Thermal Physics Concepts and Practice

Download as pdf or txt
Download as pdf or txt
You are on page 1of 309

Thermal Physics

Concepts and Practice


Allen L. Wasserman
Department of Physics
Oregon State University
Corvallis, OR 97331
SPRING 2011

ii
.

2010 Allen L. Wasserman


All rights reserved

Contents
1 Introducing thermodynamics
1.1 The beginning . . . . . . . . . . . . . . . . . . . . .
1.2 Thermodynamic vocabulary . . . . . . . . . . . . .
1.3 Energy and the First Law . . . . . . . . . . . . .
1.3.1 Thermodynamic variables defined . . . .
1.3.1.1 Macroscopic interaction energy
1.4 Quantum mechanics, the mother of theories .
1.4.1 Introduction . . . . . . . . . . . . . . . . .
1.4.2 A brief review . . . . . . . . . . . . . . . .
1.5 Probabilities in quantum mechanics . . . . . . . .
1.5.1 Expectation values . . . . . . . . . . . . .
1.6 Closing comments . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

1
1
2
3
4
5
6
6
6
8
9
9

2 A road to thermodynamics
2.1 The density operator: pure states . . . . . . .
2.1.1 Traces, expectations and information
2.2 Mixed states and macroscopic systems . . . .
2.2.1 Mixed state properties . . . . . . . . .
2.2.2 Macroscopic consequences . . . . . . .
2.2.3 Thermodynamic state functions? . . .
2.3 Thermal density operator op and entropy . .
2.3.1 Examples . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

11
11
12
13
14
15
16
17
18

3 Work, heat and the First Law


3.1 Introduction . . . . . . . . . . . . . . . . . . .
3.1.1 Work, W . . . . . . . . . . . . . . . .
3.1.2 Heat, Q . . . . . . . . . . . . . . . . .
3.1.3 Temperature, T . . . . . . . . . . . .
3.1.4 Internal energy, U . . . . . . . . . . .
3.2 Exact differentials . . . . . . . . . . . . . . .
3.3 Equations of state . . . . . . . . . . . . . . .
3.3.1 Examples I . . . . . . . . . . . . . . .
3.3.1.1 Quasistatic reversible work
3.3.1.2 Irreversible work . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

23
23
23
25
25
25
26
26
28
28
29

iii

.
.
.
.
.
.
.
.
.
.

iv

CONTENTS
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

31
32
33
33
34
35
35
36
37
37
38
40
41
43

4 A mathematical digression
4.1 Thermodynamic differentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.1 Exact differentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2 Exactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.3 Eulers criterion for exactness . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.4 Entropy A free sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.5 Eulers homogeneous function theorem . . . . . . . . . . . . . . . . . . . .
4.1.6 A cyclic chain rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.6.1 Example: Free expansion: Exactness and the cyclic chain rule
4.1.6.2 Example: Entropy and spontaneous processes . . . . . . . . . . .
4.1.7 Thermal engines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.8 Carnots thermal engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.9 The Carnot cycle entropy . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

47
47
48
48
48
49
51
52
53
57
59
59
63

5 Thermodynamic potentials
5.1 Introduction . . . . . . . . . . . . .
5.1.1 Internal energy U . . . .
5.1.2 Enthalpy H . . . . . . . .
5.1.3 Helmholtz Potential F .
5.1.4 Gibbs potential G . . . .
5.1.5 Enthalpy and throttling . .
5.1.6 Heat capacity and entropy

3.3.2
3.3.3

3.4

Heat
3.4.1
3.4.2
3.4.3

3.4.4

3.3.1.3 Quasistatic, irreversible work . .


The p-V diagram . . . . . . . . . . . . . . .
Examples II . . . . . . . . . . . . . . . . . .
3.3.3.1 A loaded rubber band . . . . . . .
3.3.3.2 A compressed metal rod . . . . . .
capacity . . . . . . . . . . . . . . . . . . . . . .
Heat transfer at constant volume . . . . . .
Heat transfer at constant pressure . . . . .
Examples III . . . . . . . . . . . . . . . . . .
3.4.3.1 Ideal monatonic gas . . . . . . . .
3.4.3.2 A steel wire . . . . . . . . . . . . .
3.4.3.3 A rapidly expanding gas . . . . .
3.4.3.4 The filling problem atmospheric
Concluding remarks . . . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

6 Knowing the unknowable


6.1 Entropy the ticket to Oz . . . . . .
6.1.1 The bridge . . . . . . . . . . .
6.1.2 Thermodynamic hamiltonians
6.1.3 Microcanonical (Boltzmanns)
6.1.4 Gibbs Canonical Theory . . .
6.1.5 Canonical thermodynamics .
6.1.6 Degeneracy and Z . . . . . . .
6.1.6.1 Examples . . . . . .

. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . .
gas filling an empty vessel
. . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

67
67
67
69
70
71
71
74

. . . .
. . . .
. . . .
theory
. . . .
. . . .
. . . .
. . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

79
79
80
81
83
86
87
88
88

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

CONTENTS
6.1.7

v
Closing comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7 The Ideal Gas


7.1 Introduction . . . . . . . . . . . . . . . . .
7.1.1 Ideal gas law . . . . . . . . . . . .
7.1.2 Quasi-classical model . . . . . . .
7.1.3 Calculating a partition function .
7.1.4 Thermodynamics of the ideal gas
7.1.5 Gibbs entropy paradox . . . . . .
7.1.6 Entropy of mixing . . . . . . . . .
7.1.7 The non-ideal gas . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

94
97
97
98
98
100
101
103
104
104

8 The 2-Level System


107
8.1 Schottky anomalous heat capacity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.1.1 Schottky Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.1.2 Two Level Systems and Negative Temperature . . . . . . . . . . . . . . . . . . . 110
9 Lattice Heat Capacity
9.1 Heat Capacity of Solids . . . . . . . . . . . . . . . . . . . . . . . . .
9.1.1 Einsteins Model . . . . . . . . . . . . . . . . . . . . . . . . .
9.1.2 Einstein Model in 1 D . . . . . . . . . . . . . . . . . . . .
9.1.2.1 Partition Function for 1 D Einstein Model . .
9.1.3 Phonons and the 1 D Oscillator . . . . . . . . . . . . . .
9.1.4 The 3 D Einstein Model . . . . . . . . . . . . . . . . . . .
9.1.4.1 Partition Function for the 3 D Einstein Model
9.1.4.2 Thermodynamics of the 3 D Einstein Model .
9.2 Debye Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
9.2.0.3 Thermodynamics of the Debye Lattice . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

113
113
114
114
116
117
117
118
118
120
122

10 Elastomers Entropy Springs


10.1 Naive 1-D Elastomer . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10.1.0.4 Elongation as an Observable . . . . . . . . . . . . . . .
10.1.0.5 Naive Elastomer Model . . . . . . . . . . . . . . . . . .
10.1.0.6 Properties of the Naive 1 D Rubber-Band . . . . .
10.1.1 Hot Rubber Bands A Thermodynamic Analysis . . . . . . .
10.1.2 Abruptly Relaxing Rubber Band . . . . . . . . . . . . . . . . . . .
10.1.3 Entropy Change in the Relaxing 1 D Elastomer . . . . . . . .
10.1.4 A Non-ideal Elastomer . . . . . . . . . . . . . . . . . . . . . . . .
10.1.5 3 D Elastomer Thermodynamics; Prelude to an Experiment
10.1.6 An Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

125
125
127
128
129
131
134
135
135
137
139

11 Magnetic Thermodynamics
11.1 Magnetism in solids . . . . . . . . . . . . . . .
11.1.1 Magnetic work . . . . . . . . . . . . . .
11.1.2 Microscopic models and uniform fields
11.1.3 Local paramagnetism . . . . . . . . . .
11.1.4 Simple paramagnetism . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

143
143
144
148
149
150

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

vi

CONTENTS
11.1.5 Local paramagnet thermodynamics . . . . . . . . . . . . . . .
11.1.6 Magnetization fluctuations . . . . . . . . . . . . . . . . . . . .
11.1.6.1 Example: Adiabatic (Isentropic) Demagnetization
11.1.7 A Model for ferromagnetism . . . . . . . . . . . . . . . . . . .
11.1.8 A mean field approximation MFA . . . . . . . . . . . . . .
11.1.9 Spontaneous magnetization . . . . . . . . . . . . . . . . . . .
11.1.10 Critical exponents . . . . . . . . . . . . . . . . . . . . . . . . .
11.1.11 Curie-Weiss magnetic susceptibility (T > Tc ) . . . . . . . .
11.1.12 Closing comment . . . . . . . . . . . . . . . . . . . . . . . . .

12 Open Systems
12.1 Variable particle number . . . . . . . . . . . .
12.1.1 Thermodynamics and particle number
12.1.2 The open system: formalities . . . . .
12.1.3 Open system thermodynamics . . . . .
12.1.4 Grand partition function . . . . . . . .
12.1.5 Grand potential . . . . . . . . . . . . .
12.1.6 G, gr and Eulers theorem . . . . . .
12.1.7 The ideal gas: A Grand example . .
12.1.8 Ideal gas thermodynamic properties .
12.2 Closing comment . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

152
155
156
157
158
159
160
161
162

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

163
163
164
164
166
166
167
168
169
170
171

13 The Amazing Chemical Potential


13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13.2 Equilibrium constraints . . . . . . . . . . . . . . . . . . . . . . .
13.2.1 Thermodynamics and diffusive equilibrium: Examples
13.2.1.1 Ideal gas in the earths gravitational field . .
13.2.1.2 Charged ideal gas between capacitor plates . .
13.2.1.3 Ideal gas in a rotating cylinder . . . . . . . .
13.2.2 Thermodynamics of Chemical Equilibrium . . . . . . .
13.2.3 A law of mass action . . . . . . . . . . . . . . . . . . . .
13.2.4 Thermodynamics of phase equilibrium . . . . . . . . . .
13.2.5 Gibbs-Duhem relation . . . . . . . . . . . . . . . . . . . .
13.2.6 Multiphase equilibrium . . . . . . . . . . . . . . . . . . .
13.2.7 The Clausius-Clapeyron equation . . . . . . . . . . . . .
13.2.8 A star is born . . . . . . . . . . . . . . . . . . . . . . . .
13.2.8.1 Thermodynamic model . . . . . . . . . . . . .
13.2.8.2 Collapse criterion . . . . . . . . . . . . . . . .
13.2.9 Van der Waals equation . . . . . . . . . . . . . . . . . .
13.2.10 Surface adsorption Langmuirs model . . . . . . . . .
13.3 Dissociative adsorption . . . . . . . . . . . . . . . . . . . . . . .
13.4 Crystalline bistability . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

173
173
174
174
174
175
176
177
178
180
180
181
183
184
184
187
188
191
193
194

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

14 Thermodynamics of Radiation
199
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
14.2 Electromagnetic eigen-energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
14.3 Thermodynamics of electromagnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201

CONTENTS
14.3.1
14.3.2
14.3.3
14.3.4
14.3.5
14.3.6
14.3.7
14.3.8
14.3.9

vii
Thermodynamics for a single photon mode . . . . . . . . .
Average photon number . . . . . . . . . . . . . . . . . . . .
Helmholtz potential and internal eneregy . . . . . . . . . .
Radiation field thermodynamics . . . . . . . . . . . . . . . .
Stefan-Boltzmann, Planck, Rayleigh-Jeans radiation laws
14.3.5.1 Example: . . . . . . . . . . . . . . . . . . . . . . . .
Wiens law . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Entropy of thermal radiation . . . . . . . . . . . . . . . . .
Stephan-Boltzmann radiation law . . . . . . . . . . . . . . .
14.3.8.1 Example: . . . . . . . . . . . . . . . . . . . . . . . .
Radiation momentum density . . . . . . . . . . . . . . . . .
14.3.9.1 Example: Star Dust Blow-out . . . . . . . . . . .

15 Ideal Fermi Gas


15.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
15.2 Ideal gas eigen-energies . . . . . . . . . . . . . . .
15.2.1 Grand partition function . . . . . . . . . .
15.2.2 Electron spin . . . . . . . . . . . . . . . . .
15.2.3 Fermi-Dirac thermodynamics . . . . . . .
15.2.4 Independent fermion model . . . . . . . .
15.2.5 The chemical potential (T 0) . . . . . .
15.2.6 Internal energy (T 0) . . . . . . . . . . .
15.2.7 Pauli paramagnetic susceptibility . . . . .
15.2.8 Electron gas model . . . . . . . . . . . . .
15.2.9 White Dwarf Stars . . . . . . . . . . . . .
15.2.9.1 White Dwarf Thermodynamics

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

201
202
202
203
203
205
206
207
208
210
210
211

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.

215
215
215
216
217
218
220
221
222
223
224
227
227

16 Ideal Bose-Einstein System


16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . .
16.2 Ideal bose gas . . . . . . . . . . . . . . . . . . . . . . .
16.2.1 Bose-Einstein thermodynamics . . . . . . . .
16.2.1.1 The quasi-classical limit . . . . . .
16.2.2 The ideal B-E gas and the B-E condensation
16.2.3 The condensate . . . . . . . . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

233
233
234
235
236
236
239

17 Thermodynamics and the CMB


17.1 Introduction . . . . . . . . . . . . . . . . . . . . . .
17.2 Thermodynamic method . . . . . . . . . . . . . . .
. . . .
17.3 Moving frame thermodynamic potential
17.3.1 Moving frame blackbody thermodynamics
17.4 Radiation energy flux . . . . . . . . . . . . . . . .

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

243
243
244
245
246
248

.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.

.
.
.
.
.

251
2
A.1 The inequality 0 < T r (op ) 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

253
B.1 Properties of F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253

viii

CONTENTS
B.1.1 The Bias Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
B.2 A Thermal Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255

257
C.1 Eulers Homogeneous Function Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

259
D.1 Occupation Numbers and The Partition Function . . . . . . . . . . . . . . . . . . . . . 259

261
E.1 Density of States Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
E.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261

265
F.1 A Lab Experiment in Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

G
.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

269
269
269
270
274

H.1 Maxwells Equations and Electromagnetic Fields . . . . . . . . . . . . . . .


H.1.1 Electromagnetic Waves . . . . . . . . . . . . . . . . . . . . . . . . . .
H.1.1.1 Solution in Free Space Periodic Boundary Conditions .
H.1.1.2 Properties of Wave Solutions . . . . . . . . . . . . . . . . .
H.1.1.3 Electromagnetic Vector Potential . . . . . . . . . . . . . .
H.1.1.4 Quantized Electromagnetic Hamiltonian . . . . . . . . . .

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

277
277
277
278
278
280
282

G.1 Static Magnetic and Electric Fields in Matter .


G.2 Thermodynamic Potentials with Magnetic Fields
G.3 Uniform External Fields . . . . . . . . . . . . . .
G.3.1 Thermodynamics With Internal Fields .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

I
I.1
J
J.1

285
Fermi-Dirac Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
289
Bose-Einstein Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
J.1.1 Integrals When = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
J.1.2 Integrals When < 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

Preface
In the preface to his book Statistical Mechanics Made Simple Professor Daniel Mattis writes:
My own experience in thermodynamics and statistical mechanics, a half century ago
at M. I. T., consisted of a single semester of Sears, skillfully taught by the man himself.
But it was a subject that seemed as distant from real physics as did poetry or French
literature.1
This frank but discouraging admission suggests that thermodynamics may not be a course eagerly
anticipated by many students not even physics, chemistry or engineering majors and at completion I would suppose that few are likely to claim it was an especially inspiring experience. With
such open aversion, the often disappointing performance on GRE2 questions covering the subject
should not be a surprise. As a teacher of the subject I have often conjectured on reasons for this
lack of enthusiasm.
Apart from its subtlety and perceived difficulty, which are probably immutable, I venture to guess
that one problem might be that most curricula resemble the thermodynamics of nearly a century
ago.
Another might be that unlike other areas of physics with their epigrammatic equations Newtons,
Maxwells or Schr
odingers providing accessibility and direction, thermal physics seems to lack a
comparable unifying principle.3 Students may therefore fail to see conceptual or methodological
coherence and experience confusion instead.
With those assumptions I propose in this book alternatives which try to address the disappointing
experience of Professor Mattis and undoubtedly others.
Thermodynamics, the set of rules and constraints governing interconversion and dissipation of energy in macroscopic systems, can be regarded as having begun with Carnots (1824) pioneering
paper on heat engine efficiency. It was the time of the industrial revolution, when the caloric
fluid theory of heat was just being questioned and steam engine efficiency was, understandably,
an essential preoccupation. Later in that formative period Rudolf Clausius introduced a First
Law of Thermodynamics (1850) formalizing the principles governing macroscopic energy conservation.
1 Daniel

Mattis, Statistical Mechanics Made Simple; World Scientific Publishing, Singapore (2003).
Record Examination: Standardized graduate school admission exam.
3 R. Baierlein, A central organizing principle for statistical and thermal physics?, Am J. Phys. 63, 108 (1995).
2 Graduate

ix

CONTENTS

Microscopic models were, at the time, largely ignored and even regarded with suspicion, to the point
where scientific contributions by some proponents of such interpretations were roundly rejected by
the editors of esteemed journals. Even with the intercession in support of kinetic models by the
respected Clausius (1857), they stubbornly remained regarded as over-imaginative, unnecessary
appeals to invisible, unverifiable detail even by physicists. A decade later when Maxwell (1866)
introduced probability into physics bringing a measure of statistical rigor to kinetic (atomic) gas
models there came, at last, a modicum of acceptance.
Within that defining decade the already esteemed Clausius (1864) invented a novel, abstract quantity as the centerpiece of a Second Law of Thermodynamics, a new principle which he named
entropy to forever change our understanding of thermal processes and, indeed, all natural processes. Clausius offered no physical interpretation of entropy leaving the matter open to intense
speculation. Ludwig Boltzmann, soon to be a center of controversy, applied Maxwells microscopic probability arguments to postulate a statistical model of entropy based on counting discrete
atomic configurations referred to as microstates.4 However Boltzmanns ideas on entropy,
which assumed an atomic and molecular reality, were far from universally embraced a personal
disappointment which some speculate led to his suicide in 1906.
Closing the book on 19th century thermal physics, J. W. Gibbs reconciled Newtonian mechanics
with thermodynamics by inventing statistical mechanics 5 based on the still mistrusted presumption
that atoms and molecules were physical realities. In this indisputably classic work, novel statistical
ensembles were postulated to define thermodynamic averages, a statistical notion later adopted in
interpreting quantum theories. Shortcomings and limited applicability of this essentially Newtonian
approach notwithstanding, it provided prescient insights into the quantum mechanics, whose full
realization was still a quarter century in the future.
Quantum mechanics revolutionized physics and defines the modern scientific era. Developing in parallel with it, and synergistically benefiting from this reshaped scientific landscape, thermal physics
has come to occupy a rightful place among the pillars of modern physics.
Quantum mechanics natural, internally consistent unification of statistics with microscopic mechanics immediately suggests the possibility of a thermodynamics derived, in some way, from microscopic
quantum averages and quantum probabilities. But thermodynamic systems are not simply the isolated quantum systems familiar from most quantum mechanics courses. Thermodynamics is about
macroscopic systems, i.e. many particle quantum systems that are never perfectly isolated from the
remainder of the universe. This interaction with the outside has enormous consequences which
when taken into account quantitatively clarifies the essence of thermodynamics.
Many thermal variables may then be approached as macroscopic quantum averages and their associated thermal probabilities as macroscopic quantum probabilities, the micro-to-macro translation
achieved, in part, by an entropy postulate. This approach gives rise to a practical organizing principle with a clear pedagogical path by which thermodynamics structure attains the epigrammatic
status of real physics.
Thermal physics is nevertheless frequently taught in the spirit of its utile 19th century origins,
minimizing both 20th and 21st century developments and, for the most part, disregarding the
4 Boltzmanns microstates suggested to Planck (1900) what eventually became the quantization he incorporated
into his theory of electromagnetic radiation.
5 J. W. Gibbs, The Elementary Principles of Statistical Mechanics, C. Scribner, New York (1902).

CONTENTS

xi

beauty, subtlety, profundity and laboratory realities of its modern rebirth which returns us to
Professor Mattis reflections. In proposing a remedy for his justifiable concerns, the opening chapter
introduces a moderate dose of quantum based content, for both review and, hopefully, to inspire
interest in and, eventually, better understanding of thermodynamics. The second chapter develops
ideas that take us to the threshold of a thermodynamics that we should begin to recognize. In
Chapter 6 thermodynamics flies from a quantum nest, ready to take on challenges of modern
physics.
Students and practitioners of thermodynamics come from a variety of disciplines. Engineers,
chemists, biologists and physicists all use thermodynamics, each with practical or scientific concerns that motivate different emphases, stress different legacies and demand different pedagogical
objectives. Since contemporary university curricula in most of these disciplines integrate some modern physics, i.e. quantum mechanics if not its mathematical details at least its primary concepts
and aims the basic thermodynamic ideas as discussed in the first two chapters should lie within
the range of students of science, engineering and chemistry. After a few chapters of re-acquaintance
with classic thermodynamic ideas, the books remaining chapters are dedicated to applications of
thermodynamic ideas developed in Chapter 6 in practical and model examples for students and
other readers.
Parts of this book first appeared in 1997 as notes for a course in thermal physics designed as a component of the revised undergraduate physics curriculum at Oregon State University. An objective
of this revision was to create paradigmatic material stressing ideas common to modern understandings and contemporary problems. Consequently, concepts and dynamic structures basic to quantum
mechanics such as hamiltonians, eigen-energies and quantum degeneracy appear and play important roles in this view of thermal physics. They are used to maintain the intended paradigm
spirit by avoiding the isolation of thermal physics from developments of the past 100 years while,
hopefully, cultivating in students and teachers alike a new perception of and appreciation for this
absolutely remarkable subject.
This work has been funded in part by NSF Grants: DUE 9653250, 0231194, 0837829.

xii

CONTENTS

The atomistic nature of matter as conceptualized by the Greeks had, by


the 19th century, been raised by scientists to a high probability. But it
was Plancks law of radiation that yielded the first exact determination
of the absolute size of atoms. More than that, he convincingly showed
that in addition to the atomistic structure of matter there is a kind of
atomistic structure to energy, governed by the universal constant h.
This discovery has almost completely dominated the development of
physics in the twentieth century. Without this discovery a workable
theory of molecules and atoms and the energy processes that govern
their transformations would not have been possible. It has, moreover,
shaken the whole framework of classical mechanics and electrodynamics
and set science the fresh task of finding a new conceptual basis for all
of physics. Despite partial success, the problem is still far from solved.
Albert Einstein, Max Planck memorial service, (1948) Original
Image, Einstein Archives Online, Jerusalem (Trans. A. Wasserman)

Chapter 1

Introducing thermodynamics

1.1

The beginning

Thermodynamics has exceeded the scope and applicability of its utile origins in the industrial
revolution to a far greater extent than other subjects of physics classical era, such as mechanics and
electromagnetism. Unquestionably this results from over a century of synergistic development with
quantum mechanics, to which it has given and from which it has gained clarification, enhancement
and relevance, earning for it a vital role in the modern development of physics as well as chemistry,
biology, engineering, and even aspects of philosophy.
The subjects fascinating history is intertwined with seminal characters who contributed much
to its present form some colorful and famous with others lesser known, maligned or merely
ignored.
Atomism (i.e. molecular models), a source of early conflict, seems to have had Daniel Bernoulli
as its earliest documented proponent when in 1738 he hypothesized a kinetic theory of gases.1
Although science thermodynamics in particular is now unthinkable without such models,
Bernoullis ideas were largely ignored and for nearly a century little interest was shown in matter as based on microscopic constituents. Despite a continuing atmosphere of suspicion, particle
1 D.

Bernoulli,Hydrodynamica, (1738).

CHAPTER 1. INTRODUCING THERMODYNAMICS

models occasionally reappeared2,3 and evolve, shedding some of their controversy (Joule 1851)
while moving towards firm, accepted hypotheses (Kronig 1856). Adoption of kinetic models by the
respected Rudolf Clausius (1857) began to erode the skeptics position and encourage the new statistical theories of Maxwell (1859) and Boltzmann (1872). At about the same time Van der Waals
(1873), theorizing forces between real atoms and molecules, developed a sucessful equation of state
for non-ideal gases.4 Nevertheless, as the 20th century dawned controversy continued only slightly
abated. J. J. Thomsons discovery of the electron (1897) should have finally convinced remaining
doubters of matters microscopic essence but it didnt. The argument continued into the twentieth
century stilled, finally, by the paradigm shift towards quantized models. It all started with Max
Planck who in 19005,6 introduced quantized energy into theories of electromagnetic radiation and
Einstein7 who used quantized lattice vibrations in his ground breaking heat capacity calculation.
But it would be another twenty-five years before the inherently probabilistic, microscopic theory of
matter quantum mechanics with its quantum probabilities and expectation values would completely reshape the scientific landscape and permanently dominate most areas of physics, providing
a basis for deeper understanding of particles, atoms and nuclei while grooming thermodynamics
for its essential role in modern physics. Thermodynamics is primarily concerned with mechanical,
thermal and electromagnetic interactions in macroscopic matter i.e. systems with huge numbers
of microscopic constituents ( 1023 particles). Although thermodynamic descriptions are generally
in terms of largely intuitive macroscopic variables, most macroscopic behaviors are, at their root,
quantum mechanical. Precisely how the classical measurement arises from quantum behavior has
been a subject of some controversy ever since quantum theorys introduction. But it now seems
clear that macroscopic systems are quantum systems that are particularly distinguished by always
being coupled (however weakly) to an environment (sometimes referred to as a reservoir) that is
also considered a quantum system. Although environmental coupling may be conceptually simple
and even uninteresting in detail, it has enormous consequences for quantum based descriptions of
macroscopic matter, i.e. thermodynamics.8

1.2

Thermodynamic vocabulary

A few general, large scale terms are used in describing objects and conditions of interest in thermodynamics.
2 J. Herapath,On the Causes, Laws and Phenomena of Heat, Gases, Gravitation, Annals of Philosophy 9 (1821).
Herapaths was one of the early papers on kinetic theory, but rejected by the Royal Society, whose reviewer objected
to the implication that there was an absolute zero of temperature at which molecular motion ceased.
3 J.J.Waterston,Thoughts on the Mental Functions (1843). This peculiar title was a most likely cause of its rejection by the Royal Society as nothing but nonsense. In recognition of Waterstons unfairly maligned achievement,
Lord Rayleigh recovered the original manuscript and had it published as On the physics of media that are composed
of free and perfectly elastic molecules in a state of motion Philosophical Transactions of the Royal Society A 183,
1 (1892), nearly 10 years after Waterstons death.
4 Van der Waals work also implied a molecular basis for critical points and the liquid-vapor phase transition.
5 Max Planck,Entropy and Temperature of Radiant Heat, Ann. der Physik 1, 719 (1900).
6 Max Planck, On the Law of Distribution of Energy in the Normal Spectrum, Ann. der Physik 4, 553 (1901).
7 A. Einstein, Plancks Theory of Radiation and the Theory of Specific Heat, Ann. der Physik 22, 180 (1907).
8 Macroscopic behavior can be quite different from the behavior of individual constituents (atoms, molecules,
nuclei, etc.) As an example, the appearance of spontaneous bulk magnetism in iron (at temperatures below some
critical temperature Tc ) is not a property of individual iron atoms but arises from large numbers of interacting iron
atoms behaving collectively.

1.3.

ENERGY AND THE FIRST LAW

System: A macroscopic unit of particular interest, especially one whose thermal properties
are under investigation. It may, for example, be a gas confined within physical boundaries or
a rod or string of elastic material (metal, rubber or polymer). It can also be matter that is
magnetizable or electrically polarizable.
Surroundings: Everything physical that is not the system or which lies outside the systems boundaries is regarded as surroundings. This may be external weights, an external
atmosphere or static electric and magnetic fields. The system plus surroundings comprise,
somewhat metaphorically, the universe.
Thermal Variables: A set of macroscopic variables that describe the state of the system.
Some variables are intuitive and familiar such as pressure, volume, elongation, tension, etc.
while others may be less intuitive and even abstract such as temperature which nevertheless, also play important roles in thermodynamics. These will be discussed in detail in this
and later chapters.
Thermal equilibrium: The final state attained in which thermal state variables that describe the macroscopic system (pressure, temperature, volume, etc.) no longer change in
time.9 It is only at thermal equilibrium that thermodynamic variables are well defined. The
time elapsed in attaining equilibrium is largely irrelevant.

1.3

Energy and the First Law

James Joules classic contribution on the mechanical equivalent of heat and his theory of energy
reallocation between a system and it surroundings10 (referred to as energy conservation) led Rudolph
Clausius11 to the historic First Law of Thermodynamics:
U = Q W,

(1.1)

Here W is mechanical work done by the system and Q is heat (thermal energy) added to the system,
both of which are classical quantities associated with surroundings. In Clausius time controversy
about reality in atomic models left U with no definitive interpretation. But being the maximum work
which could be theoretically extracted from a substance it was initially called intrinsic energy. As
kinetic (atomic) models gained acceptance (Clausius having played an influential role) U became the
mean kinetic energy of the systems microscopic constituents or, more generally, as internal energy.
Although the change in internal energy, U, is brought about by mechanical and thermal, i.e.
classical, interactions, quantum mechanics provides a clear and specific meaning to U as an average
change in energy of the macroscopic system as determined from kinetic, potential and interaction
energies of its microscopic constituents, clearly distinguishing it from other energy contributions.
The precise meaning and interrelation of this with similar macroscopic averages provides the basis
for thermodynamics.12
9 There are, nevertheless, small departures from equilibrium averages, referred to as fluctuations, whose values are
also part of any complete thermodynamic description.
10 James P. Joule, On the Existence of an Equivalent Relation between Heat and the ordinary Forms of Mechanical
Power, Phil. Mag 27 205 (1850).
11 R. Clausius, On the Moving Force of Heat, and the Laws regarding the Nature of Heat, Phil. Mag. 2 1-21,
102-119 (1851).
12 Macroscopic averages will be discussed in Chapter 2.

CHAPTER 1. INTRODUCING THERMODYNAMICS

1.3.1

Thermodynamic variables defined

Some thermodynamic concepts and macroscopic variables are familiar from classical physics while
others arise simply from operational experience. Moving beyond this, quantum mechanics provides
definitions and context for not only internal energy U, but for other macroscopic (thermodynamic)
variables, placing them within a macroscopic context that adds considerably to their meaning and
their role within thermal physics. The first law, in arraying Q and W (both classical) against U
(quantum mechanical), highlights this intrinsic partitioning of macroscopic variables into classical(C ) vis-a-vis quantum (Q).
Examples of Q-variables macroscopic variables having microscopic origins13 are:
internal energy: hop [Energy of the systems microscopic constituents kinetic plus potential ].
pressure: pop = (

hop
) [Pressure arising from a systems internal constituents.]
V

electric polarization:

Pop ,

magnetization: M op ,
elongation (length): op ,
particle number : Nop .14
Examples of C-variables classical (macroscopic) variables that exist apart from microscopic mechanics15 are:
temperature: T ,
volume: V,
static magnetic fields: B or H,
static electric fields: E or D,16
elastic tension: ,
chemical potential : .17
13 These

are defined by quantum operators


particle number operator is not a part of Schr
odingers fixed particle number theory. It makes its appearance
quite naturally in quantum field theories. Variable particle number is essential for thermodynamic descriptions of
phase transitions, chemical reactions and inhomogeneous systems. Implementing variable particle number requires:
a. operators to create and destroy particles,
b. operators to count particles.
These discussions and applications reappear in Chapters 7-16.
15 These are not defined by quantum operators.
16 Electromagnetic radiation fields are, on the other hand, representable by quantum field operators B
op and Eop
obtained from a quantum Electromagnetic Vector Potential Operator Aop . This will be discussed in Chapter 14, the
chapter on radiation theory.
17 is an energy per particle and is associated with processes having varying particle number.
14 The

1.3.

ENERGY AND THE FIRST LAW

1.3.1.1

Macroscopic interaction energy

Most Q-variables listed above appear from interaction terms added to hop . Examples of such
variables and their interactions are:
a. A tension applied to an elastic material produces its conjugate elongation , as described
by an interaction operator18

Hop
= op .

(1.2)

b. A static magnetic induction field B0 contributes an interaction operator (energy)


M
Hop
= mop B0 .

(1.3)

Here mop represents a magnetic moment operator for elementary or composite particles,
with a total magnetization 19,20
M op = mop (i)

(1.4)

representing the total magnetization.


c. A static electric field E0 contributes an interaction operator (energy)

P = p E0
Hop
op
where

(1.5)

pop represents an electric dipole moment operator and


Pop = pop (i)

(1.6)

represents the total polarization.21


d. An energy associated with creating space for the system
HW = pop V

(1.7)

with pop representing the system pressure and V the displaced volume.
e. An open system energy associated with particle creation and/or destruction
N
Hop
= Nop

(1.8)

where the chemical potential (an energy per particle) is conjugate to a particle number
operator Nop . Chemical reactions, phase transitions and other cases with variable number
of particles of different species are examples of open systems.22
18 The tension is said to be conjugate to the elongation . The variable and its conjugate comprise a thermodynamic energy.
19 The field B conjugate to m
op is the field present prior to the insertion of matter. Matter itself may contribute
0
to an effective B.
20 There are ambiguites in the thermodynamic roles of static fields, e.g. Maxwells local magnetic average B vs. an
external B0 and Maxwells local electric average E vs. external E0 .
21 The field E is conjugate to the electric dipole operator p .
0
op
22 Variable particle number is intrinsic to thermal physics even though it may be suppressed for simplicity when
particle number is assumed constant.

CHAPTER 1. INTRODUCING THERMODYNAMICS

1.4

Quantum mechanics, the mother of theories

As out of place as rigorous quantum ideas might seem in this introduction to thermodynamics it
is the authors view that they are an essential topic for an approach that strives to bring unity of
structure and calculable meaning to the subject.

1.4.1

Introduction

Macroscopic variables such as pressure p, magnetization M , elastic elongation , particle number N , etc. any (or all) of which can appear in thermodynamic descriptions of matter have
microscopic origins and should be obtainable from quantum models. The question is how?
Investigating the means by which microscopic models eventually lead to macroscopic quantities is
the aim of Chapters 2 and 6. Preliminary to that goal, this chapter reviews postulates, definitions
and rules of quantum mechanics for typical isolated systems23 . Particular attention is given to probabilities and expectation values of dynamical variables (observables).24 This provides the basic rules,
language and notation to carry us into Chapter 2 where the question What is thermodynamics? is
raised, and then to Chapter 6 where the ultimate question: How does it arise? is addressed.
In achieving this goal Chapter 2 will diverge somewhat from the familiar wave mechanics of
introductory courses and focus on a less familiar but closely related quantum mechanical object
called the density operator , op . This quantity, although derived from a quantum state function,
goes beyond state function limitations by adding the breadth and flexibility critical for answering
most of our questions about thermodynamics. It is a Yellow Brick Road that will guide us from
the land of the Munchkins to the Emerald City25 from the microscopic to the macroscopic from
quantum mechanics to thermodynamics.

1.4.2

A brief review

The review starts with Schr


odingers famous linear equation of non-relativistic quantum mechanics
(x, t) .
Hop (x, t) = ih
t

(1.9)

The solution to Eq.1.9 is the time evolving, complex, scalar wavefunction (x, t), which describes
the quantum dynamics of an isolated, fixed particle number, microscopic system.
In the widely used (and preferred) Dirac notation Eq.1.9 is identical with
,
Hop = ih
t
23 These

(1.10)

are the systems described by Schr


odinger theory.
of this review material is likely to be familiar from an earlier course in modern physics. If not, it is
recommended that you work alongside one of the many books on introductory quantum mechanics.
25 L. Frank Baum,Wizard of Oz, Dover Publications, New York (1996).
24 Most

1.4. QUANTUM MECHANICS, THE MOTHER OF THEORIES

where is the time dependent state function, or ket, corresponding to the wavefunction (x, t),
i.e.
x (x, t) .
(1.11)
Here x is an eigen-ket of the position operator with x x1 , x2 . . . xN the coordinates of an
N particle system. As indicated in Eq.1.11, the wavefunction (x, t) is merely the ket in
a coordinate representation.26 Hop is the hamiltonian operator (inspired by classical dynamics).
A hamiltonian operator for a case with only internal particle dynamics may, for example, be
written
hop =Top + Vop .

(1.12)

where Top and Vop are kinetic energy and potential energy operators, respectively.27 Additional
int
interaction terms Hop
such as Eqs.1.2 1.8 may appear, depending on the physical situation, in
which case
int
Hop = hop + Hop
.

(1.13)

The quantum state-function [See Eq.1.11] has no classical counterpart and is, moreover, not
even a measurable! But it is, nevertheless, interpreted as the generating function for a statistical
description of the quantum system including probabilities, statistical averages (expectation values) and fluctuations about these averages. It contains, in principle, all that is knowable about the
isolated, microscopic system.28
Devising an arbitrary, but reasonable, quantum knowledge scale, say
1 all knowledge

(1.14)

0 no knowledge ,

(1.15)

and using this scale to calibrate information, then Schrodingers quantum state function has
information value 1.29 Unless there is some quantum interaction with surroundings (system is no
longer isolated) will retain information value 1 indefinitely.30,31
An energy expectation value an observable is defined (in terms of the wavefunction) by
H (t) = dx (x, t) Hop (x, t)

(1.16)

where (x, t) is the complex conjugate of the wavefunction (x, t). In the more convenient Dirac
notation
H = Hop .
(1.17)
26 The

ket is an abstract, representation independent object. Its only dependence is t time.

27 V
op is assumed to also include particle-particle interactions.
28 Whereas probabilities are derivable from the system state-function,

the reverse is not the case state-functions


cannot be inferred from measurement.
29 In Chapter 2 we will introduce a more solidly based mathematical measure of information to extend this
arbitrary scale.
30 The isolated system state-function is a quantity of maximal information. The systems isolation assures that no
information will leak in or out.
31 Macroscopic complexity has relaxed absolute reliance on a detailed many-body quantum description of a macroscopic system. Even at this early point we can forsee that exact (or even approximate) knowledge of the macroscopic
(many-body) quantum state-function (even if that were possible which it is not) is unnecessary for macroscopic
descriptions.

CHAPTER 1. INTRODUCING THERMODYNAMICS

where , the conjugate to , is called a bra. The energy expectation value (sometimes called
average value) is a measurable 32 of physical interest.
But neither Eqs.1.16 nor Eq.1.17 represent thermodynamic state variables. In particular h = hop
is not macroscopic internal energy U, the centerpiece of the First Law. This is a crucial point
that will be further addressed in Chapter 2 when we inquire specifically about thermodynamic
variables.
In quantum mechanics each of natures dynamical observables (e.g. momentum, position, energy,
current density, magnetic moment, etc.) is represented by an hermitian operator. (An hermitian

operator satisfies the condition op


= op where is the symbol for hermitian conjugation. In matrix

language hermiticity means = ( )T = , where ( )T is the complex conjugate-transpose of


the matrix .)33
The quantum theory postulates an auxiliary eigenvalue problem for op (which represents a typical
hermitian operator)
op n = n n
(1.18)
with n = 1, 2, 3, .. . . ., from which the spectrum of allowed quantum observables n (called eigenvalues) is derived. The n are real numbers34 and n are their corresponding set of complex
eigenfunctions. [For simplicity we assume the eigenvalues are discrete.] The specific forms required
for dynamical hermitian operators will be introduced as the models require.

1.5

Probabilities in quantum mechanics

According to the Great Probability Postulate of quantum mechanics, the Schrodinger state function
is a generator of all probability information about an isolated quantum system with fixed number
of particles.35 Using the complete orthonormal set of eigenfunctions determined from Eq.1.18, a
state function can be expressed as a linear coherent superposition 36
= n n

(1.19)

where the coefficients


p (n , t) = n

(1.20)

are complex probability amplitudes. Coherence as used here implies phase relations among the
terms in the sum which leads to characteristic quantum interference effects.
The probability that a measurement outcome will be n is given by
2

P (n , t) = n

= n n .

(1.21)
(1.22)

32 The value of this measurable is the average of a very large number of identical measurements on identical systems
with identical apparatus.
33 In the following discussion
op is used as a generic symbol for operators representing observables.
34 The hermitian property of quantum mechanical operators assures that the eigenvalues will be real numbers.
35 The state function depends on time t so that all quantities calculated from it will in principle also depend
on time.
36 This is the famous linear superposition theorem.

1.6. CLOSING COMMENTS

The eigenfunction n acts, therefore, as a statistical projector in the sense that with some appropriate measuring device it projects out from the state function the statistical probability of
measuring n . The probabilities P (n , t) are normalized in the usual probablistic sense, i.e.
P (n , t) = 1,

(1.23)

which corresponds to state function normalization


= 1 .

1.5.1

(1.24)

Expectation values

If the probability of observing the measurable value n is P (n , t) then its expectation (average)
value is, as in ordinary statistics,
= n P (n , t)

(1.25)

= n n n .

(1.26)

For the important special case of energy of an isolated (microscopic) system, where
Hop En = En En

(1.27)

with En the allowed eigen-energies (eigenvalues) and En the corresponding eigenfunctions, the
energy expectation value for the state is
E = En En En

(1.28)

= En P (En , t)

(1.29)

which is equivalent to Eq.1.17.


The expectation value of any hermitian observable op is, for a microscopic system,
= op .

1.6

(1.30)

Closing comments

The main points in this chapter are:


1. Schr
odingers wave-functions yield quantum probabilities and quantum expectation values
(averages). But contrary to expectations, these measurables are NOT macroscopic state
variables of thermodynamics. Temporarily putting aside this disappointment, the issue will
be discussed and resolved in the next chapters.

10

CHAPTER 1. INTRODUCING THERMODYNAMICS


2. Schr
odinger state functions are capable of generating all dynamical information that can be
known about the physical system. They can be said to have information content 1 and are
called pure states. From these few clues it should be apparent that information is a term
with relevance in quantum mechanics and, ultimately, in thermodynamics.

The reader is encouraged to consult the ever growing number of fine monographs on quantum
mechanics (far too many to list here).

If someone says that he can think or talk about quantum physics


without becoming dizzy, that shows only that he has not understood
anything whatever about it.
Murray Gell-Mann, Thinking Aloud: The Simple and the Complex,
Thinking Allowed Productions, Oakland, California (1998).

Chapter 2

A road to thermodynamics
2.1

The density operator: pure states

Results from Chapter 1 can, alternatively, be expressed in terms of a density operator op which is
defined (in Dirac notation) as
op =
(2.1)
where the Dirac ket, is a pure Schrodinger state function, and the Dirac bra is its
conjugate. op is clearly hermitian with matrix elements the density matrix in some complete
basis i
ij = i j ,
(2.2)
whose diagonal elements
2

(2.3)

ii =P (i ) .

(2.4)

ii = i
are identical with the probabilities of Eq.1.21

An introduction to properties of the density operator (matrix) will demonstrate that this objects
value is not merely that of an alternative to the wave equation description in the previous chapter,
but that it is capable of defining and accommodating a broader class of quantum states in
particular states which are not pure, i.e. states for which information 1 content is less than 1.
This generalization is a huge stride towards realizing thermodynamics.2
1A

brief discussion about information appeared in Chapter 1.


the density operator (matrix) can be a measurable. R. Newton and B. Young, Measurability of the
spin density matrix, Annals of Physics 49, 393 (1968). Jean-Pierre Amiet and Stefan Weigert, Reconstructing the
density matrix of a spin s through Stern-Gerlach measurements: II,J. Phys. A 32, L269 (1999).
2 Moreover,

11

12

CHAPTER 2. A ROAD TO THERMODYNAMICS

2.1.1

Traces, expectations and information

If Aop is an operator and its matrix representation is


Aij = i Aop j ,
then its trace

(2.5)

Tr Aop is defined as the sum of its diagonal matrix elements,3


Tr Aop = i Aop i .

(2.6)

1. The expectation value in terms of the pure state density operator op is


op op
=

(2.7)

op

where is any complete set of states. Using Eq.2.6, this can be condensed to
=

Tr op op
.
Tr op

(2.8)

Since is normalized

Tr op = op

(2.9)

(2.10)

(2.11)

(2.12)

=1

(2.13)

further reducing Eq.2.8 to


= Tr op op

(2.14)

where completeness of has been applied. Using as a basis n (the eigenfunctions of


op ) Eq.2.14 becomes, as expected,
2

= n n

(2.15)

where, as discussed in Chapter 1,


2

P (n ) = n

is the probability of finding the system in an eigenstate n .


3 The

trace of a matrix is independent of the basis in which the matrix is expressed.

(2.16)

2.2. MIXED STATES AND MACROSCOPIC SYSTEMS

13

2. The probability P (n ) is easily shown to be expressible as


P (n ) = Tr {op n n } .

(2.17)

3. The brief introduction to information in Chapter 1 is now carried a step further by defining
a new quantity purity I,
I = Tr op 2

(2.18)

as a way to measure information.4 The pure state, Eq.2.1, now has purity (information)
I = Tr op op =

(2.19)

(2.20)

(2.21)

= 1,

(2.22)

in agreement with the idea of maximal information on the arbitrary scale devised in Chapter 1.
The pure state density operator op carries complete quantum information (I = 1) about an
isolated microscopic system, in the same way as does its corresponding state function , but
within a more useful framework. (The best is still to come!)
4. The density operator has an equation of motion found by using the time derivative of the
state operator, i.e. Schr
odingers Equation,
op

= ( ) + ( )
t
t
t
i
i
= ( Hop ) + ( Hop )
h
h
i
= (op Hop Hop op )
h
i
= [op , Hop ] .
h

(2.23)
(2.24)
(2.25)
(2.26)

which, in practical terms, is equivalent to Schrodingers Equation.5

2.2

Mixed states and macroscopic systems

At this point it might seem that not much has been gained by introducing the density operator. So
why bother?
Keeping in mind that our interest is in macroscopic quantum systems whose states are inextricably entangled with those of a quantum environment, the density operator provides a logical
4 This is not the only way information can be defined. A more thermodynamically relevant definition will be
expanded upon in Chapter 6.
5 The hamiltonian H
op is the full system hamiltonian.

14

CHAPTER 2. A ROAD TO THERMODYNAMICS

foundation for describing environmental mixing and its consequences.6 Although it seems that
mixing can only irreparably complicate matters it, in fact, ultimately simplifies the essence of
thermodynamics, bringing its meaning into focus.
Pure quantum states, which evolve according to Schrodingers equation, can be individually prepared and then classically combined by some (mechanical) process or apparatus.7 This combined
configuration cannot be a solution to any Schr
oedinger equation and is no longer a pure state
described by a state function, say () . It can however be described by a density operator in
particular a mixed state density operator M
op defined by
(j)
wj (j)
M
op =

(2.27)

with
wj = 1 ,

(2.28)

where the wj are real, classical, probabilities that the j th pure state density operator (j) (j)
contributes to Eq.2.27. The sum in Eq.2.27 is not an example of the superposition principle. It
is not equivalent to or related to any linear coherent superpositions (with phase relationships)
that exhibit quantum interference. Eq.2.27 represents an ensemble of systems, (j) , combined
according to classical probabilities wj with no phase relationships.

2.2.1

Mixed state properties

The mixed state density operator has several properties that coincide with those of the pure state
density operator. For example:
1. Using Eqs.2.27 and 2.28 it is easy to show that the mixed state density operator has the
normalization property
Tr M
(2.29)
op = 1.
2. The average value of the observable op for the mixed state is
= Tr M
op op .

(2.30)

3. In the mixed state the probability of measuring the observable s is


P (s ) = Tr {M
op s s }

(2.31)

4. Following the steps for deriving Eq.2.23, the mixed state density operator has an equation of
motion
M
i
op
, Hop ] .
(2.32)
= [M
h op
t
6 Even if at some initial time a state starts out pure with maximal information, interactions with the environment
quickly destroy that purity the larger the system the more rapid the loss of purity.
7 A Stern-Gerlach machine is such an apparatus. See Examples 2.3.1.

2.2. MIXED STATES AND MACROSCOPIC SYSTEMS

2.2.2

15

Macroscopic consequences

A few distinctive characteristics make a mixed state density operator the logical entity for describing
macroscopic (thermodynamic) systems.
Macroscopic systems are always coupled to their environment. This presents a complicated quantum
mechanical situation requiring consideration of a many-particle microscopic system entangled with
its quantum environment.
Expanding the pure state density operator of Eq.2.1 in the linear coherent superposition [See
Eq.2.14]
= i i
(2.33)
i

we get the density operator result


op = P (i )i i + i ij j ,
i

(2.34)

ij

in which the first sum on the right-hand-side contains diagonal terms


P (i ) {i i } ,

(2.35)

where P (i ) are the probabilities of Eq.2.16 for which


P (i ) = 1 ,

(2.36)

whereas the second sum is over off-diagonal terms


i ij j

(2.37)

ij

which are typical of quantum interference terms. The density operator formulation allows a natural
partitioning of the original pure state density operator into two parts:
a. A part (the first sum) which will be shown to have less than complete information.
b. A remaining part (the second sum) representing the lost information.
The diagonal terms in Eqs.2.34 and 2.35 constitute, by themselves, a mixed state density operator
complete with all the properties of a mixed state density operator [see Eqs.2.29 - 2.32].
If the non-diagonal interference terms could (hypothetically?) be ignored, the density operator
in Eq.2.34 would cross a boundary from pure quantum results (displaying interference) into a
classical regime with no quantum interference. This is precisely what happens when entangled
environmental coupling is introduced. Environmental coupling quicikly suppresses the off-diagonal
interference terms in Eq.2.34, leaving a mixed state thermal density operator, op ,
op = P (i ) {i i } .
i

from which in principal macroscopic averages can now be calculated.

(2.38)

16

CHAPTER 2. A ROAD TO THERMODYNAMICS

The equation of motion for the thermal density operator op is found by the method leading to
Eq.2.32 and gives
op i
= [op , Hop ] .
(2.39)
t
h
Since we are interested in thermal equilibrium, i.e. where thermodynamic variables are not, on
average, changing in time we require
op
= 0.
(2.40)
t
It therefore follows that at thermal equilibrium
[op , Hop ] = 0

(2.41)

i.e. the mixed state thermal density operator op and the quantum mechanical hamiltonian Hop
commute, guaranteeing that the thermal density operator and the hamiltonian have the same set
of macroscopic eigentstates.
Basing Eq.2.38 on the specific linear superposition constructed from
Hop Es = Es Es ,

(2.42)

where Es are macroscopic system eigen-energies and Es are corresponding eigenstates, entangled
environmental coupling reduces Eq.2.38 to a specific thermal density operator
op = P (Es )Es Es ,

(2.43)

where P (Es ) is the probability the macroscopic system has eigen-energy Es . Eq.2.43 commutes
with Hop as required.8,9
The rigorous arguments eventually leading to Eq.2.43 are called environmentally induced decoherence.10,11,12,13

2.2.3

Thermodynamic state functions?

For example, applying the hamiltonian of Eq.1.12, macroscopic internal eigen-energies14 can be
found from
hop s = s s
8 Eq.2.43

(2.44)

is also called the spectral representation of op .


op
9 In non-equilibrium thermodynamics
0 which requires evaluation of Eq.2.39. This equation can be studied
t
by a systematic thermodynamic perturbation theory.
10 E. Joos and H. D. Zeh, The Emergence of Classical Properties through Interaction with the Environment Z.
Phys. B 59, 223 (1985).
11 Maximilian Schlosshauer,Decoherence, the measurement problem and interpretations of quantum mechanics,
Rev. Mod. Phys. 76, 1267-1305 (2004).
12 W. H. Zurek, Decoherence and the Transition to Classical, Phys. Today 44 36-44 (1991).
13 Maximilian Schlosshauer,Decoherence and the Quantum to Classical Transition, Springer, New York, 2007.
14 Internal energy is generally taken to be the kinetic and potential energy of constituent particles. Total energy,
on the other hand, reflects energy contributions from all interactions.

2.3. THERMAL DENSITY OPERATOR OP AND ENTROPY

17

and with these eigen-states a thermal density operator can be constructed


op = P (s )s s

(2.45)

from which, in principle the internal energy U is then


U = Tr op hop

(2.46)

= s P (s ) .

(2.47)

Thus a thermodynamic state variable seems to be emerging from quantum theory. However
there remains a formidable obstacle. Since no measurement or collection of independent measurements can reveal the macroscopic quantum probabilities P (s ) they are unknown. Even more
discouraging they are unknowable!! Is this fatal? The question is addressed in Chapter 6 where
we argue that the answer is NO! It is not fatal.

2.3

Thermal density operator op and entropy

Macroscopic quantum systems are entangled with an environment. As a result


1. They are not isolated.
2. They are not pure Schr
odinger states.
3. They leak information to their environment.
4. As defined in Eq.2.18, for macroscopic systems15,16 purity covers the range of values
[see Appendix A],
2

0 < I = Tr (op ) 1 .

(2.48)

Beginning an adventurous ascent, purity is replaced by an alternative logarithmic scale


2

s = ln Tr (op ) ,

(2.49)

which, according to Eq.2.48, now covers the range 0 s < where


s = 0 complete knowledge

(2.50)

s complete ignorance .

(2.51)

Finally, a more significant and solidly based alternative to purity [see Appendix B] is defined
as
F = Tr op ln op
15 The
16 Zero

following result is proved in Appendix A.


purity means that all outcomes are equally likely.

(2.52)

18

CHAPTER 2. A ROAD TO THERMODYNAMICS

where is a real, positive scale constant.17 It covers the same range as s but with a different
scaling:
F = 0 complete information (totally biased) the pure state
F total ignorance no information (perfectly unbiased) ,

(2.53)
(2.54)

where it is again emphasized that no information implies that all outcomes are equally likely
there is absolutely no outcome bias. Because of its special mathematical form and properties18 F
will ultimately become associated with entropy the quantity invented by Clausius,19 interpreted
by Boltzmann20 then formalized by Gibbs and especially von Neumann.21,22
Even with all its power, quantum mechanics offers no practical recipe for finding P (s ) or, obviously,
op (see Eq. 2.43). Overcoming this remaining obstacle compells us to look outside quantum
mechanics for a well reasoned postulate whose results accord with the measurable, macroscopic,
physical universe. This is addressed in Chapter 6.

