely by the added co-author Herbert Kroemer).
usually i love examples, but i actually found a book that overdoes it. too many examples for an intro book. and because they are real-world examples, they stretch the simple intro models, so that a student would be bewildered as to how to apply the simple framework to the extremely messy example. ("if we assume the valence electrons in a conductor are like an ideal non-interacting gas..." WHAT? who in hell would have thought that was reasonable?)admirable in intent, but overkill and ultimately counterproductive. stick with simpler, non-sexy examples.
i did like some of the statistical stuff about how sharply peaked the distribution functions are, the interpretation of entropy, and the chapter on heat pumps and the carnot cycle.
i suspect the landau and lifshitz book is a better choice.
Statistical mechanics is probably the most elegant (and useful) physical theory I have encountered: you basically derive all of thermodynamics from scratch (or with the least objectionable set of fundamental postulates).
A lot of people (myself included) do not like this textbook at first but learn to appreciate its simplicity as the details of thermal physics become clearer.
It's a good book, specially after rereading it a couple of times. The style is succinct and the arguments elegant (most of the time). Maybe it works better once you know a little bit about stat mech. It lacks a bit of discussion about how minimizing a thermodynamic potential is equivalent to having the configuration with the highest probability.