2.3.1

Examples

1. Consider ions with spin +1/2 in a pure Schrodinger state 1/2 which can be expressed as a
linear coherent superposition (the superposition principle)
1/2 = +1/2 +1/2 1/2 + 1/2 1/2 1/2

(2.55)

where +1/2 and 1/2 are the complete set of spin 1/2 eigenstates quantized along the
z -axis.

1/2

1/
2

0.5 X - 1 /2

1
+ /2

1
+ /2

Figure 2.1: Stern-Gerlach apparatus with filter.


thermodynamics kB T (Boltzmanns constanttemperature.)
Appendix B.
19 R. Clausius, The Mechanical Theory of Heat, with its Applications to the Steam-Engine and to the Physical
Properties of Bodies, John Van Voorst London (1867)
20 Ludwig Boltzmann, Uber

die Mechanische Bedeutung des Zweiten Hauptsatzes der W


armetheorie, Wiener
Berichte 53,195220 (1866).
21 John von Neumann.Mathematical Foundations of Quantum Mechanics, Princeton University Press (1996).
22 Entropy was also introduced in a context of digital messaging information by Claude Shannon, A mathematical
theory of communication, Bell System Technical Journal, 27, 379-423, 623-656 (1948).
17 In

18 See

2.3. THERMAL DENSITY OPERATOR OP AND ENTROPY

19

The quantum probability an ion will be found in a +1/2 spin eigenstate is


2

P (+1/2) = +1/2 1/2

(2.56)

while the quantum probability an ion will be found in a 1/2 spin eigenstate is
2

P (1/2) = 1/2 1/2 .

(2.57)

After passing the beam through a magnetic field gradient, (Stern-Gerlach apparatus) ions
in the pure (1/2) states will emerge deflected along a pair of divergent trajectories with
probabilities w+ 12 = 21 and w 12 = 12 with w+ 21 + w 12 = 1.
Following this, each of the separated ion beams is itself deflected by, for example, a uniform
static magnetic field, with one of the beams, say the 1/2 , first passing through a purpose
designed filter that diminishes its intensity to one-half its initial value. Then, by incoherently
(classically) mixing the redirected pure +1/2 and 1/2 states, a new, single beam is
reconstituted. Clearly there is no single Schrodinger state function for the reconstituted
beam. But a mixed state density operator, as in Eq.2.27, is possible with
1
2
1
op2 = + 21 + 12 + 12 21
3
3

(2.58)

This is an incoherent (non-interfering) sum. There are no phase relationships between different
pure state contributions.
2. Consider a density matrix

=(

1
0

0
)
0

(2.59)

After matrix multiplication


() = (

1
0

0
1
)(
0
0

=(

1
0

0
)
0

0
)
0
(2.60)

its corresponding purity I = Tr () is


2

I =Tr ()
=1

(2.61)
(2.62)

corresponding to total information a pure state (totally biased).


An entropy, as introduced in Eq.2.52, is
F / = Tr ln .

(2.63)

20

CHAPTER 2. A ROAD TO THERMODYNAMICS


If f (M) is a function of the operator or matrix M, evaluating its trace

Tr f (M)

Tr
(2.64)

requires a bit more ingenuity. First, the assumption is made that the function f (M) implies
a power series in that matrix. For example
1
eM = 1 + M + M2 + .
2

(2.65)

Then, since the trace of an operator (matrix) is independent of the basis in which it is taken,
the trace can be written

Tr f (M) = mf (M) m

(2.66)

with
Mm = m m

(2.67)

where m are the eigenvalues of M and the matrix mf (M) m is in the basis that diagonalizes M. Referring to Eq.2.59 we see is already diagonal with eigenvalues {0, 1}. Therefore23

F
= {1 ln (1) + 0 ln (0)}
k
= 0.

(2.68)
(2.69)

This density operator has zero entropy, again corresponding to a pure state i.e. maximal
information (totally biased). It is equivalent to knowing the state function!
3. Now consider the density matrix
=(

1/2 0
)
0 1/2

(2.70)

for which matrix multiplication gives


2

() = (
=(

1/2 0
1/2 0
)(
)
0 1/2
0 1/2
1/4 0
).
0 1/4

(2.71)

Therefore its purity I is

Tr ()2 = 1/2
23 lim x ln x
x0

= 0.

(2.72)

2.3. THERMAL DENSITY OPERATOR OP AND ENTROPY

21

which corresponds to a mixed state with less than maximal information. Similarly the matrix
of Eq.2.70, which is already diagonal with eigenvalues {1/2, 1/2}, has entropy
F
= {1/2 ln (1/2) + 1/2 ln (1/2)}
k
= ln 2

(2.73)
(2.74)

which also implies missing information not a pure state. Moreover, the value F /k = ln 2
is the highest F /k a 2-dimensional density matrix can attain both possible states being
equally likely. Having attained this maximal value, F /k F max /k, it describes a situation
that is minimally biased and is now entitled to be called entropy [See Appendix B.]

Problems and Exercises


1. In a two-dimensional space a density matrix (hermitian with
=(

Tr = 1) is

X
)
1

a. Calculate the entropy, F /kB = Tr ln as a function of and X.


b. Find and X that minimizes entropy F /kB . What is F min /kB ?
c. Find and X that maximizes entropy F /kB . What is F max /kB ? [See Section 2.3.1.]
2. In a three-dimensional space a density matrix (hermitian with
1
= X
Y

X
2
Z

Tr = 1) is

Z
.
1 1 2

a. Find X, Y, Z, 1 and 2 that minimizes the entropy F /kB = Tr ln . What is F min /kB ?
b. Find X, Y, Z, 1 and 2 that maximize the entropy F /kB . What is F max /kB ?
3. Consider two normalized pure quantum states:

1
1 = + + 2/3
3
1
2
2 = + .
5
5

(2.75)
(2.76)

where + and are orthonormal eigenstates of the operator Sz .


Let them be incoherently mixed by some machine in equal proportions, i.e. w1 = w2 = 1/2.
a. Find the two dimensional density matrix (in the basis + , ) corresponding to this
mixed state.
b. Find the average value Sz in a measurement of the mixed states Sz spin components.

22

CHAPTER 2. A ROAD TO THERMODYNAMICS


c. Find the entropy F /kB for this mixed state.
4. Consider two normalized pure quantum states:

1
1 = + + 2/3
3
1
2
2 = + .
5
5

(2.77)
(2.78)

where + and are orthonormal eigenstates of the operator Sz .


Let them be incoherently mixed by some machine such that: w1 = 1/3, w2 = 2/3.
a. Find the two dimensional density matrix (in the basis + , ) corresponding to this
mixed state.
b. Find the average value Sz in a measurement of the mixed states Sz spin components.
c. Find the entropy F /kB for this mixed state.

... there is in the physical world one agent only, and this is called
Kraft [energy]. It may appear, according to circumstances, as motion,
chemical affinity, cohesion, electricity, light and magnetism; and from
any one of these forms it can be transformed into any of the others.
Karl Friedrich Mohr - Zeitschrift f
ur Physik (1837).

Chapter 3

Work, heat and the First Law


3.1

Introduction

Taking to heart Professor Murray Gell-mans epigraphic observation in Chapter 2, a respite from
quantum mechanics seems well deserved. Therefore, continuing the discussion begun in Chapter 1
the First Law of Thermodynamics is restated:
U = Q W ,

(3.1)

where U is a thermodynamic state function with quantum foundations. In this chapter the First
Laws classical terms W (work) and Q (heat) are discussed, followed by a few elementary applications.

3.1.1

Work, W

Work, W is energy transferred between a system and its surroundings by mechanical or electrical
changes at its boundaries. It is a notion imported from Newtons mechanics with the definition
W = F dr
(3.2)
where F is an applied force and dr is an infinitesimal displacement. The integral is carried out
over the path of the displacement.
Work is neither stored within a system nor does it describe any state of the system. Therefore
?
W = Wf inal Winitial has absolutely no thermodynamic meaning!

Incremental work, on the other hand, can be defined as d


W, with finite work being the integral

W = d
W

23

(3.3)

24

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

The bar on the d is a reminder that d W is not a true mathematical differential and that the
integral in Eq.3.3 depends on the path.1
Macroscopic parameters are further classified as:
A. extensive proportional to the size or quantity of matter (e.g. energy, length, area, volume,
number of particles, polarization, magnetization)
B. intensive independent of the quantity of matter (e.g. force, temperature, pressure, density,
tension, electric field, magnetic field.)
Incremental work has the generalized meaning of energy expended by/on the system due to an
equivalent of force (intensive) acting with an infinitesimal equivalent of displacement (extensive)
which, in accord with Eq.3.2, is written as the conjugate pair,
extensive

dW =

Fj

dxj

(3.4)

intensive

where Fj is some generalized force and dxj is its conjugate differential displacement. If work takes
place in infinitesimal steps, continually passing through equilibrium states, the work is said to be

quasi-static and is distinguished by the representation d


WQS .
Examples of conjugate work pairs () are
1.

pressure (p) volume ( dV ) ,

2.

tension ( ) elongation ( d) ,

3.

stress () strain ( d) ,

4.

surface tension () surface area ( dAS ) ,

5.

electric field (E) polarization ( dP) ,

6.

magnetic field (B) magnetization ( dM ).2

7.

chemical potential () particle number ( dN )

[It is often helpful in thermodynamics to think of mechanical work as the equivalent of raising and
lowering weights external to the system.]
Finite work, as expressed by Eq.3.3, depends on the the integration path the details of the process
and not just on the process end points.
The sign convention adopted here3 is

d
W > 0 system performs work

gas
(d Wby
= p dV )

d
W < 0 work done on the system

gas
(d
Won
= p dV )

is not a function of independent thermal variables so that dW cannot be an exact differential.


fields used in representing electric and magnetic work will be further examined in Chapter 11.
3 In the case of pressure and volume as the conjugate pair.
1 Work

2 The

3.1. INTRODUCTION

3.1.2

25

Heat, Q

Heat, Q is energy transferred between a system and any surroundings with which it is in diathermic
contact. Q is not stored within a system and is not a state of a macroscopic system. Therefore
?
Q = Qf inal Qinitial has no thermodynamic meaning! On the other hand, incremental heat is
defined as d Q, with finite heat Q

Q= d
Q.

(3.5)

Finite heat depends on the details of the process and not just the integrals end points.

The bar on the d is an essential reminder that d


Q is not a function of temperature or any

d
Q
other independent variables. It is not a true mathematical differential so derivatives like
dT
have dubious meaning.

Processes for which Q = 0, either by thermal isolation or by extreme rapidity (too quickly to
exchange heat), are said to be adiabatic.

3.1.3

Temperature, T

Temperature is associated with a subjective sense of hotness or coldness. However in thermodynamics temperature has empirical (objective) meaning through thermometers e.g. the height
of a column of mercury, the pressure of an ideal gas in a container with fixed volume, the electrical
resistance of a length of wire all of which vary in a way that allows a temperature scale to be defined. Since there is no quantum hermitian temperature operator a formal, consistent definition
of temperature is elusive. But it is achievable in practice through thermal measurements4 showing,
with remarkable consistency, how and where it thermodynamically appears and that it measures
the same property as the common thermometer.

3.1.4

Internal energy, U

As discussed in Chapter 1, internal energy U is a state of the system. Therefore U does have
meaning and dU is an ordinary mathematical differential5 with
B

U = U(B) U(A) = dU .

(3.6)

where A and B represent different macroscopic equilibrium states. Clearly, U depends only on the
end points of the integration and not on any particular thermodynamic process (path) connecting
them.
4 This
5 It

subject will be discussed in Chapter 6.


is an exact differential.

26

3.2

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Exact differentials

All thermodynamic state variables are true (exact) differentials with change in value as defined in
Eq.3.6. Moreover, state variables are not independent and can be functionally expressed in terms
of other state variables. Usually only a few are needed to completely specify any state of a system.
The precise number required is the content of an important result which is stated here without
proof.
The number of independent parameters needed to specify a thermodynamic state is
equal to the number of distinct conjugate pairs [see Subsection 3.1.1] of quasi-static
work needed to produce a differential change in the internal energy plus ONE6
For example if a gas undergoes mechanical expansion or compression, the only quasistatic work is

mechanical work, d
WQS = p dV . Therefore only two variables are required to specify a state. We
could then write the internal energy of a system as a function of any two other state parameters.
For example we can write U = U (T, V ) or U = U (p, V ). To completely specify the volume of a
gas we could write V = V (p, T ) or even V = V (p, U) . The choice of functional dependences is
determined by the physical situation and computational objectives.
If physical circumstances dictate particle number changes, say by evaporation of liquid to a gas
phase or condensation of gas to the liquid phase, or chemical reactions where atomic or molecular

species are lost or gained, a thermodynamic description requires chemical work d


WQS = dN
where dN is an infinitesimal change in particle number and is the (conjugate) chemical potential .
Chemical potential is an intensive variable denoting energy per particle while particle number
N is obviously an extensive quantity. For cases where there is both mechanical and chemical
work,

d
WQS = p dV dN ,
(3.7)
three variables are required and, for example, U = U (T, V, N ) or U = U (p, V, ).
Incorporating the definitions above, an incremental form of the First Law of Thermodynamics
is

dU = d
Qd
W
(3.8)

3.3

Equations of state

Equations of state are functional relationships among a systems equilibrium state parameters
usually those that are most easily measured and controlled in the laboratory. Determining equations
of state, by experiment or theory, is often a primary objective in practical research for then
thermodynamics can fulfill its role in understanding and predicting state parameter changes for
particular macroscopic systems.
The following are samples of equations of state for several different types of macroscopic systems.
Their application is central in the Examples subsections below.
6 The one of the plus one originates from the conjugate pair T and dS, where S is Entropy, about which much
will be said throughout the remainder of this book.

3.3. EQUATIONS OF STATE

27

1. Equations of state for gases (often called gas laws) take the functional form f (p, V, T ) = 0 where
p is the equilibrium internal gas pressure, V is the gas volume, T is the gas temperature and
f defines the functional relationship. For example, a gas at low pressure and low particle
density, n = N /V , where N is the number of gas particles (atoms, molecules, etc.), has the
equation of state
p (N /V ) kB T = 0
(3.9)
or, more familiarly (the ideal gas law )
pV = N kB T

(3.10)

where kB is Boltzmanns constant.7,8


2. An approximate equation of state for two and three dimensional metallic materials is usually
expressed as g(, , T ) = 0 where = /A is the stress,  = (` `0 )/ `0 is the strain9 , T is the
temperature, A is the cross-sectional area and is a characteristic constant. An approximate
equation of state for a metal rod is the Hookian behavior
 = .

(3.11)

3. An equation of state for a uniform one-dimensional elastomeric (rubber-like) material has the
form (, , T ) = 0 where is the equilibrium elastic tension, is the sample elongation, T
is the temperature and is the functional relationship. For example, a rubber-like polymer
that has not been stretched beyond its elastic limit can have the simple equation of state
K T = 0,

(3.12)

K T  = 0

(3.13)

or, alternatively

where K and K are material specific elastic constants.10,11


4. Dielectrics with no permanent polarization in weak polarizing fields have an equation of state
Pi = ij (T ) Ej

(3.14)

where Pi is an electric polarization vector component, ij (T ) is the temperature dependent


electric susceptibility tensor and Ej is an electric field component.12
7 For

a derivation see Chapter 7.


gases under realistic conditions Eq.3.10 may be inadequate requiring more complicated equations of state.
Nevertheless, because of its simplicity and zeroth order validity, the ideal gas law is pervasive in models and
exercises.
9 ` is the stretched length and ` is the un-stretched length
0
10 H. M. James and E. Guth, Theory of Elastic Properties of Rubber, J. Chem. Phys. 11, 455-481 (1943).
11 Comparing these equations of state with Eq.3.11 suggests a physically different origin for rubber elasticity. This
will be explored in Chapter 10.
12 Is this an external electric field or some average local internal electric field? How do Maxwells fields fit in here?
This will be addressed in Chapter 11. For the present, we assume that the sample is ellipsoidal in which case the
internal and external electric fields are the same.
8 For

28

CHAPTER 3. WORK, HEAT AND THE FIRST LAW


In stronger fields equations of state may acquire nonlinear terms13 e.g.
(2)

(1)

Pi = ij (T )Ej + ijk (T )Ej Ek .


j

(3.15)

j,k

5. Non-permanently magnetic materials which are uniformly magnetized in weak magnetic fields
are described by the equation of state
M = M (T ) B0

(3.16)

where M is the magnetization vector, M (T ) is a temperature dependent magnetic susceptibility tensor and B0 is the existing magnetic induction field prior to insertion of the
sample.14,15

3.3.1

Examples I

3.3.1.1

Quasistatic reversible work

Figure 3.1: Quasi-static, reversible work. A piston loaded down with a pile of fine sand.
A gas is confined in a cylinder by a frictionless, massless piston of negligible mass and cross-sectional
area A. The gas pressure p is counteracted by a pile of fine sand of total mass m resting on the
piston [see Figure 3.1] so that it is initially at rest. Neglecting any atmospheric (external) force
13 Robert

W. Boyd, Nonlinear Optics, Academic Press, San Diego CA (2003).


and magnetic work will be discussed in Chapter 11.
15 Solid state physics text books often express Eq.3.16 in terms of an external magnetic field H.
14 Magnetism

3.3. EQUATIONS OF STATE

29

acting on the piston, the downward force exerted by the sand is, initially, F0sand = mg. Therefore
the initial equilibrium gas pressure, pgas
0 , is
pgas
=
0

mg
.
A

(3.17)

Sand is now removed, one grain at a time allowing the gas to expand in infinitesimal steps,
continually passing through equilibrium states. Moreover, it is conceivable that at any point a
single grain of sand could be returned to the pile and exactly restore the previous equilibrium
state. Such a real or idealized infinitesimal process is called quasi-static. A quasi-static process,
by definition, takes place so slowly that the system (in this example a gas) always passes through
equilibrium states so that an equation of state always relates system pressure, volume, temperature,
tension, etc.
Note: If, as in this case, the previous state can be restored by an infinitesimal displacement with
no dissipation of energy the process is also said to be reversible.16
Assuming the quasi-static expansion also takes place isothermally at temperature TR , what work
is done by the gas in raising the piston from h1 to h2 ?
At all stages of the quasi-static expansion equilibrium is maintained with pressure pgas and temperature TR well defined. Therefore incremental work done by the gas (assumed ideal) is

d
WQS = pgas dV
N kB TR
dV .
=
V

(3.18)
(3.19)

Integrating Eq.3.19 along the isothermal path [as pictured in Figure 3.4] gives
V2 =A h2

WQS = N kB TR

V1 =A h1

dV
V

= N kB TR ln (h2 /h1 ) .
3.3.1.2

(3.20)
(3.21)

Irreversible work

Instead of a restraining sand pile the piston is now pinned a distance h1 above the bottom of the
cylinder [See Figure 3.2]. At the same time a constant atmospheric pressure P atm exerts an external
force F atm on the piston,
F atm = P atm A .

(3.22)

Initially force equilibrium is maintained with


A pgas
= Fpin + A P atm
0

(3.23)

where Fpin is the pinning force, so that pgas


> P atm .
0
16 Not all quasi-static processes are reversible but all reversible processes are quasi-static. Reversibility implies no
friction or other energy dissipation in the process.

30

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Figure 3.2: Irreversible work. A gas filled piston initially pinned at height h1 .
Quickly (and frictionlessly) extracting the pin frees the piston allowing the gas to expand rapidly
and non-uniformly. After things quiet down the gas reaches an equilibrium state, with the piston
having moved upward a distance ` = h2 h1 with a final gas pressure pgas
= P atm . This isothermal
f
process characterized by a finite expansion rate and system (gas) non-uniformity is neither
quasi-static nor reversible.17,18
What isothermal work is done by the rapidly expanding gas in raising the piston a distance `?
The gas expands irreversibly. Therefore during expansion a gas law does not apply. But initially
the gas has equilibrium pressure pgas
with volume V0
0
V0 =

N kB T
.
pgas
0

(3.24)

After the piston has returned to rest (equilibrium),


pgas
= P atm
f
17 All

(3.25)

real processes in nature are irreversible. Reversibility is a convenient idealization for thermodynamic model-

ing.
18 Irreversible as used here, is not intended as an absolute, since surroundings can always be altered to recover the
initial equilibrium state. But this is not the same as the sequence of infinitesimal steps associated with reversibility.

3.3. EQUATIONS OF STATE

31

with volume
Vf =

N kB T
.
P atm

(3.26)

While expanding irreversibly the gas lifts the piston against a constant force F atm (equivalent to
raising a weight mg = F atm in the earths gravitational field.)19 The incremental, irreversible work
done by the confined gas (the system) is

d
W = P atm dV

(3.27)

which, with constant P atm , is integrated to give


W = P atm (Vf V0 ) .

(3.28)

With Eqs. 3.24 and 3.26


W = N kB T [1

3.3.1.3

P atm
].
pgas
0

(3.29)

Quasistatic, irreversible work

As above, the gas is initially confined by a piston exposed to the atmosphere with its initial position
secured by a pin. However the pistons movement is restricted by a frictional wedge which,
together with the external atmosphere, slows the pistons upward movement, enabling it to be
approximated as quasi-static but irreversible. [See Figure 3.3.]
Assuming the process is isothermal, what work is done by the expanding gas?
The expanding gas (the system) raises the piston against atmospheric force Fatm and the force of
friction Ff riction , both taken to have constant value.
Therefore the work performed is

d
Wsys =

(Fatm + Ff riction )
dV
A

(3.30)

which after integration becomes


Wsys = (P atm + Ff riction /A) (Vf V0 )

(3.31)

where V0 is the initial equilibrium volume


V0 =

19 The

N kB T
Fpin /A
P atm (1 + atm )
P

piston starts at rest and returns to rest, so there is no change in the pistons kinetic energy.

(3.32)

32

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Figure 3.3: Quasi-static, irreversible work. A piston slowed by a frictional wedge.


and Vf is the final equilibrium volume
Vf =

N kB T
.
P atm

(3.33)

Therefore
Wsys =

3.3.2

atm
+ Ff riction /A)
N kB T Fpin (P
(
)
atm
atm
P
A
(P
+ Fpin /A)

(3.34)

The p-V diagram

In performing quasi-static work a system passes through equilibrium states, so that it is possible to
draw a continuous curve on a p V diagram describing the process. A constant temperature work
path is represented by the isotherm in Figure 3.4, where the work is the shaded area under the
isotherm.

p1

33

pressure

3.3. EQUATIONS OF STATE

T=constant

p2

p=0

volume
v1

v2

Figure 3.4: p V diagram for an isothermal, quasi-static process.

3.3.3

Examples II

3.3.3.1

A loaded rubber band

A mass 2 m affixed to the end of a vertically stretched, massless rubber band, is suspended a distance
h1 (rubber band length `1 ) above a table. When half the mass is suddenly removed the rubber band
rapidly contracts causing the remaining mass m to rise. After oscillating a while it finally comes to
rest a distance h2 , (h2 > h1 ) (rubber band length `2 ) above the table. [See Figure 3.5.]
1. What is the work done by the rubber band?
After returning to rest the work done (not quasi-static) by the rubber band (work done by
the system) in raising the weight m (against gravity) is
W = mg`

(3.35)

where ` = `2 `1 < 0.
2. What is the change in the rubber bands internal energy?
Since the process takes place rapidly it is irreversible. Moreover there is not enough time for
the rubber band to lose heat to the surroundings and the process can be regarded as adiabatic.
Using the First Law
U = +mg ` .
(3.36)
The rubber band loses internal energy.

34

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Figure 3.5: A loaded rubber band, when released, rapidly contracts.


3. This time, in counteracting rubber elasticity the total mass 2 m is restrained to rise slowly
(quasi-statically) but isothermally. What is the work done by the contracting rubber band?
We can safely use an equilibrium tension (an elastic state parameter) which acts to raise the
weight. The incremental quasi-static elastic work done by the system is

d
WQS = d

(3.37)

where d is the differential of elastic extension. Using the rubber equation of state, Eq.3.12,
and integrating over the range of extension
0

WQS = K T d

(3.38)

= K T (`)
3.3.3.2

(3.39)

A compressed metal rod

A metal rod, cross sectional area A, is suddenly struck by a sledgehammer subjecting it to a


transient but constant compressive (external) force F which momentarily changes the rods length
by ` = `f `0 < 0.

3.4. HEAT CAPACITY

35

1. What is the work done on the rod during the interval that its length decreases by `?
Sudden compression of the rod implies that it does not pass through equilibrium states so an
equilibrium internal stress in the bar = f /A is not defined. But the mechanical work done
on the rod by the constant compressive force is, however
W = F ` .

(3.40)

2. On the other hand, if the rod could be compressed infinitely slowly (perhaps by a slowly
cranked vise) the rods internal stress is a well defined equilibrium property and the quasistatic incremental work done by the system (i.e. the rod) is

d
WQS = V d

(3.41)

where V is the rods volume, is the internal stress and d = ( d`/`) defines the differential
strain. We could now apply an approximate equation of state, such as Eq.3.11, and get
f

WQS = V (/) d

(3.42)

0

= (V /2) (2f 20 )

3.4

(3.43)

Heat capacity

An infinitesimal of thermal energy d


QQS quasi-statically entering or leaving a system generally
produces an infinitesimal change in temperature dT

d
QQS = C dT

(constant ) .

(3.44)

(Exceptions are certain first-order phase transitions, such as boiling of water or melting of ice.) The
proportionality constant between the two infinitesimals C , together with the list of state variables,
, that are fixed during the quasi-static process, is called heat capacity at constant .20 The units
of heat capacity are energyK 1 . Specific heats are alternative intensive quantities, having units
energyK 1 per unit mass or energyK 1 per mole.

3.4.1

Heat transfer at constant volume

If heat is transferred at constant volume 21 , by the definition in Eq.3.44

d
QQS = CV dT

(constant volume).

(3.45)

20 The term heat capacity is a holdover from early thermodynamics when heat was regarded as a stored quantity.
Although we now know better, the language seems never to have similarly evolved.
21 and constant particle number

36

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Heat capacities can be equivalently expressed in terms of partial derivatives of state parameters.
For example, if a gas performs only quasi-static mechanical work, the infinitesimal First Law for
the process is

dU = d
QQS p dV
(3.46)
where p dV is quasi-static incremental work done by the gas (system.) Since for this case U can be
expressed in terms of two other thermal parameters, choose U = U (T, V ) so that its total differential
is
U
U
) dV + (
) dT
(3.47)
dU = (
V T
T V
and Eq.3.46 can then be rewritten

d
QQS = [(

U
U
) + p] dV + (
) dT .
V T
T V

(3.48)

Thus, for an isochoric (constant volume) path, i.e. dV = 0,


U
) dT
T V

(3.49)

U
) .
T V

(3.50)

d
QQS = (

and from Eq.3.45


CV = (

3.4.2

Heat transfer at constant pressure

Thermodynamic experiments, especially chemical reactions, are often carried out in an open atmosphere where the pressure is constant. In such cases we define

d
QQS = Cp dT

(constant pressure).

(3.51)

Expressing Cp as the partial derivative of a state parameter requires introducing a new, abstract,
state variable. Here again the infinitesimal First Law description of the quasi-static process is

dU = d
QQS p dV .

(3.52)

d (pV ) = p dV + V dp

(3.53)

d
QQS = d (U + pV ) V dp .

(3.54)

H = U + pV

(3.55)

But this time, using

Eq.3.52 can be rewritten as

Defining

3.4. HEAT CAPACITY

37

where H also a state function is called Enthalpy.22 This permits recasting a new differential
Law

d
QQS = dH V dp .

(3.56)

Then, expressing H in terms of T and p, i.e. H = H (T, p) so that a total differential is


dH = (

H
H
) dp + (
) dT ,
p T
T p

(3.57)

for an isobaric (constant pressure) path, i.e. dp = 0,

d
QQS = (

H
) dT
T p

(3.58)

and we have the definition of Cp in terms of state variables


Cp = (

H
) .
T p

(3.59)

Heat capacity measurements play a surprisingly important role in exploring microscopic properties
of matter. It was Einsteins curiosity about the unexpectedly small low temperature heat capacity
of diamond that led him to apply Plancks quantum theory to the solid state. Equations of state,
together with heat capacities, can offer a complete thermal picture of most matter.

3.4.3

Examples III

3.4.3.1

Ideal monatonic gas

An ideal monatomic gas, volume V0 , temperature T0 , is confined by a frictionless, movable piston


to a cylinder with diathermic walls. The external face of the piston is exposed to the atmosphere
at pressure P0 . A transient pulse of heat is injected through the cylinder walls causing the gas to
expand against the piston until it occupies a final volume Vf .
1. What is the change in temperature of the gas?
At the beginning and end of the expansion the system is in equilibrium so that an equation
of state applies, i.e. initially P0 V0 = N kB T0 and finally P0 Vf = N kB Tf . Therefore
T =

P0 V
N kB

(3.60)

where T = Tf T0 and V = Vf V0 .
22 Enthalpy is one of several defined state variables that are introduced to simplify specific thermodynamic constraints. In this case the constraint is constant pressure. Others will be introduced as needed.

38

CHAPTER 3. WORK, HEAT AND THE FIRST LAW


2. How much heat Q was in the pulse? The gas will expand turbulently, i.e. irreversibly, against
constant atmospheric pressure P0 (an external weight) so that incremental work done by
the expanding gas is

d
W = P0 dV .
(3.61)
Applying the First Law
Q = U + P0 V .

(3.62)

and using Eqs.3.47, 3.59 and 3.60 gives the internal energy change for this ideal gas
U = CV T
CV P0 V
=
N kB

(3.63)
(3.64)

Thus the injected heat is


Q=(

CV
+ 1) P0 V
N kB

(3.65)

where Eqs.3.62 and 3.64 have been used. Using another ideal gas property23
Cp CV = N kB

(3.66)

= Cp /CV ,

(3.67)

P0 V .
1

(3.68)

and the definition

Eq.3.65 can be condensed to


Q=

3.4.3.2

A steel wire

A steel wire, length L and radius is held with zero tension between two firmly planted rigid
vertical posts, as shown in Figure 3.6 . The wire, initially at a temperature TH , is then cooled to
a temperature TL .
Find an expression for the tension in the cooled wire.
The thermodynamic state variables used for describing elastic behavior in the steel wire (or rod) are
stress = /A where A is the wires cross sectional area, and strain  = (` `0 )/`0 where ` and `0
are the wires stretched and unstretched length. These are the same variables referred to in elastic
work [see Eq.3.41] and in the elastic equation of state [see Eq.3.16].
In the present situation the rigid posts constrain the length of the wire to remain constant so that
a decrease in temperature is expected to increase the strain (tension) in the wire. Referring to
23 This

well known result is stated, for now, without proof.

3.4. HEAT CAPACITY

39

Figure 3.6: A wire held between firmly planted rigid vertical posts.
Section 3.2, consider a functional dependence of strain  = (T, ), which are the physically relevant
state variables for the wire, and take the total differential
d = (



) dT + ( ) d .
T
T

(3.69)

The partial derivatives represent physical properties of the system under investigation for which
extensive numerical tables can be found in the literature or on the WEB.24 In Eq.3.69
(


) = L
T

(3.70)

where L is the coefficient of linear thermal expansivity, and


(


1
) =
T ET

(3.71)

where ET is the isothermal Youngs modulus.


Thus we can write

1
d ,
ET
and since the length of the wire is constant, i.e. d = 0, we have
d = L dT +

0 = L dT +

1
d .
ET

(3.72)

(3.73)

Finally, assuming L and ET are average constant values, integration gives


= (TL ) (TH ) = L ET (TL TH ) ,

(3.74)

and from the definition of stress, (TH ) = 0


(TL ) = AL ET (TL TH )
2

= r L ET (TH TL )
24 See

for example; http://www.periodictable.com/Properties/A/YoungModulus.html

(3.75)
(3.76)

40
3.4.3.3

CHAPTER 3. WORK, HEAT AND THE FIRST LAW


A rapidly expanding gas

A gas is allowed to rapidly expand from an initial state (p0 , V0 , T0 ) to a final state f.
What is the final state (pf , Vf , Tf )?
This rapid process can be regarded as taking place adiabatically, Q = 0. Applying the infinitesimal
First Law we have,

d Q = 0 = dU + d
W.
(3.77)
Since U is a state variable the integral
U = dU

(3.78)

is path (i.e. process) independent. Therefore Eq.3.77 is integrated to become

0 = U + d
W

(3.79)

which depends only on equilibrium end points. So the equilibrium end points can be connected by
a unique, quasi-static process
pf ,Vf ,Tf

0=

[ dU + p dV ] .

(3.80)

p0, V0, T0

Then with dU as in Eq.3.47


pf ,Vf ,Tf

0=

{[(

p0, V0, T0

U
U
) + p] dV + (
) dT } .
V T
T V

(3.81)

So far the result is general (any equation of state). But for simplicity, assume the gas is ideal. Using
Eq.3.49 and applying the ideal gas property25
U
) =0
V T

(3.82)

together with the ideal gas equation of state [see Eq.3.10] gives
pf ,Vf ,Tf

0=

{[

p0, V0, T0

N kB T
] dV + CV dT }
V

(3.83)

which implies
N kB T
] dV + CV dT .
V
This differential equation separates quite nicely to give upon integration
0=[

N kB ln (
25 This

Vf
T0
) = CV ln ( )
V0
Tf

useful ideal gas result is the basis of a later problem.

(3.84)

(3.85)

3.4. HEAT CAPACITY

41

or
CV

Vf N k B
T0
)
=( )
V0
Tf

(3.86)

which with Eqs.3.66 and 3.67 becomes

3.4.3.4

Vf 1
T0
)
=( ).
V0
Tf

(3.87)

The filling problem atmospheric gas filling an empty vessel

An insulated, rigid, very narrow necked vessel is evacuated and sealed with a valve. The vessel
rests in an atmosphere with pressure P0 and temperature T0 . [Figure 3.7.] The valve is suddenly
opened and air quickly fills the vessel until the pressure inside the vessel also reaches P0 . If the air
is treated as an ideal gas, what is the temperature Tf of the air inside the vessel immediately after
equilibrium is attained?
In this problem we introduce the strategy of replacing an essentially open system (the number of
molecules inside the vessel is not constant) with a closed system (the gas is conceptually bounded
by a movable piston that seals the neck and, ultimately the closed vessel below, as in Figure 3.7).
This notional replacement is a common trick in thermodynamics and we take the opportunity to
illustrate it here.

Figure 3.7: Evacuated, insulated vessel in an atmosphere with pressure P0 and temperature T0 .
Physically, the narrow neck separates the air inside the vessel from the air outside the vessel, the
latter maintained at constant pressure P0 by the infinite atmosphere. The strategy is to replace the
external atmosphere with a piston that forces outside air across the narrow neck into the vessel by
applying the constant external pressure P0 , as shown in Figure 3.8.

42

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

N T

Tf v f

Figure 3.8: Evacuated vessel filling under pressure P0 .


Focus attention on a small, fixed amount of air being forced into the vessel by the conjectured
piston, say N molecules at pressure P0 and temperature T0 occupying a small volume v0 . Therefore
in forcing these molecules through the valve the piston displaces the volume v0 . The work done on
the N molecules by the constant pressure piston is

Wby the piston = d


W

(3.88)

= P0 v

(3.89)

= P 0 v0 .

(3.90)

Therefore the work done by the N molecules of gas is


Wby the gas = P0 v0 .

(3.91)

Once the N molecules get to the other side of the valve and enter the evacuated, rigid walled vessel,
they do NO work.26 Therefore the total work done by the gas is
Wtotal = P0 v0 + 0

(3.92)

Applying the First Law to the N molecules, while approximating this sudden process as adiabatic 27
0 = U + Wtotal
26 This
27 Too

is evident since no weights are raised or lowered.


fast to allow heat exchange with the external atmosphere or outside the vessel.

(3.93)

3.4. HEAT CAPACITY

43

or
f

0 = P0 v0 + dU .

(3.94)

Choosing to write U = U (V, T ) the total differential is


dU = (

U
U
) dT + (
) dV .
T V
V T

(3.95)

Since
(

U
) = CV
T V

(3.96)

and for an ideal gas

U
) =0
V T

(3.97)

we have
P0 v0 = CV (Tf T0 ) .

(3.98)

Then with the ideal gas equation of state


N kB T = CV (Tf T0 )

(3.99)

and the ideal gas expressions Eqs.3.66 and 3.67, this becomes
Tf = T0 .

(3.100)

At room temperature (T0 = 300K) and for an ideal monatomic gas ( = 5/3), Eq.3.100 gives Tf = 500K,
indicating a substantial rise in temperature for the gas streaming into the vessel.

3.4.4

Concluding remarks

In this hiatus from quantum mechanics we have resumed the discussion of classical thermodynamics,
emphasizing the distinction between U and purely classical Work W and Heat Q.
Chapter 4 pauses to develop some mathematical skills essential for thermodynamic applications.

44

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

Problems and exercises


1. Isothermal compressibility T is defined as
T =

1 V
(
)
V p T

(3.101)

a. Describe in words the meaning of T .


b. Find T for an ideal gas.
c. A more descriptive gas law is the van der Waals equation
(p +

aN 2
) (V N b) = N kB T
V2

(3.102)

where the constant b describes volume excluding short range repulsion between gas molecules
and the constant a takes account of long range intermolecular attraction. Find T for a
van der Waals gas.
2. A thick rubber band diameter d is fastened between two rigid vertical posts a distance L
apart. When the temperature is T0 the rubber band has positive tension 0 .
Assuming rubber has an equation of state
= K T

(3.103)

at what temperature T1 will the tension be halved?


3. An ideal gas expands adiabatically from the state (V0 , p0 ) to the state (Vf , pf ). Show that
the work done by the expanding gas is
W=

(p0 V0 pf Vf )
1

(3.104)

Cp
.
Cv

(3.105)

where
=

4. The constant volume heat capacity for a hypothetical nearly ideal gas is, at low temperature,
CV = a + bT

(3.106)

where a and b are constants. Show that for an adiabatic, reversible expansion of the gas
carried out at low temperature
pV exp {(b/a) T }

(3.107)

with
=

Cp
.
Cv

(3.108)

3.4. HEAT CAPACITY

45

5. Although the heat capacity at constant volume is


CV = (

U
) ,
T V

(3.109)

show that for the heat capacity at constant pressure


Cp (

U
)
T p

(3.110)

by proving that
Cp = (

U
) + pV
T p

(3.111)

where is the volume thermal expansivity


=

1 V
(
) .
V T p

(3.112)

Note: The heat capacity at constant pressure is


H
)
T p

(3.113)

H = U + pV .

(3.114)

Cp = (
where Enthalpy H is defined as

46

CHAPTER 3. WORK, HEAT AND THE FIRST LAW

To those who do not know Mathematics it is difficult to get across a


real feeling as to the beauty, the deepest beauty of nature. ... If you
want to learn about nature, to appreciate nature, it is necessary to
understand the language that she speaks in.
Richard Feynman, The Character of Physical Law, M.I.T. Press,
1967

Chapter 4

A mathematical digression
4.1

Thermodynamic differentials

In thermodynamics the equilibrium state of a system is described by macroscopic variables, say


U, p, T, V, .., which can roughly be classified as
1. Variables that are readily measurable in a laboratory, e.g. p, V, T, Cv , etc.
2. Variables that are less accessible to direct measurement, e.g. U, H , and others, soon to be
introduced.
The choice of variables is generally from among those that appear in thermodynamic laws, those
that can be measured or controlled in laboratories (such as appear in equations of state) and those
defining process constraints, i.e. isothermal, isobaric or adiabatic. Since physics of interest involves
changes in equilibrium macroscopic states, we must study differential relationship among them.
Assuming a state variable (x, y, z), where x, y, z are also (independent) state variables, these
relationships have the mathematical form
d (x, y, z . . .) = (

)
dx + ( )
dy + ( )
dz . . .
x y,z,...
y x,z,...
z x,y,...

(4.1)

where
(

)
,
x y,z,...

)
,
y x,z,...

)
z x,y,...

(4.2)

are partial derivatives, i.e. derivatives with respect to x, y, z . . . , with the remaining independent
variables held constant. In thermodynamics the variables that are held constant dictate specific
experimental conditions and must be retained in all expressions. They are not merely inconvenient
baggage that one is free to carelessly misplace.
Several mathematical rules associated with differentials and partial differentiation especially useful
in thermodynamics are introduced in this chapter and then applied in sample scenarios.
47

48

4.1.1

CHAPTER 4. A MATHEMATICAL DIGRESSION

Exact differentials

Finite changes in thermal state parameters, e.g. U, depend only on final and initial state parameter
values and not on the path or process driving the change. This independence defines exactness
of a differential. More generally, exactness of any thermal parameter, say Y, means that the closed
circuit integral dY satisfies the closed path condition
dY = 0 .

(4.3)

Even though integrals corresponding to finite work d


W = W and finite heat d
Q = Q have macroscopic meaning, as stressed earlier W and Q are not state parameters and do not have exact

differentials, i.e. d
W 0 and d
Q 0. That is the meaning of the bar on the d.

4.1.2

Exactness

A class of thermodynamic identities called Maxwell Relations play an important role in problem
solving strategies by replacing unfriendly partial derivatives i.e. those that do not obviously
represent accessible measurables with equivalent friendly partial derivatives that contain heat
capacities or state variables that are part an equation of state. For example, measured values of the
following partial derivatives, called response functions, are found in tables of physical properties of
materials1 and can be placed in the friendly category:
p
)
isothermal bulk modulus
V T
1 V
)
isothermal compressibility
T = (
V p T
1 V
)
coefficient of thermal expansion .
p = (
V T p

T = V (

(4.4)
(4.5)
(4.6)

Several elastic coefficients previously encountered in Chapter 3 may also be found in tables. Partial
derivatives in the unfriendly category will appear in our analyses and will certainly be recognized.
Methods for dealing with them are dealt with later in the chapter.

4.1.3

Eulers criterion for exactness

If M = M (x, y) is a differentiable function of the independent variables x and y whose total differential is
dM = A (x, y) dx + B (x, y) dy ,
(4.7)
where, by definition
A (x, y) = (

M
)
x y

(4.8)

1 See for example: D.R. Lide, (Ed.), Chemical Rubber Company Handbook of Chemistry and Physics, CRC Press,
Boca Raton, Florida, USA, 79th edition, 1998.

4.1. THERMODYNAMIC DIFFERENTIALS

49

and
B (x, y) = (

M
) .
y x

(4.9)

According to Greens theorem


dM = [A (x, y) dx + B (x, y) dy]
= [(
R

A
B
) (
) ] dx dy .
x y
y x

(4.10)
(4.11)

where R is a two-dimensional region completely enclosed by the integration path. Then if


(

B
A
) =(
)
x y
y x

(4.12)

dM = 0

(4.13)

it follows that
i.e., M is exact. Conversely, if M is exact, i.e. is a thermodynamic state function, then
(

4.1.4

B
A
) =(
)
x y
y x

(4.14)

Entropy A free sample

Consider the integral Eq.3.83 in subsubsection 3.4.3.3 of Chapter 3 which is rewritten for convenience;2
pf ,Vf ,Tf

dQQS = 0 =

{[

p0 ,V0 ,T0

N kB T
] dV + CV dT } .
V

(4.15)

In that case, since the integrals value depends only on thermodynamic end points the integrand
could be taken as a quasi-static process.
This time we first test the integrand for exactness with Eulers criterion to discover that
(

N kB
N kB T
[
]) =
T
V
V
V
and
CV
(
) = 0 (for the assumed ideal gas) .
V T

So the integrand is not exact and cannot be the differential of a state function. This is no surprise
since d Q 0 always holds. But what may be a surprise is that exactness can be forced by
the simple expedient of introducing an integrating factor 1/T . In that case Eq.4.15 becomes
pf ,Vf ,Tf

dQQS /T = 0 =

p0 ,V0 ,T0

2 The

{[

N kB
] dV + CV /T dT } .
V

ideal gas with fixed particle number has been assumed.

(4.16)

50

CHAPTER 4. A MATHEMATICAL DIGRESSION

Now applying Eulers criterion


(

N kB
[
]) = 0
T
V
V
and
(

CV /T
) =0
V
T

(for the assumed ideal gas) .

The revised integrand in Eq.4.16, is exact and can be done equivalently over any path in the
T, V plane. Introducing the integrating factor has resulted in a new thermodynamic state function
S, called entropy 3

d
QQS
= dS .
(4.17)
T
Moreover, since

d
QQS
= dS = S
(4.18)

the quasi-static, adiabatic process, i.e. d


QQS /T = 0, is equivalent to S = 0 an isenthalpic process.
Having identified the new state function S (even if only for this ideal gas example4 ), for a fixed
number of particles the incremental First Law becomes

d
QQS
dU p
= dS =
+ dV
T
T
T

(4.19)

dU = T dS p dV

(4.20)

or equivalently

which is called the thermodynamic identity, from which follows


(

U
) =T
S V

(4.21)

U
) = p.
V S

(4.22)

and

Furthermore, writing S = S(U, V ) and taking its total differential


dS = (

S
S
) dU + (
) dV ,
U V
V U

(4.23)

and then comparing with Eq.4.19 gives


(

S
1
) =
U V T

(4.24)

3 Recall from Chapter 2 that the subscript QS on


dQQS and dWQS means that the processes are carried out
quasi-statically.
4 Its generality (independence of the working substance) will be established in Section 4.5.1

4.1. THERMODYNAMIC DIFFERENTIALS

51

and
(

p
S
) = .
V U T

(4.25)

Eq.4.24 may be used to define temperature, even though it is not an especially intuitive or revealing
definition.5
Nothing about this informal emergence of entropy reveals its meaning a subject with a long
and perplexing history nor does it prove a generality beyond this special ideal gas example. But
the crucial issue of generality is established, first within a thermodynamic setting in Section 4.5.1
and then in Chapter 6 with an axiomatic rendering that assures generality and assigns meaning
and purpose to entropy.
With the introduction of the thermodynamic identity [Eq.4.20] many unfriendly partial derivatives
can now be tamed, i.e. recast into more practical forms involving only variables that appear in
an equation of state or heat capacities, as a few examples will show.
The definintion of enthalpy, Eq.2.44, evolves into something resembling Eq.4.20, with

d
QQS
dH V
= dS =
dp
T
T
T

(4.26)

dH = T dS + V dp .

(4.27)

or equivalently

Finally, if the internal energy differential in Eq.4.20 is replaced by its more general First Law

expression, i.e. dU = d
Q d W, then the entropy differential is
dS =

d
WQS d
W
1
dQ + [
].
T
T

(4.28)

Since quasi-static (reversible) work is the maximum work available in a thermodynamic process,

i.e. d
WQS d W, for any real (irreversible) processes
dS

d
Q
.
T

(4.29)

which is, in essence, the Second Law of Thermodynamics.

4.1.5

Eulers homogeneous function theorem

A function (x, y, z) is homogeneous of order n if


(x, y, z) = n (x, y, z) .
5 It

is, however, useful for defining negative temperature, which will be discussed in Chapter 9.

(4.30)

52

CHAPTER 4. A MATHEMATICAL DIGRESSION

Eulers homogeneous function theorem states: if (x, y, z) is homogeneous of order n then


x(

) + y ( ) + z ( ) = n (x, y, z)
x y,z
y x,z
z x,y

(4.31)

This seemingly abstract theorem6 has several important thermodynamic implications. For example
consider the internal energy U as a function of the extensive variables S, V and N [see Eq.3.7] and
write U = U (S, V, N ). Since S, U, V and N are extensive they scale according to
U ( S, V, N ) = U (S, V, N ) .

(4.32)

i.e. U (S, V, N ) is a homogeneous function of order 1. Applying Eulers homogeneous function


theorem, [see Eq.4.31],
S(

U
U
U
)
+V (
)
+N (
)
= U (S, V, N ) .
S V,N
V S,N
N S,V

(4.33)

Then, from Eqs.4.21 and 4.22


U = S T V p + N

(4.34)

where
=(

U
)
N S,V

(4.35)

is the chemical potential. , p and T are intensive state variables conjugate to the extensive state
variables N , V and S.

4.1.6

A cyclic chain rule

In addition to the familiar chain rule of partial differentiation, i.e.


(

z
x
z
) ( ) =( )
x s y s
y s

(4.36)

there is the following so-called cyclic chain rule.


Consider a thermal variable M (x, y) and its differential
dM = (

M
M
) dx + (
) dy .
x y
y x

(4.37)

In the special case that M = constant we have dM = 0 and Eq.4.37 becomes


0=(
6A

proof of the theorem is in Appendix C.

M
M
) dx + (
) dy ,
x y
y x

(4.38)

4.1. THERMODYNAMIC DIFFERENTIALS

53

which is reorganized as
( M )
y x
x
( ) =
.
M
y M
(
)
x y

(4.39)

This is often written as a chain rule.7

4.1.6.1

x
M
y
) (
) (
) = 1 .
y M x y M x

(4.40)

Example: Free expansion: Exactness and the cyclic chain rule

A gas, assumed ideal with a fixed number of particles and initially at temperature T0 , is confined
to volume V0 in a partitioned insulated rigid cylinder. [See Figure 4.1]. The unoccupied part, also
with volume V0 , is initially evacuated.

Figure 4.1: Free expansion chamber.


The partition spontaneously begins to leak and the gas surges into the formerly unoccupied part,
eventually coming to equilibrium by uniformly occupying the entire cylinder.
a. What is the change in temperature of the gas?
Since the cylinder is rigid the surging gas does no work. [The expanding gas does not raise or
lower any weights in the surroundings.] Moreover, the cylinder is insulated so that during the
process, Q = 0. (Regarding the expansion as rapid will lead to the same conclusion.) With
Q=0
7 This

and W = 0

simple (but not obvious) bit of mathematics is useful in thermodynamic problem solving.

(4.41)

54

CHAPTER 4. A MATHEMATICAL DIGRESSION


the First Law, U = Q W, says
U = 0 ;

(4.42)

i.e., internal energy is constant.


This constant U process is called a free expansion. Since U is constant it follows that dU = 0, so
taking U = U(T, V ) its total differential is
dU = (

U
U
) dT + (
) dV = 0
T V
V T

(4.43)

and the temperature change we are looking for is


dT =

( U )
V T
dV .
U
(
)
T V

(4.44)

But a little more work is needed to give T in terms of practical thermal variables. Applying
the cyclic chain rule [see Eq.4.40] to this result we get
dT = (

T
) dV
V U

(4.45)

The partial derivative on the right hand side of Eq.4.45 is one of the unfriendly kind referred
to earlier. The objective now is relating it to a friendly partial derivative that corresponds to
laboratory measurables.
Using the thermodynamic identity apply the Euler exactness criterion to the differential dS,
[see Eq.4.34], in the following way;
( 1 )
( p )
T

= T .
V

U
U

(4.46)

After carrying out the differentiations it becomes

p T
1 p
1 T
(
) = 2(
) + (
) .
T 2 V U
T
U V T U V

(4.47)

Applying the ordinary chain rule to the last term on the right hand side and then substituting
Eq.2.41 (CV ), Eq.4.47 becomes
(

T
1
p
) =
[p T (
) ]
V U CV
T V

(4.48)

which is now in terms of measurable quantities our objective all along. Finally, applying the
ideal gas equation of state,8
1
p
[p T (
) ] = 0,
(4.49)
CV
T V
and we see from Eq.4.45 that an ideal gas undergoing a free expansion remains at constant
temperature.
8 The

analysis can be applied to any gas equation of state.

4.1. THERMODYNAMIC DIFFERENTIALS

55

b. What is the change in entropy of the expanded gas?


Focusing directly on S we write S = S (U, V ) and then take the total differential
dS = (

S
S
) dU + (
) dV .
U V
V U

(4.50)

Since this is a free expansion, U = constant ( dU = 0) and


dS = (

S
) dV .
V U

(4.51)

p
dV .
T

(4.52)

N kB
dV .
V

(4.53)

Then from Eq.4.25


dS =
Then using the ideal gas equation of state
dS =

Since the expansion is from V0 2V0 , this is integrated to give


2V0

S =S (2V0 ) S (V0 ) = N kB
V0

dV
V

2V0
=N kB ln
V0
=N kB ln 2 .

(4.54)
(4.55)
(4.56)

If the gas was more realistically described by Van der Waals equation of state,
(p +

N 2a
) (V N b) = N kB T
V2

(4.57)

where a and b are constants characteristic of the gas, would T and S be different?9
Example: Exactness and line integration Returning to Subsection 5.1.2.2, the integral in
Eq.4.16, which for convenience is reproduced here,
pf ,Vf ,Tf

0=

p0 ,V0 ,T0

{[

N kB
CV
] dV +
dT } .
V
T

(4.58)

is exact. Since all paths connecting the end points are equivalent (the virtue of exactness) this can
now be line integrated along any path connecting the integration end points (V0 , T0 Vf , Tf ). It is,
however, good strategy to choose a simple pair of straight lines as shown in Figure 4.2 integrating
9 This

is assigned as a problem.

56

CHAPTER 4. A MATHEMATICAL DIGRESSION

first along constant T and then integrating along constant V , as shown. Along constant T we have
dT = 0 and what survives is the integral (the first leg of the path)
Vf ,T0

I1 =
V0 ,T0

N kB
dV
V

(4.59)

Vf
V0

(4.60)

= N kB ln

T
Figure 4.2: Line paths of the exact integral.
The second leg of the path is along constant V as shown in the diagram. Along this leg dV = 0, so
this contribution is
Vf ,Tf

I2 =
Vf ,T0

CV
dT
T

= CV ln

Tf
.
T0

(4.61)

(4.62)

Summing the two contributions


I1 + I2 = N kB ln

Vf
Tf
+ CV ln
=0
V0
T0

which is identical to the result, Eq.3.83, in Chapter 3.

(4.63)

4.1. THERMODYNAMIC DIFFERENTIALS


4.1.6.2

57

Example: Entropy and spontaneous processes

An uncovered mug holding 250g of hot coffee at 95C sits in a room with air temperature 20C. The
coffee slowly cools, eventually coming to room temperature.
What is the change of entropy for the universe (coffee + surroundings)?
1. First, what is the change in entropy of the coffee?
Note that this (the coffee) entropy change takes place at constant pressure (the air in the room
remains at constant pressure.) Knowing the initial coffee temperature and final equilibrium
universe temperature, we wish to find S in this constant pressure process. With the
typical strategy of expressing what we want to know, i.e. S, in terms of what we can know,
i.e. T and p, we choose to write S = S(T, p) and take the total differential
dS = (

S
S
) dT + (
) dp .
T p
p T

(4.64)

With constant pressure, i.e. dp = 0, we only need to evaluate


dS = (

S
) dT .
T p

(4.65)

Knowing that
S
) = Cp
T p

(4.66)

dS =

1
Cp dT
T

(4.67)

dS =

1
mcp dT
T

(4.68)

T(
Eq.4.65 is

or

where m is the mass and cp is the specific heat per unit mass. Assuming that cp is an average
value independent of temperature
Scoffee = mcp ln (

Tf
)
T0

= 0.25kg 4184J kg 1 K 1 ln (
= 238J kg 1

(4.69)
293K
)
368K

(4.70)
(4.71)
(4.72)

The entropy of the coffee decreases.

58

CHAPTER 4. A MATHEMATICAL DIGRESSION


2. Coffee + surroundings = the universe; so having found the entropy change of the coffee we
need to find the entropy change of the surroundings. Apart from receiving heat from the coffee,
the surroundings are an infinite reservoir with volume, temperature and pressure remaining
constant. So we can apply the First Law to the constant volume (zero work) surroundings
Qsur = Usur .

(4.73)

Taking Qcoffee as the heat leaving the coffee, Qcoffee = Qsur . Then writing for the surroundings U = U(V, S) and taking the total differential (with dV = 0)
Usur
) dSsur
Ssur V
= Tsur dSsur

(4.74)

dUsur = (

(4.75)

where Eq.4.24 has been used. After integration we have


Ssur =

Qcoffee
.
Tsur

(4.76)

To find Qcoffee the First Law is applied to the coffee where we get, assuming negligible liquid
volume change,
Qcoffee = Ucoffee

(4.77)

= Cp (T0 Tf )

(4.78)

= mcp (T0 Tf )

(4.79)

Finally, using this result in Eq.4.76


Ssur = mcp

Tf T0
Tf

(4.80)

= 0.25kg 4184J kg 1 K 1 75K/293K


= 268J K

(4.81)
(4.82)

and
Suniv = Scoffee + Ssur
238J K
= 30J K

+ 268J K

(4.83)
1

(4.84)
(4.85)

The entropy of the universe has increased in this (and any other) spontaneous process. Cooling of a mug of hot coffee by leaving it in a cold room is a spontaneous process, i.e. the
entropy increase of the universe is a quantitative measure of its feasibility without outside
intervention. The reverse process the now cool coffee taking energy from the room to restore
its original hot state without some outside intervention can never happen! That is the
substance of the Second Law.

4.1. THERMODYNAMIC DIFFERENTIALS

4.1.7

59

Thermal engines

A thermal engine is a device that operates in a cycle, during which:


1. A working substance absorbs heat QH from an energy source with which it is in thermal
contact, resulting in thermodynamic changes to the working substance.
2. The working substance exhausts heat QL to an energy sink, also resulting in thermodynamic
changes to the working substance.
3. External mechanical work W can be realized.
4. The working substance returns to its initial thermodynamic state, so the change in any
state function per working substance cycle is zero, e.g. U = 0.
Earliest 18th century engines of various designs used steam exclusively as the working substance
by condensing the steam to cause a partial vacuum, assisting external atmospheric pressure in
driving a piston. Steams only active role was to force the piston back to its starting position.
In 1799 Richard Trevithick introduced a more efficient and compact high pressure steam engine
to directly drive the piston, which was the forerunner of the portable transportation engine. The
steam driven Industrial Revolution had begun.
However, engines of all designs were very inefficient, requiring large quantities of coal for little useful
work output.
From the quest for improved efficiency evolved a new science thermodynamics created in 1824
by Sadi Carnot10,11
Because of friction and turbulence, real engines are complicated irreversible devices. But their cycle
can be modeled as ideal with accuracy sufficient to estimate their efficiency of operation , defined
as the ratio of mechanical work done WQS to the heat absorbed from an energy source QH , in each
quasi-static engine cycle
=

WQS
.
QH

(4.86)

Typical modern engines are of the combustion design in which burning fuel injects heat into a
working substance which then expands against a piston to perform work. To return to its initial
state, the expanded working substance must, at some point in the cycle, exhaust heat to a sink
(usually the atmosphere or a nearby body of water). Exhausting heat to a sink is the step in the
cycle which guarantees that the efficiency of any thermodynamic engine is less than one.

4.1.8

Carnots thermal engine

The ideal Carnot engine [see Figure 4.3] operates between an energy source, in which the working
substance (in this case an ideal gas) absorbs heat QH isotherthermally at high temperature TH
(short dashed curve) and an energy sink, into which heat QL is isothermally exhausted at a low
10 S.
11 S.

Carnot, Reflections on the Motive Power of Fire. Dover (1960).


S. Wilson, Sadi Carnot, Scientific American 245, 102114 (1981).

60

CHAPTER 4. A MATHEMATICAL DIGRESSION

temperature TL (solid curve). The long dash curves which close the Carnot cycle are adiabats in
which no heat enters or leaves the working substance.

Figure 4.3: An ideal Carnot engine cycle drawn in the P V plane.

4.1. THERMODYNAMIC DIFFERENTIALS

61

To calculate the engine efficiency [see Eq.4.86] the work done in a cycle is determined by dividing
the cycle into its four component paths:

12

23

34

41

Work from an Ideal Gas Carnot Cycle


Apply the First Law

d QH = dU + d
W12 .
Take
U
U
)V dT + ( V
)T dV .
dU = ( T
Along an isotherm dT = 0 and for an ideal gas
U
( V
)T = 0 .
Therefore along the isotherm
dU = 0
and
W12 = QH .
Apply the First Law

d Q23 = dU + d
W23 .
Along the adiabat
d Q23 = 0 .
For the ideal gas
dU = CV dT
From this segment
W23 = CV (TH TL ) .
Apply the First Law

d Q34 = dU + d
W34 .
Following the arguments used along 1 2 isotherm
W34 = QL .
Apply the First Law

d Q41 = dU + d
W41 .
Following the 2 3 adiabat arguments
W41 = CV (TL TH ) .

Work performed in the cycle is therefore

W = W12 + W23 + W34 + W41


= QH QL .

(4.87)
(4.88)

with the Carnot efficiency [see Eq.4.86]

=1

QL
.
QH

(4.89)

62

CHAPTER 4. A MATHEMATICAL DIGRESSION

In terms of working substance state variables (assume an ideal gas),


V2

QH = p dV

(4.90)

V1
V2

= N kB TH
V1

= N kB TH ln

dV
V

V2
.
V1

(4.91)
(4.92)

and
V4

QL = p dV

(4.93)

V3
V4

dV
V

(4.94)

V4
.
V3

(4.95)

= N kB TL
V3

= N kB TL ln
Therefore the efficiency is
=1+

TL ln (V4 /V3 )
.
TH ln (V2 /V1 )

(4.96)

But along the adiabats 2 3


TL

CV
TH

V3

dT
= N kB
T

V2

dV
V

(4.97)

dV
,
V

(4.98)

and 4 1
TH

CV
TL

V1

dT
= N kB
T

V4

which after integrating is


V4 V3
=
,
V1 V2

(4.99)

with the Carnot efficiency of Eq.4.96 becoming


=1

TL
.
TH

(4.100)

This result is consistent with the Kelvin-Plank statement of the Second Law of Thermodynamics:

4.1. THERMODYNAMIC DIFFERENTIALS

63

It is impossible to construct an engine such that heat is cyclically removed from a reservoir at
high temperature with the sole effect of performing mechanical work. (In some part of the cycle
heat must be exhausted.)
A hypothetical Carnot engine with operating temperatures TH 600K and TL 300K would
have an efficiency 50%.12

4.1.9

The Carnot cycle entropy

Any reversible engine cycle can be divided into infinitesimal Carnot segments (adiabats and isotherms)
as shown in Figure 4.4.

Figure 4.4: An arbitrary quasi-static (reversible) engine cycle is divided into a series of infinitesimal
Carnot cycles, a single one of which is shown on the left.
Now, combining Eqs.4.89 and 4.100 we have the result for a Carnot engine cycle

12 The

QH QL
+
=0
TH
TL

Carnot engine is the most efficient engine operating between two fixed temperatures.

(4.101)

64

CHAPTER 4. A MATHEMATICAL DIGRESSION

which applied to the infinitesimal Carnot engine cycle in the subdiagram of Figure
comes

QL
d
QH d
+
=0
TH
TL

4.4 be-

(4.102)

Summing this expression for the infinitesimal Carnot cycles (which reconstruct the pictured arbitrary reversible cycle) gives
d QQS
= 0.

(4.103)

Thus, T1 is an integrating factor that makes d


QQS exact, again summoning, for an arbitrary substance, the redefinition

d
QQS
= dS .
T

(4.104)

For the complete working substance cycle S = 0 .

Problems and exercises


1. A cube of lead 1m on each side is to be slowly lowered from the sea surface 3 103 m down to
the site of a deep drilling oil spill. The surface water temperature is 23C while the sea water
temperature at the oil well head is 4C. The mean density of sea water is 1035kg m3 .
Find the volume of the lead cube when it reaches well head depth.
Isothermal compressibility of Pb
T =

1 V
(
) = 2.3 106 atm1
V p T

(4.105)

1 L
(
) = 2.9 105 K 1 .
L T p

(4.106)

Linear thermal expansivity of Pb is


=

2. A van der Waals gas at temperature T0 , is confined by a partition to half of an insulated


rigid cylinder with total volume 2V [See Figure 5.1]. The unoccupied part with volume V , is
initially evacuated.
The partition spontaneously self-destructs and the gas surges into the formerly unoccupied
part, eventually coming to equilibrium by uniformly occupying the entire cylinder.
(a) What is the change in temperature of the van der Waals gas?
(b) What is the change in entropy of the van der Waals gas?

4.1. THERMODYNAMIC DIFFERENTIALS

65

3. The isothermal compressibility of a gas is defined as


T =

1 V
(
)
V p T

(4.107)

while the adiabatic compressibility is defined as


S =

1 V
(
) .
V p S

(4.108)

Cp
S .
CV

(4.109)

Show that
T =
4. The velocity of sound in a gas is given by
c2s =

BS

(4.110)

where is the density of the gas


=

N
V

(4.111)

and BS is the adiabatic bulk modulus defined by


BS = V (

p
) ,
V S

(4.112)

with N the number of gas molecules, the mass per molecule and S the entropy.
Show that for an ideal gas at temperature T the sound velocity is given by
c2s =

kB T

(4.113)

Cp
.
CV

(4.114)

where
=

5. N molecules of an ideal gas undergo the cycle A B C A as shown in the p V diagram


of Figure 4.5. The curve A B is an isotherm at temperature Th . When the gas is in the
state (p2 , V1 ) the temperature is Tc .
a. Show that the net heat Q added to the gas during the cycle is
Q = N kB [Th ln (

Th
) (Th Tc )] .
Tc

(4.115)

66

CHAPTER 4. A MATHEMATICAL DIGRESSION


b. Find the engine efficiency.

p
A

p1

Th
Tc
p2

V2

V1

Figure 4.5: p V diagram for problem 5.


6. A body of mass m with specific heat Cp at temperature 500K is brought into contact with
an identical body at temperature 100K , with both bodies isolated from their surroundings.
After a while the systems come to thermal equilibrium. Show that the change in entropy S
of the total system is
S = mCp ln (9/5) .

(4.116)

Rudolph Clausius obtained distinction by mathematical calculations


based on the dynamical theory of heat calculations which, it is
claimed, show the necessity of a Creator and the possibility of miracles.
Obituary notice: New York Times 26 Aug. 1888

Chapter 5

Thermodynamic potentials
5.1

Introduction

Thermodynamic processes can by design or by nature take place under conditions in which particular macroscopic parameters e. g. temperature, pressure, volume, magnetic field, entropy,
particle number are kept constant. For an experimentalist, holding some variables constant, such
as entropy or internal energy, is inconvenient, while other variables, such as pressure and temperature are easier or more natural to fix. Therefore, in addition to an energy potential U (internal
energy) which accommodates constant extensive variables S, V and N , other potential functions
are introduced which are adapted to fixing other variables.

5.1.1

Internal energy U

The differential
dU = T dS p dV + dN,

(5.1)

a consequence of the First Law of Thermodynamics, is called a fundamental equation or the thermodynamic identity. With U = U (S, V, N ), the total differential is
dU = (

U
U
U
)
dS + (
)
dV + (
)
dN .
S V,N
V S,N
N S,V

(5.2)

Correspondences between Eqs.5.1 and 5.2 give the following:


U
)
=T,
S V,N
U
(
)
= p
V S,N
(

67

(5.3)
(5.4)

68

CHAPTER 5. THERMODYNAMIC POTENTIALS


magnetic
electric
elastic
surface
strain
particle number
earths gravity

B0 dM
E 0 dP
d
dAS
d
dN
mgz dN

Table 5.1: Here B0 is the magnetic field prior to insertion of matter, M is the magnetization, E 0
is an electric field prior to insertion of matter, P is electric polarization, is elastic tension, is
elastic elongation, is surface tension, AS is surface area, is stress,  is strain, is the chemical
potential, N is the number of particles, m is the mass of a particle, g is the acceleration due to
gravity and z is the particles distance above the earths surface.

and
(

U
)
= .
N S,V

(5.5)

Noting that partial derivatives of U with respect to S, V and N yield accessible thermal properties
of the system (in this case T , p and ) they are called natural variables of U. Applying Eulers
criterion to Eqs.5.3 and 5.4 gives the identity

T
p
)
= (
)
V S,N
S V,N

(5.6)

which is called a Maxwell Relation 1 .


Eq.5.1 contains only quasi-static mechanical work, p dV and particle work, dN . But thermodynamics also applies to other conjugate work pairs which can be included in the fundamental
equation Eq.5.1.
For example, from Table 5.1
dU = T dS p dV + d + B 0 dM + E0 dP + dAs + d + dN .
Since U, S, V , , M ,
can be derived

P, As ,  and N are extensive variables, the corresponding Euler equation

U = T S pV + + B 0 M + E0 P + As +  + N .
1 The

(5.7)

constant N is usually suppressed.

(5.8)

5.1. INTRODUCTION

5.1.2

69

Enthalpy H

Constant pressure is so common a thermodynamic constraint that, as in Eq.3.55, enthalpy


H = U + pV

(5.9)

is introduced to simplify isobaric analysis. Enthalpy is a total energy in the sense of being the sum
of energy of the constituent particles and the mechanical energy pV needed to provide physical
space for these constituents.
Enthalpy may have been in use since the time of Clausius under different names and symbols but
Gibbs Heat function, =  + pV , seems to be its first formalization. As for the name Enthalpy,
physicists are so conditioned to associate Kamerlingh Onnes with liquifying helium and discovering superconductivity that it might be a surprise to learn that Enthalpy is also his creation
(1901).2 The symbol H was proposed twenty years later by Alfred W. Porter3 to represent heat
content.4
Taking the total differential of Enthalpy we have
dH = dU + p dV + V dp .

(5.10)

When combined with the fundamental equation, Eq.5.1,


dH = T dS + V dp + dN

(5.11)

With H = H (S, p, N ), the total differential is


dH = (

H
H
H
)
dS + (
)
dp + (
)
dN .
S p,N
p S,N
N S,p

(5.12)

Comparing this with Eq.5.11


H
)
=T
S p,N
H
(
)
=V
p S,N
(

(5.13)
(5.14)

and
(

H
) =
N S,p

(5.15)

which affirms S, p and N as the natural variables of H .


Finally, applying Eulers criterion to Eqs.5.13 and 5.14
(

V
T
)
=(
)
S p,N
p S,N

which is another Maxwell Relation.


2 From

the Greek enthalpos, which translates as to warm within.


W. Porter, Transactions of the Faraday Society 18, 139 (1922).
4 Irmgard K. Howard, H is for Enthalpy, Journal of Chemical Education 79 697-698 (2002).

3 A.

(5.16)

70

5.1.3

CHAPTER 5. THERMODYNAMIC POTENTIALS

Helmholtz Potential F

The Helmholtz potential (Helmholtz free energy) F is formed to accommodate constant T , V and
N processes and is defined as
F = U TS .
(5.17)
The nature of this quantity is clarified by taking the total differential
dF = dU T dS S dT

(5.18)

and replacing dU from the fundamental equation, Eq.5.1, to arrive at


dF = S dT p dV + dN.

(5.19)

For a fixed number of particles at constant temperature


dF = p dV

(5.20)

where the right hand side is quasi-static work. Since reversible (quasi-static, friction free) work is
the maximum work that can be done by a system, the decrease in Helmholtz potential at constant
temperature is the maximum work a system can perform. Moreover, for a fixed number of particles
at constant T and V
dF = 0
(5.21)
which is the extremal condition for equilibrium of a system under those conditions, i.e. when
dF = 0, F = Fmin , the equilibrium state.
Expressing F = F (T, V, N ), the total differential dF is
dF = (

F
F
F
)
dT + (
)
dV + (
)
dN .
T V,N
V T,N
N T,V

(5.22)

Comparing this with Eq.5.19


F
)
= S
T V,N
F
(
)
= p
V T,N
(

(5.23)
(5.24)

and
(

F
)
=
N T,V

(5.25)

affirming T , V and N as natural variables for F .


Applying Eulers criterion to Eqs.5.23 and 5.24

(
a third Maxwell Relation.

p
S
)
=(
)
V T,N
T V,N

(5.26)

5.1. INTRODUCTION

5.1.4

71

Gibbs potential G

The Gibbs potential (Gibbs free energy) G, defined as


G = U T S + pV ,

(5.27)

accommodates the experimental constraints on T , p and N . Differentiating Eq.5.27


dG = dU T dS S dT + p dV + V dp .

(5.28)

and substituting dU from Eq.5.1 its total differential is


dG = S dT + V dp + dN.

(5.29)

Writing G = G (T, p, N ), a total differential is


dG = (

G
G
G
)
dT + (
)
dp + (
)
dN
T p,N
p T,N
N T,p

(5.30)

which, on comparison with Eq.5.29 gives


G
)
= S
T p,N
G
)
=V
(
p T,N
(

(5.31)
(5.32)

and
(

G
) =
N T,p

(5.33)

confirming T , p and N as natural variables of G. Applying Eulers criterion to Eqs.5.31 and 5.32

V
S
)
= (
)
p T,N
T p,N

(5.34)

which is the fourth Maxwell Relation.

5.1.5

Enthalpy and throttling

A gas that is slowly forced under high pressure across a very narrow, perfectly insulated porous
plug (or valve), emerging on the other side at reduced pressure, is said to be throttled. The process
is applied in commercial refrigeration and liquefaction of gasses. To simplify the discussion, the
normally open system (gas comes from a high pressure reservoir and escapes to a low pressure
reservoir) is replaced by a closed system consisting of two frictionless pistons, one maintaining
the constant high pressure P0 and the other the constant lower pressure Pf [See Figure 5.1]. Fixing

72

CHAPTER 5. THERMODYNAMIC POTENTIALS

Figure 5.1: Gas forced across a porous plug.


attention on a small amount of gas, say N molecules, the high pressure piston forces the gas through
the plug onto the low pressure piston. At the higher constant pressure P0 , the N molecules have
a temperature T0 and occupy a volume V0 . Therefore the work W0 done on the gas by the high
pressure piston in moving the N molecules through the plug is
W 0 = P0 V 0 .

(5.35)

At the lower pressure Pf the N molecules have temperature Tf and occupy a volume Vf so that
the work done by the N molecules as they emerge on the other side pushing the piston outward
is
Wf = Pf Vf .
(5.36)
Applying the First Law to this adiabatic process the net work is
Wf W0 = Uf U0

(5.37)

U0 + P0 V0 = Uf + Pf Vf ,

(5.38)

H0 = Hf .

(5.39)

and therefore

or in terms of enthalpy,

The throttling process is isenthalpic.


Of principle interest in throttling is the temperature change it produces in the gas if any. To
investigate this quasi-static process, first write the temperature (a thermodynamic parameter) as
T = T (H , P ) and take the total differential
T
T
) dH + (
) dP .
H P
P H
The good news is that since dH = 0 we have, so far,
dT = (

dT = (

T
) dP .
P H

(5.40)

(5.41)

Unfortunately the unfriendly partial derivative (T /P )H = J must be evaluated.5 Typically,


5J

is the Joule-Thomson coefficient.

5.1. INTRODUCTION

73

there is more than one way to do this. For now, start by writing Eq.5.11 as

dS =

1
V
dH dP .
T
T

(5.42)

Then, since S is exact, apply Eulers criterion

( 1 )
T
P

(V )
T
=
H

(5.43)

which becomes, after differentiation,

T
V
T
) =T(
) V (
)
P H
H P
H P
T
V
) V ](
)
= [T (
T P
H P
V
1
) V ]( )
= [T (
T P
CP

(5.44)
(5.45)
(5.46)

It is easily verified that J = 0 for an ideal gas and, consequently, T = 0. On the other hand for any
real gas, e.g. a van der Waals gas, J can be positive or negative. if J > 0 the gas will cool whereas
if J < 0 the gas will warm! Specifically, below some inversion temperature Tinv (characteristic of
each gas) a gas will cool upon throttling and above Tinv it will warm! At atmospheric pressure
nitrogen (N2 ) has Tinv = 621K and oxygen (O2 ) has Tinv = 764K. On the other hand neon (N e)
has Tinv = 231K and hydrogen (H2 ) has Tinv = 93K. [See Figure 5.2]

74

CHAPTER 5. THERMODYNAMIC POTENTIALS

Figure 5.2: Joule-Thomson coefficients J for various gases at atmospheric pressure.


He4 has an inversion temperature of 40K. To liquify He4 , one must first cool it below 40K and
then throttle it. At room temperature N2 and O2 will cool upon throttling whereas He and N e will
warm.

5.1.6

Heat capacity and entropy

Resuming the discussion begun in Chapter 4, low temperature heat capacities are among those
macroscopic phenomena that require quantum explanations. For that reason they stand high on the
list of important thermodynamic experiments. Low temperature values in diamond first suggested
to Einstein (1905) that nature was not as classical as was generally believed at the time, and he
applied quantization as proposed by Planck (1900) to develop the first quantum theory of solids
25 years before the arrival of a generally accepted quantum mechanics.
The appearance of entropy in thermodynamics brings a welcome and important uniformity to heat
capacity definitions. Whereas in Chapter 2 we derived the result
CV = (

U
) ,
T V

(5.47)

5.1. INTRODUCTION

75

we can instead, by applying the simple chain rule


(

U
U
S
) =(
) (
)
T V
S V T V

(5.48)

U
S
) =T (
)
T V
T V

(5.49)

followed by Eq.5.3, arrive at


(
so that
CV =T (

S
) .
T V

(5.50)

Similarly
H
)
T p
H
S
=(
) (
)
S p T p
S
=T(
) .
T p

Cp = (

(5.51)
(5.52)
(5.53)

In general there is a single, convenient heat capacity definition:


C = T (

S
) .
T

(5.54)

Example: A relationship between Cp and CV


The entropy based relations Cp = T (S/T )p and CV = T (S/T )V put this problem on a level
footing. Considering the final objective, expressing the entropy as S = S(T, p) is one sensible choice.
Then
S
S
dS = (
) dT + (
) dp
(5.55)
T p
p T
which already contains a piece of the answer, i.e.
dS =

Cp
S
dT + (
) dp
T
p T

(5.56)

An equally sensible choice is S = S(T, V ), in which case


S
S
) dT + (
) dV
T V
V T
CV
S
=
dT + (
) dV .
T
V T

dS = (

(5.57)
(5.58)

Then combining Eqs.5.56 and 5.58 gives


(
Next:

Cp CV
S
S
) dT = [(
) dV (
) dp] .
T
V T
p T

(5.59)

76

CHAPTER 5. THERMODYNAMIC POTENTIALS


1. Apply the pair of Maxwell Relations, Eq.5.26 and Eq.5.34, to get
(

Cp CV
p
V
) dT = [(
) dV + (
) dp] .
T
T V
T p

(5.60)

2. Apply the cyclic chain rule to the first partial derivative on the right hand side
(

Cp CV
V
p
) dT = (
) [(
) dV dp] .
T
T p V T

(5.61)

3. Noting that
dp = (

p
p
) dV + (
) dT
V T
T V

(5.62)

we are finally home with


(

Cp CV
V
p
)=(
) (
) .
T
T p T V

(5.63)

N kB
V
) =
T p
p

(5.64)

p
N kB
) =
T V
V

(5.65)

Cp CV = N kB

(5.66)

For an ideal gas


(
and
(
which gives
which is Eq.3.66, stated there without proof.
Eq.5.63 is a result that can be applied to any gas law for example, to the van der Waals equation.6
This type of result is not limited to gases. We will later derive a relation between CB (constant magnetic induction) and CM (constant magnetization) for magnetic systems and between C (constant
tension) and C (constant elongation) for linear elastic systems.7

Problems and exercises


1. N molecules of an ideal gas at temperature T0 and volume V0 undergo an isothermal expansion
to a volume V1 . What is the change in:
a. Internal energy of the expanding gas?
6 Assigned
7

as a problem.
= op is the mean elongation.

5.1. INTRODUCTION

77

b. Enthalpy of the expanding gas?


c. Helmholtz potential of the expanding gas?
d. Gibbs potential of the expanding gas?
2. Show that for an ideal gas
a. (

CV
) = 0.
p T

b. (

Cp
) = 0.
V T

3. If a gas is not ideal but has the equation of state


p=

N kB T
N
[1 + B (T )]
V
V

(5.67)

find expressions for:


a. (

CV
) .
p T

b. (

Cp
) .
V T

c. Cp CV .
4. It is possible for a specific gas to emerge on the low pressure side of a throttling plug at either
a lower or higher temperature than it had entering on the high pressure side.
Show that for the van der Waals gas, whose equation of state is
(p +

aN 2
) (V N b) = N kB T ,
V2

(5.68)

where a and b are gas specific temperature independent constants, that at temperature T0
kB T0 =
throttling produces no temperature change.

2a V N b 2
(
)
b
V

(5.69)

78

CHAPTER 5. THERMODYNAMIC POTENTIALS

Suppose we were asked to arrange the following in two categories


distance, mass, electric force, entropy, beauty, melody. I think there are
the strongest grounds for arranging entropy alongside beauty and
melody and not with the first three. Entropy is found only when the
parts are viewed in association, and it is by hearing or viewing the
parts in association that beauty and melody are discerned...The reason
why this stranger can pass itself off among the aborigines of the
physical world is that... it has a measure number associated with it.
A. Eddington- The Nature of the Physical World, Cambridge
University Press (1928).

Chapter 6

Knowing the unknowable


6.1

Entropy the ticket to Oz

Entropy S was introduced into thermodynamics by Rudolph Clausius1 as a cryptic state function
without a clear intuitive or physical interpretation. Still, he maintained it to be the basis of a
sweeping new, abstract, universal physical principle:
In all spontaneous processes the entropy of a closed system (system + environment)
tends towards a maximum, i.e.
S 0.

(6.1)

It was in his series of memoirs (1856-1865) on irreversible heat loss, i.e. inefficiency of heat engines,
that he chose the name entropy (Greek for transformation) because it sounded similar to energy,
under the mistaken impression that the two were related.
Clausius principle, in which a particular state variable entropy always increases (or remains
constant) as a system and its surroundings (i.e.the universe) evolve from one equilibrium state
to the next, is called the Second Law of Thermodynamics. The Second Law is not a conservation
rule. Moreover entropy is unlike objective observables so familiar in physics, such as mass, length,
velocity, energy, acceleration, etc., although emotive analogies or proxies such as disorder, chaos
and disorganization which point in the right direction have been applied.
1 R Clausius, On the Moving Force of Heat, and the Laws regarding the Nature of Heat itself which are deducible
therefrom (English translation of original) Phil. Mag. 2, 121, 102119 (1851).

79

80

CHAPTER 6. KNOWING THE UNKNOWABLE

The following encounter, said to have taken place in 1948, could serve to acknowledge entropys
subtle character.
Over half a century ago while creating a fledgling Information Theory, Claude Shannon (of Bell
Laboratories) sought advice from Princetons famous mathematician-physicist John von Neumann
in naming the function central to his [Shannons] (information) arguments:2
My greatest concern was what to call it. I thought of calling it information, but the
word is overly used. So I decided to call it uncertainty. When I discussed it with John
von Neumann, he had a better idea. Von Neumann told me You should call it entropy,
for two reasons. In the first place your uncertainty function has been used in statistical
mechanics under that name, so it already has a name. In the second place, and more
important, even after nearly one hundred years nobody knows what entropy really is,
so in a debate you will always have the advantage.
Although entropy is not a part of quantum mechanics, it plays a key role in postulating the bridge
to thermodynamics.

6.1.1

The bridge

As a consequence of environmental decoherence a thermodynamic density operator, op


op = P (Es ) Es Es ,

(6.2)

emerges as potentially the quantity essential in the evolution to thermodynamics. For example,
using the macroscopic internal eigen-energies from
hop s = s s

(6.3)

the average internal energy U enshrined in the First Law, could in principle be expressed
as
U = Tr op hop

(6.4)

U = s P (s ) .

(6.5)

or, equivalently, as
s

This looks promising but, alas, P (s ) are unknown and, except in trivial cases, unknowable (gasp!)
The emerging thermodynamic theory is not complete! Can it ever be?
An alternative is to replace the unknowable P (Es ) by inferred surrogates P (Es ) with which to
construct a surrogate thermal density operator op
op = P (Es )Es Es

(6.6)

that behaves in accord with the macroscopic physical universe.


2 M. Tribus and E. McIrvine,Energy and Information, Scientific American 225, p 179-188 (1971). Reproduced
with permission. Copyright 1971 SCIENTIFIC AMERICAN, a division of Nature America, Inc. All rights reserved.

6.1. ENTROPY THE TICKET TO OZ

6.1.2

81

Thermodynamic hamiltonians

A simple dynamical situation would be the just discussed


Hop = hop .

(6.7)

in which hop represents the internal energy (kinetic + potential + interaction) of, say, a gas.

If gas particles also have a magnetic moment and are in a local magnetic field B0 the field prior
to insertion of matter4 a macroscopic hamiltonian is
Hop = hop M op B0 ,

(6.8)

where M op represents macroscopic magnetization


M op = mop (i) .

(6.9)

If the particles are molecular dipoles with electric dipole moments


then

pop in an electric field E0 ,

Hop = hop Pop E0 ,


where

(6.10)

Pop represents the macroscopic electric polarization.

Eulers fundamental equation Eq.5.8, shows how other thermodynamic hamiltonians can be formulated by adding physically relevant macroscopic operators [see Table 5.1]. A more general result
is then5,6
Hop = hop + pop V op op B0 M op E0 Pop Asop Nop .

(6.12)

Since the extensive parameters describe a macroscopic system completely, Eq.6.12 contains all
relevant information to enable a model macroscopic systems average total energy to be expressed
as
Hop = Tr op Hop .

(6.13)

But the inescapable question remains what criteria can be devised for constructing the best
P(Es )? This question has no a priori answer apart from a postulate that corresponds to time-tested
thermodynamics and agrees with a measurable universe.7
Towards this goal:
3 The

gas may be atoms or cold molecules (no internal modes) or even electrons.
subtle but important point will be discussed in detail in Chapter 11.
5 Schr
odingers non-relativistic wave mechanics is based on closed systems (fixed number of particles.) It makes
no provision for destroying and creating particles. Therefore a number operator is not accommodated within that
theory. Particle creation and destruction is, however, integral to quantum field theories so that number operators
play a natural role.
6 The particle number operator N
op defines the eigenvalue equation
4 This

Nop N = n N

(6.11)

with eigenvalues n = 0, 1, 2, 3 . . ..
7 This not unlike Ludwig Boltzmanns motivation in postulating the classical microstate based theory of entropy
S = kB ln W.

82

CHAPTER 6. KNOWING THE UNKNOWABLE


1. Minimize
Hop =Tr op Hop

(6.14)

(Es )Es
=P

(6.15)

(Es ).8,9,10
with respect to the P
(Es ) that minimizes Eq.6.15, they will be chosen
2. With no a priori basis for choosing the P
without fear or favor, i.e. with minimum bias 11,12,13 (maximum uncertainty).
Recalling the function(al)14 F of Eq.2.52,
F [] = Tr op ln op ,

> 0,

(6.16)

as a measure of bias in a collection of probabilities,15 the function(al)


(Es ) ln P
(Es ) ,
(E)] = P
F [P

>0

(6.17)

(Es ) and is maximized. [See Appendix


is taken to be a measure of bias in the probabilities P
B.]
(Es ) are normalized
3. The P
(Es ) = 1 .
P

(6.18)

To meet these requirements a thermal Lagrangian is formed


(Es ) ln P
(Es ) + P
(Es )Es } 0 P
(Es ) ,
L = { P
s

(6.19)

(Es ) are varied until L is maximized.16 In which case,


where 0 is a Lagrange multiplier and the P
with kB T [see Appendix B], a thermal Lagrangian is
= {kB P
(Es ) ln P
(Es ) + T 1 P
(Es )Es } 0 P
(Es )
L [P]
s
8 L.

(6.20)

H. Thomas, The calculation of atomic fields, Proc. Cambridge Phil. Soc. 23, 542548 (1927).
Fermi,Un Metodo Statistico per la Determinazione di alcune Priopriet dellAtomo. Rend. Accad. Naz.
Lincei 6, 602607 (1927).
10 E. H. Lieb, Thomas-Fermi and related theories of atoms and molecules, Rev. Mod. Phys. 53, 603-641 (1981).
Errata 54, 311 (1982).
11 J. N. Kapur,Maximum Entropy Models in Science and Engineering, John Wiley and Sons, New York (1989).
12 E. T. Jaynes, Information theory and statistical mechanics, Phys. Rev. 106, 620 (1957).
13 R. Balian, Incomplete descriptions and relevant entropies, Am. J. Phys. 67, 1078 (1999).
14 F is a function of functions, which is called a functional.
15 See Appendix B.
16 This is equivalent to minimizing the Helmholtz potential as in Subsection 5.1.3.
9 E.

6.1. ENTROPY THE TICKET TO OZ

83

P (least bias)
and with P
S = kB P (Es ) ln P (Es )

(6.21)

the entropy of thermodynamics.

6.1.3

Microcanonical (Boltzmanns) theory

At the time Boltzmann published his ingenious entropy postulate (1864) skepticism about microscopic models continued unabated and classical mechanics was the only operative particle paradigm.
His accomplishments were so radical and ahead of his time that it would take another 25 years for
them to be reformulated by Max Planck as quantization.
With our 150 year advantage, we illustrate a thermodynamic density operator op approach to
Boltzmanns fundamental thermodynamic contribution.
Consider a many particle, closed system with:
1. A fixed number of particles N ,
2. A fixed volume V ,
3. An isolated g0 -fold degenerate macroscopic eigen-energy 0,
0, {0,1 = 0,2 = 0,3 . . . = 0,g0 }

(6.22)

.
These effectively constitute Boltzmanns microcanonical conditions. For this model system the
average hamiltonian (internal energy) is
U=

PB (s, ) s, s,0

(6.23)

s
=1,2,g0

where PB (s, ) is the probability that the system is in one of the degenerate macroscopic particle
states with eigen-energy s, . The Kronecker s, selects from within this spectrum a g0 -fold
degenerate single value 0, . Therefore after summing over the index s, Eq.6.23 becomes
g0

U = PB (0, ) 0, .

(6.24)

=1

The remaining sum over covers the g0 -fold degeneracy.17


B ] is
To determine the surrogate probabilities PB (0, ) (i.e. op ) a Thermal Lagrangian L [P
B (0, ), minimizes both macroscopic
postulated, which when maximized with respect to the P

B (0, ). In particular
energy [T rop Hop ] and any bias [see Eq.6.20] in the choice of P
g0

g0

g0

=1

=1

=1

B ] = kB P
B (0, ) ln P
B (0, ) T 1 P
B (0, ) 0, 0 P
B (0, )
L [P
17 Boltzmann

(6.25)

pictured microstates (classical configrations) contributing to a single classical internal energy.

84

CHAPTER 6. KNOWING THE UNKNOWABLE

where 0 is a Lagrange Multiplier that insures normalized surrogate probabilities.18 The Thermal
Lagrangian used here, and in the remaining examples of this book, is discussed and justified in
Appendix B.
If results from this procedure are in agreement with an observable macroscopic universe and with
time honored thermodynamics whose validity is not in doubt the yellow brick road will have
guided us, at last, to the Emerald City.
[The procedure resembles a maximum entropy formalism of E.T. Jaynes.19,20 Jaynes formulation
is, fundamentally, a statistical method in cases of partial information (Bayesian statistics.) It
adopts Gibbs-Shannon21 communication (information) theory [see Section 6.1] and is the basis of
the Information theory approach to statistical mechanics applied by several authors.22,23,24 Here,
rather than pursuing alternative classical statistics, a reasoned path from quantum mechanics via
decoherence to op and hence to Thermal Physics is postulated.]
Differentiating25 the Thermal Lagrangian with respect to the PB (0, ) and setting the resulting
expressions to zero we have
kB (ln PB (0, ) + 1) T 1 0, 0 = 0 ;

= 1, 2, 3 . . . g0

(6.26)

where g0 is the degeneracy of the eigen-energy 0, . Solving for each of the probabilities PB (0, )
we find

(T 0, 0 )
1 .
(6.27)
PB (0, ) = exp

k
B

Applying normalization

1
g0

(T 0, 0 )

B
exp

P
(
)
=

1
=1

0,

k
B

=1
=1

g0

(6.28)

which after summing gives

(T 0 0 )

g0 exp
1 = 1 .

kB

(6.29)

From Eq.6.27 the Boltzmann probabilities are therefore


PB (0, ) =

1
g0

(6.30)

18 The WEB is a rich source of tutorials on this powerful mathematical technique that is widely used in
physics and other technical fields. See for example: www.slimy.com/steuard/teaching/tutorials/Lagrange.html
and www.cs.berkeley.edu/ klein/papers/lagrange-multipliers.pdf
19 E. T. Jaynes, Information Theory and Statistical Mechanics. Phys. Rev. 106 620 (1957).
20 E. T. Jaynes, Information Theory and Statistical Mechanics II. Phys. Rev. 108 171 (1957).
21 Claude E. Shannon, Prediction and entropy. Bell System Tech. J., 30 50 (1951).
22 A. Katz, Principles of Statistical Mechanics, W.H. Freeman, San Francisco (1967).
23 R. Baierlein, Atoms and Information Theory, WH Freeman, San Francisco (1971).
24 Myron Tribus, Thermodynamics and Thermostatics, D. Van Nostrand Company Inc., New York, (1961).
25 The Lagrangian is a functional of the P() so the required process is functional differentiation as is done in
Lagrangian (classical) mechanics. However, in this uncomplicated case functional differentiation is a subtlety we can
pretend to be unaware of.

6.1. ENTROPY THE TICKET TO OZ

85

and from Eqs.6.24 and 6.30 the (internal) energy is


g0

U = PB (0, ) 0,

(6.31)

=1
g0

1
0,
g
=1 0

(6.32)

=0 .

(6.33)

U is identified with the g0 f old degenerate macroscopic Boltzmann eigen-energy 0 . Finally, from
Eqs.6.30 and 6.21 the entropy is
g0

S = kB PB (0, ) ln PB (0, )

(6.34)

=1
g0

1
1
ln
g0
=1 g0

= kB

(6.35)

=kB ln g0 .

(6.36)

An imaginary trip to Viennas Zentralfriedhof cemetery finds a tombstone bearing Boltzmanns


stony visage. Above the sculpture is carved his enduring formula for entropy S,
S = k log

(6.37)

where k was later to become kB , the Boltzmann constant26,27 and log is the natural logarithm.
Comparing Eqs.6.36 and 6.37, W (Boltzmanns number of microstates)28 corresponds to g0 the
degeneracy of the eigen-energy 0 .29,30
If Boltzmanns equation for entropy, Eq.6.37, is regarded here as the fundamental equation of
motion of thermodynamics then, in much the same way that the Lagrangian of classical mechanics generates Newtons Laws, the postulated Thermal Lagrangian, Eq.6.25, generates Boltzmanns
equation of motion. This effectively provides a central organizing principle whose absence in
Thermal and Statistical Physics has already been duly noted by Baierlein.31
Eqs.6.36 and 6.37 provide relatively easily grasped insights into entropy:
High degeneracy more statistically equivalent states means greater uncertainty (i.e., less
information) about any particular outcome less bias (higher entropy).
In evolving from configurations of high predictability (large bias) to ones of low predictability (minimum bias) the entropy increases the essence of the second law.
26 Boltzmanns

constant is universally abbreviated kB .


Planck seems to have been the first to write entropy in this form and k was initially called Plancks constant.
Yes! Another one!
28 W represents the German wahrscheinlichkeit (probability), i.e. W = 1 .
g0
29 The logarithm in Boltzmanns theory is consistent with the Thermal Lagrangian postulate.
30 Of course, Boltzmann knew nothing about quantum eigen-energies. That time was nearly half a century away.
His idea of degeneracy was microstate counting, i.e. classical microscopic configurations which corresponded to
a single classical internal energy.
31 R. Baierlein, Am J. Phys. 63, 108 (1995).
27 M.

86

6.1.4

CHAPTER 6. KNOWING THE UNKNOWABLE

Gibbs Canonical Theory

J.W. Gibbs canonical theory32 may be seen as an extension of the microcanonical case for a
closed system (constant particle number) with many internal eigen-energies
hop s = s s

(6.38)

some of which may be degenerate.


A Thermal Lagrangian will again be taken as
= kB P
(s ) ln P
(s ) T 1 P
(s ) s 0 P
(s ) .
L [P]
s

(6.39)

by (functionally) differentiating with respect to P


(s ) and setting the results to
Maximizing L [P]
zero we get the set of equations (one for each eigen-index s)
kB [ln P (s ) + 1] T 1 s 0 = 0

s = 1, 2, 3, . . .

(6.40)

whose solutions are


P (s ) = exp {(
Defining =

T 1 s 0
) 1} .
kB

(6.41)

1 and then imposing normalization to eliminate gives the result


0
kB T
P (s ) =

exp (s )
.
exp { }

(6.42)

Replacing the denominator in Eq.6.42 with


Z = exp ( ) ,

(6.43)

which is Gibbs partition function, we can write


P (s ) =

exp (s )
.
Z

(6.44)

Note again that the sums are over all states, some of which may be degenerate. Explicitly accounting
for degeneracy in carrying out the the partition function sum33
Z = g( )exp ( )

(6.45)

where g( ) is the degeneracy of the state  .


32 This is commonly called the Canonical Ensemble. But quantum mechanics already brings with it an ensemble
interpretation, so an extra use of that term seems redundant.
33 The commonly used symbol Z arises from the German Zustandssumme the sum over states, a reminder of
what has to be summed!

6.1. ENTROPY THE TICKET TO OZ

6.1.5

87

Canonical thermodynamics

1. Internal Energy:
U = hop
= s P (s )

(6.46)
(6.47)

which is identical with

U =

ln Z
.

(6.48)

2. Entropy:

S = kB P (s ) ln P (s )

(6.49)

= kB
s

exp (s )
[s ln Z]
Z

U
+ kB ln Z .
T

(6.50)
(6.51)

Then with Eq.6.48


S = kB

ln Z + kB ln Z

(6.52)

and finally
S = kB 2

1
( ln Z) .

(6.53)

3. Pressure:
Since the pressure operator is pop = (

hop
) thermodynamic pressure is
V T

p = (
s

s
) P (s )
V T

(6.54)

where s (V ) is assumed known. Applying Eq.6.43 this is identical to


p=

1 ln Z
(
) .

V T

(6.55)

88

CHAPTER 6. KNOWING THE UNKNOWABLE


4. Helmholtz potential F :
From Eq.6.51 and the definition of the Helmholtz potential [see Eq.5.17] it follows that

F =

1
ln Z

(6.56)

which from Eqs.6.53 and 6.55 follow the usual thermodynamic results
F
)
T V
F
p = (
) .
V T

S = (

(6.57)

Although Z is not a measurable, it is among the most important quantities in statistical thermodynamics.

6.1.6

Degeneracy and Z

The sum in Eq.6.45 is over discrete energies. State multiplicity (degeneracy) is taken into account
by the factor g(s ).
There are, generally, two sources of degeneracy:
1. Microscopic degeneracy:
Generally arising from solutions to the eigenvalue problem
Hop Es = Es Es .

(6.58)

in which several eigenstates, e.g. E1 , E2 , E3 . . . correspond the same eigen-energy. This


source of degeneracy will be referred to as internal degeneracy. For example the spin independent s-states of the hydrogen atom are 1-fold degenerate (this is called non-degenerate).
If spin is included the s-states of the hydrogen atom are 2-fold degenerate.
2. Macroscopic (configurational) degeneracy: Usually associated with systems composed of large
numbers of macroscopic, energetically equivalent, configurations, e.g. lattices. These degeneracies can usually be determined by combinatoric arguments.
6.1.6.1

Examples

(a) A simple model of a crystal consists of N sites, each occupied by an atom34 with spin 1/2.
With no local magnetic field (B = 0) the two spin states, identified with quantum numbers
34 Spin spin

interaction is neglected.

6.1. ENTROPY THE TICKET TO OZ

89

sz = 21 (or as cartoons, , ), have identical eigen-energies, i.e. they are internally two
fold degenerate.
In this case there is a combination of internal and configurational degeneracy. Since
each atom is 2-fold degenerate, for N identical atoms the degeneracy factor is
g =2 2 2 2 . . .
N

=2 .

(6.59)
(6.60)

If, for definiteness, this state is taken to have energy = 0, then according to Eq.6.45
Z =2N e 0
=2N ,

(6.61)
(6.62)

Since the entropy of a Gibbs system is


S = kB 2

1
( ln Z)

(6.63)

we have the result

S =N kB ln 2

(6.64)

which is, of course, identical with Boltzmanns S = kB ln W .


If the atoms had higher internal degeneracy the entropy would increase less statistical
knowledge (greater uncertainty). If the atoms were spin = 0 (non-degenerate) then g = 1
and the entropy would be zero! i.e. statistical certainty and the smallest possible
entropy.

Figure 6.1: Zeeman effect with spin

1
2

atoms in a weak magnetic field.

(b) Assuming now that the S = 12 atoms are immersed in a weak but uniform local35 magnetic
field B so that the degeneracy is lifted [see Figure 6.1] with E = B and E = B.36
35 The
36

field at the atom site is taken to be the same as the field external to the sample before matter is inserted.
= S g B with S the intrinsic spin of the atom, g the electron g f actor and B the Bohr magneton.

90

CHAPTER 6. KNOWING THE UNKNOWABLE


If n are in the state E and n are in the state E , with N = n +n the total macroscopic
system energy for any configuration {n , n } is
E = n B n B .

(6.65)

Although the states E and E are no longer internally degenerate, the N atom system
has configurational degeneracy which is determined by counting the number of unique
ways N atoms can be arranged with n in the state E and n in the state E .
Among the N total atoms there are N ! permutations, no matter what the spin state.
But not all the permutations are unique arrangements of spins on sites. In fact, n ! n !
are redundant and must be divided out. Therefore the total configurational degeneracy
for the particular macroscopic energy E in Eq.6.65 is
g(N, n , n ) =

N!
.
n !n !

(6.66)

Consequently the partition function (sum over states) for the system in a magnetic field
is
Z=

n ,n
n +n =N

N!
exp [ (n n ) B]
n n

(6.67)

which is just the binomial expansion of


N

Z = (eB + eB )

(6.68)

The internal energy of the system is

ln Z

eB eB
=N B B
e
+ eB

U =

(6.69)
(6.70)

and from Eq.6.53 the entropy is


S = N kB {ln (eB + eB ) B (

eB eB
)} .
eB + eB

(6.71)

(c) A simple model of a linear elastomer37 consists of a chain with N identical links each of
length ` [See Figure. 6.2]. The elastomer can be stretched under tension until its maximum length is Lmax = N `. (A thin rubber band is an imperfect but helpful illustration.)
In this simplified model the linkage energies are very weak so that each molecule (link)
can point either to the left or the right and only to the left or right with negligible
energy difference.38 Referring to Eq.6.45, the partition function for this model depends
only on congurational degeneracy.
37 an

elastic polymer
not without significant deciencies, the model is useful for understanding unusual elastic and thermodynamic properties of rubber and other elastic polymers.
38 Although

6.1. ENTROPY THE TICKET TO OZ

91

Figure 6.2: Elastomer completely stretched under tension.

The fully extended polymer with length L = N ` [See Figure 6.2] has only one possible
conguration (actually two, since the links can all point to the left or all to the right).
If the tension is relaxed the elastomer will retract to length L < Lmax as depicted in
Figure 6.3.
Let nL be the number of left pointing links and nR be the number of right pointing links.
For this configuration the polymer length is
L = ` nR nL

(6.72)

where the magnitude defines the 2 f old left-right symmetry. The total number of links
is obviously
N = nR + nL

(6.73)

Figure 6.3: Depiction of retraction in an elastic chain.


There are N ! permutations among the N links. But not all are unique since any arrangements with nR right pointing links and nL left pointing links are equivalent. So
the number of distinct elastomer configurations is
g(nL , nR ) = 2

N!
nL ! nR !

(6.74)

where the factor 2 accounts for the left-right symmetry. Thus the partition function is

92

CHAPTER 6. KNOWING THE UNKNOWABLE


simply
N

Z=

g (nL , nR )

(6.75)

N!
nL ! nR !

(6.76)

nL ,nR
nL +nR =N
N

=2

nL ,nR
nL +nR =N

= 2 2N

(6.77)

and in accord with Eq.6.63 as well as Boltzmanns fundamental result (applying Stirlings
approximation, Eq.7.26, to evaluate the factorials in Eq.6.74)
S = (N + 1) kB ln 2 .

(6.78)

(d) Consider a dilute (non-interacting) system of N two level atoms with non-degenerate
quantum energy levels 0 = 0 and 1 = . When n0 atoms are in the ground state 0 and
n1 atoms are in the excited state 1 the macroscopic eigen-energies are
E = n0 0 + n1 .

(6.79)

The number of distinct congurations (degeneracy) associated with n0 atoms the in the
lower energy state and n1 atoms in the excited energy state is
N!
n0 ! n1 !

g=

(6.80)

so that the partition function is


N

Z=

n0 ,n1
n0 +n1 =N

N!
exp (n1 )
n0 !n1 !
N

= (1 + e )

(6.81)
(6.82)

(e) Ignoring spin, if the excited state of the N impurities in the last example was a hydrogenic
4 f old degenerate (2s + 2p) configuration and the ground state was a non-degenerate
1s state, what would be the degeneracy g?
Generalizing to the case where n0 impurity atoms are in an r f old degenerate ground
state and n1 are in a s f old degenerate excited state, with n0 + n1 = N ,
N ! n0 n1
r s
n0 !n1 !

(6.83)

N!
n
n
(1) 0 (4) 1 ,
n0 ! n1 !

(6.84)

g=
so that for the hydrogenic impurity
g=

6.1. ENTROPY THE TICKET TO OZ

93

and the partition function is


N

Z=

n0 ,n1
n0 +n1 =N

N!
n
n
(1) 0 (4) 1 en0 (0) en1
n0 ! n1 !
N

= (1 + 4e )

(6.85)
(6.86)

Note: In the case of an arbitrary number of states 1 , 2 , . . . m with degeneracies


s1 , s2 , . . . sm and site occupancies n1 , n2 , . . . nm
g=

N!
sn1 sn2 . . . snmm
n1 ! n2 ! . . . n m ! 1 2

(6.87)

which is a generalized multinomial coefficient39 with a corresponding partition function


N

Z=

n1 ,n2 ,...nm
n1 +n2 +...nm =N

N!
sn1 sn2 . . . snmm e1 n1 e2 n2 . . . em nm
n1 ! n2 ! . . . n m ! 1 2
N

= (s1 e1 + s2 e2 . . . sm em )

(6.88)
(6.89)

(f) Consider a lattice of N identical 1-dimensional quantum harmonic oscillators, each oscillator with allowed eigen-energies
0 (ni + 1 )
Ei = h
2

(6.90)

where 0 is the oscillators natural frequency, i = 1, 2, 3, . . . N and ni = 1, 2, 3, are


the ith oscillators allowed quantum numbers. The eigen-energies of the entire oscillator
lattice are
N

E(n) = Ei

(6.91)

i=1
N
0N
h
0 ni
+ h
2
i=1
h0 N
0n
+ h
=
2

(6.92)
(6.93)

where n is the total lattice quantum number,40


N

n = ni .

(6.94)

i=1
39 This result and its widely used generalizations can be found in M. Abramowitz and I. Segun, Handbook of
Mathematical Functions, Dover 1965, Chapter 24.
40 Here we have the opportunity to introduce the idea of quasi-particles, which are notional particles used to
represent the energy excitations of fields. In this case the harmonic oscillator is represented by quasi-particles
identified with a phonon field, i.e. the quantum numbers n. Clearly these quasi-particles must be notional in
the sense that, even though there are a fixed number N of vibrating atoms, there is no similar quasi-particle number
n conservation.

94

CHAPTER 6. KNOWING THE UNKNOWABLE


In principle, using Eq.6.93 the partition function could be written

0(
Z = g (N, n) exp [ h
n=0

N
+ n)]
2

(6.95)

where g (N, n) is the degeneracy of the lattice eigen-energy [see Eq.6.93].

Figure 6.4: Occupation representation for a lattice of N identical oscillators.


In Fig. 6.4 the N oscillators are represented with compartments created by N 1 partitions (shaded bars) while the quantum numbers ni are represented by n quasi-particles
(grey bullets in the boxes). The n quasi-particles are distributed in all possible ways in
the N boxes. The Figure shows an arrangement of two kinds of objects partitions
and quasi-particles. The number of distinct ways these41 can be distributed is
g (N, n) =

(N 1 + n)!
(N 1)! n!

(6.96)

which gives the partition function

(N 1 + n)!
0 ( N + n)]
exp [ h
(N

1)!
n!
2
n=0

N h0 (N 1 + n)!

0)
2
=e
exp (n h

n=0 (N 1)! n!
Z=

(6.97)
(6.98)

This sum is not a binomial expansion but may be approximated using Stirlings asymptotic expansion [see Eq.7.26]. (An alternative exact evaluation using the Euler -function

(n) = (n 1)! = tn1 et

(6.99)

is left to Chapter 9 where the harmonic oscillator is discussed.)

6.1.7

Closing comments

Boltzmanns pre-quantum insight linked degeneracy (multiplicity) with entropy. But as indicated
in this chapter, Boltzmanns conjecture follows, in principle, by applying the Thermal Lagrangian
41 There

are n quasi-particles, N 1 partition walls and N 1 + n total objects.

6.1. ENTROPY THE TICKET TO OZ

95

postulate introduced in this chapter in the special circumstance of a single, isolated, degenerate
macroscopic eigen-energy. Applying the same rule to a continuous spectrum of macroscopic eigenenergies leads to Gibbs Canonical case. With the Thermal Lagrangian (least-biased postulate)
thermal physics becomes a direct consequence of macroscopic quantum theory.
Thermodynamics was a serious and practical science long before there was a quantum mechanics.
Doubtless, it will continue to be widely applied in its pragmatic, time honored classical form by
engineers and chemists in which the role played by quantum mechanics may be of only academic
interest or of NO interest at all.
But quantum mechanics lends modern thermodynamics its distinctive and profound character in
which microscopic models can be used as a basis for even the most exotic macroscopic systems.
This union has fashioned one of the most versatile and widely applied paradigms in modern science.

Problems and exercises


1. A hypothetical macroscopic system hamiltonian hop has eigen-energies and eigenstates given
by
hop j = j j

(6.100)

where the eigenvalues are


j = j,

j = 0, 1, 2, . . . .

(6.101)

Here a positive constant and each eigenstate j is j f old degenerate.

a. Find Pj , the least biased, normalized ( Pj = 1) probability of the macroscopic outcome


j=1

j = j.
b. What is the entropy of the hypothetical system?
c. Find the average value hop = T rop hop .
d. Find the mean uncertainty
2

(j ) = (j hop ) Pj .

(6.102)

j=1

2. DNA consists of a pair of molecular strands curled to form a double helix. With increasing temperature the strands dissociate, unraveling the helix i.e. the DNA melts. [see
Figure 6.5.] A simple model of DNA melting42 consists of two linked strands whose links
starting from only one end of the chain break-up successively. For a link to break all links
to the left of it must already be broken like a single-ended zipper! The only interior link
42 C.

Kittel, Phase Transition of a Molecular Zipper, Am. J. Phys. 37, 917 (1969).

96

CHAPTER 6. KNOWING THE UNKNOWABLE


that can break is one immediately adjacent to one that has already broken. The last link is
considered unbreakable.
3.
N96

p90
1.

Figure 6.5: The melting DNA double helix showing p broken links in a chain of N total bonds.
Assuming the links 1, 2, . . . p 1 are already broken, the energy required to break the pth link
is +. Unbroken links are taken to have energy 0. Each broken link has degeneracy due to
the multitude of spatial orientations it can assume.
(a) Show that the partition function is
1 xN
1x

(6.103)

x = exp ()

(6.104)

Z=
where

with
=

1
kB T

(6.105)

(b) In this model an order parameter is defined as the average number of broken bonds, p.
Show that
=

N xN
x

.
xN 1 x 1

(6.106)

(c) Plot vs. x for N = 1000 to demonstrate the transition and critical point at x = 1.
(d) Show that for N  << 1 the entropy of zipper DNA is
S = ln .

(6.107)

Hint:
SN 1 = 1 + x + x2 + . . . xN 1

(6.108)

1x
.
1x

(6.109)

J. Willard Gibbs is probably the most brilliant person most people have
never heard of.
Bill Bryson, A Short History of Nearly Everything, Broadway
Publishing, NY (2003).

Chapter 7

The Ideal Gas


7.1

Introduction

An ideal gas is a collection of randomly moving (uncorrelated) identical particles with negligible
forces of interaction.1
Even before there was a fully accredited quantum mechanics, W. Pauli2 conjectured that only one
electron can occupy a single-particle energy eigen-state a restriction called the Pauli exclusion
principle (PEP).
The following year, E. Fermi3 and P. Dirac4 further showed that quantum mechanics required all
particles depending on their intrinsic spin S, which can be integer or half-integer belong to one
of two possible classes:
a. Particles with half-integer spin (S = 1/2, 3/2, 5/2, . . .) obey the Pauli exclusion principle and
are called Fermi-Dirac (FD) particles, or fermions. Members of this class are quarks, electrons,
protons, neutrons and neutrinos, as well as their antiparticles, and many composite systems. For
example He3 is a composite of 2 protons, 2 electrons and 1 neutron, each with spin 1/2. This
odd number aggregate insures that He3 has 1/2-integer spin. Electrons (together with a uniform
positive background) are modeled as a gas of spin 1/2 Fermi-Dirac particles.
b. Particles with integer spin (S = 0, 1, 2, . . .) have no restriction on single-particle eigen-state
occupation. Integer spin particles are called Bose-Einstein (BE) particles, or bosons. Although
most bosons are composite systems, e.g. the H atom (1 proton and 1 electron), He4 (2 protons,
1 Just

enough interaction to eventually reach thermal equilbrium.


Pauli, On the Connexion between the Completion of Electron Groups in an Atom with the Complex
Structure of Spectra, Z. Phys. 31, 765 (1925).
3 E. Fermi, Sulla quantizzazione del gas perfetto monoatomico, Rend. Lincei 3: 145 (1926).
4 P. Dirac, On the Theory of Quantum Mechanics, Proc. Royal Soc, A 112: 661 (1926).
2 W.

97

98

CHAPTER 7. THE IDEAL GAS


2 electrons and 2 neutrons) and mesons (2 spin 1/2 quarks), there are also elementary particle
bosons, e.g. W and Z particles which carry the weak nuclear force and Gluons which are
associated with the strong nuclear force. Also included is the particle (photon) and the
feverishly sought Higgs boson H 0 . But in this era of condensed boson physics, composite boson
systems currently occupy center stage, such as atoms of Li7 , N a23 , K 41 , Rb85 and Rb87 .

In the case of many-particle systems, e.g. He3 (fermions) and He4 (bosons), there are enormous
differences in their low temperature behavior. These distinctions will be discussed in Chapters 15
and 16.

7.1.1

Ideal gas law

In the limit of low particle density and high temperature, ideal FD and BE gasses nearly shed their
quantum properties so that century-and-a-half old kinetic (classical) gas models can describe much
of their thermodynamic behavior, in particular the Equation of State
pV = N kB T

(7.1)

where p is the gas pressure, N is a fixed number of particles, V the confining volume, T the gas
temperature and kB is Boltzmanns constant.5 However in the final analysis the quasi-classical
theory is internally inconsistent and in the end must be rescued by quantum corrections.
Taking the opportunity to apply a Thermal Lagrangian to a physical problem whose results (and
contradictions) are well known, gaseous atoms and molecules near room temperature and atmospheric pressure are modeled as ideal andquasi-classical. The ideal gas is so important in
thermodynamic pedagogy as well as for many low order approximations that we forego absolute
rigor and accept heuristic arguments for these results.

7.1.2

Quasi-classical model

The single particle microscopic hamiltonian operator for non-interacting particles is


2
h
2
2m

(7.2)

hop s = s s

(7.3)

hop =
with an eigenvalue equation

with s the microscopic eigen-energies.


Consider now a low density system consisting of N independent particles each with mass m confined
to a volume V (no FD or BE occupation restrictions). Each particle can occupy any of the allowed
microscopic eigen-energies 1 , 2 , 3 . . . with the corresponding macroscopic eigen-energies
E (n1 , n2 , n3 , . . .) = n1 1 + n2 2 + n3 3 + . . .

(7.4)

5 A gas which obeys Eq.7.1 can be used as an operational thermometer in which T defines the absolute (K)
temperature scale.

7.1. INTRODUCTION

99

where n1 particles are in the state 1 , n2 particles are in the state 2 , etc., with a fixed total number
of particles
n1 + n2 + n3 . . . = N .

(7.5)

Guided by discussions in Chapter 6 the Thermal Lagrangian is


N

L [P] = kB

P (n1 , n2 , . . .) ln P (n1 , n2 , . . .)

n1 ,n2 ,n3 ...=0,1,2...


n1 +n2 +...nN =N
N

T 1

P (n1 , n2 , . . .)[n1 1 + n2 2 + n3 3 + . . . nN N ]

n1 ,n2 ,n3 ...=0,1,2...


n1 +n2 +...nN =N
N

P (n1 , n2 , . . .) .

(7.6)

n1 ,n2 ,n3 ...=0,1,2...


n1 +n2 +...nN =N

With = 1/kB T , maximizing L [P]


exp {(n1 1 + n2 2 + . . . nN N )}

P (n1 , n2 , . . . nN ) =

(7.7)

exp {(n1 1 + n2 2 + . . . nN N )}

n1 ,n2 ,n3 ...=0,1,2...


n1 +n2 +...nN =N

The denominator is the partition function Z and the sums are over all states. Following the
procedure for calculating a partition function
N

Z=

n1 ,n2 =0,1,2,...
n1 +n2 +...=N

gN (n1 , n2 , ) exp {(n1 1 + n2 2 + . . . nN N )}

(7.8)

where gN (n1 , n2 , ) is the configurational degeneracy of an N -particle state6 with energy given
by Eq.7.4
N!
gN (n1 , n2 , . . .) =
,
(7.9)
n1 ! n2 ! n3 ! . . .
so that
N

Z=

N!
exp {(n1 1 + n2 2 + . . . nN N )}
n
!n
n1 ,n2 =0,1,2,... 1 2 !

(7.10)

n1 +n2 +=N

which is the expansion of


N

Z = (e1 + e2 + e3 + . . .)

(7.11)

Unfortunately Eq.7.11 is not correct! For at its heart this is a problem in many-particle quantum
mechanics with no rigorous semi-classical argument.7 Indeed, J. W. Gibbs quickly realized that
after applying Eq.6.53 to Eq.7.11 the result was not an extensive entropy. To correct the partition function an ad hoc 1 prefactor (known as the Gibbs correction) was inserted.8 With this
N!
6 See

Chapter 6
issue will be addressed in Chapter 15.
8 The correction is a consequence of normalizing quantum many-particle state functions.
7 This

100

CHAPTER 7. THE IDEAL GAS

correction, the partition function becomes

Z=

7.1.3

N
1
(e1 + e2 + e3 + ) .
N!

(7.12)

Calculating a partition function

Practical ideal gas results can be obtained from Eq.7.12 by applying the quantum mechanical energy
states of a free particle confined to a cube9,10 of side L (volume V = L3 ),
 =

2 2
h
( ) (x2 + y2 + z2 )
2m L

(7.13)

where x , y , z = 1, 2, 3, . . .. The eigen-energies in Eq.7.13 correspond to particles with purely


translation modes, e.g. a monatomic gas or a polyatomic gas without vibrational, rotational or
electronic excitations.11
Using the energies in Eq.7.13 and the partition function, Eq.7.12,
N

2
h
2
2
2

)
(

(
)
+

1
x
y
z

e 2m L
Z=

N ! x ,y ,z =1

(7.14)

2 3N

2
h ( x )

1
.
Z=
e 2m L

N ! x =1

(7.15)

which is

Now substitute the counting variable


kx =

x
,
L

(7.16)

which implies
L
kx = x .

(7.17)

9 Periodic boundary conditions are almost universally applied for free particles and fields. But in this problem we
choose the equivalent and probably more familiar large 3 D box boundary conditions.
10 Lower dimensional systems, which are of considerable importance in modern materials science, might be dealt
with in an analogous way.
11 These eigen-energies are also appropriate for ideal FD and BE quantum gases so long as FD and BE occupation
number restrictions are observed. FD and BE quantum gases have strikingly different partition functions, even with
the same allowed eigen-energies. [See Chapters 15 and 16.]

7.1. INTRODUCTION

101

Since the quantities j , . . . are the integers 1, 2, . . . it follows that all j 1 and the partition
function is

2 x 2 3N
2 3N
h
h


(
)

k2
L
1
1
kx e 2m x .

2m
L
Z=

N ! x =1
N ! kx

(7.18)

When the cell becomes large, i.e. L , it follows that kx dkx , and the sum over kx in
Eq.7.18 can be replaced by the equivalent integral

2 3N
h

kx2
1 L
.
2m
dk
e
Z=
x

N !

(7.19)

Using the standard Gaussian integral result

x
=
dx e
0

1
2

(7.20)

the partition function is

3N
1 L
2m
Z=
[
2 ]
N ! 2
h
1
N
=
(nQ V )
N!

(7.21)
(7.22)

where
nQ = (

3/2
m
)
2
2 h

(7.23)

is the quantum concentration a particle density parameter that measures the gas degree of
dilution. In particular, if n/nQ << 1, where n = N /V is the particle concentration, then the gas is
sufficiently dilute for the quasi-classical ideal gas result Eq.7.21 to apply. If, on the other hand,
n/nQ >> 1 quantum considerations are inescapable and the arguments that brought us this far are
inadequate.12,13

7.1.4

Thermodynamics of the ideal gas

This ideal gas model is replete with familiar thermodynamic results:


12 See

Chapters 15 and 16
also appears in the partition function yet
addition to the quantum mechanical 1/N !, Plancks constant h
another intrusion of quantum mechanics. Eq.7.21 shows that even the classical ideal gas often imagined as tiny
billiard balls banging about in a box has quantum imprinting.
13 In

102

CHAPTER 7. THE IDEAL GAS


a. Internal energy U:
With P (n1 , n2 , . . .) from Eq.7.7, the internal energy is
N

U=

P (n1 , n2 , . . .) [n1 1 + n2 2 + n3 3 + nN N ] ,

(7.24)

n1 ,n2 ,n3 ...=0,1,2...


n1 +n2 +...nN =N

which is identical with


U =

ln Z
.

(7.25)

Using Eq.7.21 together with Stirlings Approximation


n! nn en

n >> 1

ln n! n ln n n

(7.26)

n >> 1 ,

we have
3N
2
3
= N kB T .
2

(7.27)

U=

(7.28)

b. Heat capacity CV : With only translational (kinetic) energy


U
)
T V

(7.29)

3
= N kB .
2

(7.30)

CV = (

c. Entropy S:
1

) ( ln Z)

5 3
m
N
=N kB [ + ln ( 2 ) ln ( )]
2 2
2 h
V
nQ
5
=N kB [ + ln (
)]
2
n

S =kB 2 (

(7.31)
(7.32)
(7.33)

which is called the Sakur-Tetrode equation.14,15


14 The inclusion of 1/N ! the Gibbs correction is a fundamental quantum mechanical result derived from the
requirement that under permutation of identical particles the many-particle state vector changes, at most, by a phase
factor.
?

Gibbs 1/N ! correction makes an essential contribution to the entropy. Without it S = N kB [ 32 + ln (V nQ )]


which does not satisfy Eulers homogeneity condition S (V ) = S (V ). S would not be extensive.
15 The

7.1. INTRODUCTION

103

d. Helmholtz Potential F :
Applying Stirlings approximation to Eq.6.56, where it was shown that
1
F = ln Z ,
(7.34)

we find
F =

nQ
N
(ln
+ 1) .

(7.35)

e. Equation of State:
The expression for pressure is
1 ln Z
(
)

V T
F
)
= (
V T

p=

(7.36)
(7.37)

giving the ideal gas law.


p=

7.1.5

N kB T
.
V

(7.38)

Gibbs entropy paradox

Consider a cylinder that is partitioned into two sections with volumes V1 and V2 so that the total
volume is V = V1 + V2 . N1 and N2 absolutely identical ideal gas molecules occupy the volumes V1
and V2 , respectively same quantum concentration, same temperature T , same concentration n.
If the partition magically dissolves the identical gases will spontaneously mix and occupy the total
volume V with the equilibrium temperature and concentration unchanged.
Objectively, nothing has happened. There is no way to distinguish between the before and
after. There is no change in the state of knowledge. For this process mixing has, apparently, no
objective meaning! Is there a change in entropy?
The Sakur - Tetrode equation gives the entropy of the original system (before the partition dissolves)
as
S =S1 + S2

(7.39)

=N1 kB [ln (nQ /n) + 5/2] + N2 kB [ln (nQ /n) + 5/2]

(7.40)

=N kB [ln (nQ /n) + 5/2]

(7.41)

which is identical to the entropy of the system after the partition disintegrates entirely in accord
with physical intuition and our interpretation of entropys informational meaning.16
If, however, the partition function of Eq.7.22 had not been quantum corrected by the Gibbs 1/N !
there would be an increase in the entropy after the partition disintegrates, which is nonsense!
16 Note, that when calculating thermodynamic averages such as U or p the Gibbs 1/N ! correction cancels. The
same cancellation does not occur in entropy calculations.

104

7.1.6

CHAPTER 7. THE IDEAL GAS

Entropy of mixing

Consider the situation just described, but now with two different ideal gases, a and b, with
unique quantum concentrations naQ and nbQ . Initially, species a occupies a volume Va and species
b occupies a volume Vb . The partition function for gas a is
Na

Za =

(Va naQ )

(7.42)

Na !

and the partition function for gas b is


Nb

Zb =

(Vb nbQ )

(7.43)

Nb !

Applying the Sakur-Tetrode equation to each species, the total initial entropy is
(7.44)

S =Sa + Sb
=Na kB [5/2 + ln (naQ /na )] + Nb kB

[5/2 + ln (nbQ /nb )]

(7.45)

where na = Na /Va and nb = Nb /Vb .


After the chamber partition magically dissolves and the gases mix, the total entropy of the mixed
system is
Sa+b = Na kB {ln [(Va + Vb ) naQ /Na ] + 5/2} + Nb kB {ln [(Va + Vb ) nbQ /Nb ] + 5/2} .

(7.46)

Defining an entropy of mixing as Smixing = Sa+b (Sa + Sb ) then


Smixing = Na kB ln (

Va + Vb
Va + Vb
) + Nb kB ln (
)>0
Va
Vb

(7.47)

The mixing entropy Smixing has the same value even if the partition functions had not been Gibbs
(quantum) corrected by 1/N !. Moreover, Eq.7.47 now says nonsensically that a finite mixing
entropy remains even after the limit
naQ nbQ

(7.48)

has been taken! Mixing entropy is obviously not a continuous function of nQ . This, physically, may
be attributable to the quantization of mass units.

7.1.7

The non-ideal gas

and the Gibbs correction factor


The idea of a classical ideal gas is clearly a fiction since h
1/N !, both from quantum mechanics, must appear in the ideal gas entropy. Moreover, atoms and
molecules in a real gas do interact and with significant consequences. An approximate equation of
state that improves upon the ideal gas law is due to Van der Waals (1873) who took account of
long range attraction and short-range repulsion between particles. Van der Waals equation not only

7.1. INTRODUCTION

105

improves descriptions of gas phase thermodynamics, it also suggests gas-liquid phase transitions and
gas critical points as consequences of interactions. This eponymous Equation of State is:17
2

[p + (N /V ) a] (V N b) = N kB T

(7.49)

where a and b are coefficients characteristic of specific gases.18


The van der Waals equation, by quantitatively providing two parameters that characterize real
molecules, would seem to have settled a bitter ongoing debate of the times did gases, and matter
in general, have microscopic structure? But it didnt.

Problems and exercises


1. Assume an N -particle ideal gas in 2 dimensions with eigen-energies
En =

2 2
h
( ) (n2x + n2y )
2m L

(7.50)

where
nx , ny = 1, 2, 3, ,

(7.51)

A =L2 .

(7.52)

The confining surface area is

(a) Write an expression for the partition function Z2D of the 2-dimensional ideal gas.
(b) Find the internal energy U2D of the gas.
(c) Find the heat capacity CA for the 2 D gas.
(d) Find the entropy S2D of the gas.
2. The quantity
T = (

U
)
V T

(7.53)

is called internal pressure and expresses the role of particle-particle interactions in the behavior
of a gas. Find T for
(a) An ideal gas.
(b) A van der waals gas.
(c) A gas with the equation of state
p=
17 A

N kB T
N
[1 + B (T )]
V
V

(7.54)

derivation of the van der Waals equation is given in Chapter 13.


R. C. (Ed.), Handbook of Chemistry and Physics (53rd Edn.), Cleveland:Chemical Rubber Co., 1972.

18 Weast.

106

CHAPTER 7. THE IDEAL GAS

A theory is the more impressive the greater the simplicity of its


premises, the more different kinds of things it relates and the more
extended is its area of applicability.
A. Einsteina
a Reproduced by permission of Open Court Publishing Company, a division
of Carus Publishing Company, from Albert Einstein: Philosopher-Scientist,
edited by Paul Arthur Schilpp (The Library of Living Philosophers Volume
VII), 1949, 1951 and 1970 by The Library of Living Philosophers.

Chapter 8

The 2-Level System


8.1

Schottky anomalous heat capacity

When insulators or semiconductors contain impurities with closely spaced electronic or nuclear
ground states1 heat capacities CV show an anomalous low temperature peak the so called Schottky
anomaly.2 It is an example of microscopic quantum effects having macroscopic (thermodynamic)
signatures.

8.1.1

Schottky Model

A simple model for the Schottky anomalous heat capacity consists of a crystal containing a
dilute3 concentration of N identical atomic impurities, each having the same pair of closely spaced,
non-degenerate energy levels a lower state 1 and a higher state 2 . If at some temperature T , n1
impurity atoms are in the low energy state and n2 are in the higher energy state the macroscopic
eigen-energies are
E (n1 , n2 ) = n1 1 + n2 2

(8.1)

1 Usually degenerate low energy states whose degeneracy is lifted by crystalline electric fields, magnetic fields,
spin-orbit coupling or hyperfine interactions.
2 The phenomenon is named for Walter Hermann Schottky (1886-1976).
3 Diluteness insures negligible inter-atomic effects.

107

108

CHAPTER 8. THE 2-LEVEL SYSTEM

with the Thermal Lagrangian


L = kB

P (n1 , n2 ) ln P (n1 , n2 )

n1 ,n2
n1 +n2 =N

T 1

(8.2)
P (n1 , n2 ) (n1 1 + n2 2 ) 0

P (n1 , n2 ) ,

n1 ,n2
n1 +n2 =N

n1 ,n2
n1 +n2 =N

where P (n1 , n2 ) is the probability that n1 impurity atoms are in the state 1 and n2 are in the state
2 . The sums are over all states. The method of Lagrange multipliers gives the probabilities
P (n1 , n2 ) =

e(n1 1 +n2 2 )
Z

(8.3)

Z=

e(n1 1 +n2 2 ) .

(8.4)

with the partition function

n1 ,n2
n1 +n2 =N

which again represents a sum over all states. One way to calculate this sum is by explicitly including
a configurational degeneracy [see Chapter 6]
N!
n1 ! n2 !

(8.5)

N ! (n1 1 +n2 2 )
e
n1 ! n2 !

(8.6)

g(n1 , n2 ) =
to give
N

Z=

n1 ,n2
n1 +n2 =N

which is the binomial expansion of4


N

Z = (e1 + e2 ) .

(8.7)

The thermodynamic properties that follow from Eq.8.7 are:


A. Internal energy:

ln Z

1 e1 + 2 e2
=N
e1 + e2

U =

(8.8)
(8.9)

4 Some authors prefer to calculate the partition function - one atom at a time - and then take the product of
N -single atom partition functions. This avoids the mathematics of combinatoric analysis in obtaining g. But it also
evades the macroscopic character of the physics.

8.1. SCHOTTKY ANOMALOUS HEAT CAPACITY

109

B. Entropy:

S = kB P (n1 , n2 ) ln P (n1 , n2 )

(8.10)

= kB 2

1
( ln Z)

(8.11)

(1 e1 + 2 e2 )

=N kB ln (e1 + e2 ) +

(e1 + e2 )

(8.12)

C. Heat capacity: In laboratory measurements an external thermodynamic constraint on


solids is usually constant atmospheric pressure. A measured heat capacity is then Cp .
However thermodynamic models more conveniently produce CV , i.e.
CV = T (

S
)
T V

(8.13)

which in solids is usually assumed not to significantly differ from Cp . Then taking Eq.8.12
with 1 = 0, 2 1 = 
e

CV = N kB ()

(e + 1)

(8.14)

which in terms of a dimensionless plotting coordinate X =  becomes


CV
eX
.
= X2
2
N kB
(eX + 1)

(8.15)

B T = 0.417. Although
Plotting this in Figure 8.1 a Schottky peak appears at X 1 = k
the low temperature heat capacity of insulating crystals usually has a dominant T 3
contribution from lattice vibrations, the sharp anomalous peak can be distinguished above
this lattice baseline.

In the high temperature limit, X << 1, Eq.8.15 reduces to


CV
X2

N kB
4
 2
=(
) .
2kB T

(8.16)
(8.17)

At low temperature, X >> 1, Eq.8.15 becomes


CV
X 2 eX
N kB
=(

 2 kT
) e B
kB T

(8.18)
(8.19)

110

CHAPTER 8. THE 2-LEVEL SYSTEM

B T to emphasize
Figure 8.1: Schottky heat capacity anomaly. The abscissa is taken to be X 1 = k
the characteristic low temperature Schottky peak.

which has the limiting Schottky behavior


lim CV = 0 .

(8.20)

T 0

8.1.2

Two Level Systems and Negative Temperature

Writing internal energy, Eq.8.9, as


U =N


1 + e

(8.21)

a dimensionless internal energy = NU can be defined. Then the entropy of Eq.8.12
S = N kB [ln (1 + e ) +


],
e + 1

(8.22)

can be re-expressed in terms of as


S
= ( 1) ln (1 ) ln
N kB

(8.23)

which is plotted in Figure 8.2.


From the thermodynamic identity written as
T dS = N  d + pdV

(8.24)

8.1. SCHOTTKY ANOMALOUS HEAT CAPACITY

111

Figure 8.2: S vs.. The region 0.5 < 1 corresponds to negative temperature.
or
dS =

N 
p
d + dV ,
T
T

(8.25)

the slope at any point is [See Figure 8.2]


1 S

(
) =
N N
T

(8.26)

which displays the properties:


a. For < 0.5 Slope> 0 temperature is positive
b. At = 0.5 Slope= 0 temperature is infinite
c. For > 0.5 Slope< 0 temperature is negative
Negative temperature means that a population of excited atoms is unstably inverted, i.e. the
system is not in a state of thermodynamic equilibrium and is, at the smallest perturbation, ready
to dump its excess energy into the ground state. In this sense negative temperatures are hotter
than positive temperatures. To achieve negative temperatures the excitation energies of the system
must have an upper bound, a condition satisfied by the isolated two level Schottky model.
The typical two level system is however rarely isolated i.e. phonons, radiation, magnetic excitations, etc. are available for interaction. Therefore quasi-stable negative temperature conditions are
unusual. There are, however, special circumstances in which strongly prohibitive optical selection
rules prevent triggers from acting except on long time scales. Population inversion in lasers is
one example in which a negative temperature is maintained until stray photons stimulate the
inverted population into a downward avalanche.

112

CHAPTER 8. THE 2-LEVEL SYSTEM

Problems and exercises


1. A large system of N weakly interacting atoms is in thermal equilibrium. Each atom has
only three possible eigen-states with energies E = 0, E =  and E = 3, all of which are
non-degenerate.
a. Find the partition function Z for the N atom system.
b. Find U/N , the internal energy per atom, when  << 1.
2. A system in thermal equilibrium at temperature T consists of a large number of atoms N0
each of which can exist in only two quantum energy states: E1 which is non-degenerate, and
E2 which is p-fold degenerate, with E2 E1 =  > 0. The macroscopic eigen-energies are
E (n1 , n2 ) = n1 E1 + n2 E2 .

(8.27)

with n1 atoms in state E1 and n2 atoms in state E2 .


a. Find the partition function Z for the N0 atom system.
b. Find an expression for the internal energy U.
c. Show that in the high temperature limit the internal energy is
p
.
1+p

(8.28)

S = N0 kB ln (1 + p) .

(8.29)

U = N0
and the entropy is

d. Find an expression for the high temperature heat capacity CV .


3. Suppose each atom in a lattice of N atoms has three states with energies , 0 and +.
The states with energies  are non-degenerate whereas the state with energy 0 is two-fold
degenerate.
a. Find the partition function Z of the system.
b. Calculate the internal energy of the system as a function of temperature.
c. Show that the heat capacity per atom CV /N is
CV /N =

2 2
.
1 + cosh ()

(8.30)

I believe that no one who has won, through long years of experience, a
reasonably reliable sense for the not always easy experimental
evaluation of a theory, will be able to contemplate these results without
immediately becoming convinced of the huge logical power of the
quantum theory.
W. Nernst, Z. fur Elektrochem. 17, p.265-275 (1911). [Trans. A.
Wasserman]

Chapter 9

Lattice Heat Capacity


9.1

Heat Capacity of Solids

The nearly 200 year old Dulong-Petit rule for molar heat capacities of crystalline matter cv ,
predicts the constant value
3
cv = NA kB
2
=24.94J/mole ,

(9.1)
(9.2)

where NA is Avagadros number. Although Dulong-Petit, which assumes solids to be dense, classical, ideal gases [see Eq.7.29], is in amazingly good agreement with the high temperature ( 300K)
molar heat capacities of many solids, it fails to account for the observed rapid fall in cv at low
temperature. An especially large effect in diamond caught Einsteins (1907) attention and with
extraordinary insight he applied Planks quanta to an oscillator model of an atomic lattice to
predict a universal decline in cv as T 0K. Several years later, when low temperature molar
heat capacities could be accurately measured, they were indeed found to behave in approximate
agreement with Einsteins prediction.1 It was this result that ultimately succeeded in making the
case for quantum theory and the need to radically reform physics to accommodate it.

1 c in metals has an additional very low temperature contribution T from conduction electrons which, of course,
v
Einstein could not account for.

113

114

9.1.1

CHAPTER 9. LATTICE HEAT CAPACITY

Einsteins Model

Einsteins model assumes a solid composed of N atoms, each of mass M , bound to equilibrium sites
within a unit cell by simple harmonic forces. The potential energy of each atom is
1
V (R ) = M 02 R R
2

(9.3)

+ 2 R = 0
R
0

(9.4)

with a clasical equation of motion

where R = R R,0 is the displacement vector of the th ion from its origin R,0 . Einsteins
independent oscillator model ignores any interactions between ions so there is only a single mode
with frequency 0 . From a modern perspective Einsteins intuitive harmonic assumption is correct,
since atoms in a solid are bound by a total potential energy V (R ) consisting of:
1. A short range repulsive component arising from the screened coulomb interaction between
positively charged ion cores
V,

Z Z expR R
e2
,

2 ,
R R

(9.5)

2. A long range attractive component arising from quantum mechanical electron-electron correlations and ion-electron interactions.
The two potential energy components are shown in Figure 9.1 together with their sum which has a
nearly harmonic2 minimum near R0 .

9.1.2

Einstein Model in 1 D

Einsteins groundbreaking calculation of the heat capacity of a crystal lattice the first application
of a quantum theory to solids is based on an independent oscillator model. Its simplest form
considers N independent oscillators in a one-dimensional lattice with the th oscillator potential
obtained by expanding V (r) about the solid curves minimum R0, [see Figure 9.1].
1
V (R ) = V (R0, ) + M 02 R 2
2

(9.6)

R = R R0, .

(9.7)

with

Quantizing the model results in coresponding quantum energy levels


0 (n + 1 )
E(n ) = h
2

(9.8)

2 With increasing displacement from R anharmonicity (departures from harmonicity) has significant physical
0
consequences.

9.1. HEAT CAPACITY OF SOLIDS

115

V(r)

Figure 9.1: Long dashed line is the screened (short-range) coulomb repulsion between ion cores.
Short dashed line is the effective ion-ion attraction due to quantum mechanical electron correlations
and ion-electron interactions. The solid line is a nearly harmonic sum of the two contributions, with
a potential minimum at R0 .

where = 1, 2, N and 0 is the natural oscillator frequency. The constant term V (R0, ) is
ignored. For the 1 D oscillator the quantum number has the integer values n = 0, 1, 2, . The
integer lattice energy levels represent quasi-particles called phonons. Although they are not real
particles they bear all relevant kinetic attributes of real particles (energy, momentum, etc.) except
they are not number conserved. The lack of number conservation has the consequence that they
always have chemical potential = 0.
With all N atoms contributing, the macroscopic eigen-energies are
N
0
N h
0 n
+ h
2
=1
0
N h
0n
=
+ h
2

E (n) =

(9.9)
(9.10)

with
N

n = n
=1

(9.11)

116

CHAPTER 9. LATTICE HEAT CAPACITY

Figure 9.2: Cartoon array of 1-D Einstein harmonic oscillator potentials showing the equally spaced
energy levels of Eq.9.8.
9.1.2.1

Partition Function for 1 D Einstein Model

From the thermal lagrangian

L = kB P (n) ln P (n)
n=0

N
1 N
0 ( N + n)] 0 P (n)
P (n) [ h
T n=0
2
n=0

(9.12)

it follows that
0(
exp [ h
P (n) =

N
+ n)]
2

(9.13)

Z
Emphasizing the role of degeneracy in performing the partition funtcions sum over all states,
Eqs.6.96 and 6.97 are used. The partition function is therefore written

(N 1 + n)!
0 ( N + n)]
exp [ h
(N

1)!
n!
2
n=0

N h0 (N 1 + n)!

0) .
2
=e
exp (n h

n=0 (N 1)! n!

Z=

(9.14)
(9.15)

The resulting sum is not entirely obvious, although it can be approximated using the Stirling
Approximation, Eq.7.26. But with the integral representation of the f unction

(n) = (n 1)! = dt tn1 et

(9.16)

we can make the replacement

(N + n 1)! = dt t(N +n1) et

(9.17)

so the partition function becomes


Z =e

0
N h
tN 1 et 1 n n h
0
2
.
t e
dt
(N 1)! n=0 n!
0

(9.18)

9.1. HEAT CAPACITY OF SOLIDS

117

Now summing over n (to get an exponential) and then integrating over t (again using the
f unction)
0
N h
2
Z=
.
N

(e h0 1)
e

(9.19)

Evaluation of the 1 D oscillator partition function in this way is not especially difficult, but it is
not obviously extensible to higher dimensionality or to more physically interesting models.

9.1.3

Phonons and the 1 D Oscillator

A preferable route using Eq.9.10 for the eigen-energies, has the 1 D oscillator partition function

Z = . . . exp [ (
n1 =0 n2 =0

nN =0

N
0
N h
0 n )]
+ h
2
=1

nN =0

=1

0 n ] ,
=eN h0 /2 . . . exp [ h
n1 =0 n2 =0

(9.20)
(9.21)

where the sum over states is equivalent to the sum over all numbers of phonons n . Explicitly
summing over gives a product of N identical geometrical series

Z =e

0
N h

0 nN ])
0 n2 ]) ( exp [ h
0 n1 ]) ( exp [ h
2
( exp [ h

(9.22)

nN =0

n2 =0

n1 =0

hence

Z =eN h0 /2 ( e h0 n )

(9.23)

n=0

which is summed to give the result

eN h0 /2
,
N

(e h0 1)

(9.24)

identical to Eq.9.19.

9.1.4

The 3 D Einstein Model

The 3 D Einstein model, although suggesting a better approximation, still falls short of what is
physically observed. [See the Debye model discussion later in the chapter.] Here the coordinate components x, y, z,
of the th independent oscillator displacements are taken into account with the eigen-energies
0 (n,x + n,y + n,z + 3/2)
E (n,x , n,y , n,z ) = h

(9.25)

118

CHAPTER 9. LATTICE HEAT CAPACITY

where again = 1, 2, 3, N and n,x , n,y , n,z = 0, 1, 2, , with 0 is the oscillator frequency.
The N oscillator lattice has the macroscopic eigen-energies
Ex,y,z =

9.1.4.1

N
0
3N h
0 (n,x + n,y + n,z )
+ h
2
=1

(9.26)

Partition Function for the 3 D Einstein Model

Using the result of Eq.9.26 the thermal lagrangian is as Eq.9.12 and the partition function written
0
3N h

0 (n,x + n,y + n,z )]}


2
Z =e
(9.27)

. . . exp { [h
n1,x =0 n2,x =0
n1,y =0 n2,y =0
n1,z =0 n3,z =0

nN,x =0
nN,y =0
nN,z =0

=1

The sum is managed, as in section 9.1.3, by first explicitly summing over in the exponential.
Then, because the three coordinate sums (n,x , n,y , n,z ) are identical, what remains is
0
3
3
3
3N h

2
Z =e
( exp [ h0 n1 ]) ( exp [ h0 n2 ]) ( exp [ h0 nN ])

(9.28)

nN =0

n2 =0

n1 =0

and finally

Z =e

0
3N
3N h

2
( e h0 n ) ,

(9.29)

n=0

where the remaining sum gives


0
h
3N

1
Z = e 2 (
) .

1 e h0

9.1.4.2

(9.30)

Thermodynamics of the 3 D Einstein Model

Following the steps from previous chapters, the internal energy is

ln Z

0 [ 1 + n]
=3N h
2

U =

(9.31)
(9.32)

where3
n =

1
.

(e h0 1)

(9.33)

3 n is the average phonon (quasi-particle) occupation number. It is also looseley referred to as a Bose-Einstein
function.

9.1. HEAT CAPACITY OF SOLIDS

119

Therefore for the 3 D model


CV = (

U
)
T V
U
)
V
0
0 )2 e h
( h

= kB 2 (
= 3N kB

0
(e h
1)

(9.34)

It is conventional to replace the oscillator frequency 0 with an Einstein temperature E


0,
kB E = h

(9.35)

so that
(E /T ) eE /T
2

CV = 3N kB

(eE /T 1)

(9.36)

with which specific materials can be characterized by fitting to experimental data.


V

Figure 9.3: 3 D Einstein model heat capacity CV vs. T . Note the sharp exponential drop as
3N kB
E
T 0
In the low temperature limit Eq.9.36 becomes
lim CV 3N kB (E /T ) eE /T
2

T 0

(9.37)

as shown in Figure 9.3. This steep exponential decline in CV is never observed. What is universally
observed is CV T 3 .

120

CHAPTER 9. LATTICE HEAT CAPACITY

Einstein was aware that a single oscillator frequency model was bound to be inadequate and he tried
to improve it, without success. But he achieved his primary objective to apply Plancks quantum
theory and show that it explained poorly understood low temperature heat capacities.
At high temperature the Einstein result lim CV = 3N kB , is in accord with Dulong-Petit.
T

9.2

Debye Model

Atom-atom interactions were added to Einsteins theory by Debye.4 Their effect is to introduce
dispersion into the oscillator frequencies, which is precisely the correction Einstein sought but never
achieved.
As a result of atom-atom interactions:
1. Translational (crystal) symmetry is introduced, with a new (wave vector ) quantum number
k, sometimes called crystal momentum, with values
kj =

2
j
Nj aj

j = x, y, z ,

(9.38)

where aj is the lattice spacing in the j th crystal direction, Nj the number of atoms in the j th
periodic cell direction and where j = 1, 2, 3 . . . Nj .5,6
2. As shown in Figure 9.4, rather than a single Einstein lattice frequency 0 there is now a range
of oscillator frequencies7 which Debye assumed varied linearly with k
= (k)
=cs k

(9.39)
(9.40)

where cs is an average speed of sound in the crystal.8


3. The infinitely sharp Einstein phonon density of states shown in Figure 9.4
DE () = N ( 0 )
is replaced in the Debye model, by

DD () =
4 P.

(9.41)

V 2
2 2 cs 3

(9.42)

Debye, Zur Theorie der spezifischen Waerme, Annalen der Physik (Leipzig), 39, 789 (1912).
a 3-dimensional crystal the lattice atom displacements R are in 3-mutually perpendicular directions (polarizations) such that k R = 0.
6 In solid state physics it is conventional to choose N < N which, in this example, would define a single
2
2
Brillouin Zone.
7 One might say that the atom-atom interactions have lifted the degeneracy among single atom oscillator frequencies 0 .
8 There are 3 different sound velocities corresponding to the 2 independent transverse phonons and 1 longitudinal
phonon. This turns out to be an approximation that accurately replicates the small k behavior of lattice vibrations
in many 3 D crystals.
9 See Appendix E.
5 In

9.2. DEBYE MODEL

121

4. Since in a finite crystal the quantum number k is bounded the range of Debyes oscillator
frequencies is also bounded, i.e.
0 < D .

(9.43)

Within the Debye model D is found from

dDE () = dDD () ,
0

(9.44)

i.e. the total number of modes is the same in both models, so that
D = (6 2 cs 3 N /V )

1/3

(9.45)

0
Einstein oscillator frequency 0

Figure 9.4: Mode dependent frequencies. The solid curve represents an approximate result for a
real lattice. kmax is the Brillouin zone boundary for the crystal (kmax < k kmax ). The dashed
line represents Debyes linear approximation. The slope of the dashed line is the average speed of
sound in the crystal. D is the Debye approximations highest attained frequency. The horizontal
fine grey line at = 0 represents the dispersion of an Einstein lattice.

122

CHAPTER 9. LATTICE HEAT CAPACITY

Whereas the Einstein internal energy UE is [see Eq.9.32]


Einstein density of states


{N ( 0 )}

d
UE =3h
0

1
1
]
[ + h

2 e
1

1
1
=3N 0 [ + h
]
2 e 0 1

(9.46)
(9.47)

the changes introduced by Debye give instead the internal energy UD [see Appendix E],
Debye density of states


3V 2
{ 2
}
2 cs 3

d
UD =h
0

D
h

3V
= 33

2 h cs 3 4
0

9.2.0.3

1
1
[ + h
]
2 e 1

1
1
dx x3 ( + x
)
2 e 1

(9.48)

(9.49)

Thermodynamics of the Debye Lattice

D << 1, the integral in Eq.9.49 can be approximated by expanding


At high temperature, h
ex 1 + x. Then using Eq.9.44 the Debye internal energy is
3

lim UD =

V
D
(
) kB T
2
2 cs

=3N kB T

(9.50)
(9.51)

D >> 1, and defining


consistent with the Dulong-Petit rule. At low temperature, h
D =kB D ,
h

(9.52)

with D the Debye temperature, the internal energy integral can be approximated as

lim UD

T 0

3V
1
3 1
)
dxx ( + x
3
3
3
4

2 h cs
2 e 1

(9.53)

from which, with Eq.9.45, follows the low temperature heat capacity
lim CV

T 0

4
6kB
2
3

h c

=N kB (

4 3
T
15

(9.54)

12 4
T 3
)(
) .
5
D

(9.55)

The low temperature heat capacity CV T 3 is almost universally observed in solids. Examples of
Debye temperatures are given in Table 9.1

9.2. DEBYE MODEL

123

aluminum
cadmium
chromium
copper
gold
iron
lead
Nickel
platinum
silicon
titanium
zinc
carbon

D
428K
209K
630K
343.5K
165K
470K
105K
450K
240K
645K
420K
327K
2230K

Table 9.1: Debye temperatures for some typical solids.

Problems and Exercises


1. Two-dimensional materials are now widely fabricated and investigated. Within the framework
of Debyes model for lattice vibrations find the low temperature lattice contribution to the
constant area heat capacity CA for a two dimensional crystalline system.
2. Ferromagnetic systems have very low temperature collective spin excitations which are called
magnons.10 Their quasi-particle behavior is similar to collective lattice phonon excitations
except that their energy spectrum in zero magnetic field is
k = D k2
h

(9.1)

D = 2SJa2

(9.2)

where

with J a magnetic moment coupling constant, S the spin of a magnetic moment and a is a
lattice constant.
Show that the low temperature magnon heat capacity CV is proportional to T 3/2 .
3. In microscopic lattice models that include anharmonic contributions (phonon-phonon interactions), phonon frequencies change with volume V. Using
p=

1 ln Z
(
)

V T

(9.3)

where Z is the Debye partition function, find the pressure of a Debye solid. Therefore in
taking the volume derivative, Eq.9.3, cs and hence D depend on volume.
10 T. Holstein and H. Primakoff, Field Dependence of the Intrinsic Domain Magnetization of a Ferromagnet,
Phys. Rev.58, 1098-1113 (1940).

124

CHAPTER 9. LATTICE HEAT CAPACITY


Express the pressure in terms of the dimensionless Gr
uneisen parameter

V
dD
(
).
D
dV

(9.4)

Mr. McGuire: I want to say one word to you. Just one word.
Benjamin: Yes, sir.
Mr. McGuire: Are you listening?
Benjamin: Yes, I am.
Mr. McGuire: Plastics.
The Graduate (1967)1

Chapter 10

Elastomers Entropy Springs

10.1

Naive 1-D Elastomer

Crystalline matter and polymeric materials exhibit considerable difference in elastic behaviors.
Under applied stress all solids experience strain (elongation) . But crystalline materials also
exhibit a property of stiffness due to accompanying large increases in bond energy. These materials
also completely lose their elastic properties at small values of strain,  0.001. Polymeric materials,
such as a common rubber band, are neither stiff nor, generally, do they lose elasticity even under
large strain.
To see that something quite distinctive must account for elasticity differences between polymers
and crystalline materials one need only compare values of isothermal Youngs moduli
ET = (

) .
 T

(10.1)

Values for crystalline materials are given in Table 10.1 and for polymeric materials in Table 10.2.
Youngs moduli in crystalline matter can be as much as four orders of magnitude larger than for
polymeric materials. What makes them so different is that elastic polymers are softly linked
chains of chemical monomers whose rubbery elasticity is associated with freedom for monomer
units to bend, rotate and randomly stack, with negligible energy cost. Figure 10.1 illustrates the
structure of the polymer polyethylene in which ethylene molecules (monomers) are joined in chains.
Elastomer fexibility is enabled by unhindered unit rotation about a carbon back-bone. Truly
1 Courtesy

of The Internet Movie Database (http://www.imdb.com). Used with permission.

125

126

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

Table 10.1: Isothermal Youngs moduli for some typical crystalline materials.

aluminum
steel
chromium
tungsten
titanium
copper
glass
diamond

7
20.7
29
40
40
12
7.5
110

ET
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2

Table 10.2: Isothermal Youngs moduli for some elastic polymers.

rubber
teflon
polypropylene
nylon
low density polyethylene
polystyrene

0.001
0.05
0.1
0.3
0.02
0.3

ET
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2
1010 N m2

rubbery materials retain their elastic behavior even when subject to strains of  1 3, which is
much larger than for metals.

Figure 10.1: Polyethylene segment. Polyethylene tends to crystallize at room temperature and is
not the best example of an elastomer.
Another polymer with a simple carbon backbone is polyvinyl chloride [see Figure 10.2].
Other polymers can be made which restrict the monomers from swinging so freely. An example
of a less flexible polymer is polystyrene [Figure 10.3] which is typically used in a rigid glassy
state.
Elasticity in polymers is primarily a result of stacking arrangement multiplicity for the constituent

10.1. NAIVE 1-D ELASTOMER

127

Figure 10.2: Polyvinyl chloride polymer has a carbon backbone similar to polyethylene, but with a
chlorine atom in the side-chains.

Figure 10.3: Polystyrene has large aromatic rings in the side chains which severely restrict rotation
about the carbon backbone.
monomers. Less tangled arrangements (elongated chains together with large elastic tension) imply
fewer stacking possibilities, i.e. less uncertainty in describing those arrangements. More tangled arrangements (shorter chains and smaller elastic tension) correspond to greater stacking possibilities,
i.e. more uncertainty. Entropy being a measure of (configurational) uncertainty, an arrangement
with less uncertainty has smaller thermodynamic entropy while one with greater uncertainty is a
higher entropy configuration. Elastomers tend to resist elongation (spring shut) in order to achieve
greater entropy. For this reason elastomers are sometimes called entropy springs.
In this chapter we first hypothesize a microscopic 1 D model of elastomer elasticity and, using
thermal lagrangians derive an elastomer Equation of State. Then using standard thermodynamic
methods as introduced in Chapters 2, 3 and 4, together with the derived Equation of State, we
discuss the elastomer models macroscopic physical behaviors.
10.1.0.4

Elongation as an Observable

Real elastomers generally function in three dimensions. When they are unidirectionally stretched,
their cross-sectional areas shrink proportionately (necking) to maintain approximately constant
volume. The change in cross-sectional area corresponds to the action of internal tensions that tend
to squash the elastomer in directions transverse to its elongation. Metallic rods, although also
elastic, due to much higher Youngs moduli hardly elongate. Therefore in metal rods there is little
change in the rods cross-sectional area and transverse tensions can be safely ignored.
Even though real elastomers should be treated as three dimensional to better account for elastic

128

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

properties, the naive, one-dimensional model discussed here does satisfactorily illustrates the
overwhelming effect of entropy in its elastic properties as well as other unusual empirical features
of a rubber band.
We consider a naive 1 D model of polymer elasticity similar to that introduced in the discussion
of degeneracy in Chapter 6. We assume here, as well, that any conformation of links is only weakly
energy dependent, i.e. any elongation assumed by the polymer has the same (internal) bond energy.
The Lagrangian used in implementing least bias for this elastic model includes in the hamiltonian
a macroscopic, observable elastomer elongation, the average difference between the polymers end
coordinates. The thermodynamic hamiltonian is therefore
Hop = h0 op

(10.2)

where is elastic tension the variable conjugate to the elongation operator, op . h0 is an internal
bond energy hamiltonian which is neglected in the model.

10.1.0.5

Naive Elastomer Model

The naive one-dimensional polymer is assumed to be an assemblage of N links, each of length a,


that can point to the left or to the right with no energy difference for either orientation.

Figure 10.4: One-dimensional model of monomer folding.


In this naive model we begin by assuming h0 = 0 so that the hamiltonian becomes
Hop = op

(10.3)

and the average macroscopic polymer elongation op is


op = T r op op .

(10.4)

The elongation operators macroscopic eigen-values are = a (nR nL ), where nL is the number
of left directed monomers, nR is the number of right directed monomers and nR + nL = N , is the
total number of monomers. Therefore macroscopic eigen-energies of the system (eigen-energies of
the hamiltonian in Eq.10.3) are
E (nR , nL ) =
= a (nR nL ) .

(10.5)
(10.6)

10.1. NAIVE 1-D ELASTOMER

129

Forming the Thermal Lagrangian


N

L = kB

P (nL , nR ) lnP (nL , nR )

nR ,nL
nR +nL =N

nR ,nL
nR +nL =N

we find, with =

(10.7)

P (nL , nR ) T 1

P (nL , nR ) [ a (nR nL )]]

nR ,nL
nR +nL =N

1 ,
kB T
e a(nR nL )
Z

(10.8)

g (nL , nR ) e a(nR nL ) .

(10.9)

P (nL , nR ) =
where Z , the partition function, is
N

Z =

nL ,nR
nL +nR =N

The degeneracy g (nL , nR ) is inserted to remind us to sum over all states. Applying the degeneracy
for this model from Chapter 6,
g (nL , nR ) = 2

N!
nL ! nR !

(10.10)

the partition function is


N

Z =

nL ,nR
nL +nR =N

N!
e a(nR nL )
nL ! nR !
N

= 2 (e a + e a )
= 2 [2 cosh ( a)]
10.1.0.6

(10.11)
(10.12)
(10.13)

Properties of the Naive 1 D Rubber-Band

1. Elongation: The average elongation of rubbery materials, op , shows some empirically


distinctive features. Using Eq.10.8 op is
N

op =

nL ,nR
nL +nR =N

N!
e a(nR nL )
[a (nR nL )]
nL !nR !
Z

(10.14)

or
op = (

1
) [ ln Z ]

(10.15)

130

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS


which gives the elastomer Equation of State.
op = N a tanh ( a) .

(10.16)

Examining this result in Figure 10.5, note the highly non-linear response to large tension,
op
diverging as
1. The elastic limit, where elasticity is permanently lost, is in that
Na
region.
At small tension, a << 1, the elongation is linear, i.e. Hookes Law,
op =

N a2
kB T

(10.17)

or
=

kB T
op ,
N a2

(10.18)

but with the distinctive feature that the elastic constant of a rubber band increases with
increasing temperature.2

Figure 10.5: Average extension, op /N a, of Rubber as a function of a . At small tension,


kB T
a << 1, the behavior is linear.
kB T

2 Finding

the thermal expansivity of the 1 D rubber band is assigned as a problem.

10.1. NAIVE 1-D ELASTOMER

131

2. Entropy: Rubber band entropy can be found from


N

S = kB

P (nL , nR ) ln P (nL , nR )

(10.19)

nL ,nR
nL +nR =N
N

= kB

nL ,nR
nL +nR =N

e a(nR nL )
[ a (nR nL )] ln Z }
Z

= kB { op

1
ln Z }

(10.20)

(10.21)

which is identical to

1
( ln Z )
T

1
=kB 2
( ln Z ) ,

S =

(10.22)
(10.23)

with the result (N >> 1)


S = N kB [ln 2 + {ln [2 cosh ( a)] a tanh ( a)}] .

(10.24)


Recasting the entropy in terms of average elastomer extension X = Nopa [see Eq.10.16]
S
2
= ln [
] X tanh1 X .
2
Na
1X

(10.25)

Extension vs. entropy is shown in Figure 10.6. In this model energy plays no role in elastic
behavior. It is evident then, that rubbers restorative springiness is due to favoring an
entropy increase in the more compact configuration.

10.1.1

Hot Rubber Bands A Thermodynamic Analysis

If a thick real rubber band is quickly stretched and then abruptly relaxed several times in rapid
succession it cannot fail to be noticed that the rubber band warms. Even without sensitive thermometry the temperature rise can be detected by placing the rubber band against the center of
your forehead or against your upper lip, both of which are sensitive to temperature. The thermodynamics of this successive pair of processes can be qualitatively studied using the 1 D polymer
model.3,4
If the elastomer is stretched so quickly that there is insufficient time for it to exchange heat with

the surroundings, then the process can be described as adiabatic (d


Q = 0). In that case,

0 = dU + d
W.

(10.26)

3 The elastic properties of a typical elastomer have, in principle, weak (but not zero) internal energy dependence.
In the interest of clarity this energy dependence is, in the 1 D model, entirely ignored. To continue the present
discussion a dose of reality must be restored. This comes in the form of a non-zero internal energy. But as will be
seen, its microscopic origin and its magnitude are irrelevant.
4 For simplicity, the average extension will now be replaced by L.
op

132

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

Figure 10.6: Average extension per monomer, op /N a, of an entropy spring as a function of S.


Since U is a state variable5 the integral
B

dU = U (B) U (A) ,

(10.27)

only depends on end points and not path. Then W also depends on end points, and a quasi-static
path can be chosen, i.e.

d
WQS = dL .

(10.28)

The remaining problem-solving steps are typical of those in Chapters 3-5 but adapted to the elastic
model.
Starting with the First Law

d
Q = dU dL

and applying the adiabatic (fast stretching) condition we have,

(10.29)
6

0 = dU dL .

(10.31)

U
U
) dT + (
) dL
T L
L T

(10.32)

Taking the differential of U = U (T, L)


dU = (
5 Note

that now hop 0 so that U 0.


it might be better to evaluate

6 Alternatively

T
) .
L S

(10.30)

10.1. NAIVE 1-D ELASTOMER

133

Eq.10.31 becomes
0=(

U
U
) dT + [(
) ] dL .
T L
L T

(10.33)

U
)
T L

(10.34)

However, from Eq.10.29


CL = (
which leaves only the partial derivative
(

U
)
L T

(10.35)

to be evaluated. A way forward7 is by using the Helmholtz potential F = U T S and taking the
constant temperature partial derivative
(

F
S
U
) =(
) +T (
) .
L T
L T
L T

(10.36)

This is handled by noting that combining the (elastic) thermodynamic identity


T dS = dU dL

(10.37)

with the differential of Helmholtz potential


dF = dU T dS S dT

(10.38)

dF = S dT + dL

(10.39)

F
) =.
L T

(10.40)

gives

and the identification


(

To simplify the last term in Eq.10.36, apply the Euler criterion to the exact differential dF , i.e.
Eq.10.39,
(

) =(
)
L T
T L

(10.41)

which is a Maxwell relation. Finally Eq.10.36 becomes


U

) = T (
)
L T
T L

(10.42)

CL
dT = (
) dL .
T
T L

(10.43)

(
and Eq.10.33 reduces to

7A

different but equivalent method was used on a similar partial derivative in Chapter 2.

134

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

Using the approximate (linearized) elastomer Equation of State, Eq.10.17, we find

kB L
) =
T L N a2

(10.44)

CL
kB
dT =
L dL
T
N a2

(10.45)

L2f L20
Tf
= exp (
)
T0
2

(10.46)

kB
1
(
).
2CL N a2

(10.47)

(
arriving at the differential equation

which is easily integrated to give

where8
2 =

Thus, as a consequence of adiabatic stretching, i.e. Lf > L0 , the elastomer warms!

10.1.2

Abruptly Relaxing Rubber Band

When the real rubber band is suddenly allowed to relax under no effective restraint (snaps shut) we

still have a fast process with d


Q = 0. But relaxing without restraint means that no work is done by

the rubber band, i.e. dW = 0. This process is thermodynamically identical to the Free Expansion of
a confined gas as studied in Chapter 3. Therefore, from the First Law, there is no internal energy
change, dU = 0.
This suggests that to find the temperature change in the rapidly relaxed rubber band, we write,
as in Chapter 3, the infinitesimal
dT = (

T
) dL .
L U

(10.48)

We are again temporarily blocked by an unfriendly partial derivative. But with the skill gained
in Chapters 3-5 we start by applying the cyclic chain rule
(

T
U
U 1
) = (
) (
)
L U
L T T L

(10.49)

and using Eqs.10.34 and 10.42 from above to give


dT =

[T (
) ] dL
CL
T L

(10.50)

which, together with the linearized polymer equation of state, shows that the temperature of an
unrestrained, rapidly relaxed elastomer does not change! Therefore a perceptible temperature increase can accumulate by repeated adiabatic stretching and abrupt (unrestrained) relaxation of the
rubber band. Try it!
8 Assuming

an average value of CL .

10.1. NAIVE 1-D ELASTOMER

10.1.3

135

Entropy Change in the Relaxing 1 D Elastomer

A rubber band snaps shut when the elongating constraint (tension) is suddenly released. Of course
it does. What else would you expect from a rubber band? What governs this spontaneous process?
As alluded to above, it is an expected increase in the entropy of the universe (i.e., the rubber
band.)9 To calculate the change in entropy in this constant U process write
dS = (

S
) dL
L U

(10.51)

What seems to be an unusual partial derivative has been considered previously in Chapter 3 where
we first applied the cyclic chain rule to get
(

U
U 1
S
) = (
) (
) .
L U
L S S L

(10.52)

Then from the thermodynamic identity, Eq.10.37, we find both


(

U
) =
L S

(10.53)

U
) =T
S L

(10.54)

and

so that Eq.10.51 reduces to


dS =

dL .
T

(10.55)

Using the linearized Equation of State this further reduces to the differential expression
kB
L dL
N a2

(10.56)

kB
(L2 L2f ) > 0 .
2N a2 0

(10.57)

dS =
which is integrated to give
S =

Spontaneous, unrestrained contraction of a stretched rubber band is driven by increased entropy of


the universe.

10.1.4

A Non-ideal Elastomer

Instead of the 1-D elastomer consider a model elastomer as a three dimensional bundle of roughly
parallel (say along the zdirection) but weakly interacting chains. The general picture is that each
chain is a system of N links (N 1,) each of length a, which can swivel around the bond at any
9A

pretty modest sized universe, to be sure.

136

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

polar or azimuthal angle , with respect to the primary zaxis. However, now an interchain energy,
assumed to be proportional to sin , is introduced. But to simplify the model we assume that these
links can only point in transverse {x, y} directions as well as the longitudinal zdirection.
All links pointing parallel to the primary zaxis are taken to have identical energy, E = 0, while
rotationally hindered links bending in {x, y} directions (subject to zero mean displacements in the
x and y directions) acquire an average energy E = . Such an interaction introduces non-ideality
in a simple but plausible way reminiscent of cross-linking that takes place in vulcanization of
rubber. In the limit  0 transverse links are as easily created as longitudinal links so that links
meander randomly in three dimensions. In this limit the elastomer may be thought of as ideal
in an analogous sense to a three dimensional non-interacting gas being ideal. [More will be said
about this comparison below.] For  0 the transverse link interaction energy causes meandering
to become biased and the elastomer is no longer ideal.
As before, we formulate a Thermal Lagrangian suited to the model described above, obtain surrogate probabilities and a partition function for the system. The partition function yields all
necessary thermodynamic detail from which the microscopic interaction parameter  can, presumably, be found by a suitable experiment.
The macroscopic hamiltonian is
Hop = h0 X op

(10.58)

where h0 accounts for the macroscopic cross-linking bond energy, X op is the length vector operator whose average X op has components in any of three mutually perpendicular directions and
is the elastic tension vector conjugate to the displacement operator.
Within the framework of the 3D elastomer model the macroscopic eigen-energies of Hop are
(nx , ny , nz ) = (n+x + nx + n+y + ny ) 
E
(10.59)
x a (n+x nx ) y a (n+y ny ) z a (n+z nz )
where nq are the discrete number of links in the q directions, with q = x, y, z the mutually perpendicular coordinate directions and a the length per link. The Thermal Lagrangian L for this
model is therefore
L = kB P (n) ln P (n)
n

T 1 [{ P (n) [(n+x + nx + n+y + ny ) ]}


n

(10.60)

{ P (n) [x a (n+x nx ) + y a (n+y ny ) + z a (n+z nz )]}]


n

0 P (n)
n

with T the absolute temperature, and where


P (n) P (nx , ny , nz )

(10.61)

is the probability there are nx , ny , nz links of each type. The sum



n

n+x ,nx
n+y ,ny
n+z ,nz

(10.62)

10.1. NAIVE 1-D ELASTOMER

137

is over the integers {nx , ny , nz } so that, for example, the average value of the longitudinal
component of the elastomer length vector Lz is
Lz = a P (n) (n+z nz )

(10.63)

U = P (n) (n+x + nx + n+y + ny ) 

(10.64)

and the internal energy U is


n

with the ideal elastomer characterized by lim U = 0. Maximizing the lagrangian L with respect
0

to the P (n) we obtain


P (n) =

e(n+y +ny +n+x +nx ) eaz (n+z nz )


ZX

(10.65)

where x = y = 0 (as required by the conditions Lx = Ly = 0) have been imposed. The


normalizing denominator
ZX = e(n+y +ny +n+x +nx ) eaz (n+z nz )

(10.66)

is the partition function. Eq. 10.66 is summed by using the degeneracy g (n) of the macroscopic
eigen-energies
N!
(10.67)
g (nx , ny , nz ) =
n+x !nx !n+y !ny !n+z !nz !
to give the partition function
ZX =
n

N!
e(n+y +ny +n+x +nx ) eaz (n+z nz )
n+x !nx !n+y !ny !n+z !nz !

(10.68)

the sum now being over distinct integer values restricted by the fixed total
n+x + nx + n+y + ny + n+z + nz = N.

(10.69)

Eq. 10.68 is just the multinomial expansion of


N

ZX = [ez a + e+z a + 4e ] .

10.1.5

(10.70)

3D Elastomer Thermodynamics; Prelude to an Experiment

Substituting probabilites of Eq. 10.65 into


N

S = kB P (n) ln P (n)

(10.71)

gives the thermodynamic entropy


S = kB (U z Lz +

1
ln ZX ) .

(10.72)

138

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

Defining GX , an elastic Gibbs potential, as


GX =

1
ln ZX

(10.73)

Eq. 10.72 is rewritten as


GX = U T S z Lz .

(10.74)

Taking the total differential of GX and using the fundamental law expression
T dS = dU z dLz ,

(10.75)

where dLz is an infinitesimal change in elastomer length, we also find


dGX = S dT Lz dz

(10.76)

z
z
z
z

Figure 10.7: Model Elastomer Stress-Strain relationships as described by Eq. 10.78. Curves with
z =  < 0 are typically observed in rubber-like elastomers.
so that from the partition function we find the elastomer length, i.e. the Equation of State,
Lz = [
=

ln ZX ]
(z )
T

N a sinh (az )
.
cosh (az ) + 2e

(10.77)
(10.78)

Plotting a vs. Lz /Lmax in Figure 10.7, various elastic polymer behaviors obtained from the
Equation of State, Eq. 10.78, are shown. The the curve for z =  > 0 is characteristic of an

10.1. NAIVE 1-D ELASTOMER

139

elasticity dominated by coiling and uncoiling of individual chains. On the other hand the curve for
z =  < 0 corresponds to a tendency to form cross linkages,10 in which case rubber initially strongly
resists stretching then exhibits a region of easier deformation (possibly associated with observed
necking) until, as Lz /Lmax 1, stretched rubber suddenly ruptures. The case z =  < 0
parallels observed characteristics of rubber.
Using Eq. 10.76 or, equivalently, Eq. 10.71, the elastomer entropy is
S = (

GX
)
T z
(10.79)

1
= kB [
( ln ZX )]

z
2

which evaluates to

[2 az e sinh (z a)]

S = N kB ln 2 + ln [2e + cosh (z a)] +


.


2 + e cosh (z a)

(10.80)

Finally, from Eqs. 10.73, 10.74, 10.78 and 10.80 the internal energy is
U=

10.1.6

2N
.
2 + e cosh (az )

(10.81)

An Experiment

The typical undergraduate thermal physics course seems to lack realistic hands-on laboratory
exercises, especially ones that might explore the synergy between statistical theory, microscopic
parameters and experiment. Perhaps this a result of historical precedence in which gases with
easy kinetic theory visualizations enjoy perceived pedagogical simplicity. Although having a
disproportionate role in thermodynamic instruction they are not especially convenient subjects
for non-trivial laboratory studies in a classroom setting nor are statistical theories of non-ideal
gases which should be the basis for exploring any experimental synergy easily accessible at the
undergraduate level.
In Appendix J macroscopic thermodynamic (lab) measurements are used to find a microscopic
interaction energy  as modeled in the 3 D elastomer above.

Problems and Exercises


1. Find the linear thermal expansivity
=
for the naive 1 D rubber band.
10 Such

as happens in vulcanization of rubber.

1 L
(
)
L T

(10.1)

140

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

2. Find the mean fluctuations in the average 1 D polymer length


2

() = ( )2

(10.2)

for the naive polymer model.


3. A macromolecule is composed of a chain of N chemically identical molecular monomers for
which two distinct states are accessible. At equilibrium N0 units are in the molecular ground
state with energy E0 and N1 units are in a an excited state with energy E1 . Both states are
non-degenerate. The ground state monomer length is ` = while in the excited state it has
length ` = (1 + ) , 0 < << 1.
a. Write an expression for the polymer partition function Z .
b. Find an equation of state describing the total length L as a function of chain tension
and temperature T , L = L (, T ).
4. A one dimensional polymer that emulates piezoelectric behaviors is modeled as N = (nR + nL )
molecular links of which nR are right pointing (+x) and nL are left pointing (x). Each link
has length and fixed electric dipole moment p . When placed in a uniform electric field E0,x
each right pointing link has an energy
R = p E 0

(10.3)

while each left pointing link has an energy


L = + p E 0 .

(10.4)

The implied piezoelectric behaviors are:


i. A mechanical tension applied along the longitudinal +x direction, results in an electric
polarization P in that direction.
ii. An electric field E0 applied in, say, the +x direction, results in a mechanical deformation
` in that direction.
The thermodynamic hamiltonian for this model is taken to be
= h0 op E0 P op
h

(10.5)

so that the corresponding macroscopic eigen-energies are


= N 0 (nR nL ) E0 p (nR nL )
E

(10.6)

with the average length


op = nR nL

(10.7)

and the average polarization


P op = p nR nL .
It is assumed that h0 is field independent and 0 0.

(10.8)

10.1. NAIVE 1-D ELASTOMER

141

a. Write a thermal Lagrangian corresponding to the model.


b. Write a partition function for this piezoelectric polymer in the form
Z = g (E)eE

(10.9)

and evaluate the expression.

p E0 .
d. Find the Helmholtz potential F in terms of and p E0 .
c. Find the entropy of the model in terms of and

e. Find the average polymer length op in terms of and

P E0 .

A characterising property of a piezoelectric is the coefficient


=(
f. Describe in words.
g. Find in terms of and

p E0 .

P
)
.
,E

(10.10)

142

CHAPTER 10. ELASTOMERS ENTROPY SPRINGS

The nation that controls magnetism controls the universe.


Diet Smith, from Chester Goulds Dick Tracy comic strip (1960).

Chapter 11

Magnetic Thermodynamics
11.1

Magnetism in solids

Lodestones fragments of magnetic1 F eO + F e2 O3 (F e3 O4 ) although known to the ancients


were, according to Pliny the Elder, first formally described in Greek 6th century B.C.E. writings.
By that time they were already the stuff of myth, superstition and amazing curative claims, some
of which survive to this day. The Chinese used lodestones in navigation as early as 200 B.C.E and
are credited with inventing the magnetic compass in the 12th century AD.
Only in the modern era has magnetism become well understood, inspiring countless papers, books2
and more than a dozen Nobel prizes in both fundamental and applied research. Among the forms
of macroscopic magnetism are:
1. Paramagnetism: In an external magnetic field B the spin-state degeneracy of local (atomic)
or itinerant (conduction) electronic states is lifted (Zeeman effect). At low temperature this
results in an induced macroscopic magnetic moment whose vector direction lies parallel to
the external field. This is referred to as paramagnetism.3
For most materials removing the external field restores spin-state degeneracy, returning the
net moment to zero.
2. Diamagnetism: Macroscopic magnetization may also be induced with a magnetization vector anti-parallel to the external field, an effect called diamagnetism. In conductors diamagnetism arises from the highly degenerate quantum eigen-energies and eigenstates (referred
to as Landau levels.)4 formed by interaction between mobile electrons and magnetic fields.
Diamagnetism is also found in insulators, but largely from surface quantum orbitals rather
1 Named, as one story goes, for Magnus, the Greek shepherd who reported a field of stones that drew the nails
from his sandals.
2 See e.g. Stephen Blundell, Magnetism in Condensed Matter Oxford Maser Series in Condensed Matter Physics
(2002); Daniel C Mattis, Theory of Magnetism Made Simple, World Scientific, London (2006); Robert M. White,
Quantum Theory of Magnetism: Magnetic Properties of Materials, 3rd rev. ed., Springer-Verlag, Berlin (2007).
3 Itinerant (conduction) electron paramagnetism is referred to as spin or Pauli paramagnetism.
4 L. Landau Diamagnetism of Metals, Z. Phys. 64, 629 (1930).

143

144

CHAPTER 11. MAGNETIC THERMODYNAMICS


than interior bulk states.5 Both cases are purely quantum phenomena, leaving macroscopic
diamagnetism without an elementary explanatory model.6,7 All solids show some diamagnetic
response, but it is usually dominated by any paramagnetism that may be present.
At high magnetic fields and low temperatures very pure metals exhibit an oscillatory diamagnetism called the de Haas-van Alfen effect whose source is exclusively the Landau levels.8

3. Permanent Magnetism
Permanent magnetism refers to macroscopic magnetization taking place without any external magnetic fields. Several examples of this phenomenon are:
Ferromagnetism: Ferromagnetism is an ordered state of matter in which local paramagnetic moments interact to produce an effective internal magnetic field with collective
alignment of moments throughout distinct regions called domains. Due to these internal
interactions, domains can remain aligned even after the external field is removed.
Ferromagnetic alignment abruptly disappears above a material specic temperature called
the Curie temperature Tc , at which point ordinary local paramagnetism returns.
Antiferromagnetism: At low temperatures, interactions between adjacent identical
paramagnetic atoms, ions or sub-lattices can induce anti-alignment of adjacent paramagnets, resulting in a net zero magnetic moment.
Ferrimagnetism:, At low temperatures interactions between unequivalent paramagnetic
atoms, ions or sub-lattices can produce anti-alignment of moments, resulting in a small
residual magnetization.
In both ferrimagnetism and antiferromagnetism, increasing temperature weakens antialignment with the collective induced moments approaching a maximum. Then, at a
material specic temperature called the N eel temperature TN , anti-alignment disappears
and the materials becomes paramagnetic.
In this chapter general concepts in the thermodynamics of magnetism and magnetic fields are
discussed as well as models of local paramagnetism and ferromagnetism.

11.1.1

Magnetic work

Central to integrating magnetic fields and magnetizable systems into the First Law of Thermodynamics is a formulation of magnetic work. Using Maxwells fields,9 the energy generated within
a volume V in a time t, by an electric field E acting on true charge currents J Joule Heat
5 D.

Ceresoli, et. al. Orbital magnetization in crystalline solids, Phys. Rev. B 74, 24408 (2006).
Bohr, Studier over Metallernes Elektrontheori, Kbenhavns Universitet (1911).
7 Hendrika Johanna van Leeuwen, Problmes de la thorie lectronique du magntisme, Journal de Physique et le
Radium, 2 361 (1921).
8 D. Shoenberg, Magnetic Oscillations in Metals, Cambridge University Press, Cambridge (1984).
9 Maxwell fields in matter and free space are the local averages that appear in his equations of electromagnetism.
6 Niels

11.1. MAGNETISM IN SOLIDS

145

is10
W M = t J E dV .

(11.1)

Therefore the quasi-static and reversible 11 magnetic work done by the system is
M
WQS
= t J E dV .

(11.2)

Using Maxwells Equation (in cgs-Gaussian units)12


H =

4
1 D
J+
c
c t

(11.3)

the work done by the system is


M
WQS

1
D

c
E dV .
= t ( H ) E dV

4
4
t

V
V

(11.4)

Using the vector identity


U V = (V U ) + V U

(11.5)

this becomes

c
M
WQS
= t

(H E) dV + H E dV 1 D E dV
.
4

V
V
V

(11.6)

The first integral on the right can be transformed by Gauss theorem into a surface integral. But
since the fields are static (non-radiative), they fall off faster than 12 so that for a very distant surface
r
the surface integral can be neglected. Then, with the Maxwell Equation (Faradays Law)
1 B
E = ( )
c t

(11.7)

incremental work done by the system is

B
1
D
1

M
WQS
= t H
dV +
E dV

4
t
4
t

1
1

= H B dV +
,
D E dV

4
4

V
V

(11.8)

(11.9)

10 Since heat and work are not state functions they do not have true differentials, so the wiggly s are used instead
to represent incremental work in an interval of time t.
11 In specifying reversibility non-reversible hysteresis effects are excluded.
12 Even though they have fallen out of pedagogical favor in E&M text books, cgs units offer unrivaled clarity in
presenting the subtle issues involved in thermodynamics of magnetic and electric fields.

146

CHAPTER 11. MAGNETIC THERMODYNAMICS

where the integrals are over the volume of the sample and surrounding free space. [Note: The
wiggly deltas in B(x), H (x) and D(x) represent functional changes, i.e. changes in the fields
(states) not the coordinates. The fields themselves are functions of coordinates x and are not just
simple variables.]
Limiting the discussion to magnetic phenomena, the magnetic contribution to quasi-static work
done by the system is therefore

M
WQS
= H B dV ,

(11.10)

so that a magnetic fundamental thermodynamic equation for U becomes


U = T S p dV +

1
H B dV .
4

(11.11)

From the Helmholtz potential, F = U T S,


F = U T S S dT ,

(11.12)

which when combined with Eq.11.11 gives the change F


1
H B dV .
4

F = S dT pdV +

(11.13)

Defining magnetic enthalpy H as


H = U + pV

1
B H dV ,
4

(11.14)

gives, with Eq.11.11, an enthalpy change H


H = T S + V dp

1
B H dV .
4

(11.15)

Finally, a magnetic Gibbs potential G is defined as


G = F + pV

1
B H dV ,
4

(11.16)

which with Eq.11.13 gives the Gibbs potential change G


G = S dT + V dp

1
B H dV .
4
V

(11.17)

11.1. MAGNETISM IN SOLIDS

147

Magnetization density13 M and polarization density14 P are inserted by the linear constitutive
relations
H = B 4M

(11.18)

D = E + 4 P

(11.19)

and

allowing quasi-static magnetic work to be written

1
M1
WQS
= B B dV M B dV

V
V

(11.20)

1
M2
= H H dV + H M dV ,
WQS

V
V

(11.21)

or

and quasi-static electric work to be written

P1
WQS
= D D dV + P D dV

(11.22)

P2
= E E dV E P dV .
WQS

(11.23)

or

The first terms in all alternatives are total field energies integrals over all space, both inside and
outside matter. The second terms are integrals over V which include only the volume of magnetized
(polarized) matter. Since magnetic (electric) thermodynamics is primarily concerned with matter
that is magnetized (polarized), one practice is to bravely ignore the total field energies. Another is
to absorb the field energies into the internal energy U. Since neither option is entirely satisfactory
a third way is discussed in Subsection 11.1.2 below.
Nevertheless, these results in terms of local average fields are general and thermodynamically
correct.15 But they are not convenient to apply. Nor are they the fields that appear in microscopic
magnetic (electric) quantum hamiltonians. In quantum magnetic (electric) models the hamiltonians for individual magnetic (electric) moments depend on only the the field before the sample is
introduced. After the sample is introduced internal fields can additionally result from:
13 Total

magnetic moment per unit volume.


electric dipole moment per unit volume
15 In both cases the second of these, W M2 and W P 2 has the correct form for work intensive extensive.
QS
QS
14 Total

148

CHAPTER 11. MAGNETIC THERMODYNAMICS


a. Interactions between internal moments as additional terms in the hamiltonian. These interactions may be approximately treated as effective fields acting in addition to the external
field. [See, for example, Subsection 11.1.5, below],
b. Internal demagnetizing fields arising from fictitious surface poles induced by B0 ,16
c. Internal currents induced by the applied field (especially in conductors).17

11.1.2

Microscopic models and uniform fields

Microscopic models of magnetic and electric hamiltonians are expressed in terms of uniform applied fields (B0 , E0 ) present before matter is introduced. This emphasis on applied fields (rather
than average Maxwell fields within matter) results in thermodynamic relations somewhat different
from Eqs.11.11 - 11.17 above. Focusing on magnetic effects, in the absence of internal magnetic
interactions quasi-static magnetic work done by the system is [see Eq.G.23 in Appendix G],

M
d
WQS
= B0 dM .

(11.24)

Since B0 is uniform, a total macroscopic magnetization vector M has been defined as:
M = M dV ,

(11.25)

with M the average magnetization per unit volume.18


Therefore thermodynamic differential relations become [see Eqs.G.24, G.26, G.25 and G.27]
T dS = dU + p dV B0 dM
T dS = dH V dp + M dB0
dF = SdT p dV + B0 dM

(11.27)

dG = S dT + V dp M dB0

16 Demagnetizing

fields introduce sample shape dependence into the magnetic properties.


Holstein, RE Norton and P Pincus, de Haas-van Alphen Effect and the Specific Heat of an Electron Gas,
Physical Review B 8 2649 (1973).
18 Similarly
17 T

P
dWQS
= E0 dP .

(11.26)

11.1. MAGNETISM IN SOLIDS

149

where the starred potentials are


1
2
B0 dV ,
8

(11.28)

1
2
B0 dV ,
8

(11.29)

1
2
B0 dV ,
8

(11.30)

1
2
G =G+
B0 dV .
8

(11.31)

U = U +

H = H +

F = F +

11.1.3

Local paramagnetism

The classical energy of a magnetic moment m in an average local (Maxwell) magnetic field B
is
E = m B .

(11.32)

For fundamental magnetic moments (electrons, protons, neutrons, etc.) quantum mechanics postulates an operator replacement m mop , and a quantum paramagnetic hamiltonian
HM = mop B0 ,

(11.33)

where B0 is the uniform field present before matter is introduced.19


The paramagnetic hamiltonian for a solid consisting of N identical moments fixed at crystalline
sites i is
N

Hop = h0 B0 mop (i)

(11.34)

i=1

where mop (i ) is a magnetic moment operator20 and h0 is a non-magnetic part of the hamiltoG includes a discussion of the implications and limitations of using B0 in the thermodynamics.
is proportional to an angular momentum (spin) operator S op , with

19 Appendix
20 m

op

mop =

gB
S op .
h

(11.35)

eh
Here g is the particle g-factor and B = 2mc
is the Bohr magneton (cgs-Gaussian units.)
For a spin = atom the quantum mechanical z-component spin operator Sz , is taken with two eigenstates and two
eigenvalues
1
1
h
(11.36)
Sz =
2
2 2

150

CHAPTER 11. MAGNETIC THERMODYNAMICS

nian.21,22 The magnetization operator (total magnetic moment per unit volume) is
Mop =

1 N
mop (i ) .
V i=1

(11.37)

One objective is to find the macroscopic Equation of State M = M (T, B0 ) where M is the
average magnetization per unit volume
M = T r op Mop .

11.1.4

(11.38)

Simple paramagnetism

Consider a spin = atom which in the absence of a magnetic field has the pair of degenerate states
with energy E0 23,24,25
mz,op
=

gB

(11.39)

In a uniform magnetic field, B0,z , the degeneracy of each atom state is lifted, creating a pair of
non-degenerate states of energy
E = E0 B0,z

and E+ = E0 + B0,z

(11.40)

g
with = 2B [See Fig. 11.1].
The macroscopic N -moment eigen-energies are
E (n+ , n ) = N E0 + (n+ n ) B0,z

(11.41)

where n+ is the number of atoms in a state with eigen-energy


E+ = E0 + B0,z

(11.42)

and n is the number of atoms in a state with eigen-energy


E = E0 B0,z ,

(11.43)

with n+ + n = N .
is assumed that there are no interactions corresponding to internal fields, Bint .
2
1
(pop qc Aop ) , where Aop is the vector potential operator, are ignored.
current terms HA J = 2m
23 h of Eq.11.34, and hence E , is assumed to make no magnetic contribution, either from interacting moments,
0
0
internal currents or other internal fields.
24 An eigenvalue equation for atomic spin J, m
z,op is
21 It

22 Field-particle

mz,op
mJ = gB mJ
mJ
where
mJ are the eigenstates and gB mJ the eigenvalues, with J mJ J .
25 To preserve simplicity, discussion is confined to spin magnetic moments.

11.1. MAGNETISM IN SOLIDS

151

Figure 11.1: Lifting the s = 1/2 spin degeneracy with a magnetic field B0,z
The probabilities P (s ) required for the thermodynamic density operator
op = P (s )Es Es

(11.44)

are found by constructing the Thermal Lagrangian


L = kB

P (n+ , n ) lnP (n+ , n ) 0

n+ ,n =0

n+ ,n =0

P (n+ , n )
(11.45)

n+ ,n =0

P (n+ , n ) [N E0 + (n+ n ) B0,z ]

using the N -atom macroscopic eigen-energies of Eq.11.41. The resulting probabilities are
P (n+ , n ) =

[N E0 + (n+ n ) B0,z ]
e
ZM

(11.46)

where = 1/kB T , and the denominator (the paramagnetic partition function) is


N

ZM =

n+ ,n
n+ +n =N

g (n+ , n )e

[N E0 + (n+ n ) B0,z ]

(11.47)

The sum over all states is accounted for by the configurational degeneracy
g (n+ , n ) =

N!
n+ ! n !

(11.48)

so that
N

ZM =

n+ ,n
n+ +n =N

N!
[N E0 + (n+ n ) B0,z ]
e
n+ ! n !

(11.49)

152

CHAPTER 11. MAGNETIC THERMODYNAMICS

which is the binomial expansion of


N

ZM =eN E0 (e B0,z + e B0,z )


= [2eE0 cosh ( B0,z )]

11.1.5

(11.50)

(11.51)

Local paramagnet thermodynamics

Using Eq.11.46, simple paramagnet thermal properties for the spin = system are found:
1. The average total magnetization is
N

,n

M =

n +
n+ +n =N

N!
[ (n+ n )] e [N E0 + (n+ n ) B0,z ]
n+ ! n !
(11.52)
ZM

or
M =

1
( ln ZM )
B0,z

= N tanh ( B0,z ) .

(11.53)
(11.54)

Eq.11.54 is called the Langevin paramagnetic equation. Note in Figure 11.2 that the magnetization saturates as 1/2 B0 where
tanh ( B0,z ) 1

(11.55)

M N .

(11.56)

with a saturation value

The linear region where B0 << 1 is called the Curie regime. In that case
tanh ( B0,z ) B0,z

(11.57)

M N 2 B0,z

(11.58)

and

2. The average energy (including the magnetization-independent energy) is


N

Hop =

n+ ,n

N!
[N E0 + (n+ n ) B0,z ] e [N E0 + (n+ n ) B0,z ]
n+ ! n !
ZM

(11.59)

so that internal energy is


U =

ln ZM + B0,z M

(11.60)

which is simply
U = N E0 .

(11.61)

11.1. MAGNETISM IN SOLIDS

153

Figure 11.2: Magnetization vs. B0


3. Comparing Eq.11.53 with Eq.11.27 we see that the uniform field Gibbs potential is found from
the uniform field partition function, Eq.G.23,
GM =

1
ln ZM .

(11.62)

as discussed in Appendix G.
4. From Eq.11.27 the entropy is
GM
)
T p,B0,z
GM
= kB 2 (
)
p,B0,z

(11.63)

S = (

(11.64)

= N kB {ln 2 + ln [cosh ( B0,z )] B0,z tanh ( B0,z )}

(11.65)

The entropy is represented in Figure 11.4. Note that as B0,z 0 the entropy attains its
maximum value Smax = kB ln 2, reflecting the original zero field 2 f old degeneracy of the
atom states.
5. The relevant heat capacities for paramagnets are those for which B or M are maintained
constant. As can be derived from Eq,11.27 the heat capacity at constant B is
CB = (

H
)
T B

(11.66)

or in terms of entropy S

CB = T (

S
) .
T B

(11.67)

154

CHAPTER 11. MAGNETIC THERMODYNAMICS


Using Eqs.11.65 and 11.67
2

CB = N kB ( B0,z ) sech2 ( B0,z )

(11.68)

a result which is shown in Figure 11.3. For B0 << 1 this assumes quadratic behavior
2

CB N kB ( B0 ) .

(11.69)

Figure 11.3: Constant field heat capacity CB /N kB vs. B0 .


The heat capacity at constant M , as derived from Eq.11.27, is
CM = (

U
)
T M

(11.70)

or

CM = T (

S
) .
T M

(11.71)

This time Eqs.11.70 and 11.61 are used and obviously give26
CM = 0
26 Alternatively,

(11.72)

B
)M ( M
)
the general relation CM CB = T ( T
T B
2

) ] [( M
) ] with its more straightforward partial derivatives,
which can be simplified to CM CB = T [( M
T B
B T
confirms the zero result. Derivation of these results is assigned as a problem.

11.1. MAGNETISM IN SOLIDS

155

Figure 11.4: Entropy vs. B0 .

11.1.6

Magnetization fluctuations

State variables may exhibit variation about equilibrium average values. These variations are called
fluctuations and are assigned the symbol .27,28 For example, magnetization fluctuations
(M ) are
(M ) M M

(11.73)

and mean square magnetic fluctuations (uncertainty) are


2

(11.74)

(11.75)

[ (M )] = (M M )
2

= M M
where
N

M 2 =
27 Generally

n ,n+

N!
2
[ (n+ n )] e [N E0 + (n+ n ) B0,z ]
n+ ! n !
.
ZM

(11.76)

speaking, vanishingly small fluctuations assure meaningful thermodynamic descriptions.


are associated with thermal state functions that have quantum operator representations. For
example, temperature does not have well defined fluctuations since there is no quantum temperature operator. See,
e.g. C Kittel, Temperature Fluctuation: An Oxymoron, Physics Today, 41(5), 93 (1988).
28 Fluctuations

156

CHAPTER 11. MAGNETIC THERMODYNAMICS

Taking this result together with ZM and M (as calculated in Eq.11.52), the mean square fluctuations are
2

M 2 M =

1 2
ln ZM
2 B02

(11.77)

= N 2 sech2 ( B0 )

(11.78)

Expressed as dimensionless root mean square fluctuations

M 2 M

1
=
N sinh ( B0.z )

(11.79)

which decrease rapidly with increasing field B0 , decreasing temperature T and with increasing
N.

11.1.6.1

Example: Adiabatic (Isentropic) Demagnetization

A paramagnetic needle immersed in liquid He4 initially at temperature T0 > T , is placed in a weak
external field B0 directed along the needles long axis.29 The magnetic field is suddenly lowered to
a value B` .
What is the change in temperature of the paramagnetic needle? This sudden process corresponds
to an adiabatic (isentropic) demagnetization too fast for immediate heat exchange.
The solution to the problem is found in Figure 11.4 where at any constant value of S the product
B0 is a constant. Therefore for the adiabatic process
B0
= constant ,
T

(11.80)

so that as B0 falls, T falls along with it. This is also called the magnetocaloric effect.
Formally, we begin with
dT = (

T
T
) dB0 + (
) dS
B0 S
S B0

(11.81)

which for this isentropic process pares down to


dT = (

T
) dB0 .
B0 S

(11.82)

29 In this configuration the demagnetization factor = 0 which simplifies the situation. [See Section A-5 of the
Appendix G.]

11.1. MAGNETISM IN SOLIDS

157

Applying the cyclic chain rule [see Chapter 4]


S
(
)
T
B0 T
(
) =
S
B0 S
(
)
T B0

T S
(
) .
CB B0 T

(11.83)

(11.84)

Using the Gibbs potential expression as given in Eq.11.27, and taking cross derivatives we have a
Maxwell relation
(

S
M
) =(
)
B0 T
T B0

(11.85)

so that Eq.11.82 is now


dT =

T M
(
) dB0 .
CB T B0

(11.86)

Inserting CB from Eq.11.69 and M from Eq.11.58


dT
dB0
=
T
B0

(11.87)

B`
) T0 ,
B0

(11.88)

which is integrated to finally give


Tf = (
i.e. the needle cools.

11.1.7

A Model for ferromagnetism

In the previous section, paramagnetism is modeled as N independent local magnetic moments. But
in general these moments produce, approximately, an internal magnetic field acting in addition to
the external field to produce an average total effective field B .
Short range, nearest-neighbor magnetic moment coupling is frequently described by the Heisenberg
exchange interaction,30
1
Hex = mop (i ) Ki,i mop (i )
2

(11.89)

30 W. Heisenberg, Mehrk
orperproblem und Resonanz in der Quantenmechanik, Zeitschrift f
ur Physik 38, 441,
(1926).

158

CHAPTER 11. MAGNETIC THERMODYNAMICS

where mop (i ) is the magnetic moment operator for the i th site and Ki,i is an interaction which couples the moment at i with the moment at a nearest neighbor site, i . The macroscopic Heisenberg
hamiltonian31 is taken to be
N

H = B0 mop (i)
i=1

1 N z

mop (i) Ki,i mop (i ) ,


2 i=1 i =1

(11.90)

where z is the total number of nearest neighbor moments. The double sum32 includes only terms
with i i .
Apart from 1 D or 2 D, this many-body problem has, generally, no analytic solution. But an
approximation a Mean Field approximation33 can be applied to replace the many-body model by
an effective one-body model and plausibly account for the phenomenon of ferromagnetism.

11.1.8

A mean field approximation MFA

In preparation for applying the MFA, rewrite Eq.11.90 as


N

H = {B0 +
i=1

1 z

Ki, i mop (i )} mop (i)


2 i =1

(11.91)

where, assuming an isotropic system, all z nearest neighbors can be treated as identical, i.e.
Ki,i K.
The essence of a mean field approximation is the identity:
mop (i ) mop (i) = [mop (i ) mop ] [mop (i) mop ]
+mop (i ) mop + mop mop (i) mop mop

(11.92)

where mop is the average magnetic moment, i.e. M /N , the magnetization per site. The MFA
neglects the first term, i.e. the product of fluctuations around the magnetization, whereas the
last term contributes a constant value. The hamiltonian can then be written in its mean field
form34
N
1
H = zN Kmop 2 {B0 + z mop K} mop (i) ,
2
i=1

(11.93)

which includes the constant term from the MFA, [see Eq.11.92]. The external field B0 is now
supplemented by an internal field, Bint
Bint =z mop K
z
= MK
N

(11.94)
(11.95)

31 Curiously, the Heisenberg magnetic interaction does not originate from magnetic arguments. Its source is
strictly interatomic electronic interactions, in particular from the electron exchange interaction in the hydrogen
molecule.
32 The factor 1/2 compenstates for the ultimate double counting by the double sum.
33 P Weiss, Lhypoth`
ese du champ moleculaire et la propri`
ete ferrmognetique. J. Phys. (Paris) 6, 661 (1907).
34 The doubled mean field sum over i and i (terms 2 and 3 in Eq.11.92), cancels the factor 1 .
2

11.1. MAGNETISM IN SOLIDS

159

so a total effective field B at the ith site is


B (i) = B0 (i) + Bint (i)
z
= B0 (i) + K M (i)
N

(11.96)
(11.97)

giving rise to a Mean Field hamiltonian


N
1
H = M Bint B (i) mop (i) .
2
i=1

(11.98)

This has the effect of replacing B0 in the paramagnet partition function of Eq.11.51 by B , in which
case
N

ZM = e 2 M Bint [2 cosh ( B )]
1

(11.99)

with a Gibbs potential35


= 1 ln ZM
G

N
1
= M Bint ln (2 cosh B ) .
2

(11.100)
(11.101)

In the absence of an external field, i.e. B0 = 0, we find from Eqs.11.53,11.97 and 11.99
M = N tanh (

zK
M) .
N

(11.102)

which is a transcendental equation in M that describes the possibility of finite magnetization even
in the absence of an external field.

11.1.9

Spontaneous magnetization

Rewriting the S =
Parameter M ,

1
2

result, Eq.11.102, as a self consistent expression in a dimensionless Order

M =

M
,
N

(11.103)

Eq.11.102 becomes
M = tanh (

Tc
M)
T

(11.104)

where
Tc =

2 z K
.
kB

(11.105)

35 The constant term from the MFA has interesting thermodynamic consequences to be discussed in appendix
G.[See Eq.G.30.]

160

CHAPTER 11. MAGNETIC THERMODYNAMICS

Figure 11.5: A graphical solution of the self-consistent equation, Eq.11.104. The sharp decline of
the Order Parameter M as T Tc (the Curie temperature) is followed by a slope discontinuity
at T = Tc . This is the general characteristic of a magnetic phase transition. When T > Tc the only
solution to Eq.11.104 is M = 0.
Tc is called the Curie temperature. Solving Eq.11.104 graphically we see that for T < Tc , M > 0
and magnetic moments spontaneously align (ferromagnetism). For T > Tc , M = 0 and spontaneous
alignment is destroyed, characterizing Tc as the transition temperature at which a phase transition
from an ordered (M > 0) to a disordered (M = 0) state takes place.36 [See Figure 11.5.]

11.1.10

Critical exponents

As T approaches Tc with T < Tc , the magnetic order parameter shows the power-law behavior
M (

c
Tc
1) ,
T

(11.106)

where c is called a critical exponent. The value of c from the MFA is found by first inverting
Eq.11.104
Tc
M = tanh1 M
T
36 This

is referred to as Symmetry Breaking.

(11.107)

11.1. MAGNETISM IN SOLIDS

161
Curie Temperature K
Fe
1043
Co
1388
Ni
627
Gd
293
Dy
85
CrBr3
37
EuO
77
M nAs
318
M nBi
670
F e2 B
1015
GdCl3
2.2

Table 11.1: Sample values of Curie Temperatures in Ferromagnetic materials.


and then expanding tanh1 M for small M
Tc
1
M = M + M3 + ...
T
3

(11.108)

1/2
Tc
1) .
T

(11.109)

to give
M
The S =

1
2

3(

mean field critical exponent is therefore


c = 1/2 .

11.1.11

(11.110)

Curie-Weiss magnetic susceptibility (T > Tc )

At T > Tc and with no external field, i.e. B0 = 0, nearest neighbor interactions are no longer
sufficient to produce spontaneous magnetization. However, upon re-introduction of an external
field B0 induced paramagnetic moments will still contribute internal fields, so that within a MFA a
total internal field is again B , as in Eq.11.97.
But for T >> Tc Eq.11.102 can be expanded37 and solved for M to give
M =

2
kB

(T Tc )

B0 ,

(11.111)

where Tc is as defined in Eq.11.105. With a magnetic susceptibility M defined as38


M = M B0
37 tanh (x)

(11.112)

x 31 x3 .
magnetization, magnetic susceptibility has no strict thermodynamic definition. In the case of non-linear
M
materials an isothermal susceptibility M = ( B
) is a more practical definition.
38 Unlike

162

CHAPTER 11. MAGNETIC THERMODYNAMICS

we have
M =

2
kB

(T Tc )

(11.113)

which is called the Curie-Weiss Law. When T >> Tc ,39 this is a satisfactory description for magnetic
susceptibility. But it fails near T Tc , where the formula displays a singularity.
The Curie-Weiss law is often expressed in terms of
Tc .

11.1.12

1
which is (advantageously) linear in T
M

Closing comment

The study of magnetic matter remains a vast and varied topic that drives contemporary research,
both fundamental and applied. The examples discussed in this chapter (paramagnetism and ferromagnetism) are but introductory samples of the role played by quantum mechanics in understanding
macroscopic magnetism.

39 T experimentally determined from Curie-Weiss behavior is usually higher than T from the ferromagnetic phase
c
c
transition.

If to any homogeneous mass in a state of hydrostatic stress we suppose


an infinitesimal quantity of any substance to be added, the mass
remaining homogeneous and its entropy and volume remaining
unchanged, the increase of the energy of the mass divided by the
quantity of the substance added is the (chemical) potential for that
substance in the mass considered.
J. W. Gibbsa
a Willard Gibbs, On the Equilibrium of Heterogeneous Substances, Trans.
Conn. Acad., vol. III, 108-248 (1876).

Chapter 12

Open Systems
12.1

Variable particle number

J. W. Gibbs ingeniously extended thermodynamics to include processes involving systems with


variable particle number1,2 processes that include:
1. Phase transitions Macroscopic matter transforms abruptly into different physically distinctive phases, usually accompanied by changes in symmetry and discontinuities in properties.
Among examples of phase changes are:
(a) melting: Asolid Aliquid
(b) magnetization:
(c) evaporation:

paramagnet ferromagnet

Aliquid Avapour

(d) sublimation: Asolid Avapour


(e) NormalSuperconductor
1 Willard Gibbs, On the Equilibrium of Heterogeneous Substances, Trans. Conn. Acad., vol. III, 343-524
(1878).
2 Willard Gibbs, On the Equilibrium of Heterogeneous Substances, Amer. Jour. of Sci. (3), vol. XVI, 441-458
(1878).

163

164

CHAPTER 12. OPEN SYSTEMS


(f) Normal He4 Bose-Einstein Condensate

2. Atoms, molecules or particles combine and recombine as different chemical units:


(a) Chemical reactions:
nA A + nB B . . . nx X + nY Y . . .

(12.1)

where nA , nB , . . . nX , nY , . . . are integer numbers of the chemical participants, A, B, . . . X, Y . . ..


(b) Fundamental particle interactions electron-positron annihalation:
e+ + e +

12.1.1

(12.2)

Thermodynamics and particle number

The hamiltonians H and state functions in Schrodingers quantum mechanics define microscopic
dynamics for systems with a fixed number of particles (closed systems). These, till now, have been
our quantum reference in understanding macroscopic systems. However, for many systems or
processes this is inadequate or mathematically unwieldy. By introducing into the hamiltonian a
particle number operator Nop and its eigenvalue equation
= N N

Nop N

(12.3)

, together with a particle number fluctuation


with eigenvalues N = 0, 1, 2, and eigenfunctions N
operator, defined as
op (N ) = Nop Nop ,

(12.4)

quantum mechanics (and thermodynamics) can logically and formally be extended to include variable particle number the defining property of open systems. The number operator is not part of
Schr
odinger mechanics but belongs to Quantum Field Theory, a quantum formulation far beyond
any plan for this book. Yet, a number operator and its eigenvalue property [see Eq.12.3] are the
basis of Gibbs prescient Grand Canonical thermodynamics, a topic discussed and applied in this
and several remaining chapters.

12.1.2

The open system: formalities

op , which constitutes the basis for open system thermodynamics, is


The hamiltonian H
op = hop Nop
H

(12.5)

which now includes particle work dWN = dNop where is the chemical potential for a single
component species. In the case there are M chemically distinct components
M

i
op = hop i Nop
H

(12.6)

12.1. VARIABLE PARTICLE NUMBER

165

where i is a chemical potential for each of the chemical species.


While not significantly modifying the canonical formalism in particular the commutation relation
op ] = 0
[
op , H

(12.7)

is still satisfied introducing Nop does have formal and thermodynamic consequences.
The single species open system Thermal Lagrangian which incorporates Eq.12.5 now becomes,

L = kB

[s (N ) , N ] ln P
[s (N ) , N ]}
{P

N =0,1,2... s

1
T

N =0,1,2...

[s (N ) , N ] [s (N ) N ]} 0
{ P
s

N =0,1,2...

[s (N ) , N ]} . (12.8)
{ P
s

[s (N ) , N ] is the normalized probability for


where P
1. N particles,
2. N particle eigen-energy s (N ) .
distinguishes it from closed system (canonical) probabilities.
The hat worn by P
Maximizing

[s (N ) , N ] gives
L with respect to variations in P
e[s (N )N ]

[s (N ) , N ] =
P

N =0,1,2,...

(12.9)

{ e[s (N )N ] }
s

where the normalizing denominator


Zgr =

{e[s (N )N ] }

N =0,1,2,...

(12.10)

is the Grand Partition Function. As in the canonical case, the s-sum covers all N -particle eigenstates s (N ) whereas the newly introduced N -sum covers all particle numbers N = 0, 1, 2, . . ., (i.e.
eigenvalues of Nop ).
Re-parsing Eq.12.10 we can equivalently write
Zgr =

eN {e s (N ) }

N =0,1,2,...

(12.11)

which emphasizes the underlying form


Zgr =

eN Z (N ) ,

(12.12)

N =0,1,2,...

where
Z (N ) = e s (N )
s

is an N particle canonical partition function.

(12.13)

166

CHAPTER 12. OPEN SYSTEMS

12.1.3

Open system thermodynamics

The quantum formulation having been adapted for variable particle number, resulting in probabil [s (N ) , N ], its consequences for thermodynamics can now be examined.
ities P

12.1.4

Grand partition function

1. Average system particle number Nop ,

Nop =

N =0,1,2,...

N e[s (N )N ]
s

e[s (N )N ]

(12.14)

N =0,1,2,... s

which can be written in terms of Eq.12.10, the Grand Partition Function, as3

Nop =

1
(
ln Zgr )
.

T,V

(12.15)

2. System internal energy U,


s (N )e[s (N )N ]

U=

N =0,1,2,... s

e[s (N )N ]

(12.16)

N =0,1,2,... s

which a little algebra transforms into

U = (

ln Zgr )
+ Nop .

,V

(12.17)

3. Assuming a volume dependence  (N, V ), pressure p is


[

p=

N =0,1,2,... s

s (N ) [s (N )N ]
]e
V
e[s (N )N ]

(12.18)

N =0,1,2,... s

which in terms of the Grand Partition Function is


p=
3 N
op

1
(
ln Zgr )
.
V
T,

is a continuous variable, in contrast to integer N of the canonical picture.

(12.19)

12.1. VARIABLE PARTICLE NUMBER

167

4. Entropy S,
S = kB

[s (N ) , N ] ln P
[s (N ) , N ]} ,
{P

(12.20)

N =0,1,2,... s

becomes, with Eqs.12.9 and 12.10,

S = kB

{[

N =0,1,2,... s

= kB (U

e(s +N )
] [ ln Zgr (s + N )]}
Zgr

1
ln Zgr + Nop )

(12.21)
(12.22)

and then with Eq.12.17

S = kB 2 [

12.1.5

1
( ln Zgr ) ] .

,V

(12.23)

Grand potential

In addition to the Grand Partition Function an open system Grand Potential , gr , is defined

gr =

1
ln Zgr .

(12.24)

Using gr Eq.12.22 can be rearranged as


gr = U T S Nop .

(12.25)

With the Helmholtz potential F = U T S we also have


gr = F Nop

(12.26)

and in terms of the Gibbs potential G = U T S + pV


gr = G Nop p V .

(12.27)

Summarizing the Grand Potentials role in the thermodynamics of Open Systems, Eqs.12.15, 12.17,

168

CHAPTER 12. OPEN SYSTEMS

12.23 and 12.19 can be re-expressed as:


gr
)
T,V

( gr )]
+ Nop
U =[

,V
gr
)
p = (
V T,
gr
S = kB 2 (
)
,V

Nop = (

(12.28)

Finally, differentiating gr = gr (T, V, )


d gr = (

gr
gr
gr
)
dT + (
)
dV + (
)
d
T V,
V T,
T,V

(12.29)

implies, from Eq.12.28, the thermodynamic relation


d gr = S dT p dV Nop d .

12.1.6

(12.30)

G, gr and Eulers theorem

Combining Eqs.12.27 and 12.30 the differential of the Gibbs potential is


dG = S dT + V dp + dNop

(12.31)

so that its natural variables are T, p and Nop . With Nop the only extensive variable among them
we write G (T, Nop , p) in the Euler form,
G (T, Nop , p) = G (T, Nop , p) ,

(12.32)

denoting the fact that G is homogeneous in Nop of degree 1. Applying Eulers Theorem [see
Chapter 4] gives
Nop (

G
)
= G (T, Nop , p) .
Nop p,T

(12.33)

But from Eq.12.31


(

G
)
=
Nop p,T

(12.34)

which leads to
G (T, Nop , p) = Nop .

(12.35)

12.1. VARIABLE PARTICLE NUMBER

169

For a multispecies system, say A1 , A2 , A3 , . . ., this is generalized as


G = Ai Nop Ai .

(12.36)

i=1

According to Eq.12.30 the natural variables of gr are T, V and , of which only V is extensive.
Writing gr = gr (T, V, ) [see Eq.12.30] in its Euler form
gr (T, V, ) = gr (T, V, )

(12.37)

(denoting that gr is homogeneous in V of degree 1) so that applying Eulers Homogeneous Function Theorem
V(

gr
)
= gr .
V T,

(12.38)

Upon substituting from Eq.12.28 this becomes


gr = pV .

(12.39)

This as we will see in the next section is a useful result for deriving equations of state.

12.1.7

The ideal gas: A Grand example

The Grand Partition Function is initially applied to the ideal gas. Then examples are given
that build on the ideal gas to model diffusion and inhomogeneity in more complex physical systems.
Using results from Chapter 7 together with Eq.12.12,
Zgr =

eN Z (N )

(12.40)

N =0,1,2,...

eN

N =0,1,2,...

N =0,1,2,...

1
N
(nQ V )
N!

1
N
(nQ V )
N!

(12.41)
(12.42)

where
= e

(12.43)

is called the fugacity and where, as defined in Chapter 7,


3/2

nQ = (

m
)
2
2 h

(12.44)

170

CHAPTER 12. OPEN SYSTEMS

Summing Eq.12.42 gives the ideal gas Grand Partition Function 4


Zgr = enQ V

(12.45)

and the ideal gas Grand Potential


gr =

12.1.8

nQ V
.

(12.46)

Ideal gas thermodynamic properties

1. From Eqs.12.28 and 12.46 the average particle number is


Nop = nQ V ,

(12.47)

which is solved to obtain the ideal gas chemical potential5

1 Nop
ln
.

nQ V

(12.48)

2. Similarly, from Eq.12.28 the internal energy U is

U=

3Nop
.
2

(12.49)

3. Again using Eq.12.28 the entropy is


Nop
5
S = Nop [ kB ln
]
2
nQ V

(12.50)

which is the Sakur-Tetrode equation of Chapter 7.


4. Using Eqs.12.35 and 12.48 the Gibbs potential for the ideal gas is
Nop Nop
ln

nQ V

(12.51)

G (T, Nop , p) = Nop .

(12.52)

G (T, Nop , p) =
which is identical with

4 The ideal gas, even in this Grand Canonical picture, is is still semi-classical requiring the ad-hoc corrections of
Chapter 7.
5 Since n

Q (the quantum concentration) depends on h, even the so-called classical ideal gas has a quantum
mechanical marker, suggesting classical thermodynamics is an oxymoron.

12.2. CLOSING COMMENT

171

5. Finally, using Eq.12.39 and the ideal gas value for gr , i.e. Eq.12.46,
nQ V
= pV .

(12.53)

which with = e and from Eq.12.48, gives


pV = Nop kB T

(12.54)

the ideal gas law.

12.2

Closing comment

Beginning with a Thermal Lagrangian, Eq.12.8, the Grand Canonical Picture unfolds, introducing
the indispensable chemical potential and bringing a workable methodology to open systems physical
situations in which particle number is variable. The next chapter discusses the chemical potentials
role in diffusion within inhomogeneous systems and extends s application to include equilibrium
in the presence of particle interactions, chemical reactants and distinct physical phases.

172

CHAPTER 12. OPEN SYSTEMS

Gibbs Equilibrium of Heterogeneous Substances has practically


unlimited scope. It spells out the fundamental thermodynamic theory of
gases, mixtures, surfaces, solids, phase changes, chemical reactions,
electrochemical cells, sedimentation and osmosis. It has been called,
without exaggeration, the Principia of Thermodynamics.
William H. Cropper, Great Physicists, Oxford University Press US,
2004.

Chapter 13

The Amazing Chemical


Potential
13.1

Introduction

The Chemical Potential , a creation of J. W. Gibbs1 , is the essential state variable for studying the
thermodynamics of open systems, in particular chemical reactions, phase transitions, non-uniform
systems, surfaces and other cases which benefit from varying particle number.2,3,4,5 Although
sometimes regarded as vague,6 when Schrodingers fixed particle number theory is extended by
including Hopen = Nop , provides both utility and clarity. Thermal Lagrangians now lead to
intuitive and manageable rules governing a variety of diffusive processes.
In order to improve understanding of the thermodynamics of reactions, phase transformations
and heterogeneous diffusion, this chapter includes supplementry examples and applications of the
chemical potential.
1 J. W. Gibbs, A Method of Geometrical Representation of the Thermodynamic Properties of Substances by
Means of Surfaces, Transactions of the Connecticut Academy (1873)
2 G. Cook and R. H. Dickerson, Understanding the chemical potential, Am.J.Phys. 63 73742 (1995).
3 T. A. Kaplan, The chemical potential, J. of Stat. Physics. 122, 1237-1260 (2006).
4 R Baierlein, The elusive chemical potential, Am.J.Phys. 69, 423 (2001).
5 G. Job and F. Herrmann, Chemical potential a quantity in search of recognition, Eur. J. Phys. 27 353-371
(2006).
6 In the preface to Introduction to Solid State Physics, Wiley New York 1971, C. Kittel writes: A vague discomfort
at the thought of the chemical potential is still characteristic of a physics education. This intellectual gap is is probably
due to the obscurity of the writings of J. Willard Gibbs who discovered and understood the matter 100 years ago.

173

174

13.2

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

Equilibrium constraints

Consider a single system (equivalently referred to as a phase) at temperature T and pressure p


composed of, say, two distinct species A and B. Allow the species to initially be diffusively isolated
and chemically non-reactive. This species composite can be described in terms of an open system
macroscopic hamiltonian
A
B
composite = hop [A Nop
]
H
+ B Nop

(13.1)

with a corresponding Thermal Lagrangian


1
T

ln P
0 P
hop +
L = kB P

1 A A
B
[ Nop + B Nop
]
T

(13.2)

where, for total species isolation A B and where abbreviations


ln P
=
P

N =0,1,2...

=
P

N =0,1,2...

[s (N ) , N ] ln P
[s (N ) , N ]}
{ P

(13.3)

[s (N ) , N ]}
{ P

(13.4)

are used.

13.2.1

Thermodynamics and diffusive equilibrium: Examples

If the two species A and B are in complete diffusive contact, at equilibrium the individual Nop
B
A
B
and Nop
are no longer distinguishable being replaced by the cumulative Nop = Nop
+ Nop

and, therefore, the corresponding Thermal Lagrangian


ln P
0 P

L = kB P

A
B
+ Nop
] ,
hop + [Nop
T
T

(13.5)

i.e., if species A and B freely diffuse, A and B must have the same chemical potential, A = B = .
13.2.1.1

Ideal gas in the earths gravitational field

In the following examples of diffusive equilibrium the the ideal gas provides an easily understood
model for illustrating the versatile role of the chemical potential.
At a distance z above but close to the earths surface, ideal gas molecules experience a quasiclassical gravitational potential energy per particle P E = mgz. Here g is the earths gravitational
constant and m the molecular mass. As introduced in Chapter 6 [see Table 6.1] and again in
Chapter 13, a Thermal Lagrangian, and hence Zgr , can be written to reflect this additional gravitational potential (work) contribution
Zgr =

N =0,1,...

e(mgz)N Z (N )

(13.6)

13.2. EQUILIBRIUM CONSTRAINTS

175

where Z (N ) is the N -particle (canonical) partition function for the ideal gas [see Chapter 7]
1
N
(nQ V ) .
N!

(13.7)

Zgr = exp [e(mgz) nQ V ] ,

(13.8)

Z (N ) =
Upon summing N Eq.13.6 becomes

and then, as also shown in Chapter 12 [see Eq.12.15], the average particle number at a distance z > 0
above the earths surface is
1

1
=

Nop (z) =

ln Zgr )

T,V

(mgz)
( )e
nQ V

= e emgz nQ V .

(13.9)
(13.10)
(13.11)

The chemical potential of gas molecules at any height z is, therefore,


(z) =

1 Nop (z)
ln
+ mgz .

nQ V

(13.12)

Gas particles (at height z) can freely diffuse to any adjacent level, say zz, so that in diffusive equilibrium
must be uniform throughout the atmosphere, i.e. independent of z. At ground level, i.e. z = 0,
the chemical potential for ideal gas molecules is
(0) =

1 Nop (0)
ln
,

nQ V

(13.13)

so in diffusive equilibrium Eqs.13.12 and 13.13 are equated to give


Nop (z) = Nop (0)emgz

(13.14)

the well known barometric equation.


13.2.1.2

Charged ideal gas between capacitor plates

Ideal gas molecules, each with charge q, lie between a pair of large, parallel capacitor plates placed
a distance d apart and charged to a potential V = Ex d, where Ex is a uniform electric field between
the plates. An electric potential energy per particle, P E = qxEx adds, quasi-classically, to
the N -particle ideal gas eigen-energies so that, as in Eq.13.11, the number of charged molecules a
distance x from the positive plate is
Nop (x) = e eqxEx nQ V .

(13.15)

1 Nop (x)
ln
q x Ex .

nQ V

(13.16)

Solving for the chemical potential,


(x) =

176

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

When x = 0 (the positive plate)


1 Nop (0)
ln
.

nQ V

(0) =

(13.17)

For diffusive equilibrium, in the region between the plates is independent of x and
Nop (x) = Nop (0)eqxEx .

(13.18)

Nop (d) = Nop (0)eqV

(13.19)

At x = d, i.e. the negative plate,

where V = Ex d is the electric potential difference between the plates.


13.2.1.3

Ideal gas in a rotating cylinder

Ideal gas particles, mass m, in a cylinder rotating with angular frequency , acquire a rotational
potential energy per particle (in the rotating frame) which, quasi-classically, is P Erot = 12 m 2 r2 .
Here r is a particles radial distance from the cylinder axis. The grand partition function is
1

Zgr =

e(+ 2 m

r )N

2 2

Z (N )

(13.20)

N =0,1,...

where Z (N ) is, as in Eq.13.7, the N particle (canonical) ideal gas partition function. After
summing N , at a distance r from the axis of rotation the average particle number is therefore
1

1
=

Nop (r) =

ln Zgr )

T,V
2 2
1

( ) e( 2 m r ) nQ V

= e e 2 m

2 2

nQ V .

(13.21)
(13.22)
(13.23)

Hence, the chemical potential at r is


(r) =

1 Nop (r) 1
ln
m 2 r2 ,

nQ V
2

(13.24)

1 Nop (0)
ln
.

nQ V

(13.25)

whereas the chemical potential at r = 0 is


(0) =

In diffusive equilibrium must be uniform throughout the cylinder, i.e. independent of r, giving
the radial distribution of particles in a centrifuge

Nop (r) = Nop (0) exp ( m 2 r2 ) .


2

(13.26)

13.2. EQUILIBRIUM CONSTRAINTS

13.2.2

177

Thermodynamics of Chemical Equilibrium

Consider the concrete example of a system at temperature T containing three species A, B and
AB. Initially allow all three species to be non-interacting so that a Thermal Lagrangian is
1
T

ln P
0 P
Hop +
L = kB P

1 A A
B
AB
[ Nop + B Nop
+ AB Nop
]
T

(13.27)

where A , B and AB are the chemical potentials for each of the isolated species.
Now let the species react according to the chemical equation
A + B AB .

(13.28)

In such a chemical reaction all reactants and products coexist with small fluctuations about mean
equilibrium concentrations of the participants.7 This balanced chemical equation defines atom
diffusion such that at equilibrium only the average atom sums
A
AB
Nop
+ Nop
= A

(13.29)

B
AB
Nop
+ Nop
= B

(13.30)

and

are still identifiable. These constraints coordinate with a chemically diffusive equilibrium condition
A + B = AB

(13.31)

to give the Thermal Lagrangian


1
T

ln P
0 P
Hop +
L = kB P

1 A
A
AB
B
AB
[ (Nop
+ Nop
) + B (Nop
+ Nop
)] .
T

(13.32)

As a second example consider a hypothetical reaction among the diatoms A2 , B2 and the molecule
A B,
A2 + B2 2 AB .

(13.33)

Assuming uniform reaction temperature T and pressure p, consider the chemical species as initially
non-reactive. In that case the Thermal Lagrangian is
1
T

ln P
0 P
Hop +
L = kB P

1 A2 A2
B2
AB
[ Nop + B2 Nop
+ AB Nop
] .
T

(13.34)

According to Eq.13.33 the atom-reaction constraints for this case are


A2
Nop
+

AB
Nop

= A

(13.35)

7 There are a few rare chemical reactions where concentrations of reactants and products oscillate in time so that
equilibrium is never attained.

178

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

and
B2
Nop
+

AB
Nop

= B

(13.36)

where A and B are the still identifiable equilibrium atom numbers. These reaction constraints
coordinate with chemically diffusive equilibrium
A2 + B2 = 2AB .

(13.37)

and result in the Thermal Lagrangian


1
T

ln P
0 P
Hop +
L = kB P

1 A2
1 AB
1 AB
A2
B2
[ (Nop
+ Nop
) + B2 (Nop
+ Nop
)] (13.38)
T
2
2

derived from Eq.13.34. These results can be generalized to any chemical reaction, e.g.
a1 A1 + a2 A2 + a3 A3 z1 Z1 + z2 Z2 + z3 Z3
in which case the atom reaction constraints are
a1
A1
Z3
+ Nop
Nop
= a1
z3
a2
A2
Z3
+ Nop
= a2
Nop
z3
a3
A3
Z3
+ Nop
Nop
= a3
z3
z1
Z1
Z3
Nop
Nop
= z1
z3
z2
Z2
Z3
Nop
Nop
= z2
z3

(13.39)

(13.40)
(13.41)
(13.42)
(13.43)
(13.44)

where a1 , a2 , a3 , z1 and z2 are the still identifiable atom numbers which coordinate with the
chemically-diffusive equilibrium condition
a1 A1 + a2 A2 + a3 A2 z1 Z1 z2 Z2 z3 Z3 = 0

(13.45)

to generate the Thermal Lagrangian.


1
T

ln P
0 P
Hop
L = kB P
+

a1
a2
z2
1 A1
A1
Z3
A2
Z3
Z2
Z3
[ (Nop
+ Nop
) + A2 (Nop
+ Nop
) . . . + Z2 (Nop
Nop
)] . (13.46)
T
z3
z3
z3

13.2.3

A law of mass action

Chemical reactions such as (hypothetically) defined in Eqs.13.28, 13.33, 13.39 do not usually go to
completion. Within a sealed reaction vessel are contained an equilibrium mixture of reactants
and products. For example, the gaseous reaction
3H2 + N2 2 N H3

(13.47)

13.2. EQUILIBRIUM CONSTRAINTS

179

in which at thermal equilibrium


3H2 + N2 2N H3 = 0 ,

(13.48)

yields only a small amount of N H3 (ammonia), with reactants H2 and N2 still having a substantial
presence. The equilibrium concentrations of reactants and products in the closed reactionvessel
is, however, well described by a Law of Mass Action which is a consequence of Eq.13.48. This
law is arrived at by first exponentiating
exp (3H2 ) exp (N2 )

=1

(13.49)

= 1.

(13.50)

exp (2N H3 )
or
3

[exp (H2 )] [exp (N2 )]


[exp (N H3 )]

Continuing in the spirit of the example, assume all participating gases g are ideal and at sufficiently low temperature that no internal modes of the reactive molecules (rotational, vibrational
or electronic) are excited. In which case
C g

g = ln g
n
Q
= ln {[

(13.51)

mg 3/2 g
C }
2 ]
2 h

(13.52)

where mg is the molecular mass of the gas g and


Cg =

g
Nop

(13.53)

is the gas concentration. From this follows a typical mass action result
3

[C N H3 ]
3

[C H2 ] [C N2 ]

N2
2
{nH
Q } {nQ }
2

H3
{nN
}
Q

= K (T )

(13.54)
(13.55)

where K depends only on T . Specific mass action constants K are associated with different
chemical reactions or molecular models. If, for example, in the previous model temperature is raised
sufficiently to excite molecular rotational and vibration (internal modes) the N -particle canonical
partition function becomes

Z (N ) =

1
N
N
[Zint ] [nQ V ]
N!

(13.56)

where Zint is the partition function for the internal (rotational and vibrational) modes. In that
case we redefine nQ for the molecular gas g
g
ngQ Zint
ngQ = ngint

(13.57)

180

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

so that the mass action expression becomes


[C N H3 ]

[C H2 ] [C N2 ]

= Kint (T ) ,

(13.58)

where
3

Kint (T ) =

13.2.4

N2
2
{nH
int } {nint }
2

H3
{nN
int }

(13.59)

Thermodynamics of phase equilibrium

A phase is a chemically and physically uniform macroscopic state of matter. Vapor, liquid and
solid are among the familiar examples of phases. Other examples include different coexisting crystal
structures with the same chemical composition or the more exotic He4 - Bose-Einstein condensate
phases. Phase equilibrium describes a situation in which two or more homogeneous regions (phases)
coexist in thermodynamic equilibrium within a rigid container but are separated by physical boundaries. An example are the three phases of H2 O ice, liquid water and water vapor in which any
two or all three can coexist. Distinct phases may also exist within a given state, such as in solid iron
alloys and in the several phases that can coexist for liquid states. Those phases that are possible
depend on temperature and pressure. If two coexisting phases, designated and , composed of a
single species, freely exchange particles at temperature T within a fixed volume (V + V ), then a
Thermal Lagrangian is8
ln P
+ P
ln P
] 0 P
0 P

L E,N,V = kB [P

[h
+ Nop
] , (13.60)
p (V + V ) + [Nop
op + hop ]
T
T op
T

with
T = T = T
= =
p
op

13.2.5

pop

(13.61)

= pop

Gibbs-Duhem relation

Long before he published his monumental treatise on statical mechanics Gibbs had already extended
the fundamental equation of thermodynamics to include particle transfer. For a single phase and a
single species the fundamental equation becomes
T dS = dU + p dV dNop .
8 Mechanical

energy, p V is included in the Lagrangian.

(13.62)

13.2. EQUILIBRIUM CONSTRAINTS

181

When U is expressed in terms of its natural variables


U = U (S, V, Nop )

(13.63)

all of which are extensive, then in the Euler form


U (S, V, Nop ) = U (S, V, N ) .

(13.64)

Applying Eulers homogeneous function theorem


S(

U
U
U
)
+V (
)
+Nop (
)
= U (S, V, Nop ) .
S V,Nop
V S,Nop
Nop V,S

(13.65)

Comparing this with Eq.13.62, the fundamental equation is effectively integrated to get an Euler
equation
U T S + V p Nop = 0 .

(13.66)

dU T dS S dT + V dp + p dV dNop Nop d = 0

(13.67)

Taking the total differential

and combining it with the fundamental equation Eq.13.62, we have for a single phase with one species
S dT + V dp Nop d = 0 .

(13.68)

This is called the Gibbs-Duhem Equation. In the case of a single phase with M species Gibbs-Duhem
becomes
M

S dT + V dp Nop i di = 0

(13.69)

showing that intensive variables T, p, i are not independent.

13.2.6

Multiphase equilibrium

Gibbs-Duhem opens the door to basic understanding of coexistence between different phases, such
as between liquid water and water vapor, or between ice and liquid water, etc.
Begin by rewriting Eq.13.68 in the form
d = s dT +v dp

(13.70)

where intensive quantities


s=

S
Nop

v=

V
Nop

(13.71)

182

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

have been defined. Taking advantage of p and T as natural variables, for two phases, and ,
at the same T and p and composed of only a single species (pure phases), the chemical potentials
are

() = () (p, T )

(13.72)

() = () (p, T ) .

(13.73)

If these phases are to coexist, then

() (p, T ) = () (p, T )

(13.74)

which is a system of one equation with two unknowns, T and p. Elementary algebra says although
there is no unique solution, T can be found in terms of p so that the and phases can coexist
along some curve in the {p, T }-plane.

Now suppose there are three pure phases, , and . Then for coexistence between the three
phases

() (p, T ) = () (p, T )

(13.75)

() (p, T ) = () (p, T ) ,

(13.76)

which is a system of two equations and two unknowns with a unique solution a single point in
the {p, T }-plane. The three phases can coexist at what is called a Triple Point 9 . Figure 13.1 is an
example of a phase diagram for a single component system, e.g. water10 in which temperature and
pressure form the coordinate axes. As shown in the diagram, only certain phases are possible at a
particular temperature and pressure, with each phase separated from the others by a curve in the
{p, T } plane called a phase boundary. This curve indicates the values of T and p along which two
phases can coexist.

point of water: Ttp = 273.16K, Ptp = 0.00604atm.(1atm = 101.3kP a).


experimental solid-liquid boundary for water has an anomalous negative slope so the diagram is only an
approximate representation for water. This is due to the peculiarity of ice being less dense than liquid water.
9 Triple

10 The

13.2. EQUILIBRIUM CONSTRAINTS

183

/dT
dp

is
ex
co

ce
ten

Figure 13.1: Single species phase diagram. The point where the solid-liquid and liquid-vapor phase
lines meet is called the triple point. At that point all three phases solid, liquid and vapor can
coexist.
A unique coexistence property is the point at which the phase boundary between liquid and vapor
abruptly disappears. This is called a critical point. Beyond this point liquid phase and vapor phase
lose their usual meaning, becoming instead a supercritical fluid having properties of both vapor and
liquid. 11

13.2.7

The Clausius-Clapeyron equation

For the case of two coexisting pure phases and (() = () ),


s() dT + v () dp = s() dT + v () dp ,

(13.77)

so that the slope of a coexistence curve is [see Figure 13.1]

11 Critical

dp
s() s()
]
= ()
dT coexistence v v ()
s
=
v

point of water: Tc = 647K, pc = 217.7atm. Critical point of CO2 : Tc = 304K, pc = 72.8atm.

(13.78)
(13.79)

184

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

which is called the Clausius-Clapeyron equation. As a concrete example that relates to Figure 13.1
let ` (liquid phase), and g (gas phase). Then s is the liquid-gas entropy difference per
particle and v is the liquid-gas volume difference per particle. Moreover
s =

Qvap
Nop T

(13.80)

where Qvap /Nop is the heat of vaporization per particle,12 so that


[

Qvap
dp
]
=
.
dT coexistence T Nop v

(13.81)

Qvap
is not a simple function of p and T so that integrating Eq.13.81 to obtain
T Nop v
the vapor pressure may not be straightforward. On the other hand assuming an ideal gas phase
and since, in general, v (gas) >> v (`) , the differential equation simplifies to
In general

p Qvap
dp
]
=
.
dT coexistence kB T 2

(13.82)

Since generally v (solid) < v (`) for most systems, as suggested in Figure 13.1, the solid-liquid coexistence curve is nearly vertical but positively sloped. In the unusual case of water for which
v (solid) > v (`) the curve is also nearly vertical, but now slightly negatively sloped.

13.2.8

A star is born

Star formation takes place inside cold (T 10K), dense interstellar regions of molecular hydrogen
(H2 ), carbon monoxide (CO) and dust, called giant molecular clouds (GMC). The process of star
formation is one of gravitational collapse, triggered by mutual gravitational attraction within the
GMC, to form a region called a protostar. GMCs have typical masses of 6 106 solar masses,
densities of 100 particles per cm3 and diameters of 9.5 1014 km. The process of collapse resembles
a phase transition in that its onset occurs at critical GMC values of temperature (Tc ), mass (mc )
and pressure (pc ). Once begun the collapse continues until the GMC is so hot and compact that
nuclear fusion begins and a star is born ( 10 15 106 years).
13.2.8.1

Thermodynamic model

A thermodynamic model for a GMC is an ideal gas with mutual gravitational attraction between
the gas particles. Were it not for gravitational effects the GMC equation of state would simply
be pV = N kB T . But an attractive gravitational potential (proportional to 1/ r) contributes the
additional interaction energy
1
1
Vgrav = G dr dr [ (r)
(r )]
2
r r
12 The

heat required to evaporate one particle.

(13.83)

13.2. EQUILIBRIUM CONSTRAINTS

185

with (r) the mass density of the cloud and G the universal gravitational constant. The manyparticle gravitational interaction is manageable for this application in a mean field approximation
(MFA),13,14 in which each particle sees the average potential of all the other particles as an effective
interaction, i.e.
1
Vef f = dr
(r) .
2

(13.85)

Applying a MFA [see footnote Eq.13.84] the potential (r) is


(r) = G dr

{2 (r )
}
.

r r

(13.86)

with
a uniform particle density
N mH
V
3N mH
=
4R3

(13.87)
(13.88)

and
(r) = Nop (r)/V

(13.89)

is the particle density function. Here N is the number of molecules in the GMC, with mH the
mass of a hydrogen molecule (assumed to be the dominant species in the GMC) and V the clouds
volume.
Poissons equation under the MFA
2 (r) = 4G {2 (r)
} .

(13.90)

suggests Gauss Gravitational Flux Law for integrating Eq.13.86. In this bare bones model
all GMC radial mass dependence is ignored the cloud being assumed spherical with uniform
density.15,16 Under this uniformity assumption (r)
and using Gauss Flux Law
(r) =
13 The

2G

(r2 3R2 )
3

0<rR

(13.91)

essence of a mean field approximation is the identity:


(x ) (x) = [ (x ) ] [ (x) ] + (x ) + (x)

(13.84)

where is the average density. The MFA neglects the first term, i.e. fluctuations around the average density.
14 Peter Palffy-Muhoray, The single particle potential in mean field theory, Am. J. Phys. 70, 433-437 (2002).
15 This simplification has obvious shortcomings. Nevertheless, main features of gravitational collapse are preserved.
16 W.B. Bonnor, Boyles Law and Gravitational Instability, Mon. Notices Roy. Astron. Soc. 116, 351 (1956).

186

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

so that
R

Vef f

1
=
dr 4r2 (r)
2

(13.92)

3GM 2
5R
3GM
= N mH
.
5R
As in Eq.13.6, a grand partition function may now be written
=

[ +
Zgr =

3mH M G
]N
5R
Z (N )

(13.93)
(13.94)

(13.95)

N =0,1,2...

which from Eqs.13.7 and 13.8 is17

Zgr

3mH M G

]
[ +

5R
= exp nQ V e
.

(13.96)

Using this we find:


1. The mean number of hydrogen molecules in the GMC
Nop =

1
(
ln Zgr )

T,V

3GM
[ + mH (
)]
1
5R
= ( ) nQ V e

3GM
[ + mH (
)]
5R
= nQ V e
,

(13.97)

(13.98)

(13.99)

2. The chemical potential


= mH (

1 Nop
3GM
) + ln
5R

nQ V

(13.100)

3. The pressure in the GMC


p=

which, with R = (

(13.101)

3V 1/3
)
and Eq.13.100, gives the GMC Equation of State
4
p=

17 The

1
(
ln Zgr )
V
T,

Nop kB T Nop mH 4 1/3

( ) GM V 4/3 .
V
5
3

factor 3/5 is due to assumed spherical symmetry of the GMC.

(13.102)

13.2. EQUILIBRIUM CONSTRAINTS

187

Restoring M = Nop mH the equation of state becomes


p=

13.2.8.2

Nop kB T Nop 2 m2H 4 1/3

( ) GV 4/3 .
V
5
3

(13.103)

Collapse criterion

A criterion for collapse is that the clouds adiabatic compressibility


1
V
1
=
V

S =

V
)
p S
V
(
) ,
p T
(

(13.104)
(13.105)

becomes infinite.18 At that critical point the cloud becomes gravitationally unstable, spontaneously
shrinking until some new pressure source (nuclear fusion) overcomes gravity.19 As a first step
towards visualizing Eq.13.103 set =

1
5

( 4
)
3

1/3

. Then form the dimensionless variables


4

p = {G3 M 2 (mH ) } p

(13.109)

and
3
V = (GM mH ) V

(13.110)

to give a reduced GMC equation of state


p =

.
V V 4/3

(13.111)

This is plotted in Figure 13.2. The slope of the curve is zero at Vc = 0.0794 and pc = 3.147, the
coordinate of instability.
previously defined, = Cp /CV .
is equivalent to the Jeans criterion, [J. H. Jeans, The Stability of a Spherical Nebula, Phil. Trans. Royal
Soc.(London). Series A,199, 1 (1902).] that the velocity of sound cs in the GMC becomes imaginary!!. Since the
velocity of sound in a gas is
18 As

19 This

c2s =

KS

(13.106)

where KS is the gas adiabatic bulk modulus


p
)
V S
p
= V (
)
V T

KS = V (

(13.107)
(13.108)

and is its density, the Jeans criterion is that the adiabatic bulk modulus (the reciprocal of compressibility) becomes
zero.

188

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

V
Figure 13.2: Reduced Equation of State for GMC. Vc = 0.0794. pc = 3.147.
Alternatively, from Eqs.13.103 and 13.105 the adiabatic compressibility is
1

4 4 1/3
1
2
[Nop kB T V 2
( ) Nop m2H GV 7/3 ]
kS =
V
15 3

(13.112)

so that with = Nop /V the critical radius, rc , for collapse is

45 1/2
kB T
)
rc = (
16
m2H G

(13.113)

and the critical pressure is


pc =

kB T
.
4

(13.114)

Star formation is triggered by GMC compression to p pc by collisions with other GMCs or


compression by a nearby supernova event. Alternatively, galactic collisions can trigger bursts of
star formation as the gas clouds in each galaxy are compressed.

13.2.9

Van der Waals equation

An ideal gas (classical or quantum) is defined by neglecting interactions between particles. More
realistic models (theoretical or computational simulations) are generally based on pairwise forces
with a long range attractive and short range repulsive component, the latter restricting a gas of
molecules to finite volume. In 1873 Van der Waals pursued these ideas in his Ph. D. Thesis,
replacing the ideal gas law with an improved equation of state that bears his name.20 For this he
was awarded the 1910 Nobel Prize in Physics.
20 The

microscopic reality of atoms was still being challenged by influential skeptics.

13.2. EQUILIBRIUM CONSTRAINTS

189

Apart from an interaction potential being between gas (liquid) molecules, a derivation of the van der
Waals equation is similar to the discussion in previous section. In this case we choose an interaction
U (r r) that displays hard core repulsion as well as a long range inter-molecular attraction, as
pictured in Figure 13.3.

U(r)

r0

Figure 13.3: Intermolecular potential.


Assuming an interaction energy
1
Vint = dr dr [ (r) U (r r) (r )]
2

(13.115)

with the density in units of particles per unit volume. This many-particle inter-molecular interaction
is also manageable in the MFA [see Eq.13.84], in which case
1
(r) ,
Vint = dr
2

(13.116)

with an average density

N
.
V

(13.117)

The potential function is

(r) =4 (r)U (r )r 2 dr
r0

=4 (
r0

Nop
)U (r )r2 dr
V

(13.118)

(13.119)

190

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

which, assuming gas uniformity, becomes

(r) = 4(

Nop
) U (r)r2 dr .
V

(13.120)

r0

With

a = 4 U (r)r2 dr

(13.121)

r0

we have the interaction potential


Vint = N a

Nop
V

(13.122)

and the grand partition function


[ + a
Zgr =

Nop
]N
V
Z (N )

(13.123)

N =0,1,2...

Nop

]
[ + a

V
= exp nQ V e
.

(13.124)

From this we find


Nop =

1
(
ln Zgr )

T,V

(13.125)

Nop
[ + a
]
1
V
= ( ) nQ V e

Nop
[ + a
]
V
=n V e
,
Q

(13.126)

(13.127)

and the chemical potential


= a

Nop 1 Nop
+ ln
.
V

nQ V

(13.128)

Finally, the pressure is


p=

1
(
ln Zgr )
V
T,

(13.129)
2

Nop
Nop
a(
)
V
V

(13.130)

13.2. EQUILIBRIUM CONSTRAINTS

191

and an equation of state is, thus far,


2

p + a(

Nop
Nop
) =
.
V
V

(13.131)

The final step in the van der Waals argument assigns a minimum gas volume Vmin = bNop as
the fully packed space occupied by the gas molecules, where b = 6 r03 is the hard core restricted
volume per molecule. Thus the van der Waals equation of state becomes

p + a ( Nop ) [V Nop b] = Nop .

(13.132)

[V Nop b] is sometimes labeled V , the effective volume.

13.2.10

Surface adsorption Langmuirs model

In the process of film growth or surface doping, atoms or molecules in gas phase or in dilute solution,
bind to the film surface. A theory of solid surface coverage by these molecules was formulated by
Irving Langmuir (1916), the acknowledged pioneer of surface chemistry.
Assumptions of the model are:
1. The solid surface is in contact with ideal monatomic gas atoms at temperature T and pressure
p0 .
2. There are a fixed number of sites N on the surface available for bonding.
3. Each surface site can be only singly occupied (monolayer coverage).
4. Adsorption at a given site is independent of occupation of neighboring sites (no interactions).
5. A gas molecule bound to the surface has the non-degenerate eigen-energy
Ebound = 

(13.133)

with respect to the energy of a free gas atom.


6. The energy difference between adsorbed gas atoms and free gas atoms is independent of
surface coverage.
7. Atoms bound to the surface (Abd ) are in thermal and diffusive (phase) equilibrium with gas
atoms as described by the reaction
Agas Abd

(13.134)

With Langmuirs model the fractional surface coverage


=

bd
Nop

Nmax

(13.135)

192

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

bd
can be determined, where Nop
is the average number of gas atoms bound to Nmax accessible solid
surface sites.

Diffusive (phase) equilibrium in the surface phenomenon of Eq.13.134 is described by the Thermal
Lagrangian
bd ln P
bd + P
gas ln P
gas ] 0 P
bd 0gas P
gas
L E,N = kB [P
bd

1
gas
bd
gas
[hbd
[Nop
+ Nop
] (13.136)
op + hop ]
T
T

which correlates with

gas = bd =

(13.137)

where bd is the chemical potential of atoms bound to surface sites and gas is the chemical potential
for ideal gas atoms.

Figure 13.4: Gas molecules interacting with solid surface.


bd
Finding the average number of atoms bound to surface sites Nop
follows discussions in Chapter 12
where it was shown that
Nmax

N e{[Es (N )]bd N }

bd
Nop

N =0 s
Nmax

N =0 s

(13.138)

e{[Es (N )]bd N }

with the number of bound atoms limited by Nmax and where Es (N ) = N . The denominator in
Eq.13.138 is the Grand Partition Function Zgr ,
Nmax

Zgr = ebd N Z (N )

(13.139)

N =0

where Z (N ) is the (canonical) partition function for N -bound atoms


Z (N ) = g (N, Nmax ) eN  .

(13.140)

13.3. DISSOCIATIVE ADSORPTION

193

Since the bound atoms only eigen-energy [see Eq.13.133] has the configurational degeneracy
g (N, Nmax ) =

Nmax !
(Nmax N )! N !

(13.141)

the grand partition function is


Nmax

Zgr =

N =0

Nmax !
eN (bd +)
(Nmax N )! N !

= [1 + e(bd +) ]

Nmax

(13.142)
(13.143)

and the fractional surface coverage is [see Eq.13.138]


=

bd
Nop

Nmax
1
1
=

ln Zgr
Nmax bd

ebd e
.
1 + ebd e

(13.144)
(13.145)
(13.146)

The ideal gas in contact with the surface has the chemical potential
gas =

gas
1 Nop
ln
.

nQ V

(13.147)

Equating the chemical potentials [see Eq.13.137] (diffusive equilibrium) and using the ideal gas
gas
Nop

equation of state p0 V =
gives the Langmuir fractional surface coverage

p0
0 + p0

(13.148)

nQ e

(13.149)

where 0 is the temperature dependent factor


0 =
known as the Langmuir isotherm.

13.3

Dissociative adsorption

A diatomic gas molecule, say A2 , may not simply bind to a surface site, but may dissociate, with
each component atom binding to a single site. This is referred to as dissociative adsorption and is
described by the chemical reaction
A2 2Abound .

(13.150)

194

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

Therefore, the thermal equilibrium chemical potentials satisfy


A2 2A = 0 ,

(13.151)

where A2 is the chemical potential of the ideal diatomic gas and A is the chemical potential
of surface bound atoms. The ideal diatomic gas in contact with the surface has the chemical
potential
A2 =

A
1 Nop2
ln A2 .

nint V

(13.152)

which includes the effect of internal (rotational and vibrational) modes [see Eq.13.56]. Once again
taking Ebound =  the fractional surface occupation is, according to Eq.13.146,
=

eA e
.
1 + eA e

(13.153)

Now applying Eq.13.151 the surface coverage is


=

e(1/2)A2 e
.
1 + e(1/2)A2 e

Using the ideal gas result, Eq.13.152, we finally have

p
=

D + p
where p is the gas pressure and the Langmuir isotherm is

nA2
int e .
D =

13.4

(13.154)

(13.155)

(13.156)

Crystalline bistability

In an ordered crystal it is common for atoms to be bistable, i.e. atoms at normal crystalline sites
can migrate to abnormal sites (usually accompanied by a lattice distortion), where they have a
different binding energy. If a crystal at a temperature T has normal sites and displaced sites,
the latter lying on an interstitial sub-lattice [see Figure 13.5], after a long time a certain fraction
of the atoms will occupy displaced sites. Since the total number of atoms normal plus defect
is conserved, the atom constraint is
norm
dis
Nop
+ Nop
= Nnormal (the total number of normal sites) .

(13.157)

The migration process can be interpreted as the chemical reaction


Anorm Adis .
Following details of Langmuirs model and assuming:

(13.158)

13.4. CRYSTALLINE BISTABILITY

195

Figure 13.5: Lattice atoms migrating from normal to displaced sites.


1. An equal number of normal and displaced sites, Ns .
2. Displaced atoms have energy 0 > 0 with respect to normal sites.
3. Only single occupancy of normal and dispaced sites is permitted.
The grand partition function for single occupancy of a normal site is
Ns

Ns !
}
(Ns N )!N !

norm
Zgr
= eN n {
N =0

Ns

= (1 + en )

(13.159)
(13.160)

while the grand partition function for single occupancy of a displaced site is
Ns

inter
Zgr
= eN dis {
N =0

Ns !
eN 0 }
(Ns N )!N !
Ns

= (1 + e0 edis )

(13.161)
(13.162)

From grand partition functions, Eqs.13.160 and 13.162,


1

norm
ln Zgr
norm
norm
= Ns
1 + norm

norm
Nop
=

(13.163)
(13.164)

and
1
dis
ln Zgr
dis
dis e0
= Ns
1 + dis e0

dis
Nop
=

(13.165)
(13.166)

where
norm = enorm

(13.167)

dis = edis .

(13.168)

and

196

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

At thermal equilibrium
norm = dis

(13.169)

norm = dis = .

(13.170)

or

Applying Eq.13.157 (Nnormal = Ns ) together with Eqs.13.164, 13.166, 13.170,

/
Figure 13.6: Fractional occupation number of displaced sites.

e0
=1
+
1 + 1 + e0

(13.171)

we find
= e0 /2 .

(13.172)

Therefore, according to Eq.13.166 the average fractional occupation of displacement sites is


dis
Nop

Ns
as shown in Figure 13.6.

1
1 + exp (0 /2)

(13.173)

13.4. CRYSTALLINE BISTABILITY

197

Problems and exercises


1. A narrow vertical shaft is drilled into the earth (earth radius RE and mass ME ) to a depth
d, with d << R [see figure 13.7.] The shaft fills with air and comes to thermal and diffusive
equilibrium at a temperature T .
(a) Treating air as an ideal gas, what is the chemical potential for the air within the shaft?
(b) Find an expression for the gas pressure p (r) , 0 < r d, within the shaft.

r
d

Figure 13.7: Shaft drilled into the earth.

HINT: The potential energy (r) of a gas molecule, mass m, a distance r below the top of
the shaft is
(r) =

mg
2
[(R r) R2 ]
2R

0 < r d.

(13.174)

2. Find Cp CV for the van der Waals gas.


3. An absorbed surface layer of atoms uniformly covers an area A. The atoms, although free to
move, interact with a two-dimensional potential V () where is the interatomic separation.
[See Figure 13.8]
Find the surface pressure pS of this two-dimensional gas as a function of atom density
n = Nop /A up to order n2 .

198

CHAPTER 13. THE AMAZING CHEMICAL POTENTIAL

Figure 13.8: Two-dimensional potential.


4. Atoms are bound to solid surface sites with binding energy 0 relative to vapor (free, ideal
gas) atoms with which they are in equilibrium. Bound atoms are modeled as 2-dimensional
Einstein oscillators with frequency 0 so that the eigen-energies of the th bound surface
atom are
0 (n,x + n,y )
E = 0 + h

(13.175)

where n,x , n,y = 0, 1, 2, 3, . . . .


a Find the grand potential, , that describes bound surface atoms.
b Find the chemical potential for the bound surface atoms.
c Find an expression for the pressure of the ideal gas vapor that is in equilibrium with surface
atoms.
5. A particular mesh surface is employed as a catalyst for removal of ozone O3 from exhaust
gases by converting it to oxygen molecules O2 according to the reaction
2O3 3O2 .

(13.176)

If the binding energy of an oxygen molecule to a surface site is EB =  and the ozone pressure
is pO3 what is the covering fraction
=

nO2
N

(13.177)

where nO2 is the average number of oxygen molecules bound to the surface per unit area
and N is the number of possible O2 binding sites per unit mesh area.

A new scientific truth does not triumph by convincing its opponents


and making them see the light, but because its opponents eventually die,
and a new generation grows up that is familiar with it.
Max Planck, Scientific Autobiography and Other Papers, Copyright:
Philosophical Library.

Chapter 14

Thermodynamics of Radiation
14.1

Introduction

In 1900 Max Planck discovered that the temperature dependent law of radiating bodies could not be
derived solely from Maxwellian electrodynamics according to which the energy of an electromagnetic
field is1
E=

1
2
2
dx [E (x, t) + B (x, t) ] .
2

(14.1)

Planck arrived, instead, at results consistent with the relevant electromagnetic experiments by treating radiation of a given frequency as though it consisted of packets of energy photons each
with energy h, with a corresponding electromagnetic field energy
EP lanck = nh

(14.2)

where n = 0, 1, 2, . . . is the number of photons in the packet and h is Plancks universal constant.
Plancks hypothesis2 was the initial link in the chain of 20th century discoveries that is quantum
physics.

14.2

Electromagnetic eigen-energies

Proceeding directly to the thermodynamics of radiation, Eq.14.2 is replaced by the electromagnetic


quantum eigen-energies for photons of a single mode k having polarization
Ek, = hk, [n (k, ) + 1/2]
1 In

rationalized c.g.s. units.


inspired by Boltzmanns concept of microstates.

2 Likely

199

(14.3)

200

CHAPTER 14. THERMODYNAMICS OF RADIATION

where n (k, ) = 0, 1, 2, . . . is the number of photons. k and are quantum numbers for a single
photon of frequency k, . 3
Photons are fundamental, particle-like excitations of the electromagnetic field. They have zero
mass and are their own antiparticle. They can be created and absorbed (destroyed) without number
conservation and therefore have zero chemical potential. They appear in an almost limitless variety
of atomic, molecular or nuclear processes as well as continuous radiation (synchrotron radiation) as
sociated with kinematic acceleration of particles. The photon carries spin angular momentum h
corresponding to right and left circular polarization.
The Maxwellian total electromagnetic energy [see Eq.14.1] is similarly replaced by the quantum
average
HEM = hk, [nop (k, ) + 1/2]

(14.4)

=1,2 k

where HEM is the electromagnetic hamiltonian,


HEM = hk, [nop (k, ) + 1/2]

(14.5)

=1,2 k

with the photon mode number operator nop (k, ) identified by a three components wave vector
k with k kx , ky , kz , and two mutually perpendicular polarization directions 1 , 2 . The
eigenvalues of nop (k, ) are n (k, ) = 0, 1, 2 . . . and nop (k, ) is the average number of photons
in that mode. The wave vector k is the direction of propagation of the photon and the polarization
is the vector direction of the concurrent electric field E [see Figure H.1]. In free space the
2component polarization vector is perpendicular (transverse) to k. This follows from the Maxwell
equation4 E = 0. For electromagnetic radiation in free space, the energy of a photon associated
with a particular mode (k, ) is

hk, =

h
c k
2

(14.6)

which depends only on the magnitude k and is independent of the polarization . Here c is
the speed of light in vacuum. As indicated in Appendix H, in free space nop (k, ), the average number of photons in the mode (k, ), also depends only on k and is independent of the
polarization .

3 Arriving at Plancks result from Maxwells equations is the realm of quantum field theories, the details of which
are well beyond the scope of this book. Nevertheless, for completeness an outline of the method is discussed in
Appendix H.
4 Since there is no analog to this Maxwell equation in crystalline elastic equations of motion, there remain three
phonon polarizations.

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

14.3

201

Thermodynamics of electromagnetism

Based on macroscopic electromagnetic eigen-energies given by Eq.14.3 (see Appendix H), a Thermal
Electromagnetic Lagrangian LEM is constructed
LEM = kB P (nk, ) ln P (nk, ) 0 P (nk, )
nk,

(14.7)

nk,

1
P (nk, ) {hk, (nk, + 1/2)}
T nk,
k
=1,2

where P (nk, ) is the probability that nk, photons are in the mode k with polarization . The
sum
(14.8)

nk,

ranges over nk, = 0, 1, 2, . . . for each mode k and both polarizations ( = 1, 2). The Lagrange
multiplier 0 assures normalized probabilities. Following by now familiar procedures, we find from
Eq.14.7

exp hk, (nk, + 1/2)

=1,2

P (nk, ) =

exp hk, (nk, + 1/2)

nk,

=1,2

(14.9)

where the normalizing denominator is the thermal equilibrium electromagnetic partition function,

ZEM = exp hk, (nk, + 1/2) .

nk,

=1,2 k

14.3.1

(14.10)

Thermodynamics for a single photon mode

To simplify evaluation of Eq.14.10 we first examine the partition function for a single mode k

ZEM
=

exp hk , (nk , + 1/2)

nk , =0,1,2,3...
=1,2

(14.11)

where

nk , =0,1,2,3...

(14.12)

202

CHAPTER 14. THERMODYNAMICS OF RADIATION

is the sum over all integer photon numbers for a single mode (k , ) . Explicitly summing over
both polarizations in the exponent of Eq.14.11

ZEM
= ehk exp [hk (nk ,1 )] exp [hk (nk ,2 )]
nka ,1

=e

hk

(14.13)

nka ,2

exp [hk (nk ,1 )]

nka ,1

(14.14)

where the factor ehk results from the vacuum radiation term in the eigen-energies. The sum
over photon numbers nk ,1 = 0, 1, 2, . . . is just a geometric series, which gives for the single mode
k and both polarizations

ZEM

14.3.2

hk
2

= e

h
k

1e

(14.15)

Average photon number

From Eq.14.9 the average number of photons in the single mode k with a single polarization = 1,
say nk ,1 , is
nk ,1 exp {hk ,1 nk ,1 }

nk ,1 =

nk ,1

exp {hk ,1 nk ,1 }

(14.16)

nk ,1

This is identical with


1

ln (ZEM
)] 1/2
2hka ,1
1
=
,
h
k
a ,1 1
e

nk ,1 = [

(14.17)
(14.18)

which is called a Planck distribution function.

14.3.3

Helmholtz potential and internal eneregy

The Helmholtz potential F for a single mode k , is


1

log ZEM

2
= hk + log (1 ehk )

F =

(14.19)
(14.20)

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

203

Similarly, for a single mode k (counting both polarizations) the the radiation field internal energy
U is

ln ZEM

ehk + 1
= hk ( h
)
e k 1
1
1
= 2hk ( + h
)
2 e k 1
1
= 2hk ( + nk )
2

U =

14.3.4

(14.21)
(14.22)
(14.23)
(14.24)

Radiation field thermodynamics

The thermal electromagnetic field assembles all photon modes k with both transverse polarizations
= 1, 2, requiring the full evaluation of Eq.14.10. Rewriting ZEM as
ZEM

2
2

hk2 (nk2 +1/2)


hk1 (nk1 +1/2)

e
e
=

...

nk2 =0,1,2,...
nk1 =0,1,2,...

2
1
= [ehk (
) ],
1 ehk
k

(14.25)
(14.26)

in analogy with Eqs.14.19 and 14.24 the Helmholtz potential FEM is


2
1
1
hk
(
) ]
ln [e
h
k
k
1e
2
= [hk + ln (1 ehk )]

FEM =

(14.27)
(14.28)

and the internal energy UEM is


1
UEM = 2 hk ( + nk ) .
2
k

14.3.5

(14.29)

Stefan-Boltzmann, Planck, Rayleigh-Jeans radiation laws

In the limit of macroscopic volume V the sum over k in Eq.14.29 is replaced by (see Appendix
H)

(2)

dkx dky dkz ,

(14.30)

so that ignoring the divergent, constant vacuum energy


UEM = 2

V
(2)

dkx dky dkz (

hk
).
ehk 1

(14.31)

204

CHAPTER 14. THERMODYNAMICS OF RADIATION

But in vacuum we have Eq.14.6 so that the integrals can be carried out in spherical coordinates, in
which case

dkx dky dkz = d d sin d k k

and the internal energy UEM

(14.32)

is (with h = 2 h)
UEM

1
hcV
3
).
= 2 d k k ( hck

e
1

(14.33)

The integral can be brought into a more standard form with the substitution
x
k =
hc

(14.34)

so that
UEM =

1
hcV
).
dx x3 ( x
4
e 1
2 ( hc)

(14.35)

The integral in Eq.14.35 is one of several similar integrals that appear in thermal radiation theory.
They are somewhat subtle to evaluate but they can be found in comprehensive tables or computed
with Mathematica with the result

dx x3 (

1
4
)
=
ex 1
15

(14.36)

so that the radiation energy density is


UEM
2
.
=
3
V
15 4 (hc)

(14.37)

This is called the Stefan-Boltzmann Radiation Law .


If in Eq.14.35 we substitute x = h, the radiation energy density may be written

UEM
=
V

d u ()

(14.38)

8h
3
c3 eh 1

(14.39)

where
u () =

which is the frequency distribution of thermal radiation at any temperature T . It is called the
spectral density or Plancks Radiation Law. In the high temperature limit, h << 1, the Planck
Law becomes the classical Rayleigh-Jeans Law. [See Fig 14.1.]
8kB T 2
.
c3
The cancellation of Plancks constant removes any quantum reference.
uclassical () =

(14.40)

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

205

Figure 14.1: Plank Radiation Law and Rayleigh-Jeans approximation.

14.3.5.1

Example:

What is the average number density of thermal photons all modes, both polarizations at
temperature T .
Beginning with Eq.14.17 the average photon number is
(14.41)

n = nk,
k
=1,2

2V
=
(2)

dk

1
.
ehk 1

(14.42)

ck )
In spherical coordinates this becomes (with = 2
2

2V
n =
(2)

d d sin dk
0

k2

e hck

(14.43)

V
k2
dk hck

e
1
0

(14.44)

206

CHAPTER 14. THERMODYNAMICS OF RADIATION

= x the number density is


which after the change of variable hck

x2
n
1
dx x
=
.

V
e 1
2 ( hc)
0

(14.45)

This integral resembles Eq.14.36, but unlike that integral this one has no result in terms of familiar
constants. But it can be estimated by the following series of steps.

dx
0

x2
x2 ex
= dx
x
e 1
1 ex

(14.46)

= dxx2 ex esx

(14.47)

s=0

= dxx2 ex(s+1)
s=0 0

s=0 (s + 1)

=2

s =1

1
s3

(14.48)
(14.49)
(14.50)

The last sum is called the Riemann -function.5 In this case we have (3) which can only be
evaluated by summing, term-by-term, to any desired accuracy. In this case (3) = 1.20206 . . ..
Therefore the photon number density is
n
2 (3)
=
2
3
V
( hc)

14.3.6

(14.51)

Wiens law

In stellar astronomy two revealing parameters are a stars surface temperature and its luminosity.
The surface temperature is found from its color, which corresponds to the frequency peak in the
spectral density curve of Eq.14.39. These two pieces of information determine the stars location
on the empirical but important Hertzsprung-Russel diagram [which you can read about in any
introductory astronomy text.] From this it is possible to accurately determine the chemistry of the
star, its age and its stage of evolution.
The spectral peak is determined from Eq.14.39 by differentiation, i.e.
W (h) =
5

(z) =

n=1

1
nz

3
d
( h
)=0
d e
1

(14.52)

. For even integer values of z (but not for odd values) the -function can be found in closed form.

In general, is complex.

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

207

which gives
W (h) = 3eh 3 h eh = 0 .

(14.53)

This can be solved graphically [See Fig 14.2]

Figure 14.2: Graphical determination of Wiens Law constant.


to find
hmax = 2.82 kB T .

(14.54)

The frequency max at which the peak of Plancks radiation curve is located is proportional to the
absolute temperature of the radiating source. This is called Wiens Law.

14.3.7

Entropy of thermal radiation

The entropy of thermal radiation is


SEM = kB

nk, =0,1,2,...

P (nk, ) ln P (nk, )

(14.55)

with P (nk, ) as found in Eq.14.9. This is equivalent to

ln ZEM

FEM
)
= kB 2 (
V

SEM = kB 2

(14.56)
(14.57)

208

CHAPTER 14. THERMODYNAMICS OF RADIATION

where ZEM is the partition function [see Eq,14.10] and FEM is the electromagnetic Helmholtz
potential [see, e.g. Eq.14.19]. Completing the full radiation field calculation of Eq.14.19 requires
summing over all modes. Substituting x = h into that equation and ignoring the vacuum field
contribution (it does not depend on ), the Helmholtz potential is the integral
FEM =

8V
3
4 (hc)

2
x
dx x ln (1 e ) .

(14.58)

Integrating by parts gives


FEM =

8V
3 4 (hc)

dx
0

x3
(ex 1)

V 2
3 c3 4
45h

(14.59)
(14.60)

where the integral Eq.14.36 has been used. Finally, applying Eq.14.56, the entropy is
SEM =

4 2 kB V
3
45 ( hc)

(14.61)

which is proportional to T 3 .

14.3.8

Stephan-Boltzmann radiation law

An object at temperature T radiates electromagnetic energy from its surface [see Fig 14.3]. The
energy radiated per unit area per unit time (energy current density) is the Poynting Vector S, which
in the classical theory is
Sclassical =

1
dx E B .
8

(14.62)

The Poynting vector for photons k, is

Sk, = V 1 hk nk, c k

(14.63)

is the unit vector in the direction of propagation of the mode k, and


where k
1
nk, =

ehk

(14.64)

[See Eq.14.17].
The differential radiation flux density dk, of the mode k with polarization from the element
of area dA [see Fig 14.3] is
dk, = Sk, dA

(14.65)

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

209

Figure 14.3: Photon k leaving the surface of a radiating body.


with the total radiation flux differential d
n
d = V 1 hk nk, c k
dA

(14.66)

k
=1,2

where n
is the unit vector normal to the surface element dA [see Fig 14.3]. The total radiation
flux density per unit area is therefore
d
n
= V 1 hk nk, c k

dA
k

(14.67)

=1,2

n
With k
= cos the total outward (0 < < /2) radiation flux density per unit area is
2

/2

d
2
chk 2
=
d d sin cos dk h
3
dA (2)
e
1
0
0
0

(14.68)

which is integrated to give


d 2 (kB T )
=
3 c2
dA
60h

(14.69)

or
d
= B T 4 ,
dA

(14.70)

where
2 kB 4
3 c2
60h
= 5.67 108 W atts m2 K 4

B =

is called the Stefan-Boltzmann constant.

(14.71)
(14.72)

210
14.3.8.1

CHAPTER 14. THERMODYNAMICS OF RADIATION


Example:

Surface Temperature of Exoplanet Gliese 481g.


Gliese 581, an M-class red dwarf star, has an approximate radius of 2 105 km and an effective
surface temperature about 3200K. Its most recently discovered exoplanet, designated Gliese 581g,
orbits 2.18107 km from the star. Assuming the star and the exoplanet behave as perfect absorbers
and radiators estimate the surface temperature of Gliese 581g.

fraction intercepted by planetary disk

4R2 B T4


rp2
{
}
2
4Rp


energy rate radiated from star

4rp2 B Tp4

(14.73)

(14.74)

energy rate reradiated from planet

where rp is the exoplanet radius, Tp is the exoplanet temperature, R is the star radius and Rp
is the exoplanets orbital radius.
After simplification we have
Tp = (

R2
)
2
4Rp

1/4

T .

(14.75)

In particular, for Gliese 581g


Tg =217 K ,

(14.76)

not too far off from the Goldilocks zone of supporting life.

14.3.9

Radiation momentum density

The momentum p of a photon (a massless particle) is related, by special relativity, to its energy E
by
p=

E
.
c

(14.77)

Therefore the radiated momentum flux density per unit area is


d
n
= V 1 hk nk, k

dA
k

(14.78)

=1,2

which is evaluated, as in Eqs.14.67-14.69, to give


d 2 (kB T )
=
3 c3
dA
60h

(14.79)

14.3. THERMODYNAMICS OF ELECTROMAGNETISM


14.3.9.1

211

Example: Star Dust Blow-out

When stars are formed by gravitational collapse they are initially embedded in cocoons of tiny
silicate dust particles, a dust so dense that at this stage the star can be detected only from secondary
infrared radiation emitted by the heated dust. When the newly formed star reaches a sufficiently
high temperature the radiation from the star begins blowing out the dust cocoon allowing the
star to be seen in visible light.
Assuming that the radiation absorption cross-section of a dust particle is approximately 1/2 of the
physical cross-section and the dust particles have radii a < 107 m, at what stellar temperature will
all cocoon dust particles be blown away from the star?
The total momentum density per unit time radiated by the star is
4

2 (kB T )
2
3 c3 4R
60h

(14.80)

where T is the stars temperature and R is its radius. Therefore the momentum density per unit
time intercepted by a dust particle of radius a at a distance Rdust from the star is
4

dust =

a2
2 (kB T )
2
).

4R

2
3 c3
60h
4Rdust

(14.81)

Taking into account the estimated 50% absorption cross-section, the momentum density per unit
time absorbed by a dust particle is
4

A
dust
=

2 (kB T )
a2
) (1/2) .
4R2 (
2
3
3

60h c
4Rdust

(14.82)

On the other hand, the force of gravity acting on the particle is, by Newtons Law of gravitation,
Fgrav =

GM
4a3

(
dust )
2
Rdust
3

(14.83)

where G is the gravitational constant and dust = 3 103 kg m3 is the approximate density of
silicate dust.
For the radiation-momentum absorbing dust particle to be in gravitational equilibrium
4

2 (kB T )
a2
GM
4a3
4R2 (
) (1/2) = 2
(
dust ) ,
2
3
3

60h c
4Rdust
Rdust
3

(14.84)

where M is the mass of the star, so that


T4 =

8a cGM dust

3
B R2

where B is the Stefan-Boltzmann constant [see Eq.14.71].

(14.85)

212

CHAPTER 14. THERMODYNAMICS OF RADIATION

Assuming, for numerical specificity, that the newly formed star is sun-like with 1 solar mass and 1
solar radius, i.e.
M = 1.99 1030 kg
8

G = 6.67

(14.86)

R = 6.95 10 m

(14.87)

11

(14.88)

N m kg

the dust blow-out temperature is T 6000K , independent of the distance between the star and
the dust particles, Rdust .

Problems and exercises


1. Show that the work done by photons during an isentropic expansion is
W =(

2
4
3
3 c3 ) kB (Ti Tf ) Vi Ti
15h

(14.89)

where the subscripts i and f refer to initial and final states.


2. The total energy radiated by the sun and received per unit time per unit area at the earths
orbital distance is KS , the solar constant, which has the value KS = 0.136Js 1cm 2.
The mean earth-sun distance is RES = 1.5 1011 m and the radius of the sun is approximately
R = 7 108 m.
Find the effective black body temperature of the sun.
3. Show that the average square fluctuations of blackbody internal energy,
2

(U) = (U U)

(14.90)

is equal to
2

(U) =

2
ln Z .
2

(14.91)

4. Gliese 581 is an M-class red dwarf star which lies in the constellation Libra. It has an approximate radius of 2105 km and an effective surface temperature about 3200K. Exoplanets,
have recently been discovered orbiting Gliese 581. Two of the earlier discoveries have been
designated Gliese 581c (2007) and Gliese 581d (2007). These orbit at 1.092 107 km and
3.27 107 km, respectively, from the host star. Assuming the star and the exoplanets behave
as perfect absorbers and radiators:
(a) Estimate the peak frequency at which the star radiates energy.
(b) Estimate the surface temperatures of Gliese 581c and Gliese 581d.

14.3. THERMODYNAMICS OF ELECTROMAGNETISM

213

5. Consider a volume V containing thermal equilibrium electromagnetic radiation at a temperature T . Given that the partition function for electromagnetic radiation is (neglecting
zero-point energy)

Z = {(1 e hk ) }

(14.92)

with the product over all modes k and counting both polarizations, evaluate Eq.14.91 and
show that the average square fluctuations in electromagnetic energy is
2

(U) =

4
5
3 c3 (kB T ) ,
15h

(14.93)

6. find find the average square fluctuations in blackbody radiation particle number. Hint: You
may need the integrals

3
dxx (
0

1
4
)=
x
e 1
15

dx
0

x4 ex
2

(ex 1)

4 4
15

(14.94)

7. The ideal Carnot engine operates reversibly between an energy source, in which the working
substance (usually an ideal gas) absorbs heat isothermally at high temperature, and an energy
sink, to which it exhausts heat isothermally at a low temperature. The remaining curves which
close the Carnot cycle are adiabats. The efficiency of such an engine is defined as
=

Wr
QH

(14.95)

where Wr is the work done in the cycle and QH is the heat absorbed from the high temperature
energy source.
If light is used as the working substance find the Carnot engine efficiency.
8. When nucleons collide at very high energy, say by shooting a proton beam into a metallic
target, they annihilate and produce an amount of energy W mainly in the form of short
lived subatomic pions (quark-antiquark pairs) which can be positively charged, + , negatively
charged or neutral 0 . The + and are particle-antiparticle pairs while the 0 is its
own antiparticle. The pions have ultra-relativistic energies so their rest masses are negligible,
i.e. they behave like photons with energies
k
E = hc

(14.96)

with periodic boundary conditions


2

k = k x 2 + k y 2 + k z 2

(14.97)

where
kj =

2
j ;
L

j = 0, 1, 2, . . .

(14.98)

But whereas photons have 2 polarizations (1 and 2 ), pions have three ( + , 0 , ). Assuming

214

CHAPTER 14. THERMODYNAMICS OF RADIATION


i. The pions have ZERO chemical potential,
ii. Thermal equilibrium is rapidly reached (times shorter than the pion lifetimes.)
iii. All the collision energy is produced within a volume V ,
a. Find an expression, in terms of W and V , for the temperature of the pion spark.
b. Find an expression in terms of W and V for the mean number of pions produced in the
collision.

Why is it that particles with half-integer spin are Fermi particles


whereas particles with integer spin are Bose particles? An explanation
has been worked out by Pauli from complicated arguments from
quantum field theory and relativity. He has shown that the two must
necessarily go together .... but we have not been able to reproduce his
arguments on an elementary level. This probably means we do not have
a complete understanding of the fundamental principle involved...
R. P. Feynman, R.B. Leighton and M. Sands, Feynman Lectures on
Physics, Volume 3, Chapter 4 Section 1, Addison-Wesley, Reading, MA
(1963).

Chapter 15

Ideal Fermi Gas


15.1

Introduction

Particles with 12 -integer angular momentum obey the Pauli Exclusion Principle (PEP) a restriction that no more than one such identical particle can occupy the same quantum state, i.e. a single
particle quantum state can have 12 -integer spin occupation number of only 0 or 1. This restriction
was announced by W. Pauli in 1924 for which, in 1945, he received the Nobel Prize in Physics. Soon
after Pauli, the exclusion principle was generalized by P. Dirac and E. Fermi who independently
integrated it into quantum mechanics. As a consequence 21 -integer spin particles are called fermidirac particles or fermions. PEP applies to electrons, protons, neutrons, neutrinos, quarks and
their antiparticles as well as composite fermions, such as He3 atoms. Thermodynamic properties
of metals and semiconductors are largely determined by electron behavior. Metals, for example, exhibit the uniquely fermionic low temperature heat capacity CV T , while some fermi-dirac systems
undergo a transition to a distinctive superconducting state with signature properties of magnetic
flux exclusion (no interior B field) and zero electrical resistivity.

15.2

Ideal gas eigen-energies

The eigen-energies and eigenfunctions of any ideal (non-interacting) gas are constructed from the
eigenvalue problem
hop s = s s
215

(15.1)

216

CHAPTER 15. IDEAL FERMI GAS

with s the single particle eigen-energies and


s the corresponding single particle eigenstates. hop
is the single particle hamiltonian operator for the ith particle1
hop (i) =

2
h
2 + V (i) .
2m i

(15.2)

Generally V (i) represents some charge neutralizing positive background potential, e.g. positive
ions, so obtaining a solution to the single particle problem, i.e. Eqs.15.1 and 15.2, may be computationally intensive. It is, however, common and useful to assume that the background ions are
replaced by a uniform positive charge. This is called the free electron approximation. Presuming
a solution to this reduced problem is obtained, we can proceed to investigate fermi-dirac system
thermodynamics.

15.2.1

Grand partition function

A many-particle (open system) thermodynamic hamiltonian is


op = hop Nop
H

(15.3)

op is a sum of single particle hamiltonians


where the many-particle hamiltonian H
op = {hop (i) Nop (i)} .
H

(15.4)

i=1

op are
The macroscopic eigen-energies of H
(n1 , n2 , . . . ) = (s ) ns
E

(15.5)

where ns = 0, 1, 2 . . . are number of particles occupying the single particle eigenstate


s and is
the system chemical potential. The sum over s is a sum over all states.2
However, temporarily ignoring spin, each (non-degenerate) single particle state is restricted by PEP
to have occupation number zero (no particles) or one (one particle), i.e. ns = 0 or 1. Constructing
a Thermal Lagrangian 3 L F D

L F D = kB

n1 ,n2 ,...= 0,1

1
T

F D (n1 , n2 , . . . ) ln P
F D (n1 , n2 , . . . )
P

n1 ,n2 ,...= 0,1

n1 ,n2 ,...= 0,1

1 This

(15.6)

F D (n1 , n2 , . . . ) (s )ns


P
s

F D (n1 , n2 , . . . )
P

applies to particles of either integer or half-integer spin.


single particle state can be pictured as an open system into and out of which particles diffuse so that
in thermodynamic (diffusive) equilibrium all particles have the same chemical potential.
3 See Appendix B.
2 Each

15.2. IDEAL GAS EIGEN-ENERGIES

217

where 0 insures normalization.


F D (n1 , n2 , . . . ) is the probability there are n1 particles in the single particle state
P
1 , n2 particles
in the single particle state
2 , etc. and where the occupation numbers are restricted by the PEP,
so that the sums mean
.

(15.7)

n1 =0,1 n2 =0,1

n1 ,n2 ...=0,1

F D (n1 , n2 , . . . ) we find the probabilities


Maximizing LF D with respect to P
e

F D (n1 , n2 , . . . ) =
P

(s )ns
s

(s )ns

(15.8)

n1 ,n2 ,...= 0,1

F D
The denominator in Eq.15.8 is identified as the grand partition function, Zgr
F D
Zgr
=

(s )ns

(15.9)

n1 ,n2 ,...= 0,1

It is evaluated by rearranging the sum over states s in the exponential to give


F D
Zgr
= e(s ) ns

(15.10)

F D
Zgr
= [1 + e(s ) ] .

(15.11)

s ns =0,1

and summing ns = 0, 1

15.2.2

Electron spin

Taking electron spin into account, a pair of single particle states,


s, ,
s, emerges with corresponding eigen-energies s, and s, . Assuming them initially to be non-degenerate
F D
Zgr
= [1 + e(s, ) ] [1 + e(s, ) ] .

(15.12)

However in zero magnetic field a two-fold degeneracy arises with s, = s, = s producing the
partition function
F D
Zgr
= [1 + e(s ) ]
s

(15.13)

218

CHAPTER 15. IDEAL FERMI GAS

15.2.3

Fermi-Dirac thermodynamics

D
Thermodynamic properties of Fermi-Dirac systems are determined from F
gr , the grand potential

1
F D
ln Zgr
,

(15.14)

2
(s )
]
ln [1 + e
s

(15.15)

D
F
=
gr

which, with Eq.15.13, is


D
F
=
gr

where the sum is over all states


s .
s

1. The average particle number Nop follows from Eq.15.8,


2 [ ns e(s )ns ]
s

Nop =

ns =0,1

[
s

ns =0,1

(15.16)

e(s )ns ]

the factor 2 accounting for spin degeneracy. An equivalent result is


Nop = (

D
F
gr

= 2[
s

(15.17)

)
T,V

1
]
e(s ) + 1

= 2 nF
s

(15.18)
(15.19)

s
D
where nF
is a single particle average occupation number for the state
s ,
s

D
nF
=[
s

1
],
e(s ) + 1

(15.20)

which is called the fermi-dirac function.


Using Eq.15.18, the chemical potential can be calculated as a function of Nop /V and temperature T . [See examples below.] In free electron models the average fermion (electron)
density Nop /V [see Table 15.1] is a parameter that distinguishes among different materials.4
2. The F-D internal energy UF D is found from the general result of Eq.13.16
UF D = [
4 At

D
(F
+ Nop .
gr )]

V,

T = 0 the electron chemical potential is referred to as the Fermi energy, F : i.e., (T = 0) F .

(15.21)

15.2. IDEAL GAS EIGEN-ENERGIES

219
Nop
V

Element
Cu
Ag
Au
Al
Be
Sn
Mg
Fe
Pb
Hg

8.47
5.86
5.90
18.1
24.7
14.8
8.61
17.0
13.2
8.65

Table 15.1: Average electron density Nop /V for several elemental metals. Densities are in units
of 1028 m3 .

Figure 15.1: The Fermi-Dirac function in Eq.15.20. The solid line represents ns at low temperature. The short dashed curve is representative of higher temperatures. With increasing temperature
ns exp (Es ) the semi-classical limit. At T = 0 ns is a step function with unit jump at
Es = EF .

which, using Eq.15.15, yields the intuitively apparent


D
UF D = 2 s nF
.
s
s

(15.22)

220

CHAPTER 15. IDEAL FERMI GAS

3. The equation-of-state of the Fermi-Dirac gas is


D
pV = F
gr
2
D
= ln (1 nF
) .
s
s

(15.23)
(15.24)

4. The entropy S, as discussed in Chapter 6, is


S = kB

n1 ,n2 ,...=0,1

F D (n1 , n2 , . . . ) ln P
F D (n1 , n2 , . . . )
P

(15.25)

F D (n1 , n2 , . . . ) is as in Eq.15.8. Equivalently,


where P
S = kB 2 [
=

15.2.4

D
F
gr

(15.26)

,V
D
D
D
D
2kB [(1 nF
) ln (1 nF
) + nF
ln nF
]
s
s
s
s
s

(15.27)

Independent fermion model

The independent Fermion approximation for metals usually includes interactions with a static
background of positive ions (band structure approximation) but neglects both electron-electron
repulsion (correlations) and interactions with vibrating ions (electron-phonon interactions.)5
D
Independent fermion thermodynamics begins with F
[see Eq.15.15] where, using Diracs -function6
gr

df () ( s ) = f (s ) .

(15.28)

D
F
may be strategically rewritten as
gr

D
F
=
gr

2
()
].
d ( s ) ln [1 + e

(15.29)

Identifying the density of single particle states D () [see Appendix E]7


D () = ( s )

(15.30)

s
5 These effects can often be summarized for individual systems by a few parameters, such as an effective mass m
and an electron-phonon coupling constant . The Fermi gas model has been successful in describing many metals
and semiconductors.
6 (  ) is the Dirac delta-function.
s
7 Density of single particle states defined here does not include a factor 2 for spin. See Eq.E.21.

15.2. IDEAL GAS EIGEN-ENERGIES

221

D
F
becomes
gr

D
F
=
gr

2
()
].
dD () ln [1 + e

(15.31)

Replacing eigen-energies s with a density of states enables some experimental results to be used
in assessing electronic structure calculations.8
Inserting the f unction in an identical way, Eqs.15.18 and 15.22 become

Nop = 2 d
0

D ()
e() + 1

(15.32)

D ()
e() + 1

(15.33)

and

UF D = 2 d
0

Evaluating, or even approximating, integrals Eqs.15.32, 15.33 and related types that appear in
degenerate fermion models ( 1) requires specialized techniques. The simplest of these is
Sommerfelds asymptotic approximation which is discussed in Appendix I.9,10

15.2.5

The chemical potential (T 0)

Thermodynamic properties of a degenerate fermi gas ( >> 1) can be expressed as expansions11


N
in T . In particular, the chemical potential is determined from the density Vop by expanding the
right hand side of Eq.15.32 about = F to linear order, where F (Fermi energy) is the symbol
assigned to the chemical potential at T = 0, to give

Nop = 2 d
0
8 Approximate

D ()
( F )
2
+
dD () sech [ ( F )]
(
)
F
2
2
e
+1

(15.34)

densities of states can be obtained from optical or tunneling experiments.


asymptotic approximation is not an expansion in the usual Taylor series sense. It takes advantage
D approaches the unit step function at E [see
of the property that for T 0 the Fermi-Dirac function nF
f
s
fig. 15.1].
10 With increasing temperature, << 1, the steepness of the step declines over a width k T eventually taking
B
1
the form of a smooth classical exponential ()
e() .
e
+1
11 These expansions are not the usual convergent Taylor series type but belong to a special category called asymptotic expansions. These, generally, must be handled with care.
9 Sommerfelds

222

CHAPTER 15. IDEAL FERMI GAS

The first integral on the right-hand-side can be approximated by Sommerfelds method [see Appendix I] to give

F

2
D ()
dD () + 2 D (F )
d (F )
6
e
+1

(15.35)

while the second integral is approximated by

2
dD () sech [ ( F )] D (F ) + 2 D (F )
4
2
6

(15.36)

where D () has been similarly expanded about = F . Combining Eqs.15.35 and 15.36 with
Eq.15.34, while noting that
F

2 dD () = Nop

(15.37)

gives,
2 D ( )
F
6 2
= F
2
D (F ) + 2 D (F )
6

(15.38)

which to order T 2 is

= F

2 D (F )
.
6 2 D (F )

(15.39)

In the degenerate regime (where this expansion applies) chemical potential decreases with increasing
temperature. At very high temperature (where this expansion does not apply) .

15.2.6

Internal energy (T 0)

Similar to the argument above, the internal energy, [Eq.15.33], is first expanded about = F

D ()
( F )
D ()
2
d ( )
+
d ()
d D () sech [ ( F )] .
F +1
4
2
e
+1
e
0

(15.40)

Applying Sommerfelds approximation to the first term on the right-hand-side

d
0

F

D ()
2
d
d D () + 2 { [ D ( )]}
(
)
F +1
6 d
e
=F
0

(15.41)

15.2. IDEAL GAS EIGEN-ENERGIES

223

while the second term becomes

2
d D () sech [
0

4
( F )] F D (F ) .
2

(15.42)

2
D (F )
3 2

(15.43)

Combining Eqs.15.40, 15.41, 15.42 with 15.39


F

U = 2 d D () +
0

from which the constant volume heat capacity


CV = (

U
)
T V

is
CV =

2 2 2
k D (F ) T ,
3 B

(15.44)

quite different from both the constant valued classical result and the T 3 phonon contribution.12
This particular form follows because PEP allows only an effective number of particles near F , i.e.
Nef f N kB T , to participate in low temperature thermal processes.
Macroscopic heat capacity is proportional to D (F ) the density of states of the fermi system
evaluated at the fermi energy, a microscopic quantum property.

15.2.7

Pauli paramagnetic susceptibility

In the presence of an external magnetic field B0 , the spin degeneracy of conduction electrons is
lifted (Zeeman splitting) and the single particle states in Eq.15.48 become
s, s +

(15.45)

s, s

(15.46)

=m B0

(15.47)

where

with m the electron magnetic moment. In that case


F D
Zgr
= [1 + e(s +) ] [1 + e(s ) ] .
s
12 The

single particle density of states defined in this work does not include the factor 2 for spin.

(15.48)

224

CHAPTER 15. IDEAL FERMI GAS

and
D
F
=
gr

1
(s +)
] + ln [1 + e(s ) ]}
{ln [1 + e
s

(15.49)

where the sum is over all states


s . Using the single particle density of states D (), the sums
s

are replaced by integrals13


D
F
=
gr

dD () ln [1 + e(+) ] + dD () ln [1 + e() ] .

(15.50)

For weak fields, expansions in produce

1
ln [1 + e(+) ] + ln [1 + e() ] 2 ln [1 + e() ] + 2 2 sech2 [ ( )]
4
2

(15.51)

so that keeping only the term of order 2


D
F
gr

2 2

dD () sech2 [ ( )] .

4
2

(15.52)

The integral is done as discussed in Appendix I, giving in the degenerate limit ( >> 1)
D
F
2 D (F )
gr

(15.53)

where from Eq.15.39 only the leading term is used. With


M = (

F D )
B0 gr
T,

(15.54)

= 2m2 B0 D (F )

(15.55)

M = 2m2 D (F ) .

(15.56)

the Pauli susceptibility, M = M /B0 , is14

15.2.8

Electron gas model

The electron gas model is used to describe the behavior of simple metals. It assumes a macroscopic
collection of non-interacting spin particles, neutralized only by a uniform positive background,
with a single particle hamiltonian for the ith particle
hop (i) =
13 The
14 If

2
h
2 .
2m i

(15.57)

reason for not having incorporated the spin factor 2 into the density of states should now be apparent.
the density of states is defined to include the factor 2 for electron spin, as some authors prefer, M = m2 D (F )

15.2. IDEAL GAS EIGEN-ENERGIES

225

Choosing periodic boundary conditions [see Appendix G], the single particle eigen-energies are
 (k) =

2 k2
h
2m

(15.58)

where kj (j = x, y, z) are quantum numbers which assume the values


kj =

2
j
L

(15.59)

where
j = 0, 1, 2, 3, . . . .

(15.60)

Here L is the length of one side of a notional periodic cube with macroscopic volume V = L3 .
From Appendix E the density of states, D (), corresponding to this model [see Eq.15.58] is
D () = [  (k)]
k

V
2m 3/2
,
( 2 )
2
4
h

(15.61)

which is the distinctive 3 dimensional electron gas result.15


With Eq.15.61 the grand potential [see Eq.15.31] is
D
F
gr

V
2m 3/2
= 2 ( 2 ) d ln [1 + e() ]
2 h

(15.62)

which is integrated by parts to give

D
F
=
gr

3/2
V
2m 3/2
.
( 2 ) d ()
2
h
3
e
+1

(15.63)

Since
p = (

gr
)
V T,

(15.64)

the fermion gas pressure is


F D

1 2m 3/2
3/2
= 2 ( 2 ) d ()
.
3
h
e
+1

(15.65)

15 The

mass can be replaced by an effective mass, m m , where m approximates the effect of a crystal lattice.

226

CHAPTER 15. IDEAL FERMI GAS

Figure 15.2: Density of states for 3 D electron gas model. [See Eq.15.61].
In the degenerate case ( >> 1), Eq.15.65 is expanded in powers of T by applying the steps in Sec.
15.2.3.2, starting with an expansion of the integrand about = F ,

d
0

3/2
e() + 1

= d
0

3/2
e(F ) + 1

( F )
3/2
2
d sech [ ( F )] .
4
2

(15.66)

The first integral on the right-hand-side is approximated using Sommerfelds method

d
0

3/2

2 5/2 2 1/2

 + 2 F
4
e(F ) + 1 5 F

(15.67)

while the second integral is

3/2
2
d sech [
0

4 3/2
( F )] F .
2

(15.68)

Therefore Eq.15.66, together with Eqs.15.67, 15.68 and 15.39, gives the pressure of a degenerate
( >> 1) Femi-Dirac gas model
pF D =

2
2m 3/2 5/2
5 2
(
)

(1
+
).
F
2
15 2 h
12 2 2F

(15.69)

Unlike a classical ideal gas for which p T , the pressure of a degenerate Fermi-Dirac gas is nearly
independent of temperature.
Finally, in the Fermi gas model we have [see Eq.15.37]
F

Nop
1 2m 3/2
= 2 ( 2 ) d
V
2
h

(15.70)

1 2m 3/2 3/2
= 2 ( 2 ) F
3
h

(15.71)

15.2. IDEAL GAS EIGEN-ENERGIES

227

or
2/3
2
h
2 Nop
F =
(3
)
2m
V

(15.72)

permitting thermodynamic properties to be expressed in terms of the fermion density [e.g. Table15.1].

15.2.9

White Dwarf Stars

The thermodynamics of stars is a study in self-gravitating systems a titanic struggle between


gravity, which compresses stellar matter, and outward gas pressure usually associated with high
temperatures within stellar interiors.16 As a result of nuclear and chemical evolution, equilibrium
between these forces is in constant flux, resulting in continuous stellar transformations.
Stars that are not too massive, say on the order of our suns mass, eventually burn nearly all their
core hydrogen resulting in a cooler stellar core richer in helium (and heavier nuclei.) Consequently
gas pressure is reduced which, seemingly, throws the dynamical advantage to gravitational forces,
suggesting the stars continued collapse, perhaps without limit. Although ordinary gas pressure
is no longer capable of resisting gravitational collapse, star stability can now be maintained by
quantum behavior of degenerate (but not necessarily relativistic) electrons, whose source is the now
fully ionized He atoms.
These stars can collapse to radii the order of the earths while maintaining masses on the order of
the sun a density of 106 times that of water. Such stars were discovered by British astronomer
William Herschel (1783) who failed to realize their significance. These relatively nearby (on a
galactic scale) but intrinsically dim stars are called White Dwarfs. They are resident corpses in the
evolutionary stellar graveyard for stars of about a solar mass and less.

15.2.9.1

White Dwarf Thermodynamics

A white dwarf can be modeled as an aggregate of He nuclei densely packed at the stars hot central
region. The He nuclei intermingle with a degenerate ( >> 1) electron gas whose origin is the
fully ionized He atoms.
The degenerate electron gas behavior is governed by the PEP, so using Eqs.15.69 and 15.72,
the nearly temperature independent outward pressure of the degenerate electron gas component
is
2/3

p =

(3 2 )
5

5/3
e
2
Nop

h
)(
) ,
me
V

(15.73)

16 High temperatures result from nuclear fusion at the stars center in which hydrogen burns to helium and,
eventually, heavier nuclei including carbon and oxygen.

228

CHAPTER 15. IDEAL FERMI GAS

where me is the electron mass and


approximately

e
Nop

is the electron number density. The stellar mass, M , is

e
M = 2Nop
mHe .

(15.74)

where mHe is the mass of a He nucleus.


The stars dense He core exerts inward gravitational pressure
pcore = (

core
gr
V

(15.75)

T,

where, for bosonic He nuclei,

core
gr

V
2mHe 3/2
3/2
= 2 ( 2 ) d (+W )
3
h
e
1

(15.76)

with W the mean field gravitational potential energy per He nucleus [see Chapter 13.] Assuming
a spherically distributed uniform mass
M
V
3M
=
,
4R3

(15.77)
(15.78)

where R is the star radius, we find, as in Section 13.2.2,


W = mHe

3GM
.
5R

(15.79)

where G is the universal gravitational constant. In the extreme non-degenerate (classical) limit
( << 1), where boson-fermion statistics are irrelevant, Eq.15.76 becomes

core
gr =

V
2mHe 3/2
( 2 ) d 3/2 e(+W )
2
3
h

(15.80)

V
=
(2 3 )

1/2

5/2

mHe 3/2
( 2 ) e(W ) .
h

(15.81)

Applying Eq.15.75 the inward nuclear core pressure is


core

1
=

1/2

(2 3 )

5/2

1/3
1 GM 2 mHe 3/2
2
mHe 3/2 (W )
( 2 ) e

( 2 ) ( 2 ) e(W ) .
h
3
5 V 1/3 h

(15.82)

Next, applying Eq.15.17 to Eq.15.81, the average number of He nuclei is


mHe 3/2
V
He
Nop
= ( 2 ) e(W ) .
2 h

(15.83)

15.2. IDEAL GAS EIGEN-ENERGIES

229

Therefore, substituting for the chemical potential , the nuclear core pressure becomes
pcore =

He
Nop

1 4 1/3 GM 2
( )
5 3
V 4/3

(15.84)

where the first term is the classical ideal He gas pressure, which is negligible for this cool star. The
total pressure i.e. the equation of state for the white dwarf star, is therefore
2/3

ptotal =

(3 2 )

5/3
e
2
Nop

h
1 4 1/3 GM 2
)(
) ( )
.
me
V
5 3
V 4/3

(15.85)

At mechanical equilibrium ptotal = 0, giving


(3 2 )
5

2/3

5/3
e
2
Nop

h
1 4 1/3 GM 2
(
) = ( )
)(
me
V
5 3
V 4/3

(15.86)

which, with
e
He
Nop
= 2Nop
(charge neutrality)
M
=
2mHe

(15.87)
(15.88)

yields the remarkable (non-relativistic) white dwarf mass-radius condition


[

2
34/3 2/3 h
] = RM 1/3 .
5/3
8G(mHe ) me

(15.89)

Problems and exercises


1. Eventually the sun will pass through a red giant phase and collapse into a white dwarf star
( 6 109 years). Neglecting the shedding of mass in the red dwarf stage, what fraction of its
present radius will the sun be?
2. Consider a white dwarf star that is in an advanced stage of evolution with nearly all its fully
degenerate electrons behaving ultra-relativistically
cp
h

(15.90)

k.
where c is the speed of light and p = h
(a) Find the single particle density of states D ().
(b) Find an expression for the stars pressure.
(c) Find a critical mass (in solar mass units) beyond which the star is no longer stable and
will gravitationally collapse into a neutron star or a black hole.

230

CHAPTER 15. IDEAL FERMI GAS

3. An uncharged degenerate, spin 1/2 fermi gas (say He3 ) at pressure p and temperature T is
confined to one half of a rigid insulated chamber, of total volume V . The other half of the
chamber, is empty (vacuum). The partition separating the two halves suddenly dissolves and
the fermi gas freely and adiabatically flows to occupy the entire chamber (free expansion.)
Find the change in temperature of the fermi gas.
4. A cylinder is separated into two compartments by a sliding frictionless piston. Each compartment is occupied by an ideal fermi gas, say A and B, which differ only in that A is composed of
particles of spin 1/2 while B is composed of particles of spin 3/2. Find the relative equilibrium
density of the two gases
a. At T = 0K
b. As T
5. A degenerate fermi gas of non-interacting electrons moves in a uniform positive background
confined to one dimension. (An example is the conducting polymer trans-polyacetylene.)
a. Find an expression for the single particle density of states D () of independent particles
in one dimension.
b. Find the leading non-zero term for the chemical potential of a one-dimensional conductor
as a function of temperature in the limit of high degeneracy >> 1.
6. In E. Fermis model for ultra-high energy particle collisions17 a microscopic fireball is created
consisting of:
(a) Ultra-relativistic nucleon-antinucleon pairs,
i. Protons (two up quarks and one down quark)
ii. Antiprotons (two up antiquarks and one down antiquark)
iii. Neutrons (one up quark and two down quarks)
iv. Antineutrons (one up antiquark and two down antiquarks)
(b) Ultra-relativistic pions , + and 0 ,
i. ; negatively charged (down quark and anti-up quark)
ii. + ; positively charged (up quark and anti-down quark)
iii. 0 ; uncharged and is its own antiparticle (up quark - anti-up quark down quark anti-down quark superposition.)
All nucleons are spin = 1/2 fermi-dirac particles. All pions are spin = 0 bose-einstein particles.
At these high energies particles have negligible rest masses, i.e. they behave like photons
(except for their statistics) with ZERO chemical potential and energies
k
E = hc
17 E.

Fermi,High energy nuclear events, Prog. Theoret. Phys. (Japan), 5, 570-583 (1950).

(15.91)

15.2. IDEAL GAS EIGEN-ENERGIES

231

with periodic boundary conditions


2

k = k x 2 + k y 2 + k z 2

(15.92)

where
kj =

2
j ;
L

j = 0, 1, 2, . . . .

(15.93)

Dynamic assumptions are:


i. Interaction between pions and nucleons is so strong that statistical equilibrium is
attained in times less than pion lifetimes.
ii. Energy and charge and are conserved.
iii. The energy available in the center of mass system, W , is released in a small volume

h
V = 43 R3 , where R is the compton wavelength R
.
c
iv. Angular momentum conservation is ignored.
a. Find an expression for the temperature T of the fireball in terms of W and V .
b. Find the average density of pions and of nucleons produced.
c. The fireballs brightness (total radiated energy per unit time.)
Consider two alternative scenarios:
(1) The fireball is entirely composed of ultra-relativistic fermions
(2) The fireball is entirely composed of photons
Show that the internal energy per unit volume for the fermion fireball is exactly 7/2 the
internal energy per unit volume for the photon (EM radiation) fireball.
Note: The four distinct fermion varieties give rise to eight-fold degeneracy, i.e. each of the
four created particles has two spin polarizations.
You may need the following integrals:

3
dxx (
0

1
7 4
)
=
ex + 1
120

3
dxx (
0

1
4
)
=
ex 1
15

(15.94)

232

CHAPTER 15. IDEAL FERMI GAS

From a certain temperature on, the molecules condense without


attractive forces; that is, they accumulate at zero velocity. The theory is
pretty, but is there also some truth to it?
Albert Einstein, Letter to Ehrenfest (Dec 1924). Abraham Pais, Subtle
Is the Lord: The Science and the Life of Albert Einstein, Oxford
University Press, New York 1982.

Chapter 16

Ideal Bose-Einstein System


16.1

Introduction

For over 50 years the low temperature liquid state of uncharged, spinless He4 was the only system
in which a Bose-Einstein condensation was considered experimentally realized.1,2,3 In that case,
cold (< 2.19K) liquid He4 passes into an extraordinary phase of matter called a superfluid, in
which the liquids viscosity and entropy become zero.
With advances in atomic cooling4,5 (to 109 K) number of bose systems which demonstrably pass
into a condensate has considerably increased. These include several isotopes of alkali gas atoms as
well as fermionic atoms that pair into integer spin (boson) composites.6
Although an ideal bose gas does exhibit a low temperature critical instability, the ideal B-E gas
theory is not, on its own, able to describe the B-E condensate wave state7 In order for a theory
of bosons to account for a condensate wave state, interactions between the bosons are included.
Nevertheless there is much interesting physics to be extracted from an ideal bose gas model.

Kapitza, Viscosity of Liquid Helium below the P oint, Nature 141 74 (1938).
F. Allen and Don Misener, Flow of Liquid Helium II, Nature 141 75 (1938).
3 Allan Griffin, New light on the intriguing history of superfluidity in liquid 4 He, J. Phys.: Condens. Matter
21 164220 (2009).
4 M.H. Anderson, J.R. Ensher, M.R. Matthews, C.E. Wieman and E.A. Cornell, Observation of BoseEinstein
Condensation in a Dilute Atomic Vapor, Science 269 198 (1995).
5 Eric A. Cornell and Carl E. Wieman, The Bose-Einstein Condensate, Scientific American 278 40 (1998).
6 C. A. Regal, M. Greiner and D. S. Jin, Observation of Resonance Condensation of Fermionic Atom Pairs,
Phys. Rev. Lett. 92 040403 (2004).
7 V.L. Ginzburg and L.D. Landau, Zh. Eksp. Teor. Fiz. 20, 1064 (1950).
1 Peter
2 John

233

234

16.2

CHAPTER 16. IDEAL BOSE-EINSTEIN SYSTEM

Ideal bose gas

The Bose-Einstein (B-E) system allows any occupation number, n = 0, 1, 2, 3, . . . , so that the
Thermal Lagrangian is

n1 =0

nN =0

BE (n1 , n2 , . . .) ln P
BE (n1 , n2 , . . .)
L BE = kB P

(16.1)

1
BE (n1 , n2 , . . .) [(Es ) ns ]}
{ P
T n1 =0 nN =0
ns =0

n1 =0

nN =0

BE (n1 , n2 , . . .)
0 P
1 ,
BE (n1 , n2 , . . . ) is the probability8 there are n1 particles in the single particle state E
where P

n2 particles in the single particle state E2 , etc. Maximizing the Thermal Lagrangian gives the
Bose-Einstein probabilities

BE (n1 , n2 , . . .) =
P

ns (Es )

n1 =0

ns (Es )

(16.2)

nN =0

The denominator is the B-E Grand Partition Function

n1 =0

nN =0

BE
Zgr
= e

ns (Es )
s

= ens (Es ) .

(16.3)
(16.4)

s ns =0

The infinite sum over ns converges only for e(Es ) < 1. Thus if the minimum value in the Es
energy spectrum is taken to be zero, convergence is guaranteed only if < 0. As will soon be shown,
this convergence criterion has astonishing physical implications!
Assuming convergence of the ns sum, the B-E Grand Partition Function is
BE
Zgr
= (
s

1
1 e(Es )

(16.5)

and the corresponding grand potential is

BE =

8 Surrogate

probability.

1
Es
).
ln (1 e e
s=0

(16.6)

16.2. IDEAL BOSE GAS

16.2.1

235

Bose-Einstein thermodynamics

Applying the probabilities from Eq.16.2:


1. Average number of particles in the pth single particle state is

np enp (Ep )

np BE =

np =0

enp (Ep )

(16.7)

np =0

or
np

BE

1
=

np (Ep )

ln e
.

np =0

(16.8)

After carrying out the sum and differentiation in Eq.16.8 the B-E average occupation number
for the single particle state Ep is

np BE =

1
.
e(Ep ) 1

(16.9)

Similarly, the average total particle number is


BE
Nop
= np BE

(16.10)

p=0

BE

gr

(16.11)

1
.
(Ep ) 1
e
p=0

(16.12)

But Eq.16.9 presents a problem. Since the lowest value of Ep is taken to be ZERO, if > 0
the average number of particles in state Ep is negative which is nonsense. It clearly must
be that
BE 0 .

(16.13)

This constraint (already mentioned above) has implications that will be discussed in the next
sections.
2. Internal Energy of the ideal B-E system is the intuitive result
U = Ep np BE
p=0

(16.14)

236
16.2.1.1

CHAPTER 16. IDEAL BOSE-EINSTEIN SYSTEM


The quasi-classical limit

The distinction between np BE and np F D i.e. the sign in the denominator has enormous
physical consequences at low temperature. But in the limit , i.e. the quasi-classical limit,
where n/nQ << 1, all distinctions disappear,
np BE np F D e eEp .

16.2.2

(16.15)

The ideal B-E gas and the B-E condensation

Using the free-particle eigenstates and density of states previously applied in Fermi-Dirac thermodynamics the spin = 0 (no spin degeneracy) B-E grand potential [see Eq.16.6] is

BE
gr =

2m 3/2
1 V
( 2 ) ( 2 ) d ln (1 e e )
4
h

(16.16)

with 0. Integrating by parts

3/2
2m 3/2
V
= 2 ( 2 ) d ()
6
h
e
1

BE
gr

(16.17)

and then applying Eq.16.11 gives

BE
Nop

1/2
2m 3/2
V
.
= 2 ( 2 ) d ()
4
h
e
1

(16.18)

Now recalling the Fermi-Dirac expression

FD
Nop

V
2m 3/2
1/2
= 2 ( 2 ) d ()
,
2
h
e
+1

(16.19)

and the quasi-classical expression

QC
Nop

V
2m 3/2
= 2 ( 2 ) d 1/2 e() ,
2
h

(16.20)

the variation of with T can, in each case, be plotted to show any distinctions between the
three kinds of statistics. At high temperature all three curves begin to merge as indicated in
Eq.16.15. As temperature falls the B-E, F-D and quasi-classical curves begin to diverge, as shown
in Figure 16.1, with F-D and quasi-classical curves rising above = 0. However the B-E curve
cannot rise above = 0 and since the curve is single valued, there is some temperature Tc given
by
(

BE
Nop

=
T Tc+

1 2m 3/2
1/2
( 2 ) d
,
2
4
h
e c 1
0

(16.21)

16.2. IDEAL BOSE GAS

237

Figure 16.1: B-E, F-D and quasi-classical result for vs. T .


1 , at which it continues for T < T as a horizontal line that coincides with = 0, (see
c
kB Tc
Figure 16.1). Although the B-E curve is continuous, its slope is discontinuous. At temperatures
T < Tc the chemical potential assumes the fixed value = 0 and the average B-E particle density
becomes
where c =

BE
Nop

)
T <Tc

1 2m 3/2
1/2
(
d
.
)

2
4 2 h
e 1

(16.22)

Comparing Eq.16.21 with Eq.16.22 we have the result that

BE
Nop

T <Tc
=0

<(

BE
Nop

T Tc
<0

(16.23)

Bosons are missing! Where are they? They have, in fact, thrown in their lot with an entirely
new state of matter having its own wavefunction the Bose Einstein condensate. For T < Tc ,
as T Tc , ordinary bosons begin to reappear. Thus for T < Tc there are two coexisting phases
normal bosons and a B-E condensate.
The transition temperature, Tc , at which the B-E condensate begins to form is found by integrating
Eq.16.21. A procedure is to first expand the denominator

dx
0

xs
= dxxs e(n+1)x
x
e 1
n=0
0

(16.24)

238

CHAPTER 16. IDEAL BOSE-EINSTEIN SYSTEM

and then integrate term-by-term

dx
0

1
xs
=

(s
+
1)

s+1
ex 1

=1

(16.25)

where the -function is defined by

(r) = dz z r1 ez

Re r > 1 .

(16.26)

The remaining sum is the Riemann -function

1
s
n=1 n

(16.27)

xs
= (s + 1) (s + 1)
1

(16.28)

(s) =

which finally gives

dx
0

ex

and therefore
(

BE
Nop

T Tc+

(3/2) (3/2) 2m 3/2


( 2 ) .
4 2
c h

(16.29)

Solving for Tc ,
2/3
2/3
BE
2
Nop

4 2
h
[
] (
)
Tc =
2mkB (3/2) (3/2)
V
T T +

(16.30)

where (3/2) = 0.886 and (3/2) = 2.612. Therefore

Tc = 3.31 (

2/3
BE
2
Nop

h
)(
) .
mkB
V

(16.31)

BE
For the case of 4 He where Nop
/V = 2.18 1022 cm3 we find Tc = 3.14 K. This may be compared
to the temperature T = 2.19 K (the point) at which liquid 4 He actually undergoes a transition
to the remarkable superfluid phase.

16.2. IDEAL BOSE GAS

16.2.3

239

The condensate

As noted above, at temperatures 0 < T < Tc the boson chemical potential remains = 0 generating
the missing boson paradox. From what we know about the electromagnetic fields photons as
well as elastic fields phonons whose chemical potential is inherently zero particles are created
and destroyed without regard to conservation. Similarly, at temperatures T < Tc Bose-Einstein
particles assume the property that they can be created and destroyed without regard to number
conservation. Here physical particles are freely disappearing into and emerging from a condensate
that no longer keeps track of particle number condensate particle number has large uncertainty
with the uncertainty relation ncos 1/2 imposing a trade-off. The condensate wavefunction
is a superposition of states with different particle number the condensate is, instead, a coherent
phase of matter.9 An emergent property of the condensate is classical phase. In analogy with laser
light (and superconductors) which surrender particle number certainty to acquire good phase
the condensate behaves like a classical macroscopic wave with definite (or nearly definite) phase.
Also in analogy with the photon vacuum, the condensate acts like a Bose-Einstein particle vacuum,
i.e. as temperature increases particles return from the condensate (vacuum) into the ideal B-E gas
phase and visa-versa.
The density of missing particles (the number subsumed into the condensate) is

BE
Nop

=(

superfluid

N
N
)
( )
.
V T >Tc
V T <Tc

(16.32)

Therefore integrating Eq.16.22 to get

BE
Nop

T <Tc

(3/2) (3/2) 2m 3/2


( 2 )
4 2
h

(16.33)

then taking its ratio with Eq.16.29 gives

BE
Nop

/(

BE
Nop

T <Tc

= (T /Tc )

3/2

(16.34)

T >Tc

The mystery of the missing bosons has been solved. Specifically,

BE
Nop

)
superfluid

=(

BE
Nop

[1 (
T >Tc

T 3/2
) ].
Tc

(16.35)

At T = 0 all the bosons are subsumed by the condensate, whereas at T = Tc the condensate is empty
(there is no condensate.) [See Figure 16.2]
A behavior such as Eq.16.35 occurs in certain phase transitions. [See e.g. Chapter 12]. The
exponent 3/2 in the temperature dependence is this phase transitions critical exponent.
9 L.

Susskind and J. Glogower, Quantum mechanical phase and time operator, Physica 1, 49-61 (1964).

240

CHAPTER 16. IDEAL BOSE-EINSTEIN SYSTEM

Figure 16.2: Fractional Condensate occupation (Eq.16.35) vs. T /Tc .

Problems and exercises


1. When an ideal Bose-Einstein gas passes through the critical temperature Tc the heat capacity
CV remains constant but its temperature derivative is discontinuous, with
(

dCV
dCV
) (
) = c .
dT Tc+
dT Tc

(16.36)

Find the value of .


2. Consider an ultra-relativistic Bose-Einstein gas with single particle eigen-energies
k .
 = hc

(16.37)

a. Show that the single particle density of states is


D () =

V
2
3 c3  .
2 2 h

(16.38)

b. Find the critical temperature Tc of the B.E. gas.


3. The volume of an n-dimensional sphere with radius k is
Vn =

2 n/2
kn
n (n/2)

(16.39)

16.2. IDEAL BOSE GAS

241

where
n

k 2 = ki2 .

(16.40)

i=1

(a) Assuming an n-dimensional Bose-Einstein gas with single particle eigenvalues


=

2 k2
h
2m

(16.41)

show that the single particle density of states is


D () =

n n/2 2m (n/2)1
( 2 ) 
.
(n/2) h

(b) Find the critical temperature Tc for an n-dimensional B.E. gas.


(c) Show that the 2-dimensional B.E. gas exhibits no Bose-Einstein condensation.

(16.42)

242

CHAPTER 16. IDEAL BOSE-EINSTEIN SYSTEM

We are quantifying the cosmos in a different way to open up a new


window for understanding the universe in its earliest times.
Researchers have long been working to test cosmological models for how
the universe developed, but previously have not been able to collect data
that could provide a picture this clear.
Lyman Page - Princeton University, WMAP co-investigator (2006).

Chapter 17

Thermodynamics and the CMB


17.1

Introduction

The once super hot but now cold afterglow from the creation of the universe has been studied by
the satellites COBE1 and WMAP2 . WMAPs continuing task, and that of the European Space
Agencys Planck satellite, is to measure small-scale temperature anisotropies in the cosmic microwave background (CMB) in order to better determine cosmological parameters which are a test
of inflationary scenarios.3
The satellite moves through the microwave background with a Local Group Velocity v = 627
22km sec1 in the galactic coordinate direction (l 276 3 , b 33 3 ) relative to the CMB rest
frame, so that WMAPs local velocity parameter is v
c = .002. Relativistic effects will therefore
produce measurable temperature variations that must be subtracted from the total signal.4 This
effect was first calculated exactly by Peebles and Wilkinson (PW)5 only a few years after the
discovery of the CMB.6
The intent in this chapter is not to re-derive the well known PW result, but is solely for the purpose of discussing relativity in blackbody thermodynamics by applying the Thermal Lagrangian
formulation7 and, by happenstance, derive the PW result.
To find thermodynamic properties measured in a frame of reference moving with velocity v through
a blackbody radiation rest frame (CMB), we start with a thermodynamic hamiltonian which con1 COsmic

Background Explorer.
Microwave Anisotropy Probe.
3 M. White, and J. D. Cohn, Resource Letter: TACMB-1: The theory of anisotropies in the cosmic microwave
background, Am. J. Phys. 70, 106-118 (2002).
4 M. Kamionkowski, and L. L. Knox, Aspects of the Cosmic Microwave Background Dipole, Phys. Rev. D 67,
63001-63005 (2003).
5 P. J. E. Peebles, and D. T. Wilkinson, Comment on the anisotropy of the primeval fireball, Phys. Rev. 174,
2168 (1968).
6 A. A. Penzias and R. W. Wilson, A Measurement of Excess Antenna Temperature at 4080 Mc/s, Astrophysical
Journal, 142, 419-421 (1965).
7 See Appendix B.
2 Wilkinson

243

244

CHAPTER 17. THERMODYNAMICS AND THE CMB

tains, in addition to the formal electromagnetic quantum hamiltonian, the energy (work) required
to transport the satellite through the radiation field,
k, [a ak, + 1/2] v pop ,
Hop = h
k,
k
=1,2

(17.1)

k
=1,2

where
ak, ,
pop = hka
k,

(17.2)

ak, ak, = Nop (k, )

(17.3)

and

is the photon number operator whose eigenvectors and eigenvalues are given by
ak, ak, nk, = nk, nk, .

(17.4)

Here k is the photon wave-vector, = 1, 2 covers the two transverse photon polarizations and
nk, = 0, 1, 2 . . . is the occupation number of photons with wave-vector k and polatization .
The eigen-energies of the thermodynamic hamiltonian (including the contribution from the satellites uniform velocity v relative to the CMB rest frame) are, from Eq.17.1,
k, [nk, + 1/2] v hkn
k, .
E (n) = h
(17.5)
k
=1,2

17.2

k
=1,2

Thermodynamic method

Using the eigen-energes of Eq.17.5, the moving frame Thermal Lagrangian is

L = kB P (nk, ) ln P (nk, )
nk,

P (nk, ) hk, [nk, + 1/2] v hknk,

nk,

k
k
=1,2

=1,2

0 P (nk, ) .

(17.6)

nk,

The P (nk, ) are probabilities that nk, photons are in the mode k with polarization . Here
is the temperature in the moving frame. Maximizing the Lagrangian of Eq.17.6 with respect to
P (nk, ) (following by now familiar steps) gives, in the moving frame, the probabilities

exp [hk, (nk, + 1/2) hnk, v k cos ]

=1,2

P (nk, ) =

exp [hk, (nk, + 1/2) hnk, v k cos ]

nk,
k

=1,2

(17.7)


17.3. MOVING FRAME THERMODYNAMIC POTENTIAL

245

where = (kB ) and is the angle (in the moving frame) between the local velocity vector
v and the direction of CMB observation. The frequency k, is also in the moving frame. The
denominator in Eq.17.7 is identified as the moving frame blackbody partition function.

17.3

Moving frame thermodynamic potential

The thermodynamic entropy S, is


S = kB P (nk, ) ln P (nk, )

(17.8)

nk,

which after substituting the probabilities of Eq.17.7 is evaluated to give


1
S = U v
pop + ln Z

(17.9)

where U is the moving frame blackbody internal energy,


pop is the moving frame average radiation
momentum and Z is the moving frame blackbody partition function. We can now define a moving
frame thermodynamic potential function
= 1 ln Z

(17.10)

with
2

n
k (k v k cos )]}
Z = { exp [h
k

(17.11)

nk

where squaring accounts for the two transverse photon polarizations. Combining Eqs.17.9 and
17.10 we can write
=U
S v

pop .

(17.12)

Moreover, by direct differentiation of Eq.17.8 we find a corresponding thermodynamic identity, i.e.


a moving frame fundamental equation
dS = dU + p dV v d
pop

(17.13)

where p is the moving frame radiation pressure. Combining this with the total differential of
Eq.17.12
=
d
p dV S d
pop dv

(17.14)

246

CHAPTER 17. THERMODYNAMICS AND THE CMB

from which follow the thermodynamic results,

S = (

)
,
V ,v

) ,
V ,v

pop = (
)
v ,V
p = (

17.3.1

(17.15)

Moving frame blackbody thermodynamics

The partition function of Eq.17.11 is evaluated by first summing over nk to get


2

Z =

k (1 v/c cos )]}


1 exp { [hc

(17.16)

1
where = c k has been used. In the CMB rest frame v = 0 and , where = (kB T0 ) with
T0 the CMB rest frame temperature.

The moving frame thermodynamic potential is


= 1 ln Z

(17.17)

2
k (1 v/c cos )]}} .
ln {1 exp { [hc
k

(17.18)

In the large volume limit this is expressed as the integral


=

V
2 2

2
k (1 v/c cos )]}
d d sin d k k ln {1 exp [hc
0

(17.19)

which, with the change in variable


(1 v/c cos )
z = hc
v/c sin d
dz = hc

(17.20)
(17.21)

followed by an integration by parts, is


2 4
V 4
= kB

.
3 c3
2 2
45h
(1 v/c )

(17.22)


17.3. MOVING FRAME THERMODYNAMIC POTENTIAL

247

Using this result with Eq.17.15, the moving frame radiation pressure is
p =

4
2 kB
4

.
3 c3
2 2
45h
(1 v/c )

(17.23)

Again applying Eq.17.15, we find an expression for the moving frame entropy
S=

2 k4
4
V 3
3B3
.
2 2
3 45h c
(1 v/c )

(17.24)

2 k4
4
3B3 V0 T03
3 45h c

(17.25)

However in the CMB rest frame


S0 =

where V0 is a rest frame volume and T0 is the rest frame temperature. But entropy is not altered
by changing reference frames8 , so that S = S0 . Since in the moving frame the rest frame volume is
Lorentz contracted

2
(17.26)
V = V0 1 v/c ,
equating Eqs.17.24 and 17.25 gives

1
2 3/2

3 = T03

(17.27)

[1 v/c ]
so that

2 1/2

= T0 [1 v/c ]

(17.28)

which is the Lorentz transformation for temperature.9


Completing this part of the discussion, an application of Eq.17.15 gives

pop =

4 2
4kB
v 4 V
2 3

(17.29)

3 c5 (1 v/c )
45h

Finally, with Eqs.17.29, 17.35 and 17.22, we get from Eq.17.12 the moving frame radiation field
internal energy U
2

4 (1 + 1 v/c )
V
2 kB
3

U = 3 3
4
15h c (1 v/c2 )3
8N

(17.30)

G Van Kampen, Relativistic Thermodynamics of Moving Systems, Phys. Rev. 173, 295 (1968).
Mi, H. Y. Zhong and D. M. Tong, Different proposals for relativistic temperature transforms: Whys and
wherefores, Mod. Phys. Let. 24 73-80 (2009).
9 D.

248

CHAPTER 17. THERMODYNAMICS AND THE CMB

or

U = U0

(1 + 13 v/c )
2 1/2

(17.31)

(1 v/c )
where
U0 =

4 4
2 kB
T0 V0

15h3 c3

(17.32)

is the rest frame radiation field energy.10 Finally, since entropy is Lorentz invariant, for quasi-static
heat transfer
dS =

d
Q d
Q
=

T0

(17.33)

so that
1/2

d
Q

17.4

v2
= dQ (1 2 )
c

(17.34)

Radiation energy flux

That part of the CMB anisotropy (frequency k, and velocity v at a particular angle with
respect to the Local Group Velocity) which arises solely from satellite motion with respect to the
microwave CMB rest frame, provides a base anisotropy to be subtracted11 from the microwave
signal in order to obtain the anisotropy of the 13.7 billion year old CMB fireball itself.
For this we write the Poynting vector S (, v) (energy per unit time per unit detector area) observed
from the angle with respect to the moving frame velocity v. For a single mode k and one
polarization
d Sk, (, v)
k v
k,
= (1/V ) ch
n (k, ; , v)

dA
k v

(17.35)

where Sk, (, v) is the radiation flux for the mode k with polarization , and
n (k, ; , v) is the
moving frame average photon occupation number. Using Eq.17.7 we find directly

n (k, ; , v) =
10 C

1
.

e h (k, v k cos ) 1

Mo
/ller, The Theory of Relativity, Clarendon Press, Oxford (1952).
with other known galactic radiation corrections

11 along

(17.36)

17.4. RADIATION ENERGY FLUX

249

which is a Bose-Einstein average occupation number. With k, = c k and in the large volume
limit the sum in Eq.17.35 including both polarizations becomes, per unit solid angle d, the
integral

3
d Sk, (, v)
2V
k
d k
) c2 h
(1/
V
=
.
3
k [1 (v /c) cos ]
hc
dA d

(2)
e
1
0

(17.37)

Integrating over all modes k for a particular radiation reception angle gives
d Sk, (, v)

=
3 (1 v/c cos )4 4
dA d
60c2 h

(17.38)

4 4
kB
T0 (1 v/c)
.
3 (1 v/c cos )4
60c2 h

(17.39)

In other words, the moving observers Planck radiation function is


1

(1 v/c cos )
1
u (, T0 ; , v) = exp

k
T
1

v/c

B 0

and the moving frame temperature for the admittance angle is

2
T0 1 v/c
(, v) =
(1 v/c cos )

(17.40)

(17.41)

which is the PW result.


The sensitivity of modern microwave satellite detectors assures that relativistic temperature anisotropy
effects will be observable up to several multipole orders. These have to be carefully subtracted in
order to extract the significant underlying cosmological information. Eq.17.41 is essential to such
analyses.

Problems and exercises


1. Find an expressions in terms of rest values H0 , F0 , G0 for:
(a) Moving frame blackbody Enthalpy H .
(b) Moving frame blackbody Helmholtz Potential F .
(c) Moving frame blackbody Gibbs Potential G.

250

CHAPTER 17. THERMODYNAMICS AND THE CMB

Appendix A

A.1

The inequality 0 < T r (op ) 1

For the general mixed state


2

T r (op ) = n i wi i j wj j n ,

(A.1)

i,j,n

which is rearranged to read


2

T r (op ) = wi i j wj j n n i .

(A.2)

i,j,n

Explicitly carrying out the trace operation (sum over n)


2

T r (op ) = wi i j wj j i .

(A.3)

i,j

Sharing the classical (real, positive) probabilities wk among the terms

T r (op ) = i wi j wj j wj i wi .

(A.4)

i,j

By the Schwartz inequality

i wi j wj j wj i wi
i,j

{ i wi i wi } j wj j wj

i
j

(A.5)

or
2

0 < T r (op ) (T rop ) = 1

251

(A.6)

252

APPENDIX A.

Appendix B

B.1

Properties of F

We seek a functional1 of probabilities F(P1 , P2 , P3 , . . . ) that serves as a mathematical measure of


bias (uncertainty). The possibilities for F can be considerably narrowed by imposing reasonable
properties:2
1. F(P1 , P2 , P3 , . . . ) is a continuous function of the Ps . With probabilities Ps normalized to unity,
incremental changes in any of the Ps produce only incremental changes in F(P1 , P2 , P3 , . . .).
2. F(P1 , P2 , P3 , . . .) is a symmetric function of its arguments. No change in ordering of the
arguments can change the value of the function.
3. F(P1 , P2 , P3 , . . .) is a minimum when all probabilities but one are zero (only one possible
outcome completely biased.)
4. F (P1 , P2 , P3 , . . .) is a maximum when all probabilities are equal (completely unbiased.)
5.
F (P1 , P2 , P3 , . . . , 0) = F (P1 , P2 , P3 , . . . )

(B.1)

If an event of zero probability is added to the set the value of the function remains unchanged.
6. For the case in which all probabilities Ps are equal
P1 = p, P2 = p, P3 = p, . . . , Pn = p ,

(B.2)

F (p, p, p, . . . , Pn = p) f (n) ,

(B.3)

a function f (n) is defined

that is a monotonically increasing function of n. Fewer distinct outcomes mean increased


bias.
abbreviation P (s ) Ps is used.
N. Kapur,Maximum Entropy Models in Science and Engineering, John Wiley and Sons, New York (1989).

1 The
2 J.

253

254

APPENDIX B.

7. There is an addition rule for combining bias functions of independent events. For example
consider two independent probability distributions corresponding to events P and Q,
P = {P1 , P2 , P3 , . . . , Pn }

(B.4)

Q = {Q1 , Q2 , Q3 , . . . , Qm }

(B.5)

with bias functions


FP (P1 , P2 , P3 , . . . , Pn )

(B.6)

FQ (Q1 , Q2 , Q3 , . . . , Qm ) .

(B.7)

Let there be a set of joint events


P Q = {P1 Q1 , P1 Q2 , . . . P2 Q1 , P2 Q2 , P2 Q3 . . . Pm Pn }

(B.8)

with bias function


FP Q (P1 Q1 , P1 Q2 , . . . P2 Q1 , P2 Q2 , P2 Q3 . . . Pm Pn ) .

(B.9)

The Addition rule is:


FP Q (P1 Q1 , P1 Q2 , . . . P2 Q1 , P2 Q2 , P2 Q3 . . . Pm Pn )
= FP (P1 , P2 , P3 , . . . , Pn ) + FQ (Q1 , Q2 , Q3 , . . . , Qm )

(B.10)

The addition rule excludes, for example, forms like purity I.


8. In the special case that all probabilities are equal, applying Eq.B.3 together with the addition
rule Eq.B.10 gives
f (n m) = f (n) + f (m)

(B.11)

and
f (my ) = yf (m) ;

B.1.1

y>0

(B.12)

The Bias Function

From Eq.B.12 we can guess that


f (m) = ln m

(B.13)

and since f (m) is a monotonically increasing function of m, is a positive constant. In particular,


since there are m events with equal probability p, that probability must be p = 1/m, so
F (p, p, p, . . .) = mp ln p

(B.14)

B.2. A THERMAL LAGRANGIAN

255

which, in the present restricted case p1 = p2 = p3 = . . . pm , may be written


m

F (p, p, p, . . .) = pr ln pr .

(B.15)

r=1

Relaxing this restriction to include all P the result becomes a general bias function
m

F (P1 , P2 , P3 , . . . Pm ) = Pr ln Pr

(B.16)

r=1

which is also called the Gibbs-Shannon entropy.

B.2

A Thermal Lagrangian

The method chosen for determining the best P (s ) is to minimize bias among the possible
outcomes
= (1 , 2 , 3 , . . . n )

(B.17)

by maximizing the bias functional


n

(i ) ln P
(i )
= P
F [P]

(B.18)

i=1

while simultaneously minimizing the average total energy


n

(r ) r
H = P

(B.19)

r=1

Here r are macroscopic eigenvalues of the hamiltonian,


with respect to variations in the P.
H = h0 + Y m xm

(B.20)

m=1

with from the fundamental equation Eq.5.8, Y m and xm conjugate intensive and extensive variables,
respectively. This is done while maintaining normalization
n

(r ) = 1
P

(B.21)

r=1

a functional combining bias and average energy and using a Lagrange multiplier
Forming [P],
to enforce normalization
n

r=1

r=1

r=1

= { P
(r ) ln P
(r ) P
(r ) r } 0 P
(r ) .
[P]

(B.22)

256

APPENDIX B.

(r ). [P]
is then to be maxHere 0 is the Lagrange multiplier enforcing normalization of the P
(r ) by setting the functional derivatives / P
equal to zero
imized with respect to the P

(r ) + 1) r 0
= 0 = (ln P

r = 1, 2, . . . n

(B.23)

along with
n

(r ) = 1 .
P

(B.24)

r=1

The best P (r ) are then


P (r ) =

exp (1 r )

(B.25)

where
n

Z = exp (1 s ) .

(B.26)

s=1

with still to be determined.


Taking the constant kB T the Thermal Lagrangian functional to be maximized with respect to
P (r ) is
n

L [P] = kB P (r ) ln P (r )
r=1

n
1 n
P (r ) r 0 P (r )
T r=1
r=1

(B.27)

which is identical to determining the equilibrium thermodynamic configuration by maximizing a


negative Helmholtz potential functional F [P] [see Subsection 5.1.3] subject to normalization of
P (r ). Therefore the entropy is
n

S = kB P (r ) ln P (r )

(B.28)

r=1
1

and with = (kB T )

exp (r )
,
Z

(B.29)

Z = exp (s ) .

(B.30)

P (r ) =
with
n
s=1

Appendix C

C.1

Eulers Homogeneous Function Theorem

Theorem: If f (x1 , x2 , x3 , . . . , xN ) is a function with the property


f (x1 , x2 , x3 , . . . , xN ) = n f (x1 , x2 , x3 , . . . , xN ) ,

(C.1)

it is said to be homogeneous of order n. Then, according to Eulers Homogeneous Function theorem:


N

xi (
i=1

f
) = nf (x1 , x2 , . . . xN ) .
xi

(C.2)

Proof: Differentiating Eq.C.1 with respect to


d
d n
f (x1 , x2 , x3 , . . . , xN ) =
f (x1 , x2 , x3 , . . . , xN ) .
d
d

(C.3)

N
d
f d (xi )
f (x1 , x2 , . . . , xN ) =
d

(x
d
i)
i=1

(C.4)

But

f
xi
i=1 (xi )

(C.5)

and
d n
f (x1 , x2 , x3 , . . . , xN ) = nn1 f (x1 , x2 , x3 , . . . , xN ) .
d

(C.6)

Now, setting = 1,
N

xi (
i=1

f
) = nf (x1 , x2 , . . . xN ) .
xi

257

(C.7)

258

APPENDIX C.

Appendix D

D.1

Occupation Numbers and The Partition Function

Explicit determination of the eigenstate degeneracy g(E),1 has been used in arriving at Eq.8.6.
In some circumstances this is too limited or inconvenient for calculating a partition function.
A different method is demonstrated by reconsidering the two level Schottky problem of the last
section. The macroscopic eigen-energies may be written with a different emphasis as
E {[

N
nN
n31
n21
n11
j
j
1
3 ];...[
2 ];[
1 ];[
N ]} = (1 n1 + 2 n2 )
n2
n2
n2
n2
j=1

(D.1)

where
[

nj1
]
nj2

(D.2)

represent j th atom occupation number pairs, which for the two level example of the previous
section where only 1 or 2 can be occupied on each atom are
(

nj1 = 1
)
nj2 = 0

and

nj1 = 0
)
nj2 = 1

(D.3)

The sum over j in Eq.D.1 is over all atom occupation pairs. Using Eq.D.1 the partition function
is

Z=
exp (1 nj1 + 2 nj2 )
(D.4)

j=1
1
2
N

n1 n1 n1

,
...

n1 n2
nN

2
2
2

n1

n2
1 Similar

(1 n1 +2 n2 )
e

to Boltzmanns microstate count, i.e. his classical idea of an internal energy degeneracy.

259

(D.5)

260

APPENDIX D.

where the remaining sum is over the occupation pair


[

n1
]
n2

1
)
0

and

0
)
1

(D.6)

i.e. there are only two terms, which gives


N

Z = (e1 + e2 )

(D.7)

Appendix E

E.1

Density of States Definition

The density of states D () of a quantum system is simply the number of states at the energy level
. This number is determined by sifting through all the quantum eigen-energies s of the system
and counting how many belong in the energy bin  = . If there is degeneracy among the
eigen-energies, e.g., 1 = 2 = . . . g = all g of these degenerate energies must find their way into
the energy bin .
Finding a density of states D () is a mathematical counting process defined by
D () = ( s )

(E.1)

where ( s ) is the Dirac delta function which undertakes the task of adding s to the number
of states accumulating in bin .
The Dirac delta function has the property of a well chosen unity in the sense that

1 = d ( s )

(E.2)

In calculating a density of states the eigen-energy subscript s is usually replaced by quantum


numbers that identify the eigen-energies, so that counting in Eq.E.1 can take place over quantum
numbers.

E.2

Examples

1. Consider a one-dimensional system of independent particles with only kinetic energy

 (k) =
261

2 k2
h
2m

(E.3)

262

APPENDIX E.
where
< k < .

(E.4)

Here s in Eq.E.1 is replaced by  (k), where k is the free particle quantum number and the
density of single particle states is written
D () = [  (k)]

(E.5)

= [
k

2 k2
h
]
2m

(E.6)

with the sum over s replaced by a sum over k.


The property of the Dirac-delta function that is used in density of states calculations is
(k k0 )

(k)k=k

[f (k)] =

d
f
dk

k0

(E.7)
0

where k0 denotes the zeros of f (k)


f (k0 ) = 0

(E.8)

and the sum in Eq.E.7 is over k0 , all the zeros of f (k).


In this one-dimensional case where
D () = [f (k)]

(E.9)

with
f (k) =

2 k2
h
,
2m

(E.10)

locating its zeros we find the pair

k01 = +

2m
2
h

(E.11)

2m
2 .
h

(E.12)

and

k02 =
Next, the denominator in Eq.E.7 is evaluated
2 k01
d
h
f (k)
=
dk
m
k=k01
2

d
h k02
f (k)
=

dk
m
k=k02

(E.13)
(E.14)

E.2. EXAMPLES

263

so that, finally, the density of states is written

(k k01 ) (k k02 )
+ 2
.
D () = 2

h
2m
h
2m

2
2

h
h
m
m

(E.15)

In the case of a macroscopically long line of particles the k values become closely spaced
so that in 1 D the sum becomes the integral

L
dk ,
2

(E.16)

1,2

where L is the line length.

The integral is easily evaluated to give


D () =

m
2
2h

The density of single particle states is customarily given per unit length,

D () 1
m
=
.

L
2h2
If there is spin degeneracy a factor g may be included,

D ()
1
m
=g
2 .
L
2h

(E.19)

(E.20)

(E.21)

where g = 2 in the case spin = .


2. In the case of the 3 D Debye model3
= cs k .

(E.22)

Therefore, since the zero of Eq.E.22 is

k0 = /cs ,
1 In

(E.23)

2D

with A the 2 D area.


2 In 3 D

A
(2)

dkx dky

(E.17)

dkx dky dkz


(2)3

(E.18)

with V the 3 D volume.


3 In this example the distinction between longitudinal (1 mode) and transverse (2 modes) phonons is ignored,
using instead an average speed of sound cs and a degeneracy g = 3.

264

APPENDIX E.
we can write

1
2

D () =

(k

/c
)
k
d
k
4
s

(2) 0 cs

(E.24)

to finally give
D () =

3V 2
2 2 cs 3

(E.25)

where the approximate mode degeneracy (1 longitudinal+ 2 transverse) is accounted for by


the inserted factor g = 3.

Appendix F

F.1

A Lab Experiment in Elasticity

Here we will explore the possibility of finding the microscopic interaction energy  for a particular
elastomer sample using macroscopic thermodynamic measurements. To examine this possibility
start by writing tension as a function of length L = Lz and temperature T , i.e. = (L, T ) .
The corresponding tension differential is
d = (

) dL + (
) dT.
L T
T L

(F.1)

If we perform an experiment in which a rubber bands length L is held fixed, the tension differential
becomes

d = (
) dT.
(F.2)
T L
Furthermore, applying a Maxwell relation
(

S
) = (
)
L T
T L

(F.3)

we can replace Eq. F.2 by


d = (

S
) dT
L T

(F.4)

which demonstrates that (S/L)T , a quantity in which we have some interest [see discussion
below], is the slope on a vs.T curve acquired at fixed elastomer length. Furthermore, consider the
Helmholtz potential F
F = U TS
(F.5)
and its derivative with respect to L at constant T
(

F
U
S
) =(
) T (
)
L T
L T
L T

(F.6)

Taking the total differential of Eq. F.5 and combining it with the first law we have
dF = SdT + dL
265

(F.7)

266

APPENDIX F.

from which we note


(

F
) =.
L T

(F.8)

Figure F.1: Experimental arrangement for measurement of vs.T. Add boiling water to the glass
tube or fill the tube with acetone and dry ice 78o C. Then allow the system to approach room
temperature.
Then we can rewrite Eq. F.6 as
S
U
) +(
)
L T
L T
which is equivalent to having integrated Eq. F.4.
= T (

(F.9)

According to Eq. F.9, if we fix the length of a rubber band and measure the tension while varying
temperature T, the slope of the curve at any temperature is (S/L)T while the intercept (at
T = 0) is (U /L)T , which is another quantity of interest. In fact comparing
T (

S
) ,
L T

(F.10)

obtained from the slope, with


U
) ,
(F.11)
L T
obtained from the intercept, it is possible to assess the relative importance of entropy vs. internal
energy to an elastomers elasticity. This is one motivation for the proposed experiment, although
not the prime objective. A possible apparatus is illustrated in Figure F.1.
(

Results and Analysis


1. Using the Vernier instrumentation and Logger Pro computer software, obtain best vs. T
linear fits to the data. Identify the values of the slope and the intercept.

F.1. A LAB EXPERIMENT IN ELASTICITY

267

2. Determine if the elasticity of this rubber band is entropy or internal energy driven?
3. Now we must see what the experiment can say about the parameter . According to the
models thermodynamics as expressed in Eqs. 10.78 and 10.80 of Chapter 11,
(

S
2 sinh (a )
) = kB { +
}
L T
a [e + 2 cosh (a )]

(F.12)

Verify this result.


Also, from Eqs. 10.78 and 10.81
(

2 sinh (a )
U
) =
L T
a [e + 2 cosh (a )]

(F.13)

Verify this result.


Now expand Eqs. F.12 and F.13 for small . The resulting series have only odd powers of .
Do the same thing with the Equation of State and use that result to eliminate from each of
the expansions obtained from F.12 and F.13. This gives
(

S
) kB [1 + 2e (1 + )] (L/N a2 )
L T

(F.14)

U
) 2 e (L/N a2 )
L T

(F.15)

Verify this result


and
(
Verify this result.
4. Evaluate Eqs. F.14 and F.15 in the limit  0. Compare these results with ideal gas counterparts, i.e. (S/V )T and (U /V )T .
5. To this order the ratio of Eq. F.14 and Eq. F.15 does not contain the length parameter L/N a2
and can therefore be used to estimate the interaction energy .
6. Use the experimental results to evaluate  and (L/N a2 ) . (You may have to use Maple or
Mathematica to solve the equations.)

268

APPENDIX F.

Appendix G

G.1

Static Magnetic and Electric Fields in Matter

Magnetic and electric fields and their response functions (magnetization and polarization) are generally defined, consistent with Maxwells Equations, as local averages over microscopic regions.
Except for special sample geometries and field orientations, these fields within matter :
1. Differ from external fields,
2. Are non-uniform inside matter and in the surrounding free space.
This can make thermodynamic calculations and experimental interpretation a non-trivial problem.1

G.2

Thermodynamic Potentials with Magnetic Fields

With magnetic work, as in Eq.11.10, the fundamental equation becomes

T S = U + pdV

1
H B dV .
4

(G.1)

The Helmholtz potential differential F is

F = SdT pdV +

1
H B dV .
4

(G.2)

1 Experimentalists strive to fabricate samples which conform to special geometries for which internal demagnetizing
fields are usually but not always negligible.

269

270

APPENDIX G.

The enthalpy differential H is

H = T S + V dp

1
B H dV .
4

(G.3)

1
B H dV .
4

(G.4)

The magnetic Gibbs potential differential G is

G = SdT + V dp

These results are based on average local fields (Maxwells equations) and are valid for any mechanism
of magnetization. But, as noted in Section 11.1.1, internal interactions and field-current interactions
as expressed in Schr
odingers Equation require additional considerations.

G.3

Uniform External Fields

Microscopic based theory and experiment can be properly compared only if magnetic models use
practical expressions for thermodynamic work in the case of matter being inserted into previously
uniform external fields (B0 , E0 ).2,3,4
To the experimentalist this means:
1. Constraining free currents that create the external field H0 to remain unchanged when matter
is inserted into the field, so that H = H0 ;
2. Constraining free charges that create the external field D0 to remain unchanged when matter
is inserted into the field, so that D = D0 .
Because magnetization M and polarization P are both zero in the absence of matter, the defining
constitutive relations for H and D ,
H = B 4M

(G.5)

D = E + 4P

(G.6)

specify that before the sample is introduced H0 = B0 and D0 = E0 . After the sample is introduced
B, H , E and D become the local average fields.
From this point we examine only magnetic contributions, the electric field case being formulated
by parallel arguments.
2 LD Landau, LP Pitaevskii, EM Lifshitz,Electrodymanics of Continuous Media, 2nd Edition, Elsevier Amsterdam (1984).
3 V. Heine, The Thermodynamics of Bodies in Static Electromagnetic Fields, Proc. Camb. Phil. Soc. 52,
546-552 (1956).
4 E. A. Guggenheim, Thermodynamics: An Advanced Treatment for Chemists and Physicists, North-Holland,
Amsterdam, 1967.

G.3. UNIFORM EXTERNAL FIELDS

271

For the uniform pre-existant field (prior to insertion of matter) B0 = H0 and we can construct the
identity
1
1
B0 B dV
H0 B dV = 0 ,
4
4
V

(G.7)

and use it to construct yet another identity for the difference between magnetic work with average
Maxwellian fields [see Eq.11.9] and magnetic energy of the uniform field before matter is introduced,
namely
1
1
H B dV
B0 B0 dV
4
4
V

1
1
= (H H0 ) BdV +
B0 (B B0 ) dV .
4
4
V

(G.8)

Using the constitutive equation relating B and H


(H H0 ) = (B B0 ) 4M

(G.9)

and
(H H0 ) = (B B0 ) 4M ,

(G.10)

where B0 and H0 are the prior existing fields, allows the right hand side of Eq.G.8 to be re-expressed
in terms of magnetization as
1
1
H B dV
B0 B0 dV
4
4
V

1
1
=
(H H0 ) B dV +
B0 [(H H0 ) + 4M] dV
4
4
V

1
1
=
(H H0 ) B dV +
B0 (H H0 ) dV + B0 M dV .
4
4
V

(G.11)

Applying
B =A

(G.12)

B = A

(G.13)

B0 = A0

(G.14)

B0 = A0 ,

(G.15)

where A and A0 are static magnetic vector potentials, the first term on the final line of Eq.G.11
is
1
(H H0 ) A dV
4
V

(G.16)

272

APPENDIX G.

and the second term is


1
( A0 ) (H H0 ) dV .
4

(G.17)

U V = V U + (V U )

(G.18)

Then, with the vector identity

1. The integral G.16 is transformed


1
(H H0 ) A dV
4
V

1
1
A (H H0 ) dV
[A (H H0 )]dV .
4
4
V

(G.19)

2. Similarly, the integral G.17 is transformed


1
( A0 ) (H H0 ) dV
4
V

1
1
A0 (H H0 ) dV
[A (H H0 )] dV .
4
4
V

(G.20)

Both resulting integrals, Eqs.G.19 and G.20, are zero since:


(a) The currents that create the external field are constrained to remain unchanged when
introducing the sample into the field. Therefore in the first terms of each of these integrals,
(H H0 ) = 0 = (H H0 ).
(b) The second term of each integral can be turned into a surface integral (Gauss Theorem)
on a very distant surface where static (non-radiative) fields fall off faster than 12 .
r
With these simplifications Eq.G.11 is rearranged to become
1
1
H B dV =
B0 B0 dV + B0 MdV .
4
4

(G.21)

The left hand side is the magnetic work with average local fields while the right hand side involves:
a. The total magnetic field energy before insertion of matter,
b. The energy associated with the magnetic material in the prior existing external field, B0 .
Since B0 is constant and M = 0 outside matter, the volume in the second term V covers
magnetized material only. Therefore, whether or not M is uniform,
M = MdV
V

(G.22)

G.3. UNIFORM EXTERNAL FIELDS

273

where M is the total magnetization. The final result is therefore


V
1
2
(B0 ) + B0 dM .
H B dV =
4
8

(G.23)

Applying Eq.G.1 the resulting thermodynamic relations are


T dS = dU + pdV B0 dM

(G.24)

T dS = dH V dp + M dB0

(G.25)

dF = SdT pdV + B0 dM

(G.26)

dG = SdT + V dp M dB0

(G.27)

each having the advantage of B0 being a simple independent variable rather than a function of
2
position. (The field energies V (B0 ) /8 have been included in the internal energy and hence into
the thermodynamic potentials, which are now distinguished by superscripted stars.)
But of course, there is no free lunch. The potentials F and G , as defined in Eqs.G.26 and G.27,
are not true free energies because they do not depend upon mechanism independent (physical)
average local fields inside matter.
Since the entire system is magnetized, contributions to B from the remaining magnetized matter
will contribute at each site. Moreover, F and G as defined here carry no information about
currents and fields inside matter. For example the kinetic energy in the hamiltonian cannot simply
be p2 /2m because it ignores consequences of the external field within matter. The kinetic energy
should be
H=

2
1
e
{p (A0 + A)} ,
2m
c

(G.28)

where A0 is an external vector potential and A is the internal vector potential whose source is the
currents induced by the external field and which interacts with internal currents via the interaction
hamiltonian A p.
Free energies, F and G of Eqs.G.26 and G.27 are, nevertheless, thermodynamic state functions
(with the mathematical properties discussed in Chapter 4) and can be used in the usual thermodynamic calculations. For example, from Eq.G.27 the magnetization M is given by
M = (

G
)
,
B0 p,T

(G.29)

as applied in Eq.11.53. But it is a magnetization that takes no account of any internal


fields.

274

G.3.1

APPENDIX G.

Thermodynamics With Internal Fields

The final term in the mean field decomposition


mop (i ) mop (i) = [mop (i ) mop ] [mop (i) mop ]
+mop (i ) mop + mop mop (i) mop mop

(G.30)

constitutes an internal magnetic contribution to the Gibbs potential


1
WMFA = Bint M
2

(G.31)

so that the Gibbs potential G becomes


= G 1 Bint M
G
2
zK

=G
M M,
2N

(G.32)
(G.33)

denotes the additional dependence on the effective field , Eq.11.97. Furthermore


where G
= dG
dG

zK
M dM
N

zK
M dM
N
zK
= SdT Md (B0 +
M)
N
= SdT MdB
= SdT MdB0

(G.34)
(G.35)
(G.36)
(G.37)

which means that as a result of the internal field the magnetization is now
M = (

G
) .
B T

(G.38)

Therefore, re-examining Eq.11.99 we see a proper role for B in the following way
1

(
ln ZM )
B
T
= N tanh ( B ) .

M=

(G.39)
(G.40)

Another internal field of practical importance is the so-called called demagnetization field. Demagnetization is defined through geometrically based factors, j (demagnetization factors), which
describe a uniform field Hint,j inside an ellipsoid arising from magnetization
Hint,j = H0,j 4j,k Mk ,

(G.41)

with j = x, y, z. Formally, its source is the fictitious, classical magnetic surface poles arising from
the applied field H0 . Demagnetization factors (actually tensors) j,k depend on sample geometry

G.3. UNIFORM EXTERNAL FIELDS

275

and are obviously responsible for magnetic anisotropy. Calculating demagnetization factors for
different geometries remains a on-going activity in the field of practical magnetics.5
Applying the constitutive equation
Hint,j = Bint,j 4Mj

(G.42)

to Eq.G.41, we have (ignoring, for simplicity, the tensor character of ) the effective field
Bint,j = B0,j + 4 (1 ) Mj .

(G.43)

5 M. Beleggia, M. DeGraff and Y. Millev, Demagnetization factors of the general ellipsoid, Phil. Mag 86, 2451
(2006).

276

APPENDIX G.

Appendix H

H.1

Maxwells Equations and Electromagnetic Fields

Electric fields E (x, t) and magnetic fields B (x, t) in free space satisfy the differential Maxwell
relations1
E(x, t) = 0

(H.1)

B(x, t) = 0

(H.2)

1 B (x, t)
E(x, t) =
c
t
1 E (x, t)
B(x, t) =
c
t

(H.3)
(H.4)

Unlike Schr
oedingers wave mechanics, which retains some suggestion of classical particle dynamics, a
quantum theory of radiation is guided by Maxwells field relationships and Plancks photon postulate
which Dirac used in his theory of quantum2 fields.

H.1.1

Electromagnetic Waves

In classical theory time varying electric and magnetic fields propagate in free space as waves with
velocity c. This can be seen by combining the last of Maxwells equations, Eq.H.4, with Maxwells
first equation, Eq. H.1 to give the wave equation for an electric field
2 E (x, t) =

1 2 E (x, t)
.
c2
t2

(H.5)

Similarly, Eqs.H.2 and H.3 are combined for a corresponding wave equation for the magnetic
field
2 B (x, t) =

1 2 B (x, t)
.
c2
t2

1 In

(H.6)

rationalized c.g.s. units.


A. M. Dirac, Quantum Theory of Emission and Absorpstion of Radiation, Proc. Roy. Soc. (London),
A114, 243 (1927)
2 P.

277

278

APPENDIX H.

H.1.1.1

Solution in Free Space Periodic Boundary Conditions

Imagine a cube with side L and volume V = L3 within which a solution to Eq.H.5 is assumed
consisting of the sum of propagating waves

1
d
ikx it
Ej (x, t) =
ej (k, ) e e
V 2 k

(H.7)

where Ej is the j th component of the Maxwell electric field vector (j = x, y, z) and ej is, correspondingly, the Fourier vector component of the kth propagating wave amplitude. The vector k
in Eq.H.7 defines the spatial direction in which that particular Fourier component propagates (its
direction of propagation.) Each propagation vector k is identified as a mode index.
The solutions within the notional cubic volume are supplemented by sensible (but synthetic) boundary conditions according to which translation in any direction x, y or z by the finite side length L,
results in exactly the same electric fields, i.e.

d
1
ik(x+L) it
e
= Ej (x, t) .
Ej (x + L, t) =
ej (k, ) e
V 2 k

(H.8)

These are called periodic boundary conditions. They are used in nearly all field theories, classical
and quantum. When the cube is finally made large enough to really look like infinite free space it
hardly matter what goes on at the boundaries since the cubes surface to volume ratio is rapidly
diminishing to zero.
But in order for Eq.H.8 to hold, the following restrictions on k must apply:
kx L = 2nx ;

nx = 0, 1, 2, . . .

(H.9)

ky L = 2ny ;

ny = 0, 1, 2, . . .

(H.10)

kz L = 2nz ;

nz = 0, 1, 2, . . .

(H.11)

i.e., under periodic boundary conditions the propagation vectors can have only discrete positive
and negative values
kj =

2nj
.
L

(H.12)

However, as can be seen from Eq.H.12, in the limit of a large volume, i.e. L , the spacing
between successive values of kj becomes infinitesimal. This important result allows the sums in
Eq.H.19 (and other similar sums) to be treated as integrals.
H.1.1.2

Properties of Wave Solutions

Substituting Eq.H.19 into the wave equation Eq.H.5

d
1
d
1
ikx it
ikx it

2
e
(k,
)
e
e

=
j
[k k] ej (k, ) e e

2
2
V

k
k
V

(H.13)

H.1. MAXWELLS EQUATIONS AND ELECTROMAGNETIC FIELDS

279

and
1 2
c2 t2

d
d
1
2
1
ikx it
ikx it

=
ej (k, ) e e
[ 2 ] ej (k, ) e e

2 k
2 k
c
V
V

(H.14)

showing that Eq.H.19 is a wave equation solution only if


k2 =

2
c2

(H.15)

where
k 2 = kx 2 + ky 2 + kz 2

(H.16)

i.e., the magnitude of the propagation vector k is proportional to the frequency .


Of course, the solution must also satisfy the first of Maxwells equations, i.e. Eq. H.1, which also
has important implications. Taking the divergence of Eq.H.19 and setting it equal to ZERO

d
1
ikx it
=0
E =
[kx ex (k, ) + ky ey (k, ) + kz ez (k, )]e e
V 2 k

(H.17)

from which is concluded


k e (k, ) = 0

(H.18)

i.e., the fourier vector amplitude e (k, ) must be perpendicular to the propagation vector k. The
fourier vector amplitude e (k, ) can then be expressed in terms of two mutually perpendicular unit
vectors, (k, 1) and (k, 2) a pair for each mode k both of which are perpendicular to the
propagation vector k, i.e. (k, ) k = 0 where = 1, 2. These two transverse unit vectors are called
electric field polarization vectors.3
In terms of unit polarization vectors the electric field is written

d
1
ikx it
Ej (x, t) =
(k, ) ej, (k, ) e e
V 2 k

(H.19)

=1,2

where ej, is the projection of the j th cartesian component (j = x, y, z) of the field vectors kth
fourier component along the th polarization vector direction (k, ). [See Fig H.1.]
Each photon therefore requires two identifiers:
1. k, the mode
2. (k, ), the polarization ( = 1 or 2).
3 Vibrational modes in solids (phonons) have no governing relation similar to Maxwells first law, and so maintain
three independent polarization directions.

280

APPENDIX H.

Figure H.1: Propagation vector k and two transverse polarization unit vectors (k, 1) and (k, 2).
Each cartesian component of the fourier vector amplitude ej (k, ) can be expressed as a linear
combination of (k, 1) and (k, 2).
H.1.1.3

Electromagnetic Vector Potential

The energy eigenvalues of an electromagnetic field in free space are expressed in terms of photons
particle-like excitations first hypothesized by Planck. Quantum electrodynamics, the 20th century
refinement of Plancks theory, describes photons as quantum excitations of the vector potential field
A (x, t), where
B (x, t) = A (x, t)

(H.20)

and
E (x, t) =

1
A (x, t)
c t

(H.21)

where is a scalar potential. Because the vector potential is not unique, gauge fixing is required.
One common choice is
A (x, t) = 0
which is called the Coulomb gauge4 or, alternatively, the transverse gauge since A A .5
Coulomb gauge and in the absence of sources we can take 0.

(H.22)
6

In the

In this gauge the classical wave equation becomes


2 A
4 Because

2 A
=0
t2

Coulombs law is satised in this gauge.


gauge restricts the vector potential to have only transverse components.
6 This gauge has the same consequence as Maxwells first law, i.e. Coulombs Law.
5 This

(H.23)

H.1. MAXWELLS EQUATIONS AND ELECTROMAGNETIC FIELDS

281

The (transverse) vector potential can be expanded in a Fourier series of plane waves
A (x, t) = (
k
=1,2

1
) [
(k, ) ak, (t) eikx + (k, ) ak, (t) eikx ]
2V k

(H.24)

where ak, and ak, are complex coefficients, k = c k is the vacuum electromagnetic dispersion relation and the coefficient (2V k )

is chosen with an eye towards simplifying the final result.

It follows from the Coulomb gauge condition that


k (k, ) = 0 ,

= 1, 2

(H.25)

where the unit polarization vectors themselves are chosen to be mutually perpendicular and therefore satisfy the orthogonality condition
(k, ) (k, ) = ,

(H.26)

The classical (transverse) electric field is from Eqs.H.21 and H.24 (ignoring circular or elliptical
polarization)
E (x, t) =

1/2
ak, (t) ikx ak, (t) ikx
1
1
)

(k,
)
[
e
+
e
]
(

c k 2V k
t
t

(H.27)

=1,2

However by inserting the expansion Eq.H.24 into the wave equation Eq.H.23 the following simplifying results are obtained,
ak,
= ik ak,
t

(H.28)

and
ak,
t

= ik ak,

(H.29)

so that
E (x, t) =

i
k 1/2
) (k, ) [ak, (t) eikx ak, (t) eikx ] .
(
c k 2V

(H.30)

=1,2

The transverse magnetic field is, similarly,


B (x, t) = A
=i (
k
=1,2

1/2
1
) [k (k, )] [ak, (t) eikx ak, (t) eikx ]
2V k

(H.31)

282
H.1.1.4

APPENDIX H.
Quantized Electromagnetic Hamiltonian

The field theoretic canonical quantization procedure postulates the complex coefficients in the
expansions for Eqs.H.30 and H.31 morph into Fock space operators
ak, Z aop (k, )

ak, Z aop (k, )

(H.32)
(H.33)

where the operators fulfill equal time commutator relations


[aop (k, ) , aop (k , )] = k,k ,
[aop (k, ) , aop (k , )] = 0

[aop (k, ) , aop (k , )] = 0

(H.34)
(H.35)
(H.36)

where
aop (k, ) aop (k, ) = nop (k, )

(H.37)

with nop (k, ) the photon number operator. Now we have the vector potential operator Aop (x, t)
Aop (x, t) = (
k
=1,2

1
) (k, ) [aop (k, )eikx + aop (k, )eikx ] ,
2V k

(H.38)

the electric field operator Eop (x, t)


Eop (x, t) =

k
i
) (k, ) [aop (k, )eikx aop (k, )eikx ]
(
c k 2V

(H.39)

=1,2

and a magnetic field operator Bop (x, t)


Bop (x, t) = i (
k
=1,2

1
) [k (k, )] [aop (k, ) eikx aop (k, ) eikx ] .
2V k

(H.40)

The electromagnetic hamiltonian is


EM
Hop
=

1
2
2
dx [Eop (x, t) + Bop (x, t)] .
2

(H.41)

Now substituting Eqs.H.39 and H.40, carrying out the required integrations, using the commutator
relations and applying the results
1
ikx ikx
= k+p,0
dx e e
V

(H.42)

(k, ) (k, ) = (k, ) (k, )

(H.43)

= ,

(H.44)

H.1. MAXWELLS EQUATIONS AND ELECTROMAGNETIC FIELDS

283

and
[k (k, )] [k (k, )] = [k (k, )] [k (k, )]
2

= k ,

(H.45)
(H.46)

we find
EM
Hop
=

k [aop (k, ) aop (k, ) + aop (k, ) aop (k, )]


2 k

(H.47)

=1,2

1
= k [aop (k, ) aop (k, ) + ]
2
k

(H.48)

1
= k [nop (k, ) + ]
2
k

(H.49)

=1,2

=1,2

284

APPENDIX H.

Appendix I

I.1

Fermi-Dirac Integrals

The typical Fermi-Dirac integral in the theory of metals and semi-conductors has the form

I F D () = ds ()D ()
0

1
e() + 1

(I.1)

where D () is the density of states and s () is some function of , often a fractional or integer
power, such that D () s () is analytic (slowly varying) .
Analytic evaluation of these integrals is not possible so several approximations and numerical methods have appeared over the years.1,2,3,4
The Sommerfeld asymptotic expansion, which is the earliest published result5 exploits the approximate property that
d
1
1
[ ()
] ( )
d e

+1

(I.2)

A logical first step in Sommerfelds method is to integrate Eq.I.1 by parts to get


I

F D

() = d d s ( ) D ( ) sech2 [ ( )]

4
2

(I.3)

where for 1 the function sech2 [ 2 ( )] peaks sharply at . Now, Taylor expanding the
1 W. B. Joyce and R. W. Dixon, Analytic approximations for the Fermi energy of an ideal Fermi gas, Appl.
Phys. Lett. 31, 354 (1977).
2 A. Wasserman, Fermi-Dirac Integrals, Phys. Letters 27A, 360 (1968).
3 T. M. Garoni, N. E. Frankel, and M. L. Glasser, Complete asymptotic expansions of the FermiDirac integrals,
J. Math. Phys., 42, 1860 (2001).
4 Raseong
Kim and Mark Lundstrom, Notes on Fermi-Dirac Integrals (3rd Edition), (2008)
[http://nanohub.org/resources/5475].
5 A. Sommerfeld, Zur Elekronentheorie der Metalle auf Grund der Fermischen Statisik, Z. Phys. 47, 1 (1928).

285

286

APPENDIX I.

bracketed integral

d s ( ) D ( )

(I.4)

about = ,

d s ( )D ( )

0
0
=

1
2 d
d s ( )D ( )
+ ( )
+ ...

2
d 2

0
=

d s ( ) D ( ) = d s ( )D ( ) + ( )
0

d
d

= d s ( )D ( ) + ( ) s () D ()
0

d
1
2
+ ...
+ ( ) { [s ( ) D ( )]}
2
d
=

(I.5)

After substituting z = ( ) we have the approximate result

z
4I F D () = d s ( )D ( ) dz sech2 ( )
2
0

z
+ s () D () dz z sech2 ( )
2

d
z
1
{ [s ( ) D ( )]}
dzz 2 sech2 ( ) + . . .

2
2 d
2
=

(I.6)

where in the degenerate case ( >> 1) lower limits on the integrals have been extended,
. Since the function sech2 ( z2 ) peaks sharply at z = 0, this introduces negligible error.
Evaluating the integrals:

2 z
dz sech ( ) = 4
2

(I.7)

2 z
dz z sech ( ) = 0
2

(I.8)

4 2
2
2 z
dz z sech ( ) =
2
3

(I.9)

3
2 z
dz z sech ( ) = 0
2

(I.10)

I.1. FERMI-DIRAC INTEGRALS

287

we have the approximate result

F D

() = d s ( )D ( ) +
0

d
2
{
[s ( ) D ( )]}
+ ...
6 2 d
=

(I.11)

288

APPENDIX I.

Appendix J

J.1
J.1.1

Bose-Einstein Integrals
Integrals When = 0

In the Bose-Einstein (B-E) problem we encounter two kind of integrals. The first appears when
T < Tc so that = 0. These have the form

IBE = dx
0

xs
.
ex 1

(J.1)

They can be done by first expanding the denominator

dx
0

xs
s
e(n+1)x
dx
x
=

ex 1
n=0

(J.2)

xs
1
=

(s
+
1)

s+1
ex 1

=1

(J.3)

and then integrating term-by-term

dx
0

where the -function is defined by

(r) = dz z r1 ez

Re (r) > 1 .

(J.4)

The remaining sum is the Riemann -function

1
s
n
n=1

(s) =

289

(J.5)

290

APPENDIX J.

which finally gives

dx
0

J.1.2

xs
= (s + 1) (s + 1) .
ex 1

(J.6)

Integrals When < 0

The second class of integrals is best demonstrated by the specific example

BE
Nop
=

1
V
2m 3/2
) d 1/2 ( ()
(
)
2
2

h
4
e
1

(J.7)

V
2m 3/2
e
= 2 ( 2 ) d 1/2 (
)
h
4
1 e

(J.8)

where = e and therefore e < 1. Expanding the denominator in a binomial series

BE
Nop
=

2m 3/2
V
1/2

n en
d
e
(
)

2
4 2 h
n=0

(J.9)

and integrating term by term


BE
Nop
=

V
2m 3/2 (3/2)
2
3
( 2 ) [ 3/2 ] ( + 3/2 + 3/2 + )
2
4
h

2
3

(J.10)

2m 3/2 (3/2)
V
(
2 ) [ 3/2 ] (3/2, )
4 2 h

(J.11)

where
(s, ) = +

2 3
+
+
2s 3s

(J.12)

is a generalized -function. The function as studied by mathematicians (and in the computational


arsenal of, say, Mathematica) is Lerchs -function (also called Hurwitz-Lerch -function) defined
as

zn
s
n=0 (n + a)

(z, s, a) =

(J.13)

where any term for which the denominator is zero is excluded. But the generalized -function
appearing in Bose-Einstein integrals is

zn
s
n=1 n

(s, z) =

(J.14)

which is related to Lerchs -function by


(s, z) = z (z, s, 1) .

(J.15)

J.1. BOSE-EINSTEIN INTEGRALS

291

Therefore
BE
Nop
=

V 2m 3/2 (3/2)
( 2 ) [ 3/2 ] (, 3/2, 1) .
4 2 h

(J.16)

Using Mathematica it is a simple matter to plot (3/2, ) = (, 3/2, 1) vs. [see Figure J.1].

Figure J.1: (3/2, ) vs. .


It is useful to note that (3/2, ) when << 1 so that
BE
Nop

which is the quasi-classical result.

2m 3/2
(
2 ) (3/2)
4 2 h

(J.17)

Index
adiabatic
definition, 25
demagnetization, 156
expansion, 40
sudden process as, 33
throttling, 71
antiferromagnetism, 144
bias in probabilities
define, 255
entropy, 256
lagrange multipliers, 255
measure of, 253
minimization of, 255
Boltzmann, x
constant, 27
Microcanonical theory, 83
Bose-Einstein
condensate, 233, 237
condensate wave function, 239
condensation, 236
grand potential, 234
integrals, 289
internal energy, 235
occupation number, 235
thermodynamics, 235
bosons, 97
Carnot, ix
cycle, 63
engine, 59
chemical potential, 4, 5, 24, 26, 173
chemical reactions, 177
Clausius, ix, x, 2, 79
coherent, 8
cosmic microwave background (CMB), 243
Debye

lattice model, 120


density of states, 263
lattice heat capacity, 122
Debye temperature, 122
decoherence, 16, 80
degeneracy, 86, 88
1-D harmonic oscillator, 93
elastomer, 91
hydrogenic, 92
internal quantum, 88
macroscopic, 88
demagnetization
factor, 274
field, 274
tensor, 274
density of states
1-D crystal, 261
definition, 261
density operator, 6
definition, 11
density matrix
examples, 18
equation of motion, 13
information and the, 11
mixed state, 14
properties, 11
pure state, 11, 13
thermal
defined, 16
diamagnetism, 143
diathermic, 37
differential
exact, 26
differentials, 47
exact
Eulers criterion, 48
Greens theorem, 49
292

INDEX
Dirac
delta-function
definition, 220
property, 262
notation, 7
Einstein
1-D model, 114, 116
partition function, 116
3-D model, 117
Einstein temperature, 119
heat capacity, 119
internal energy, 118
partition function, 118
historical note, 2, 37
elasticity
1-D model, 125
3-D non-ideal, 136
elongation, 6, 24
entropy, 131
equation of state, 130
extension, 34
Lab Experiment, 265
lab experiment introduction, 139
non-deal equation of state, 138
partition function, 129
tension, 4, 24, 27
thermodynamics of, 131
elastomer, 27, 90
degeneracy, 91
electric
constitutive relation, 270
field, 4, 27
fields in matter, 269
polarization, 24, 27
nonlinear, 28
susceptibility
tensor, 27
electromagnetic
hamiltonian
quantized, 282
Maxwells equations, 277
polarizations, 280
vector potential, 280
wave equation
classical, 280

293
waves, 277
energy
microscopic system, 9
ensembles, x
enthalpy, 51
definition, 37, 69
Maxwells relation, 69
entropy, x, 17, 18, 55, 57, 79, 85
definition, 50
of mixing, 104
equation of state, 26
dielectric, 27
equilibrium, 174
thermodynamic
definition, 3
Euler
Homogeneous function theorem, 51
Eulers
theorem
proof, 257
exact differentials, 48
exoplanet surface temperature, 210, 212
Fermi energy, 218
Fermi gas model, 220
Fermi-Dirac
average particle number, 218
chemical potential, 221
temperature dependence, 222
electron gas model, 224
free electron approximation, 216
function, 218
grand partition function, 217
grand potential, 218
heat capacity, 223
integrals, 285
Sommerfeld expansion, 285
internal energy, 218
temperature dependence, 222
pressure
equation of state, 220
single particle average occupation number,
218
thermodynamics, 218
fermions, 97
ferrimagnetism, 144

294
ferromagnetism, 144
critical exponent, 160
Curie-Weiss susceptibility, 161
Gibbs potential, 159
Heisenberg exchange, 157
Heisenberg model, 157
order parameter, 159
field
energy, 273
external, 270
external magnetic, 272
internal, 274
local average, 272
magnetic
effective, 274
mean, 274
fields
demagnetizing, 269
First Law of Thermodynamics, 3
heat, 23
definition, 25
incremental, 26
work
definition, 23, 24
force, 23
free expansion, 53
functional, 82
gamma function, 238
gas law, 27
ideal, 27
derivation, 98
non-ideal
van der Waals equation, 104
pressure, 27
temperature, 27
van der Waals, 55
Gibbs, x
canonical theory, 86
correction, 99
entropy paradox, 103
partition function, 86
potential
magnetic, 274
Gibbs potential
definition, 71

INDEX
Maxwells relation, 71
Gibbs-Duhem equation, 181
grand canonical
entropy, 167
Eulers theorem, 168
ideal gas example, 169
internal energy, 166
particle number, 166
potential function, 167
pressure, 166
thermodynamics, 166
grand partition function, 165
heat, 3
heat capacity, 35
constant pressure, 36
constant volume, 35
Dulong-Petit, 113
entropy, 74
Helmholtz potential
canonical, 88
definition, 70
Maxwells relation, 70
ideal gas, 97
barometric equation, 175
charged, 175
entropy
Sakur-Tetrode equation, 102
heat capacity, 102
Helmholtz potential, 103
in earths gravitational field, 174
in rotating cylinder, 176
internal energy, 101
pressure
equation of state, 103
information, 10, 15
missing, 21
purity, 13
Shannon, 80
internal energy, 3, 4, 8, 17, 25, 33, 54
blackbody
moving frame, 247
Maxwells relation, 67
isenthalpic, 50
isochoric, 36
isotherm, 32

INDEX
isothermal, 29, 34
law of mass action, 178
line integration, 55
macroscopic, x
averages, 3
definition, 2
mixed state, 13
magnetic
constitutive relation, 270
field
internal, 274
fields in matter, 269
thermodynamic potentials, 269
magnetism, 143
constitutive relation, 147
de Haas-van Alphen effect, 144
enthalpy, 146
fundamental equation, 146
fundamental equations, 148
Gibbs potential, 146
Helmholtz potential, 146
magnetic field, 28
magnetic fields, 4
magnetic moment, 5
magnetic susceptibility, 28
magnetic work, 144, 147
magnetization, 24
magnetization vector, 148
paramagnetic partition function, 151
work in applied fields, 148
magnetization, 273
fluctuations, 155
Mattis, ix, xi
Maxwell, x
thermodynamic relations, 48
mean field approximation, 274
ferromagnetism, 158
gravitational interaction, 185
microstates, x
minimum bias, 82
negative temperature, 110
normalization
state function, 9

295
open system, 5
open systems, 173
operator
density
thermodynamic, 80
electromagnetic
quantized, 282
electromagnetic hamiltonian
quantized, 282
elongation, 4
function of, 20
hamiltonian, 7
hermitian, 8, 9
definition, 8
magnetization, 4
particle number, 4, 5
polarization, 4
pressure, 4
quantum hamiltonian, 4
trace
definition, 12
vector potential, 4
paramagentism
Gibbs potential, 153
paramagetism
internal energy, 152
paramagnetism, 143, 149
Curie regime, 152
entropy, 153
heat capacity, 153
thermodynamics, 152
partial differentiation, 47
chain rules, 52
cyclic chain rule, 52
ordinary chain rule, 52
partition function, 86
blackbody moving frame, 246
ideal gas, 99
Pauli Exclusion Principle, 215
Pauli exclusion principle, 97
Pauli susceptibility, 223
periodic boundary conditions, 225
phonons, 115
Planck
historical note, 2

296
Planck radiation function
CMB, 249
polymer
rubber, 27
rubber band, 33
Poynting vector
blackbody
moving frame, 248
pressure, 24
purity, 17, 19
quantum
average value, 9
eigenfunctions, 8, 9
eigenvalues, 8, 9
energy expectation value, 9
expectation value, 7, 9
expectation values, 7
fluctuations, 7
hamiltonian, 5, 7
microscopic, 6
mixed state, 13
probability, 8
probability amplitude, 8
probablities, 7
pure state, 10
state function, 7
pure, 11
superposition theorem, 8
wavefunction, 6, 9
quantum concentration, nQ , 101
radiation thermodynamics
moving frame, 246
Schottky 2-level model, 107
heat capacity, 109
Schr
odinger
equation, 6
Schwartz inequality, 251
Second Law of Thermodynamics, 79
specific heat, 35
stars
formation of, 184
giant molecular clouds, 184
gravitational collapse, 187
gravitational equation of state, 187

INDEX
state function
normalization, 12
Stern-Gerlach, 19
strain, 24, 38
definition, 27
stress, 24, 38
definition, 27
surface adsorption, 191
dissociative, 193
Langmuir isotherm, 193
Langmuir model, 191
temperature, 4, 25
tension
surface, 24
thermal expansivity, 39
thermal lagrangian
Bose-Einstein, 234
CMB, 244
definition, 256
Einstein 1-D model, 116
elastic, 127
elasticity, 129
Fermi-Dirac, 216
microcanonical theory, 83
open system, 165
paramagnetic, 151
radiation, 201
Schottky 2-level model, 108
semi-classical ideal gas, 99
thermal radiation
entropy, 207
hamiltonian
classical, 199
Helmholtz potential, 203
momentum density, 210
partition function, 201
Planck constant, 199
Plank quantization, 199
stardust blow-out, 211
Stefan-Boltzmann law, 204
thermodynamics of, 201
Wiens law, 206
thermodynamic identity, 50
thermodynamics
equilibrium states, 29

INDEX
fluctuations
definition, 3
surroundings, 3
definition, 3
system
definition, 3
throttling, 71
van der Waals
derivation of equation, 188
equation, 105, 191
historical note, 2
variable
conjugate pairs, 24
extensive, 24
intensive, 24
state, 26
variables
thermal, 3
white dwarf star
mass-radius constant, 229
white dwarf stars, 227
work, 3, 23
electric, 147
irreversible, 30, 31
magnetic, 147
quasi-static, 31, 34
definition, 28
reversible, 29
Youngs modulus, 39, 126
zeta function, 238

297

You might also like